Nucl - Phys.B v.785

Nuclear Physics B 785 (2007) 118
The strong coupling limit of the scaling function from

the quantum string Bethe ansatz
P.Y. Casteill a , C. Kristjansen b,
a The Niels Bohr Institute, Copenhagen University, Blegdamsvej 17, DK-2100 Copenhagen , Denmark
b The Niels Bohr Institute and NORDITA, Copenhagen University, Blegdamsvej 17, DK-2100 Copenhagen , Denmark
Received 25 May 2007; accepted 18 June 2007

Available online 27 June 2007
Abstract
Using the quantum string Bethe ansatz we derive the one-loop energy of a folded
string rotating with
angular momenta (S, J ) in AdS3 S 1 AdS5 S 5 in the limit 1 J S, z = log(S/J )/( J ) fixed.
The one-loop energy is a sum of two contributions, one originating from the HernandezLopez phase and
another one being due to spin chain finite size effects. We find a result which at the functional level exactly
matches the result of a string theory computation. Expanding the result for large z we obtain the strong
coupling limit of the scaling function for low twist, high spin operators of the SL(2) sector of N = 4 SYM.
3 log(2)
In particular, we recover the famous . Its appearance is a result of non-trivial cancellations between
the finite size effects and the HernandezLopez correction.
2007 Elsevier B.V. All rights reserved.
PACS: 11.15.-q; 11.15.Me; 11.25.Tq
Keywords: Cusp anomalous dimension; Scaling function; Strong coupling expansion; Bethe equations; AdS/CFT
correspondence
1. Introduction
Due to recent years discovery of integrable models underlying the spectral problems of both
N = 4 SYM [1] and type IIB string theory on AdS5 S 5 [2] the spectral part of the AdS/CFT
conjecture [3] can now be stated in a very pointed manner. Namely, the conjecture simply says
* Corresponding author.
E-mail address: kristjan@nbi.dk (C. Kristjansen).

0550-3213/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2007.06.011
P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118
that the S-matrix of the respective integrable models must agree [4]. Furthermore, the common
symmetry group of the two theories constrains the S-matrix up to a phase factor [5]. The formulation of the conjecture can thus be further sharpened to the statement that the phase factors of
respectively N = 4 SYM and type IIB string theory on AdS5 S 5 should be identical.
Based on educated guessing, phase factors for both N = 4 SYM and type IIB string theory
on AdS5 S 5 have been put forward. In accordance with the strong-weak coupling nature of the
AdS/CFT correspondence the gauge theory phase factor [6] is given as an infinite series in the
t Hooft coupling
constant whereas the string theory phase factor [7] is given as an asymptotic
expansion in 1/ . There exist arguments that the string theory asymptotic expansion for large
can originate from the same function as defined by the gauge theory perturbative expansion
which has a finite radius of convergence [6]. However, both phase factors are rather involved
functions and it would be reassuring to see an example of a simple observable which can be
extrapolated smoothly from weak to strong coupling.
A candidate for such an observable is the universal scaling function or cusp anomalous dimension, f (g) where g 2 = 8 2 . It is related to the anomalous dimension of low twist operators
of N = 4 SYM of the type

O = Tr D S Z J + .
(1)
Here D is a light cone covariant derivative, Z is a complex scalar, S is the spacetime spin
and J is denoted as the twist. For leading twist, i.e. J = 2, it is well known that the anomalous
dimension of such an operator for large values of the spin grows logarithmically with the spin
S = f (g) log(S),
S ,
(2)
where f (g) can be expanded perturbatively in g. The scaling function has the appealing feature
that, as opposed to other observables one could think of, it depends only on one parameter g.
For instance, it is not polluted by any additional J -dependence. The function f (g) has been determined by solid field theory calculations up to and including four-loop order [8]. Furthermore,
starting from the asymptotic gauge theory Bethe equations [9], inserting the conjectured gauge
theory phase factor [6] and taking a large-S limit it has been possible to derive an equation which
determines f (g) to all orders in g [10]. This equation, known as the BES equation, correctly
reproduces the known first four orders in g 2 . Its derivation, however, relies on the assumption
that the scaling function is the same for all operators with a finite value of the twist and that at
the same time it is permitted to take J sufficiently large so that the asymptotic Bethe equations
are correct.
On the string theory side a low twist, high spin operator corresponds to a folded string rotating with angular momentum S on AdS3 AdS5 S 5 [11]. The energy of such a string has an
expansion for large which reads

3 log(2)
1
E=
(3)
+O
log S, S .
Here the first term follows from semi-classical analysis [11] and the second one from a one-loop
computation [12]. Deriving this result from the Bethe equations would yield a very comforting
confirmation of both the integrability approach as well as of the AdS/CFT conjecture itself.
However, the strong coupling analysis of the BES equation has proved hard. For the moment
only the leading semi-classical contribution has been derived from the BES equation by analytic
means [1315]. By numerical analysis of the equation both the leading [16,17] and the next to
leading order term [16] can be reproduced with high accuracy. Furthermore, it is possible to
predict numerically the next term in the expansion which would result from a string theory twoloop computation [16]. In the present paper we shall consider an alternative way of obtaining an
expansion la (3) by Bethe equation techniques.
An operator of the type (1) for which J is not finite has a string theory dual which in addition
to the angular momentum, S on AdS3 carries an angular momentum J on S 1 S 5 . For such a
string, considering the situation

S
1 J S,
z
(4)
log
, fixed,
J
J
one finds that the semi-classical [18] as well as the one loop energy [19] can be written down
in a closed form as a function of z. Furthermore, the formula obtained for the string energy
interpolates smoothly between small and large values of z and the large-z expansion looks as (3)
just with the replacement log S log( JS ). We shall discuss this string solution and the various
expansions of its energy in more detail shortly. Subsequently, we will show how to reproduce
the precise functional dependence of the string energy on z from the string Bethe equations. In
particular, we will derive by analytic means the celebrated 3 log(2)
. Our starting point will be
the asymptotic Bethe equations, whose application is now more justified since we take J 1,
supplemented with the conjectured string phase factor. The classical string energy as a function
of z is obtained almost immediately by considering only the AFS phase [20] whereas the oneloop energy requires more work. For one we have to take into account the HL-correction [21] to
the phase and secondly we have to consider spin-chain finite size effects [22]. As we shall see
we are able to determine the contribution from each of these effects exactly as a function of z.
The 3 log(2)
results from a non-trivial cancellation between the two types of terms as z .
We start in Section 2 by recalling from reference [19] the description of the folded string
rotating on AdS3 S 1 AdS5 S 5 in the limit given by Eq. (4). In Section 3 we write down the
relevant string Bethe equations and perform the necessary expansions. After that, in Sections 4
and 5, we extract from these respectively the semi-classical and the one-loop energy. Finally,
Section 6 contains our conclusion.
2. The folded string spinning on AdS3 S 1
A folded string living in AdS5 S 5 and carrying large angular momenta S and J on respectively AdS3 and S 1 is a system which has successfully been studied in the semi-classical
approximation. Hence, its classical energy was determined in [12]. The expression for the energy simplifies considerably in the limit given in Eq. (4), i.e. [18]

S
1 J S,
z
(5)
log
, fixed.
J
J
One finds

E0 = S + J 1 + z2 .
Expanding for large z we get

S
E0 (z 1) = S +
log
+ .
(6)
(7)
Here we notice the leading strong coupling term announced earlier, cf. Eq. (3). For z 1 one
recovers what is known as the fast spinning string solution [23]

2
2 S
4 S
log
E0 (z 1) = S + J +
log
J
J
2 2 J
8 4 J 3

3
S
(8)
log6
+ .
+
J
16 6 J 5
The first logarithmic term was reproduced in the Bethe ansatz approach in [24] and the second
one is contained in the work in [18]. Later, we shall show that when the limit (5) is imposed from
the beginning in the all loop Bethe ansatz, the exact square root formula immediately appears.
Recently, an expression for the one-loop contribution to the energy in the same limit was derived
[19]. The result reads

J
1
z 1 + z2 1 + 2z2 log z + 1 + z2
E1 =
1 + z2

z2 + 2 1 + z2 log 1 + z2 1 + 2z2 log 1 + 2z2 .
(9)
It is obtained under the further assumption that
J
1 + z2 1.
Expanding (9) for small z, we get for the fast spinning case

4
42
3 S
5 S
log
+
E1 (z 1) = 3 2 log
J
J
3 J
5 5 J 4

5/2
S
log6
+ .
+
6
5
J
3 J
Taking in stead z to be large, one finds

3 log(2)
S
log
+ .
E1 (z 1) =
(10)
(11)
(12)
coefficient from the large- expansion (3). As we shall

Here we recognize the famous 3 log(2)
explain in the next section, from the Bethe equation perspective it is natural to separate E1 into a
part which is analytic in for small and one which is not. Terms which are analytic, respectively
non-analytic, in originate from terms which are odd, respectively even, in z. (The even terms
constitute the first line in Eq. (9) and the odd ones the second line.) Explicitly, we have

J
1 + 2z2
analytic
2
(E1 )string = z
(13)
log z + 1 + z
1 + z2

4 log3 ( JS ) 1 1 2
16 4
=
(14)
z +
z + ,
3 5
105
3J 2

J
1
non-analytic
(E1 )string
=
z2 + 2 1 + z2 log 1 + z2
1 + z2

(15)
1 + 2z2 log 1 + 2z2

5/2 log6 ( JS ) 1 2 2 43 4
=
z + z + .
3 3
40
6J 5
(16)
The first term in the expansion (14) of the analytic part was recovered using the one-loop Bethe
ansatz in [18]. Below we shall recover the exact functional expressions (13) and (15). It is an
important point to notice that the appearance of the 3 log(2)
term for large z is due to non-trivial
cancellations between the analytic and the non-analytic part. More precisely, we have

2 log(z) + 1 2 log(2)
S
analytic
(E1 )string
(17)
log
as z ,
J

2 log(z) 1 log(2)
S
non-analytic
log
as z .
(E1 )string
(18)
J
3. The string Bethe equations
The spectrum of strings moving on AdS3 S 1 AdS5 S 5 is encoded in the Bethe equations
of a generalized sl(2) spin chain, i.e.
+ J
S x x + 1 g 2 /2x + x
xk
k
j
k j
(19)
=
2 (xk , xj ).
+
2 /2x + x
xk
x
x
1
g
k
j
j
k
j =k
Here S and J are representation labels associated with the angular momentum of the string on
respectively AdS3 and S 1 and g is the inverse string tension
1
2 .
(20)
2
8
The indices j , k label elementary excitations and the x variables are related to the momenta
carried by these excitations via
g2 =
x+
(21)
.
x
Furthermore, the quantity (xk , xj ) is the phase factor, restricted by symmetry arguments to be
of the form [25]
exp(ip) =
(xk , xj ) = ei(xk ,xj ) ,

2 (r+s1)/2

g
(xk , xj ) =
cr,s (g) qr (xk )qs (xj ) qr (xj )qs (xk ) ,
2
(22)
r=2 s=r+1
where the charges qr (x) (with r 2) are defined by

i
1
1
qr (xk ) =
=
qr (xk ).
,
Q
r
+
r 1 (xk )r1 (xk )r1

k
(23)
In the string theory description, the cr,s coefficients are expected to have an expansion in
1 (1) 1 (2)
(0)
cr,s () = cr,s
+ cr,s
+ cr,s + ,
(24)
(i)
and the string phase factor conjecture [7] accordingly involves an explicit conjecture for the cr,s .
The first two terms can be determined by comparing to conventional string theory computations
and read [20,21]

(0)
cr,s
= r,s+1 ,
(r 1)(s 1)

(1)
cr,s
= 4 1 (1)r+s
.
(s + r 2)(s r)
(25)
(26)
In order to describe proper string states the Bethe equations must be supplemented by the level
matching or momentum condition
S +

x
k
k=1
xk
= 1,
(27)
and finally the string energy is then obtained as
Q2 .
(28)
8 2
Now, our aim is to determine the classical and the one-loop energy of a certain string configuration in the limit given by Eq. (5). For that purpose we need to expand the phase factor to two
(0)
(1)
leading orders in 1 , i.e. to take into account cr,s and cr,s above. Correspondingly, we have
E=
to expand all terms to two leading orders in J1 . In order to perform the large-J expansion we need
to express the x-variables via a rapidity variable u in the following way
x = x(u i/2),

2g 2
u u
1 2 ,
x(u) = +
2 2
u
u(x) = x +
g2
.
2x
(29)
(30)
(31)
We then rescale the variables x = x(u) and g 2 in the following way

x = x(u) J x,
(32)
2
g
g2 2 .
(33)
J
Taking the logarithm of the Bethe equations and expanding to the relevant order in J and we
obtain
1
+ 2mk
xk (1 g 2 /(2xk2 ))
=
S
S

1
1
1
2
2 g2
1
2 1 g 2 /(2x 2 )
2 /(2x 2 ))
2 /(2x 2 ) xj
J
J
(x
x
)(1
g
2x
1
g
k
j
j
k
k j =k
j
j =k
1
1
(34)
Anomaly(xk ) + Non-analytic(xk ),
J
J
where mk is a mode number coming from the ambiguity of the logarithm. The two first lines
constitute the classical Bethe equations and the last line contains the one-loop correction. The
one-loop correction consists of two terms. The term Anomaly(xk ) is a spin chain finite size
effect. It arises due to the fact that the naive expansion of the logarithm becomes invalid when
+
xj xk O(1/J ) [22]. This term is analytic in . As indicated by the notation, the other oneloop term is non-analytic in . It is the part of (xj , xk ) which originates from the 1 term
in Eq. (24), i.e. the HernandezLopez phase [21]. Notice that the leading part of (xj , xk ), i.e.
the AFS phase [20], contributes already at the classical level. Now we make the assumption
about the distribution of Bethe roots that is known to lead to the folded string solution [24],
namely we assume that the roots lie in two intervals [b, a] and [a, b] on the real axis and are
symmetrically distributed around zero. This means that the second term on the right-hand side
of Eq. (34) vanishes.1 Furthermore, we assign the mode number n to roots lying in the right
interval and mode number +n to roots lying in the left interval. Finally, we introduce a resolvent
corresponding to the roots lying in the right interval

S/2
1
(y)
1 1
dy
,
G(x) =
J
x xj 1 g 2 /(2xj2 )
xy
j =1
b
(35)
and we assume that G(x) has a well-defined expansion in

G(x) = G0 (x) +
1
J
1
,
log(S/J )
1
G1 (x) + ,
J
i.e.
(36)
where each Gi (x) is analytic in the complex plane except for a cut [a, b]. Accordingly, the
density (x) needs to have a well-defined J1 expansion
(x) = 0 (x) +
1
1 (x) +
J
(37)
with each term in the expansion having support on the interval [a, b]. The normalization condition for (x) reads
b

g2
S
dy (y) 1 2 =
,
2J
2
2y
(38)
and the string energy, E, is encoded in (y) in the following way

b
dy (y) =
S
ES J
+
.
2J
4J
(39)
If we write
G(x) = G+ (x) + xG (x),
we have
E = J + S + 2J g

2
dx
where G (x) = G (x),
(x)
= J + S 2J g 2 G (0).
x2
(40)
(41)
1 The fact that the sum in this term does not include the root at j = k is an 1/J effect which can be ignored as the term
does not have any accompanying factors of log( JS ).
Using the resolvent we can write the Bethe equation in the classical limit as
G0 (x + i0) + G0 (x i0) 2G0 (x) =
1/x
+ 2n,
1 g 2 /(2x 2 )
x [a, b].
(42)
This Eq. (42) is nothing but the saddle point equation of the O(n) model on a random lattice
for n = 2 [26] with the terms on the right-hand side playing the role of the derivative of the
potential. Its solution with the given boundary conditions can be written in various ways [27,28].
Here we shall use the formulation of [28] where the solution is given in closed form for any
potential using contour integrals. In order to find the one-loop correction to the string energy
we have to take into account also the two last terms in Eq. (34). These terms can, at the order
considered, be expressed in terms of the leading order density as follows

1
1
(x)
coth
(x)
Anomaly(x) =
(43)
0
0
0 (x)
1 g 2 /(2x 2 )
and
Non-analytic(x) =
where
(x, y) =
1
x2
x 2 g 2 /2
b

dy 0 (y)
(x, y) +
(x, y) ,
(44)

1
g 2m+2r1
1
1 (1)
cr,2m+r+1
.
2
x r y 2m+r+1 x 2m+r+1 y r
2
r=2 m=0
(45)
Notice that we have taken into account the fact that the full set of Bethe roots is distributed
symmetrically around zero by forming the combination [
(x, y) +
(x, y)].
4. The semi-classical string energy
As mentioned above the leading order Eq. (42) is nothing but the saddle point equation of the
O(n) model on a random lattice for n = 2 and its solution can conveniently be written down
using contour integrals [28]

dy V0 (y) (x 2 a 2 )1/2 (x 2 b2 )1/2
1
,
G0 (x) =
(46)
2
2i x 2 y 2 (y 2 a 2 )1/2 (y 2 b2 )1/2
C+

G0+ (x) = 2
C+
dy
y2
,
G (y) 2
2i
x y2
(47)
where the contour C+ encircles the interval [a, b] counterclockwise and where
V0 (y) =
1/y
+ 2n.
1 g 2 /(2y 2 )
The endpoints of C+ , a and b, are determined by

V0 (y)
dy
= 0,
2i (y 2 a 2 )1/2 (y 2 b2 )1/2
C+
(48)
(49)
and

C+
V0 (y)y 2
dy
g2
+
2
2
1/2
2
2
1/2
2i (y a ) (y b )
2

C+
V0 (y)ab
dy
S
= .
2
2
2
1/2
2
2
1/2
2i y (y a ) (y b )
J
(50)
The first condition expresses the fact that G(x) should tend to 0 as x tends to infinity, and the
second condition is a rewriting of Eq. (38). We need that the Bethe roots stay away from the
singularities of the potential, i.e. the points y = 0 and y = g . This means that we must have
2
g 2 < 2a 2 or 2b2 < g 2 . We choose to work with the former assumption, i.e.
g 2 < 2a 2 ,
(51)
as this will directly reproduce the result of [24] in the case g = 0. Inserting the explicit expression
(48) for the potential V0 (y) the boundary conditions read
0=
1
1
2n
,
K(k )
b
2 (a 2 g 2 /2)(b2 g 2 /2)
(52)
and
S
1
1 1
g 2 /2
+ ng 2 E(k )
= 2nbE(k )
J
2 2 (a 2 g 2 /2)(b2 g 2 /2)
a

ab
1
,
1
+
2
2
2
(a g /2)(b2 g 2 /2)
(53)
where K(k ) and E(k ) are standard elliptic integrals of the first and the second kind respectively,
with k being given by
1/2

a
.
k = 1 k2
k= ,
(54)
b
Furthermore, the expression for the semi-classical string energy takes the form

d
V ()
E0 S J = abJ g 2
2i 2 (2 a 2 )1/2 (2 b2 )1/2

= J g 2
C+

1
2n
ab
.
E(k ) + 2 1
a
g
(a 2 g 2 /2)(b2 g 2 /2)
(55)
Considering only the terms of leading order in g we reproduce the results of [24], namely
a=
1
,
4nK(k )
1
E(k ) 1
=
,
2 2K(k ) k
(56)
and

n2
K(k ) 1 + k 2 K(k ) 2E(k ) .
(57)
2 2 J
It is obvious that by means of Eqs. (52), (53) and (55) one can recursively express the semiclassical energy order by order in . This idea has been pursued f.inst. in Refs. [18,29]. Here, we
shall in stead consider the limit (4)

n
S
log
, fixed,
1 J S,
z=
(58)
J
J
E0 S J =
10
where it is possible to obtain a closed expression for the all-loop energy. We immediately see
that in this limit we have
k 0,
a 0,
b ,
(59)
and from the second boundary equation (53) we find

S
1

log
K(k ) log
.
J
k
Introducing the notation
g
g = ,
2a
we notice that the first boundary equation (52) can be written as
a=
1

,
2
4n 1 g 1 k 2 g 2 K(k )
(60)
(61)
(62)
and therefore in the limit (58) we have

g 2 =
z2
,
z2 + 1
(63)
and in particular g 2 < 2a 2 . Using Eq. (62) we can express the energy as

E(k )
1
1 g 2
E0 = S + J
.
K(k )
(1 g 2 )(1 k 2 g 2 )
From here we immediately find, in the limit given by Eq. (58),

E0 = S + J 1 + z2 ,
(64)
(65)
which agrees exactly with the string theory result of Ref. [19], cf. Eq. (6). From our result for
G0 we can extract the Bethe root distribution at leading order 0 (x) in terms of which the one
loop correction terms are expressed. One finds

x
G0 (x i0) G0 (x + i0)
0 (x) =
i

dy V (y)
1
x 2
2 1/2 2
2 1/2
b x
= x a
2i x 2 y 2 (y 2 a 2 )1/2 (y 2 b2 )1/2
C+

b

dy V (y)
1
x 2
2 1/2 2
2 1/2
b x
= x a
x 2 y 2 (y 2 a 2 )1/2 (b2 y 2 )1/2

a

2

2
2
b2 g2
x x a
x2
b

=
4n 1 2 , k
,
2
b
2b b2 x 2 x 2 g2 a 2 g 2
2
(66)
where in the last integral the principal value appears. Considering only leading order in g and
setting n = 1 we recover the expression obtained in [24]. Introducing
x
= ,
(67)
a
11
we can also write

1
2

2kn 2 1 k 2 g

2 2

K(k ) 1 k , k ,
0 () =
2
2
k12 2 g
(68)
where now the normalization condition reads

b

1/k
g2
g 2
S
dx 0 (x) 1 2 = a
d 0 () 1 2 =
.
2J
2y
(69)
We also note the formula

0 () =
2n
k
2 2 1

(1 + (1 2g 2 )k 2 )g 2 (1 g 2 )(1 k 2 g 2 )(g 2 + 2 )
2 2
k g +
+
K(k )
g 2 2
(g 2 2 )2

E(k ) .
(70)
1
k2
Taking the limit (4), we get

g 2
n 2 1
1 + 1 k 2 2
2
2
0 ()
+ log
2 log() 1 k 2
g 2
1 1 k 2 2
2n 2 1
log(),
2 g 2
(71)
where the latter expression of course needs to be treated with some care. Furthermore,
0 () =
2n (1 2g 2 )2 + g 2
log().
2 1(2 g 2 )2
(72)
5. The one-loop string energy

Including the one-loop corrections, our Bethe equations read
G(x + i0) + G(x i0) 2G(x) = V0 (x) +
1
V (x),
J 1
(73)
with V0 (x) given by Eq. (48) and with

V1 (x) = Anomaly(x) + Non-analytic(x).
(74)
By applying the solution formula (46) to this equation and expanding everything including the
interval boundaries in J1 , one derives the following formula for G1 (x)
G1 (x) =

g2
d
1
1
2ab
V1 () 2
2i
x 2 1 g 2 2
2ab
C+
2

2
1/2
2
2
1/2
( a ) ( b )
,
(x 2 a 2 )1/2 (x 2 b2 )1/2
1
2
(75)
12
where we stress that the points a and b are the same as for the leading order solution. The oneloop contribution to the energy then reads, cf. Eq. (41)
E1 = 2g 2 G1 (0).
(76)
5.1. The spin chain finite size correction

As explained above the spin chain finite size corrections will give us the analytic part of the
one-loop string energy. This contribution is determined from (76) by inserting Anomaly(x) at
the place of V1 (x) in Eq. (75). One gets
b
g2
ab

dx
1
1
0 (x) coth 0 (x)
1 g2
0 (x)
2x 2
analytic
(E1 )Bethe
analytic
(E1 )Bethe

g2
= 2 2n log()
a
g
1 ab
a

2
(x a 2 )(b2 x 2 )
.
x2
In the limit we are interested in, 0 (x) and 0 (x) are given by Eqs. (71) and (72). In particular,
since 0 (x) contains the divergent factor log() we can use the approximation coth(0 (x))
1
0 (x) = 1. In this way the integral above becomes

1
d (1 2g 2 )2 + g 2
(2 g 2 )3

d (1 2g 2 )2 + g 2
J 3
2
= 4 z 1 g
(2 g 2 )3
1

J
1 + 2z2
= z
log z + 1 + z2
1 + z2
which exactly agrees with the expression (13) obtained in Ref. [19].
(77)
(78)
(79)
5.2. The HL phase

(1)
The non-analytic contribution is given by the HL phase (22) through the coefficients cr,s
of
(26). More precisely,
1
x2
Non-analytic(x) =
x 2 g 2 /2
b

dy 0 (y)
(x, y) +
(x, y) ,
(80)
where

1
1 (1)
g 2m+2r1
1
cr,2m+r+1
(x, y) =
.
2
x r y 2m+r+1 x 2m+r+1 y r
2
r=2 m=0
(81)
Notice that we have taken into account the fact that the full set of Bethe roots is symmetrically
distributed around zero by forming the combination [
(x, y) +
(x, y)]. Let us define
y
x
= .
= ,
(82)
a
a
13
The double sum above can be carried out explicitly and gives

1
2g
(, ) = 2
a ( )( g 2 )

1
g 2
( g)(
+ g)
+
+
.
log
( + g)(
g)
( )2 ( g 2 )2
Furthermore,
(, ) +
(, )
(83)

1
g 4
g(
g 2 + 2 )
+ g
4
+
+
log
= 2
g
a ( 2 2 )( 2 2 g 4 )
(2 2 )2 ( 2 2 g 4 )2
2

2
2
2
2
4
1
g ( + g )
+
g
+
(84)
+ 2 2
.
log
2
2
2
4
2
2 ( )
+ g
( g )
The correction to the energy (76) is then given by
k g 2
non-analytic
(E1 )Bethe
= 2a
1 k g 2

C+
d
2i

2 1 2
1
k2
Non-analytic().
(85)
In the limit (4) and in the variables used here, the contour C+ encircles the real half line [1, +[.
The non-analytic part of the energy will therefore be given by the following double integral:

d 1 2
4na 2
non-analytic
2
= 2 log()g
(E1 )Bethe
2i 2 g 2
C+
d
1

2 1
(, ) +
(, ) .
2
2
g
(86)
This integration is carried out in Appendix A. The result reads

non-analytic
(E1 )Bethe

n log() 2
2g + 3 g 2 log 1 g 2 + 1 + g 2 log 1 + g 2
2 g

J
1
=
z2 + 2 1 + z2 log 1 + z2
2
1+z

1 + 2z2 log 1 + 2z2 .
(87)
Here again, our result matches perfectly with the expression (15) from Ref. [19].
6. Conclusion
We have extracted the strong coupling limit of the scaling function for low twist, high spin
operators of N = 4 SYM from the quantum string Bethe equations by applying these to a folded
string rotating with angular momenta (S, J ) in AdS3 S 1 AdS5 S 5 and considering the limit
14
1 J S,
z=

S
log
,
J
J
fixed.
(88)
It is interesting to notice that this limit which was observed in [18] and further explored in [19]
from the string theory perspective also follows naturally from the quantum string Bethe ansatz.
Namely, assuming the simplest possible analyticity structure with two cuts one is led to the
relation (51) and using the rewritings in Eqs. (61) to (63) the quantity z naturally appears.
Our computation involved first a solution of the Bethe equations at the classical level. This
part was straightforward and immediately led to the square root expression (65) for the classical
energy. Subsequently, we determined the one-loop contribution to the energy. This contribution
consisted of two parts, one originating from spin chain finite size effects and one being due to
the HernandezLopez phase. Both parts could be treated exactly and led to a total expression
for the string one-loop energy, J F (z), which agreed at the functional level with the result of
a traditional string theory computation, cf. Eqs. (13), (15), (79) and (87). Both the classical and
the one loop energy when considered as a function of z could be smoothly extrapolated to large
values of z and led to the strong coupling limit of the scaling function
f () =

3 log(2)
1
+O .
(89)
is due to a highly non-trivial cancellation between terms

We stress again that the famous 3 log(2)
originating from the HL-phase and terms due to spin chain finite size effects. More precisely, we
have

2 log(z) + 1 2 log(2)
S
analytic
E1
(90)
log
as z ,
J

2 log(z) 1 log(2)
S
non-analytic
log
as z .
E1
(91)
As mentioned earlier there exists a numerical prediction for the coefficient of the O(1/ ) term
of (89) [16]. Furthermore, a genuine string theory calculation of the same quantity seems to be
under way [30]. Given these developments it might be interesting to pursue our approach to twoloop order. It is obvious that the same strategy should be applicable and we are convinced that
the Bethe equations will once again prove their efficiency.
Acknowledgements
The authors thank Lisa Freyhult, Sergey Frolov and Matthias Staudacher for useful discussions. Both authors were supported by ENRAGE (European Network on Random Geometry), a
Marie Curie Research Training Network financed by the European Communitys Sixth Framework Programme, network contract MRTN-CT-2004-005616.
Appendix A
In this appendix, we present the evaluation of the double integral (86) which up to a factor
log() can be written as
g 2
16n
2

I=
C+

d 1 2
2 1
A(, ) + B(, ) ,
d 2
2
2
2
2i g
g
15
(A.1)
where

1
g(
g 2 + 2 )
g 4
+ g
+
+
,
log
g
( 2 2 )( 2 2 g 4 )
(2 2 )2 ( 2 2 g 4 )2
2

1
g 2 ( 2 2 + g 4 )
+ 2
g
B(, ) =
+ 2 2
,
log
2 (2 2 )2
+ g
( g 4 )2
A(, ) =
and the contour C+ encircles the real half line [1, +[. The integration of A(, ) with respect
to is straightforward:

C+
d 1 2
A(, )
2i 2 g 2

1 2
1 2
= Res
A(, ) Res
A(, )
2
2 g 2
=g 2 g 2
= g

(g 2 + 2 ) 1 g 2 2 g 4
=
2g(
g 2 2 )2

(g 4 + 2 g 2 2 2 + 4 1 g 2 2 g 4 )
+ g

log
.
g
4(g 2 2 )2 2 g 4
(A.2)
In order to integrate the last term of (A.1), one first exploits the parity of the integrand to extend
the contour C+ to C+ C (see Fig. 1). Subsequently, the resulting contour can be deformed into
the contour C0 around the cut [g,
g].
This contour integral can then be re-expressed as the finite
part of the integral along [g,
g]
with the substitution

g
log
2i.
+ g
One gets
Fig. 1. Cuts on the complex plane in the integration of B(, ).
16

C+
d 1 2
B(, )
2i 2 g 2
1
=
2

C0
d 1 2
B(, )
2i 2 g 2
2

1
1 2 1
g 2 ( 2 2 + g 4 )
+ 2
= d 2
+
2
g 2 2 (2 2 )2
( 2 2 g 4 )2
(A.3)
g
(A.4)

2

(g 2 + 2 ) 1 g 2 2 g 4
(g 2 + 2 2)
1 g 1

=
tan
2g(
g 2 2 )2
2( 2 g 2 )2 2 1
1 g 2

g 4 + (g 2 2) 2
g 3
1

tanh
+
2( 2 g 2 )2 2 g 4
+ 1 g 2 2 g 4

1 g 2 (g 2 + 2 )
log 1 g 2 .
(A.5)
2
2
2
2( g ) g
! g
Here, g means that all poles in the interval [g,
g]
should be subtracted from the integrand.
Summing (A.2) and (A.5), we obtain

d 1 2
A(, ) + B(, )
2
2
2i g
C+

2 1 g 2
1 g 2 (g 2 + 2 )
2
1 g
log 1 g 2
tanh
=
2( 2 g 2 )2 g
( g 2 )2

(g 2 + 2 2)
g 2 1
tan1
2( 2 g 2 )2 2 1
1 g 2

2
g 4 + (g 2 2) 2
1 g 1 g

tanh
.
2( 2 g 2 )2 2 g 4
2 g 4
(A.6)
The next task is to integrate (A.6) with respect to . This can be done using the same kind of
techniques as previously after rewriting the integral as a contour integral around C + . By this
strategy, one gets the following intermediate results:
+ 2
1 g 2 + 2
2 g 2
d =
,
2
2
2
2
2
8 (1 g 2 )3/2
g ( g )
(A.7)

+ 2
1
1 g
tanh
d
2 g 2 (g 2 2 )2
1

1
3 2
2
,
=
1 g + 2 log 1 g
2
32g(1
g 2 )3/2
g
(A.8)
17
2

+ 2
1 (g 2 + 2 2)
1
1 g

tan
d
1 g 2
2 g 2 (g 2 2 )2 2 1
1

3 2 1 g 4
2
=
,
log 1 g
1 g +
2
16g(1
g 2 )
g 2
g 1g 2
+ 2
(g 4 + (g 2 2) 2 ) tanh1 2 4
1
g

d
2
2
2
2
4
2 g 2
(g ) g
1

5 2 1 g 4
2
=
.
log 1 + g
1 g
2
16g(1
g 2 )
g 2
Finally, putting everything together we end up with

2
2g + 3 g 2 log 1 g 2 + 1 + g 2 log 1 + g 2 ,
I=
3
32g
(A.9)
(A.10)
(A.11)
g
which, up to the factor 16n
log(), gives back the result (87).
2
2
References
[1] J.A. Minahan, K. Zarembo, JHEP 0303 (2003) 013, hep-th/0212208;
N. Beisert, C. Kristjansen, M. Staudacher, Nucl. Phys. B 664 (2003) 131, hep-th/0303060;
N. Beisert, M. Staudacher, Nucl. Phys. B 670 (2003) 439, hep-th/0307042.
[2] G. Mandal, N.V. Suryanarayana, S.R. Wadia, Phys. Lett. B 543 (2002) 81, hep-th/0206103;
I. Bena, J. Polchinski, R. Roiban, hep-th/0305116;
I. Bena, J. Polchinski, R. Roiban, JHEP 0310 (2003) 017, hep-th/0308089.
[3] J.M. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200.
[4] M. Staudacher, JHEP 0505 (2005) 054, hep-th/0412188.
[5] N. Beisert, hep-th/0511082.
[6] N. Beisert, B. Eden, M. Staudacher, J. Stat. Mech. 0701 (2007) P021, hep-th/0610251.
[7] N. Beisert, R. Hernandez, E. Lopez, JHEP 0611 (2006) 070, hep-th/0609044.
[8] Z. Bern, M. Czakon, L.J. Dixon, D.A. Kosower, V.A. Smirnov, Phys. Rev. D 75 (2007) 085010, hep-th/0610248.
[9] N. Beisert, M. Staudacher, Nucl. Phys. B 727 (2005) 1, hep-th/0504190.
[10] B. Eden, M. Staudacher, J. Stat. Mech. 0611 (2006) P014, hep-th/0603157.
[11] S.S. Gubser, I.R. Klebanov, A.M. Polyakov, Nucl. Phys. B 636 (2002) 99, hep-th/0204051.
[12] S. Frolov, A.A. Tseytlin, JHEP 0206 (2002) 007, hep-th/0204226.
[13] A.V. Kotikov, L.N. Lipatov, Nucl. Phys. B 769 (2007) 217, hep-th/0611204.
[14] L.F. Alday, G. Arutyunov, M.K. Benna, B. Eden, I.R. Klebanov, JHEP 0704 (2007) 082, hep-th/0702028.
[15] I. Kostov, D. Serban, D. Volin, hep-th/0703031.
[16] M.K. Benna, S. Benvenuti, I.R. Klebanov, A. Scardicchio, Phys. Rev. Lett. 98 (2007) 131603, hep-th/0611135.
[17] M. Beccaria, G.F. De Angelis, V. Forini, JHEP 0704 (2007) 066, hep-th/0703131.
[18] A.V. Belitsky, A.S. Gorsky, G.P. Korchemsky, Nucl. Phys. B 748 (2006) 24, hep-th/0601112.
[19] S. Frolov, A. Tirziu, A.A. Tseytlin, Nucl. Phys. B 766 (2007) 232, hep-th/0611269.
[20] G. Arutyunov, S. Frolov, M. Staudacher, JHEP 0410 (2004) 016, hep-th/0406256.
[21] R. Hernandez, E. Lopez, JHEP 0607 (2006) 004, hep-th/0603204;
L. Freyhult, C. Kristjansen, Phys. Lett. B 638 (2006) 258, hep-th/0604069;
N. Gromov, P. Vieira, hep-th/0703191.
[22] N. Beisert, A.A. Tseytlin, K. Zarembo, Nucl. Phys. B 715 (2005) 190, hep-th/0502173;
R. Hernandez, E. Lopez, A. Perianez, G. Sierra, JHEP 0506 (2005) 011, hep-th/0502188.
[23] S. Frolov, A.A. Tseytlin, Nucl. Phys. B 668 (2003) 77, hep-th/0304255.
[24] N. Beisert, S. Frolov, M. Staudacher, A.A. Tseytlin, JHEP 0310 (2003) 037, hep-th/0308117.
[25] N. Beisert, T. Klose, J. Stat. Mech. 0607 (2006) P006, hep-th/0510124.
18
[26] I.K. Kostov, Mod. Phys. Lett. A 4 (1989) 217.

[27] I.K. Kostov, M. Staudacher, Nucl. Phys. B 384 (1992) 459, hep-th/9203030.
[28] B. Eynard, C. Kristjansen, Nucl. Phys. B 455 (1995) 577, hep-th/9506193;
B. Eynard, C. Kristjansen, Nucl. Phys. B 466 (1996) 463, hep-th/9512052.
[29] B. Basso, G.P. Korchemsky, hep-th/0612247.
[30] R. Roiban, A. Tirziu, A.A. Tseytlin, arXiv: 0704.3638 [hep-th].
Screening masses in the SU(3) pure gauge theory

and universality
R. Falcone , R. Fiore , M. Gravina , A. Papa
Dipartimento di Fisica, Universit della Calabria, and Istituto Nazionale di Fisica Nucleare,
Gruppo collegato di Cosenza, I-87036 Arcavacata di Rende, Cosenza, Italy
Received 4 May 2007; received in revised form 14 June 2007; accepted 19 June 2007
Abstract
We determine from Polyakov loop correlators the screening masses in the deconfined phase of the
(3 + 1)d SU(3) pure gauge theory at finite temperature near the transition, for two different channels of
angular momentum and parity. Their ratio is compared with that of the massive excitations with the same
quantum numbers in the 3d 3-state Potts model in the broken phase near the transition point at zero magnetic
field. Moreover we study the inverse decay length of the correlation between the real parts and between the
imaginary parts of the Polyakov loop and compare the results with expectations from perturbation theory
and mean-field Polyakov loop models.
1. Introduction
The essential features of confinement, of the deconfinement phase transition at finite temperature and of the deconfined phase in QCD can be studied without loss of relevant dynamics in the regime of infinitely massive quarks or, equivalently, in the SU(3) pure gauge theory.
The (3 + 1)d SU(3) pure gauge theory at finite temperature undergoes a confinement/deconfinement phase transition associated with the breaking of the center of the gauge group,
E-mail addresses: rfalcone@cs.infn.it (R. Falcone), fiore@cs.infn.it (R. Fiore), gravina@cs.infn.it (M. Gravina),
papa@cs.infn.it (A. Papa).
doi:10.1016/j.nuclphysb.2007.06.009
20
R. Falcone et al. / Nuclear Physics B 785 (2007) 1933
Z(3) [1,2]. This transition is weakly first order [37] and the order parameter is the Polyakov
loop [1,2].
In general, the deconfinement transition in the (d + 1)-dimensional SU(N ) pure gauge theory
at finite temperature can be put in relation with the order/disorder [8] phase transition of the
d-dimensional Z(N) spin model through the SvetitskyYaffe conjecture [9]. According to this
conjecture, when the transition is second order the gauge theory and the spin model belong to
the same universality class and therefore share critical indices, amplitude ratios and correlation
functions at criticality. There are so many numerical evidences in favor of this conjecturesee,
for instance, Refs. [1016] and, for a review, Ref. [17]that there is no residual doubt about its
validity.
In the case of first order phase transitions, as for the (3 + 1)d SU(3) pure gauge theory, the
SvetitskyYaffe conjecture is not applicable in strict sense, since the correlation length keeps finite at the transition point and a universal behavior, independent of the details of the microscopic
interactions, cannot be established. Therefore there are no a priori reasons why the critical dynamics of (3 + 1)d SU(3) and of its would-be counterpart spin model, 3d Z(3) or 3-state Potts
model [18,19], should look similar. It turns out, however, that for both these theories the phase
transition is weakly first order, which means that the correlation length at the transition point,
although finite, becomes much larger than the lattice spacing. This leads to the expectation that
the two theories may have some long distance aspects in common and that some specific issues
of SU(3) in the transition region can be studied taking the 3d 3-state Potts model as an effective
model.
Since there are no critical indices to compare, a test of this possibility could be based on
the comparison of the spectrum of massive excitations of the 3d 3-state Potts model in the
broken phase near the transition at zero magnetic field with the spectrum of the inverse decay lengths of Polyakov loop correlators in the (3 + 1)d SU(3) gauge theory in the deconfined phase near the transition. If universality would apply in strict sense, these spectra should
exhibit the same pattern, as suggested by several numerical determinations in the 3d Ising
class [2023].
In this paper we determine the low-lying masses of the spectrum of the (3 + 1)d SU(3) pure
gauge theory at finite temperature in the deconfined phase near the transition in two different
sectors of parity and orbital angular momentum, 0+ and 2+ , and compare their ratio with that of
the corresponding massive excitations in the phase of broken Z(3) symmetry of the 3d 3-state
Potts model, determined in Ref. [24].
In fact, we extend our numerical analysis to temperatures far away from the transition temperature Tt in order to look for possible scaling of the fundamental masses with temperature.
Moreover, we consider also the screening masses resulting from correlators of the real parts
and of the imaginary parts of the Polyakov loop. These determinations can represent useful
benchmarks for effective models of the high-temperature phase of SU(3), such as those based
on mean-field theories of the Polyakov loop, suggested by Pisarski [25].
The paper is organized as follows: in Section 2 we describe the methods used to extract
the screening masses through Polyakov loop correlators in (3 + 1)d SU(3); in Section 3 we
present our numerical results and compare them with expectations from the 3d 3-state Potts
model and from Polyakov loop effective models; in the last section we draw our conclusions.
21
2. Screening masses from Polyakov loop correlators

In this section we define the general procedure for extracting screening masses in (3 + 1)d
SU(3) at finite temperature from correlators of the Polyakov loop,
P (x, y, z) = Tr
Nt
U4 (x, y, z, nt a),
(1)
nt =1
where Nt is the number of lattice sizes in the temporal direction and a is the lattice spacing.
The pointpoint connected correlation function is defined as

i0 (r) = Pi Pi0 Pi Pi0 ,
(2)
where i and i0 are the indices of sites and r is the distance between them. The large-r behavior of
the pointpoint connected correlation function is determined by the correlation length of the theory, 0 , or, equivalently, by its inverse, the fundamental mass. In order to extract the fundamental
mass it is convenient to define the wallwall connected correlation function, since numerical data
in this case can be directly compared with exponentials in r, without any power prefactor.
The 0+ -channel connected wallwall correlator in the z-direction is defined as
2

G |z1 z2 | = Re P (z1 )P (z2 ) P ,
(3)
where
P (z) =
Ny
Nx
1
P (nx a, ny a, z),
Nx Ny
(4)
nx =1 ny =1
represents the Polyakov loop averaged over the xy-plane at a given z. Here and in the following,
Ni (i = x, y, z) is the number of lattice sites in the i-direction. The wall average implies the
projection at zero momentum in the xy-plane.
The correlation function (3) takes contribution from all the screening masses in the 0+ channel. In fact, the general large distance behavior for the function G(|z1 z2 |), in an infinite lattice,
is:

G |z1 z2 | =
(5)
an emn |z1 z2 | ,
n
where m0 is the fundamental mass, while m1 , m2 , . . . are higher masses with the same angular
momentum and parity (0+ ) quantum numbers of the fundamental mass. On a periodic lattice the
above equation must be modified by the inclusion of the so-called echo term:
m |z z |

an e n 1 2 + emn (Nz |z1 z2 |) .
G |z1 z2 | =
(6)
n
The fundamental mass in a definite channel can be extracted from wallwall correlators by looking for a plateau of the effective mass,
meff (z) = ln
G(z 1)
G(z)
(7)
at large distances. For the 2+ -channel, we used the variational method [26,27], which consists in
defining several wall-averaged operators with 2+ quantum numbers and in building the matrix
22
of cross-correlations between them. The eigenvalues of this matrix are single exponentials of the
effective masses in the given channel. The variational method is usually adopted to determine
massive excitations above the fundamental one in the given channel. In our case, however, it
turned out that this method improved the determination of the fundamental mass in the 2+ channel, but did not lead to determine higher excitations. Our choice of wall-averaged operators
in the 2+ -channel is inspired by Ref. [28] and reads
Pn (z) =
Ny
Nx
1
P (nx a, ny a, z)
Nx Ny
nx =1 ny =1

P (nx a + na, ny a, z) P (nx a, ny a + na, z) .
(8)
In most cases we have taken 8 operators, corresponding to different values of n, with the largest n
almost reaching the spatial lattice size Nx . We have determined the 0+ and the 2+ fundamental
2+ , over a wide interval of temperatures above the transition
masses in lattice units, m
0+ and m
temperature Tt of (3 + 1)d SU(3) and have studied
if in some region, lying strictly above Tt , a scaling law can be found for the fundamental
0+ ;
correlation length 0 = 1/m
if m
2+ has the same scaling behavior as m
0+ ;
how the ratio m2+ /m0+ compares with the ratio of the corresponding excitations in the
broken phase of 3d 3-state Potts model near the transition, m2+ /m0+ = 2.43(10), found in
Ref. [24] on a 483 lattice.
We consider also correlators of the (wall-averaged) real and imaginary parts of the Polyakov
loop, defined as

GR |z1 z2 | = Re P (z1 ) Re P (z2 ) Re P (z1 ) Re P (z2 ) ,
(9)

GI |z1 z2 | = Im P (z1 ) Im P (z2 ) .
(10)
I , can be extracted in the same way as for the
The corresponding screening masses, m
R and m
0+ mass.
We have studied the ratio mI /mR over a wide interval of temperatures above the transition
temperature Tt of (3 + 1)d SU(3) and seen how it compares with the prediction from hightemperature perturbation theory, according to which it should be equal to 3/2 [30,31], and with
the prediction from the mean-field Polyakov loop model of Ref. [29], according to which it
should be equal to 3 in the transition region. The interplay between the two regimes should
delimit the region where mean-field Polyakov loop models should be effective.
3. Numerical results
We used the Wilson lattice action and generated Monte Carlo configurations by a combination
of the modified Metropolis algorithm [32] with over-relaxation on SU(2) subgroups [33]. The
error analysis was performed by the jackknife method over bins at different blocking levels. We
performed our simulations on a 163 4 lattice, for which t = 5.6908(2) [34], over an interval
of values ranging from 5.69 to 9.0.
Screening masses are determined from the plateau of meff (z) as a function of the wall separation z. In each case, the plateau mass is taken as the effective mass (with its error) belonging
23
(a)
(b)
(c)
Fig. 1. Scatter plots of the real and imaginary part of the Polyakov loop at (a) = 5.69 t , (b) = 5.71 and
(c) = 5.74.
24
(a)
(b)
Fig. 2. (a) Typical scatter plot of the complex order parameter P for larger than 5.705 on a 163 4 lattice. There
are no states in the symmetric phase, but tunneling survives between the three broken minima. (b) Same as (a) with the
tunneling between broken minima removed by the rotation to the real phase.
to the plateau and having the minimal uncertainty. We define plateau the largest set of consecutive data points, consistent with each other within 1 . This procedure is more conservative than
identifying the plateau mass and its error as the results of a fit with a constant on the effective
masses meff (z), for large enough z. To illustrate our definition of plateau mass, let us apply it
to Figs. 3(b) and 4(a), which give respectively the m
2+ and the m
I effective masses at = 5.74
versus the separation z between walls (the cases of m
2+ and m
I are usually the most problematic
for the determination of the plateau mass). In Fig. 3(b), according to our definition, the plateau
is formed by the points at z = 3, 4 and 5, since they are compatible each other within 1 ; the
plateau mass is taken as the effective mass at z = 3, since this effective mass has the smallest
uncertainty among the three masses belonging to the plateau. In Fig. 4(a), according to our definition, the plateau is formed by the points at z = 4, 5 and 6, since they are compatible each other
within 1 ; the plateau mass is taken as the effective mass at z = 4, since this effective mass has
the smallest uncertainty among the three masses belonging to the plateau.
Just above the critical value t we find a large correlation length, which is not of physical
relevance. It is instead a genuine finite size effect [8] related to tunneling between degenerate
25
(a)
(b)
Fig. 3. Effective mass in the 0+ (a) and the 2+ (b) channel as a function of the separation between walls on the (x, y)
plane at = 5.74.
vacua. This effect disappears by going to larger lattice volumes or moving away from t in the
deconfined phase. Indeed, by increasing one can see that the symmetric phase becomes less
and less important in the Monte Carlo ensemble, up to disappearance; then, for large enough ,
also the structure with three-degenerate Z(3) sectors disappears, leaving only the real sector
of Z(3). This is illustrated in Fig. 1 where the scatter plot of the Polyakov loop is shown for
three representative values.
Tunneling can occur between the symmetric and the broken phase, and between the three
degenerate vacua of the deconfined phase. The former survives only in a short range of temperatures around t , and at 5.71 has almost completely disappeared, as we can see from
26
(a)
(b)
Fig. 4. Effective mass m
I (a) and m
R (b) as a function of the separation between walls on the (x, y) plane at = 5.74.
Fig. 1(b). When tunneling is active, the correlation function has the following expression [8]:

G |z1 z2 | a0 em0 |z1 z2 | + b0 emt |z1 z2 | ,
(11)
where mt is the inverse of the tunneling correlation length and is generally much smaller than
the fundamental mass m0 and therefore behaves as a constant additive term in the correlation
function. In (11) we have taken into account only the lowest masses in the spectrum and omitted
the echo terms. The dependence on mt in the correlation function can be removed by extracting
the effective mass by use of the combination
meff (z) = ln
G(z) G(z + 1)
.
G(z + 1) G(z + 2)
(12)
27
(a)
(b)
Fig. 5. Screening mass in the 0+ (a) and in the 2+ (b) channel as a function of .
For 5.71 the only active tunneling is among the three broken minima. However, the separation among them is so clear (see, for instance, Fig. 2(a)), that it is possible to rotate
unambiguously all the configurations to the real sector (see Fig. 2(b)) and to treat them on the
same ground.
We have performed simulations for several values of in the deconfined phase, up to = 9.0,
with a statistics of a few hundred thousand measurements at each value. Simulations have
been performed on the PC cluster Majorana of the INFN Group of Cosenza.
As discussed in the Introduction and in Section 2, our typical observable is a screening mass in
a given channel of angular momentum and parity, identified as the plateau value in the plot of the
effective mass as a function of the separation z between two-dimensional walls. The procedure to
28
(a)
(b)
Fig. 6. Screening mass m
I (a) and m
R (b) as a function of .
determine the plateau value has been illustrated in Section 2. A typical example of the behavior
of the effective mass with z is shown in Fig. 3 for the 0+ and the 2+ channels and in Fig. 4 for
the masses m
I and m
R extracted from Eqs. (9) and (10). The deviation from the plateau value at
small z can be attributed to lattice artifacts and to the possible effect of other states with excited
masses in the same channel. Our determinations refer only to the fundamental masses in a given
channel. The behavior with of the fundamental masses in the 0+ and the 2+ channels is given
in Fig. 5. In Fig. 6 we show instead the masses m
R and m
I for varying . A summary of all
our results is presented in Table 1. We observe from Table 1 that m
0+ and m
R are consistent
within statistical errors, this indicating that the Polyakov loop correlation is dominated by the
correlation between the real parts. Our results for m
0+ at = 5.70 and = 5.90 (T 3Tt /2) can
29
Table 1
I and m
R and for some mass ratios
Summary of the results for the fundamental masses in the 0+ and 2+ channels, for m
m
0+
m
2+
m2+ /m0+
m
I
m
R
mI /mR
stat.
5.67
5.68
5.69
5.695
5.70
5.705
5.71
5.715
5.717
5.72
5.73
5.74
5.75
5.76
5.77
5.78
5.81
5.85
5.89
5.90
9.0
0.2870(46)
0.2293(67)
0.175(10)
0.251(12)
0.3088(80)
0.342(17)
0.3615(30)
0.4054(51)
0.4301(46)
0.4319(69)
0.4811(92)
0.5267(52)
0.5592(50)
0.5857(65)
0.620(14)
0.650(10)
0.6742(99)
0.718(10)
0.697(29)
0.734(17)
0.671(19)
0.81(10)
1.161(38)
0.94(11)
1.09(12)
1.31(18)
1.25(30)
1.452(49)
1.500(80)
1.29(17)
1.44(15)
1.695(77)
1.849(39)
1.83(13)
1.79(12)
1.812(77)
1.83(12)
2.21(13)
2.09(18)
1.30(57)
4.61(86)
4.63(38)
3.04(45)
3.20(50)
3.63(53)
3.08(77)
3.38(15)
3.47(24)
2.69(40)
2.73(31)
3.03(17)
3.16(10)
2.95(28)
2.76(23)
2.69(15)
2.55(21)
3.17(32)
2.85(31)
1.93(91)
0.65(10)
0.883(47)
1.056(23)
1.092(54)
1.105(40)
1.252(15)
1.164(58)
1.173(39)
1.035(98)
1.06(11)
1.227(28)
1.421(26)
1.27(12)
1.39(10)
1.34(13)
1.299(90)
1.284(98)
1.10(11)
0.2793(65)
0.2226(97)
0.176(14)
0.205(37)
0.307(13)
0.352(23)
0.386(23)
0.4129(61)
0.4237(43)
0.4296(76)
0.4762(92)
0.5241(44)
0.5562(49)
0.5880(36)
0.6173(73)
0.635(10)
0.671(10)
0.7042(87)
0.686(29)
0.7415(87)
0.670(19)
3.07(87)
2.87(27)
2.86(15)
2.98(19)
2.68(14)
2.955(65)
2.71(18)
2.46(13)
1.98(20)
1.91(22)
2.087(61)
2.303(70)
2.00(23)
2.07(18)
1.90(21)
1.89(21)
1.73(15)
1.64(23)
750k
250k
770k
320k
770k
750k
770k
250k
320k
770k
170k
770k
770k
770k
60k
60k
190k
110k
190k
215k
200k
be compared with the corresponding determinations on a 4 82 16 lattice of Ref. [35], which

give 0.27+0.01
0.02 and 0.64(1), respectively. The 15% disagreement can be taken as a measure
of the systematic effects involved in the whole procedure for the determination of masses. We
cannot make a direct comparison with the determination for m
0+ at = 5.93 on a 163 4 lattice
of Ref. [36], since we did not perform simulations at this value of . We observe, however,
that the determination of Ref. [36], which gives 0.73(5) according to our definition of plateau
mass value, agrees with the results for the two adjacent values in our table, = 5.90 and
= 9.0.
R , becomes much smaller
We can see that the fundamental mass in the 0+ channel, as well as m
0+ and of m
R
than 1 at t , as expected for a weakly first order phase transition. In the cases of m
we have made some determinations below t (see Figs. 5 and 6). It turns out that masses in lattice
units take their minimum value just at t , where there is a cusp in the -dependence. Such a
behaviour was observed also by the authors of Ref. [37], whose results, when the comparison is
possible, agree with ours.
3.1. Scaling behavior and comparison with the Potts model
It would be interesting to find a scaling law for the fundamental mass in the 0+ channel,
which looks like the one which holds for a second order phase transition, with a suitable critical
exponent. There is, however, an important caveat: while the correlation length in lattice units
diverges at a second order critical point, it keeps finite at a first order transition point. Therefore,
any second-order-like scaling law, when applied to the region near a first order phase transition,
should be taken as an effective description, which cannot hold too close to the transition point.
30
Table 2
0+ () with the function ((1 t )/( t )) . The second column
Summary of the fits of the mass ratios m
0+ (1 )/m
contains the largest window of values for which the fit has a 2 /d.o.f. lower than 1
1
Window of values
2 /d.o.f.
5.75
5.74
5.73
5.72
5.717
5.715
5.71
5.705
5.70
5.695
5.705.78
5.705.78
5.6955.85
5.6955.78
5.7155.81
5.6955.78
5.725.78
5.7055.78
5.6955.90
5.6955.90
0.3619(90)
0.365(11)
0.329(12)
0.354(12)
0.3228(94)
0.358(10)
0.3951(76)
0.448(22)
0.3141(58)
0.3095(67)
0.90
0.92
0.77
0.73
0.69
0.86
0.33
0.89
0.94
0.41
Fig. 7. Comparison between the scaling function [(1 t )/( t )]1/3 and the mass ratio m0+ (1 )/m0+ () for
varying , with 1 = 5.72.
With this spirit, we have compared our data with the scaling law

1 t m
0+ (1 )
,
2 t
m
0+ (2 )
(13)
0+ (2 ) are the fundamental masses in the 0+ channel at 1 and 2 , respecwhere m

0+ (1 ) and m
tively. This scaling law (13) is an approximation of the law

+ (T1 )
T 1 Tt m
0
,
(14)
T 2 Tt
m
0+ (T2 )
valid in an interval around Tt short enough that the linear approximation T1 Tt 1 t
holds. We have considered several choices of 1 and found that for each of them there is a
31
+
Fig. 8. Ratio m+
2 /m0 as a function of in the deconfined phase. The three upper horizontal lines represent the constant
(with its error) which fits the data (see the text for details); the three lower horizontal lines represent the corresponding
mass ratio (with its error) found in the 3d 3-state Potts model [24].
wide window of values above t where the scaling law (13) works, with a dynamical
exponent . Our results are summarized in Table 2. For 1 = 5.72 and 5.73, which lie well inside
this window, the parameter is compatible with = 1/3, suggested in Ref. [38] to apply to
the standard correlation function. For 1 = 5.72 we have calculated also the 2 /d.o.f. when
is put exactly equal to 1/3, getting 2 /d.o.f. = 0.75 in the window from = 5.715 to = 5.78.
In Fig. 7 we show, for this choice of 1 , the comparison between data and the scaling function
with set equal to 1/3.
Then, we have considered the -dependence of the ratio m2+ /m0+ , shown in Fig. 8. We have
found that this ratio can be interpolated with a constant in the interval from t to = 5.77. This
constant turned out to be 3.172(65), with a 2 /d.o.f equal to 1.085. In the fit we excluded the
point at = 5.695, for which the determination of m2+ is probably to be rejected (see Fig. 5(b)).
If the point at = 5.695 is included, the constant becomes 3.214(64) with 2 /d.o.f. = 2.21. The
fact that the ratio m2+ /m0+ is compatible with a constant in the mentioned interval suggests that
m
2+ scales similarly to m
0+ near the transition. This constant turns out to be larger than the ratio
between the lowest massive excitations in the same channels in the broken phase of the 3d 3-state
Potts model, which was determined in Ref. [24] to be 2.43(10).
3.2. Comparison with Polyakov loop models
We have calculated the ratio mI /mR for ranging from 5.695 up to 9.0. We observe from
Fig. 9 that this ratio is compatible with 3/2 at the largest values considered, in agreement with
the high-temperature perturbation theory. Then, when the temperature is lowered towards the
transition, this ratio goes up to a value compatible with 3, in agreement with the Polyakov loop
32
Fig. 9. Ratio mI /mR as a function of in the deconfined phase. The vertical line corresponds to the critical value.
model of Ref. [29], which contains only quadratic, cubic and quartic powers of the Polyakov
loop, i.e. the minimum number of terms required in order to be compatible with a first order
phase transition. The same trend has been observed also in Ref. [37].
4. Conclusions and outlook
In this work we have studied some screening masses of the (3 + 1)d SU(3) pure gauge theory from Polyakov loop correlators in a large interval of temperatures above the deconfinement
transition.
In particular, we have considered the lowest masses in the 0+ and the 2+ channels of angular
momentum and parity and the screening masses resulting from the correlation between the real
parts and between the imaginary parts of the Polyakov loop.
The behavior of the ratio between the masses in the 0+ and the 2+ channels with the temperature suggests that they have a common scaling above the transition temperature. This ratio
turns to be 30% larger than the ratio of the lowest massive excitations in the same channels of the 3d 3-state Potts model in the broken phase. This can be taken as an estimate of
the level of approximation by which the SvetitskyYaffe conjecture, valid in strict sense only
for continuous phase transitions, can play some role also for (3 + 1)d SU(3) at finite temperature.
The dependence on the temperature of the ratio between the screening masses from the
correlation between the real parts and between the imaginary parts of the Polyakov loop
shows a nice interplay between the high-temperature regime, where perturbation theory should
work, and the transition regime, where mean-field effective Polyakov loop models could apply.
33
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
A.M. Polyakov, Phys. Lett. B 72 (1978) 477.

L. Susskind, Phys. Rev. D 20 (1979) 2610.
M. Fukugita, T. Kaneko, A. Ukawa, Phys. Lett. B 154 (1985) 185.
P. Bacilieri, et al., Phys. Rev. Lett. 61 (1988) 1545.
F.R. Brown, N.H. Christ, Y. Deng, M. Gao, T.J. Woch, Phys. Rev. Lett. 61 (1988) 2058.
J. Kogut, et al., Phys. Rev. Lett. 51 (1983) 869.
O. Kaczmarek, F. Karsch, E. Laermann, M. Ltgemeier, Phys. Rev. D 62 (2000) 034021.
R.V. Gavai, F. Karsch, B. Petersson, Nucl. Phys. B 322 (1989) 738.
B. Svetitsky, L.G. Yaffe, Nucl. Phys. B 210 (1982) 423.
F. Gliozzi, P. Provero, Phys. Rev. D 56 (1997) 1131.
R. Fiore, F. Gliozzi, P. Provero, Phys. Rev. D 58 (1998) 114502.
J. Engels, T. Scheideler, Nucl. Phys. B 539 (1999) 557.
S. Fortunato, F. Karsch, P. Petreczky, H. Satz, Nucl. Phys. B (Proc. Suppl.) 94 (2001) 398.
R. Fiore, A. Papa, P. Provero, Nucl. Phys. B (Proc. Suppl.) 106 (2002) 486.
R. Fiore, A. Papa, P. Provero, Phys. Rev. D 63 (2001) 117503.
A. Papa, C. Vena, Int. J. Mod. Phys. A 19 (2004) 3209.
A. Pelissetto, E. Vicari, Phys. Rep. 368 (2002) 549.
H.W.J. Blte, R.H. Swendsen, Phys. Rev. Lett. 43 (1979) 779.
W. Janke, R. Villanova, Nucl. Phys. B 489 (1997) 679, and references therein.
M. Caselle, M. Hasenbusch, P. Provero, Nucl. Phys. B 556 (1999) 575.
M. Caselle, M. Hasenbusch, P. Provero, K. Zarembo, Nucl. Phys. B 623 (2002) 474.
R. Fiore, A. Papa, P. Provero, Nucl. Phys. B (Proc. Suppl.) 119 (2003) 490.
R. Fiore, A. Papa, P. Provero, Phys. Rev. D 67 (2003) 114508.
R. Falcone, R. Fiore, M. Gravina, A. Papa, Nucl. Phys. B 767 (2007) 385.
R.D. Pisarski, Phys. Rev. D 62 (2000) 111501.
A.S. Kronfeld, Nucl. Phys. B (Proc. Suppl.) 17 (1990) 313.
M. Lscher, U. Wolff, Nucl. Phys. B 339 (1990) 222.
V. Agostini, G. Carlino, M. Caselle, M. Hasenbusch, Nucl. Phys. B 484 (1997) 331.
A. Dumitru, R.D. Pisarski, Nucl. Phys. B (Proc. Suppl.) 106 (2002) 483.
S. Nadkarni, Phys. Rev. D 33 (1986) 3738.
A. Dumitru, R.D. Pisarski, Phys. Rev. D 66 (2002) 096003.
N. Cabibbo, E. Berker, Phys. Lett. B 119 (1982) 387.
S.L. Adler, Phys. Rev. D 23 (1981) 2901.
G. Boyd, J. Engels, F. Karsch, E. Laermann, C. Legeland, M. Lutgemeier, B. Petersson, Nucl. Phys. B 469 (1996)
419.
S. Datta, S. Gupta, Nucl. Phys. B 534 (1998) 392, hep-lat/980634.
B. Grossman, S. Gupta, U.M. Heller, F. Karsch, Nucl. Phys. B 417 (1994) 289, hep-lat/9309007.
S. Datta, S. Gupta, Phys. Rev. D 67 (2003) 054503, hep-lat/0208001.
M.E. Fisher, A.N. Berker, Phys. Rev. B 26 (1982) 2507.
Five-dimensional N = 1 AdS superspace:

Geometry, off-shell multiplets and dynamics
Sergei M. Kuzenko, Gabriele Tartaglino-Mazzucchelli
School of Physics M013, The University of Western Australia, 35 Stirling Highway, Crawley W.A. 6009, Australia
Received 20 April 2007; accepted 20 June 2007
Abstract
As a step towards formulating projective superspace techniques for supergravity theories with eight
supercharges, this work is devoted to field theory in five-dimensional N = 1 anti-de Sitter superspace
AdS5|8 = SU(2, 2|1)/SO(4, 1) U(1) which is a maximally symmetric curved background. We develop
the differential geometry of AdS5|8 and describe its isometries in terms of Killing supervectors. Various
off-shell supermultiplets in AdS5|8 S 2 are defined, and supersymmetric actions are constructed both in
harmonic and projective superspace approaches. Several families of supersymmetric theories are presented
including nonlinear sigma-models, ChernSimons theories and vectortensor dynamical systems. Using
a suitable coset representative, we make use of the coset construction to develop an explicit realization
for one half of the superspace AdS5|8 as a trivial fiber bundle with fibers isomorphic to four-dimensional
Minkowski superspace.
1. Introduction
In four-dimensional N = 2 Poincar supersymmetry, there exist two powerful formalisms to
construct off-shell manifestly supersymmetric actions: harmonic superspace [1,2] and projective
superspace [36]. Both approaches make use of the superspace R4|8 S 2 and its supersymmetric
subspaces, which were introduced for the first time by Rosly [7] who built on earlier ideas due to
Witten [8]. Both approaches can naturally be extended to the case of d-dimensional supersym* Corresponding author.
E-mail addresses: kuzenko@cyllene.uwa.edu.au (S.M. Kuzenko), gtm@cyllene.uwa.edu.au

(G. Tartaglino-Mazzucchelli).
doi:10.1016/j.nuclphysb.2007.06.014
S.M. Kuzenko, G. Tartaglino-Mazzucchelli / Nuclear Physics B 785 (2007) 3473
35
metry with eight supercharges, for d 6, where the appropriate flat superspace with auxiliary
bosonic dimensions is Rd|8 S 2 . Specifically, the harmonic superspace formulations were developed in [9,10] for d = 5, and in [11] for d = 6. The projective superspace formulations were
developed in [10] for d = 5, and in [12,13] for d = 6.
In projective superspace, off-shell multiplets are reasonably short and can readily be expressed
in terms of 4D N = 1 superfields. The latter property is very appealing from the point of view of
brane(-world) models. It is also expected [14,15] that projective superspace should be relevant in
the context of hybrid string theory [16]. For these and similar possible applications, one actually
needs projective superspace techniques for supergravity. So far, to the best of our knowledge, the
projective superspace approach has been mastered only in the flat case.
In harmonic superspace, the prepotential structure of 4D N = 2 supergravity is well understood [2,17], and similar constructions are clearly applicable in five and six dimensions, see [18]
for the six-dimensional case. What is still missing here, in our opinion, is a properly incorporated
covariant formalism of differential geometry for superfield supergravity, which should be similar
in spirit to the famous WessZumino approach to (the old minimal formulation for) 4D N = 1
supergravity reviewed in [19]. In four-dimensional N = 1 supergravity, it has been recognized
for a long time that the most efficient approach to superfield supergravity occurs if one merges
together and uses, depending on a concrete application, both the covariant and prepotential techniques [20,21].
Unlike the purely prepotential approach pursued in [2,17], this paper is targeted at (making
the first step towards) developing covariant superfield techniques for supergravity theories with
eight supercharges. Our point of departure is as follows. It is known that all information about offshell supergravity formulations (including the structure of possible matter multiplets) is encoded
in the corresponding algebra of covariant derivatives. We would like to use only this input and try
to develop techniques to construct supersymmetric actions both in the harmonic and projective
settings. In this paper we consider one particular supergravity backgroundfive-dimensional
N = 1 anti-de Sitter superspace, AdS5|8 , and explicitly develop harmonic and projective formulations in a covariant fashion using only the language of differential geometry. We believe that
similar ideas should be applicable for a general supergravity background, as well as in four and
six spacetime dimensions. In particular, the case of 4D N = 2 anti-de Sitter superspace1 can be
treated similarly.
This paper is organized as follows. In Section 2 we derive the algebra of the covariant derivatives for 5D N = 1 anti-de Sitter superspace by solving the Bianchi identities. In Section 3 the
isometries of AdS5|8 are realized in terms of Killing supervectors. In Section 4 we introduce
analytic multiplets over the harmonic superspace AdS5|8 S 2 and formulate the harmonic superspace action. Various projective multiplets are defined in Section 5, as well as the projective
superspace action is formulated. A remarkable feature of this supersymmetric action is that it is
uniquely determined by two independent requirements: (i) projective invariance; (ii) invariance
under the isometry group SU(2, 2|1). Some important examples of dynamical systems in the
AdS projective superspace are given in Section 6. An explicit coset construction for one half of
AdS5|8 (Poincar chart) is elaborated in Section 7. Our 5D notation and conventions are collected
in Appendix A.
1 The 4D N = 2 anti-de Sitter superspace was studied in detail in [22] where a manifestly supersymmetric formulation
for the off-shell 4D N = 2 anti-de Sitter higher spin supermultiplets [23] was given. A few years later, some formal
aspects of this superspace were also discussed in [24].
36
2. Covariant derivatives
In this section, we develop the differential geometry of five-dimensional N = 1 anti-de Sitter superspace, AdS5|8 . This is a supersymmetric version of spaces of constant curvature and,
similar to all symmetric spaces, it can be realized as a coset space, specifically AdS5|8 =
SU(2, 2|1)/SO(4, 1) U(1). Group-theoretical aspects of AdS5|8 will be discussed in Section 7.
Let zM = (x m , i ) be local bosonic (x) and fermionic ( ) coordinates parametrizing AdS5|8 ,
where m
= 0, 1, . . . , 4, = 1, . . . , 4, and i = 1, 2. The Grassmann variables i are assumed
to obey a standard pseudo-Majorana reality condition. Since the holonomy group of AdS5|8 is
SO(4, 1) U(1), the superspace covariant derivative DA = (Da , Di ) can be chosen to have the
form
1
DA = EA + iA J + A bc Mb c = EA + iA J + A M .
2
(2.1)
Here EA = EA M (z)M is the supervielbein, with M = /zM , J the Hermitian generator of
the group U(1), Mb c the generators of the Lorentz group SO(4, 1), and A (z) and A bc (z)
the corresponding connections. The Lorentz generators with vector indices (Ma b = Mb a ) and
spinor indices (M = M ) are related to each other by the rule: Ma b = (a b ) M , see

Appendix A for more details regarding our 5D notation and conventions. The generators of the
holonomy group act on the covariant derivatives as follows:

j
J, Di = J i j D ,
(2.2)

1
M , Di = Di + Di .
(2.3)
2
The Hermitian matrix J i j should be traceless, J i i = 0, in order to preserve the pseudo-Majorana
condition enjoyed by the covariant derivatives. The latter condition is equivalent to the fact that
the isotensors2 J ij = j k J i k and Jij = ik J k j are symmetric, J ij = J j i , Jij = Jj i . The fact that
J i j is Hermitian, can be seen to be equivalent to (J ij ) = Jij .
The algebra of covariant derivatives can be reconstructed if we impose the following two requirements: (i) the torsion tensor is covariantly constant3 ; (ii) the group SO(4, 1) U(1) belongs
to the automorphism group. These requirements lead, in particular, to the anstze:
i
j
D , D = 2i ij D + x ij J + f ij M ,
(2.4)

j
Da , Di = C i j (a ) D ,
(2.5)
where

D = a Da ,
(2.6)
and x is a constant parameters, f ij = f j i , C i j is a 2 2 matrix. Eq. (2.5) can be rewritten in the

equivalent form
2 Two-component indices i, j are raised and lowered using the SL(2, C)-invariant antisymmetric tensors ij and
ij
normalized by ik kj = ji and 12 = 21 = 1.
3 Then, in accordance with Dragons theorem [26], the curvature tensor is covariantly constant.

1
j
j
j
i
i
D , D = 2C j D D + D .
37
(2.7)
Note that setting x = m = C i j = 0 gives the flat supersymmetry algebra, see. e.g., [10].
The (covariantly) constant parameters x, f ij and C i j in (2.4) and (2.5) turn out to be considerably constrained on general grounds. Firstly, the tensor f ij must be invariant under the action
of J ,

Jf ij = J i k f kj + J j k f ik = 0 f ij = mJ ij ,
(2.8)
with m a constant parameter. Secondly, we should take care of reality conditions such as
i
D F = (1) (F ) Di F ,
(2.9)
where (F ) is the Grassmann parity of F . They imply that

x = x,
and
Ci
m=m
(2.10)
is anti-Hermitian,

C = Ci j .
C = C,
(2.11)
Of course, we should also guarantee the fulfillment of the Bianchi identities, and this proves to
lead to additional restrictions on the parameters. In particular, the dimension-3/2 Bianchi identity
i j k j k i k i
j
D , D , D + D , D , D + D , D , D = 0
(2.12)
can be shown to imply

i
C i j = J i j ,
2

=

m
x .
2
(2.13)
Imposing the dimension-2 Bianchi identity

j

j

j
Da , Di , D + Di , D , Da D , Da , Di = 0
(2.14)
leads, in particular, to

i

j
ij (b ) Da , Di , D
16
j

j
i

= ij (b ) Di , D , Da D , Da , Di ,
16
[Da , Db ] =
(2.15)
and then
1
[Da , Db ] = mJ 2 Ma b ,
4
where
1
J 2 J ij Jij .
2
Another consequence of the dimension-2 Bianchi identity (2.14) is
1
= m.
4
(2.16)
(2.17)
(2.18)
38
As a result, all the parameters in (2.4) and (2.5) have been expressed in terms of . With the
above conditions taken into account, the remaining dimension- 25 Bianchi identity

Da , Db , Di + Db , Di , Da + Di , [Da , Db ] = 0,
(2.19)
and dimension-3 Bianchi identity

Da , [Db , Dc ] + Db , [Dc , Da ] + Dc , [Da , Db ] = 0,
(2.20)
are satisfied identically.

Let us summarise the results obtained. The covariant derivatives for AdS5|8 obey the algebra
i
j
D , D = 2i ij D 3 ij J 4J ij M ,
(2.21a)
i
j
Da , Di = J i j (a ) D ,
2
[Da , Db ] = 2 J 2 Ma b .
(2.21b)
(2.21c)
It is useful to rewrite (2.21b) in the equivalent form

1
j
j
j
D , Di = iJ i j D D + D .
(2.22)
As follows from (2.21c), the bosonic body of the superspace is characterised by a constant negative curvature, and therefore it is AdS5 . Indeed, since J i j is Hermitian and traceless, we have
1
1
J i j = J I (I )i j J 2 = J ij Jij = J I J J tr(I J ) = J I JI > 0,
2
2
(2.23)
where J I is a real tree-vector, with I = 1, 2, 3, and I are the Pauli matrices. In Section 7, we
will give an explicit (coset space) realization of the geometry described.
Up to an isomorphism, one can always choose J i j (3 )i j , and hence J 11 = J 22 = 0. Then,
1
it follows from (2.21a)(2.21c) that each of the two subsets of covariant derivatives (Da , D )
2
and (Da , D ) forms a closed algebra, in particular

1 1
D , D = 0,
i
1
1
Da , D = J 1 1 (a ) D ,
2
[Da , Db ] = 2 J 2 Ma b .
(2.24a)
(2.24b)
(2.24c)
Therefore, one can consistently define covariantly chiral superfields obeying the constraint
2
D = 0. Unlike the case of 4D N = 1 anti-de Sitter superspace [27], such multiplets can
transform in arbitrary representations of the Lorentz group.
In what follows, it will be useful to deal with a different basis for the spinor covariant deriva
tives. Let us introduce two linearly independent isospinors u+
i and ui ,
+
i
u+i u
i (u u ) = 0 j =

1 +i
u uj ui u+
j ,
(u+ u )
(2.25)
which do not transform under the action of J , that is J u+

i = J ui = 0. Then, defining
Di u
D
i ,
(2.26)

+
J ++ J ij u+
i uj ,
J + J ij u+
i uj ,
J J ij u
i uj ,
39
(2.27)
the relations (2.21a) and (2.21b) become

{D+
, D+ } = 4J ++ M ,
,
D
} = 2(u+ u )iD + 3(u+ u ) J
{D+
, D } = 4J M ,
{D
(2.28a)
4J + M ,
(2.28b)
(2.28c)
(a ) (J ++ D J + D+ ),
2(u+ u )
i
]=
(a ) (J D+ J + D ).
[Da , D
2(u+ u )
[Da , D+
]=
(2.28d)
(2.28e)
Eqs. (2.28d) and (2.28e) are equivalent to

i
1
[D , D+ ] = + J ++ D D
+
D
(u u )
2

i
1
+
+ J + D+ D+
+
D
(u u )
2

i
1
[D , D ] = + J + D D
+
D
(u u )
2

i
1
+
+ J D+ D+
+
D
(u u )
2
(2.29a)
(2.29b)
Under general coordinate and local SO(4, 1) U(1) transformations, the covariant derivatives
change as
DA D
= e DA e ,
A
= B (z)DB + i (z)J + (z)M .
(2.30)
This gauge freedom can be used to impose a suitable WessZumino gauge. The latter can be
chosen such that
1
Da | = a = ea m (x)m + a bc (x)Mb c ,
2
where U | means the independent part of a superfield U (x, ),
U = U (z) = U (x, ),
U | = U (x, = 0)
(2.31)
(2.32)
and a stands for the covariant derivatives of anti-de Sitter space,

[a , b ] = 2 J 2 Ma b .
(2.33)
3. Killing supervectors
Similar to the 4D N = 1 case [21], the isometry group SU(2, 2|1) of AdS5|8 is generated by
those supervector fields A (z)EA which enjoy the property

DA = + iJ + M , DA = 0,
(3.1)
40
where
1
A DA = a Da + i Di = D + i Di ,
4
(3.2)
for some real scalar (z) and symmetric tensor (z) = (z). The A (z)EA is called a
Killing supervector. The set of all Killing supervectors forms a Lie algebra with respect to the
Lie bracket. Given a Killing supervector, it generates a symmetry transformation of matter superfields, which live on AdS5|8 , defined as

= + iJ + M .
(3.3)
Using the (anti)commutation relations (2.21a)(2.21c), Eq. (3.1) can be seen to be equivalent
to

i
1 i

D + 2i i D +
J i j Di j + iJ i j + ji D
4
2

3i + iDi J 2J ij j + j + Di M ,

0=
and from here we deduce the set of Killing supervector equations

Di a = 2i a i ,
i
0 = J i j Di j iJ i j + ji ,
2
iDi = 3i ,

Di = 2J ij j + j .
(3.4)
(3.5a)
(3.5b)
(3.5c)
(3.5d)
Note that (3.5b) is equivalent to the following equations

Di i = 2 ,
(3.6a)
Di + Dj i = 8iJ ij ,

j
j
(a ) Di + D i = 4iJ ij a .
(3.6b)
(3.6c)
It is seen that the parameters of U(1) and Lorentz transformations, and , are uniquely
expressed in terms of the spinor components of the Killing supervector. As to the vector components a of ,which is also uniquely determined in terms of the spinor components of , it obeys
the standard Killing equation
D(a b) = 0.
(3.7)
To prove (3.7), it suffices to represent Da in (3.7) in the form

i a i
(3.8)
D Di ,
8
and then make use of relations (3.5a) and (3.6a).
As is seen from Eqs. (3.5c), (3.6a) and (3.6c), the components of (hence, the Lorentz parameter as well) can be expressed in terms of the scalar parameter as follows:
Da =
i =
i i
D ,
3
(3.9a)
41
i
1
j
j
Jij (a ) Di = 2 2 Jij (a ) Di D ,
2
J
3 J
1 i
i i
= D i = D Di .
2
6
a =
(3.9b)
(3.9c)
This is similar to the situation in 4D N = 2 AdS supersymmetry [23].

We should point out that Eq. (3.9c) implies
0 = Di i = Di Di .
(3.10)
Furthermore, Eqs. (3.5a), (3.5d) and (3.6b) imply

i j k

1
Jj k a
D D 2 a i ,
2
J

1

0 = Di Dj j + 2J ij j + j ,
2

1 i j
0=
D D + Dj Di + 8J ij .
3
0=
(3.11a)
(3.11b)
(3.11c)
From (3.11b) we also deduce

Di Dj j Di Dj j = 0 Di D = 0,
(3.12)
and hence
D = 0.
(3.13)
We conclude that is annihilated by the vector covariant derivatives.

For later applications, we also observe that the relation Di i = 0 and Eq. (3.5a) imply
5
j
D i = iJ i j .
2
(3.14)
4. Harmonic superspace approach

In the previous two sections, we have described the differential geometry of five-dimensional
N = 1 AdS superspace and its isometries. From now on, we turn to constructing off-shell supersymmetric theories in AdS5|8 . This section is devoted to developing a harmonic superspace
approach. To comply with the conventions generally accepted by the harmonic superspace practitioners [1,2], the isospinors u+ and u in (2.28a)(2.28e) will be chosen to obey the following
constraints:
+i
+
+
u
= u
(u
(4.1)
i , ui ) SU(2),
i , (u u ) = 1.
As a first step, it is natural to introduce analytic supermultiplets living on harmonic superspace.
4.1. Analytic multiplets
We start our analysis with the introduction of O(n) supermultiplets living in AdS5|8 . Such a
multiplet is described by a completely symmetric superfield H i1 in (z) = H (i1 in ) (z) (with the
42
symmetrization involving a factor of 1/n!) constrained to enjoy the analyticity condition4

D (i1 H i2 in+1 ) (z) = 0.
(4.2)
It follows from the algebra of covariant derivatives, that this constraint is consistent provided
the superfield is scalar with respect to SO(4, 1). If one associates with H i1 in (z) a superfield
H (n) (z, u) of harmonic charge n,
+ i1 in
H (n) (z, u) = u+
(z),
i 1 ui n H
(4.3)
the analyticity condition (4.2) can be seen to be equivalent to

H (n) (z, u) = 0,
D+
D ++ H (n) (z, u) = 0.
(4.4)
Here D ++ is one of the harmonic derivatives (D ++ , D , D 0 ),
D = ui +i ,
D 0 = u+i +i ui i ,
D ++ = u+i i ,
u
u
u
u
0
= 2D ,
[D ++ , D ] = D 0 ,
D ,D
(4.5)
which form a basis in the space of left-invariant vector fields for SU(2).
Without imposing the analyticity condition, Eq. (4.2), one can consistently define an isotensor
superfield F i1 in (z) = F (i1 in ) (z) that transforms under the action of the isometry group as
follows:
F i1 in = ( + iJ )F i1 in = F i1 in iqnJ (i1 k F i2 in )k ,
(4.6)
F i1 in .
One can associate with

where is the Killing supervector, and q is the J -charge of
+ i1 in
F i1 in (z) the harmonic superfield F (n) (z, u) = u+
u
F
(z).
The
latter obeys the algei1
in
++
(n)
braic constraint D F = 0, and its isometry transformation is

F (n) = F (n) + iq J ++ D F (n) nJ + F (n) ,
(4.7)
where it has been used the fact that u
i are inert under the action of J . It is also worth noting that
the Killing supervector can be rewritten as
= a Da + D
+ D+
.
(4.8)
It is easy to see that the constraint D ++ F (n) = 0 is preserved under the isometry transformations
D ++ F (n) = 0.
If the superfield F (n) is constrained to be analytic, D+
F (n) = 0, then the value of its J -charge
turns out to be uniquely fixed, and namely q = 1. Therefore, the isometry transformation of the
O(n) multiplet is

H (n) = ( + iJ )H (n) = a Da + D
i(J ++ D nJ + ) H (n) . (4.9)
4 In 4D N = 2 supersymmetry, off-shell superfields H (i1 in ) (z) obeying the constrains D (i1 H i2 in+1 ) (z) =
D (i1 H i2 in+1 ) (z) = 0 have a long history. In the presence of an intrinsic central charge, the cases n = 1 and n = 2
correspond to the FayetSohnius hypermultiplet [28] and the linear multiplet [29] respectively. In the absence of central
charge, the case n = 2 corresponds to the tensor multiplet [30]. The case n = 4 was discussed in [31]. The multiplets with
n > 2 were introduced in [32], in the projective superspace approach, and then re-discovered in [5]. They were called
O(n) multiplets in [33]. Their harmonic superspace description was given in [34].
43
It is not difficult to extend the above consideration to include more general multiplets. Within
the harmonic superspace approach [2], one has to deal with superfields of the form Q(n) (z, u),
with n an integer, such that (i) Q(n) (z, u) is a smooth function over the group manifold SU(2)
parametrized by u = (ui , ui + ); (ii) under harmonic phase transformations u exp(i)u ,
the charge of Q(n) (z, u) is equal to n,

Q(n) z, ei u+ , ei u = eni Q(n) (z, u+ , u ) D 0 Q(n) (z, u) = nQ(n) (z, u).
Such a superfield can be represented by a convergent Fourier series (for definiteness, we choose
n 0)
Q(n) (z, u) =
Q(i1 ik+n j1 jk ) (z)u+

i1 uik+n uj1 ujk .
(4.10)
k=0
To realise an action of the U(1) generator J on Q(n) , we define the component superfields in
(4.10) to transform by the law:
(i ik+n j1 jk )
J Qq 1
i ik+n j1 jk )r
= q(2k + n)J (i1 r Qq2
(4.11)
with the same charge q for all the component superfields. This leads to
++
D J ++ D + nJ + )Q(n)
J Q(n)
q (z, u) = q(J
q (z, u).
(4.12)
The J -charge turns out to be uniquely fixed, q = 1, if Qq is covariantly analytic, D+

Qq = 0.
(n)
To summarise, given a covariantly analytic superfield Q (z, u),
(n)
D+
Q(n) = 0,
(n)
(4.13)
the infinitesimal isometry transformation acts on it as follows:

Q(n) = ( + iJ )Q(n)

= a Da + D
+ i(J D ++ J ++ D + nJ + ) Q(n) .
(4.14)
Given two covariantly analytic superfields Q(n) and Q(m) , their product Q(n) Q(m) is covariantly
analytic and transforms as Q(n+m) . In addition, the superfield D ++ Q(n) can be seen to be covariantly analytic and transform as Q(n+2) .
4.2. Harmonic action principle
After having introduced various analytic multiplets in AdS5|8 S 2 , let us turn to constructing
a supersymmetric action. It is worth recalling that in the flat global case ( = 0), the action
principle in 5D harmonic superspace naturally generalizes the original 4D action rule [1,2] and
is given by [10]

1

5
d x du (D )4 L(+4) , (D )4 = D
(4.15)
DDD,

96
i
(+4) is a real analytic Lagrangian
where D
= Di u
i , D are the flat covariant derivatives, and L
+ (+4)
= 0.
of harmonic charge +4, D L
We would like to generalize the flat action to the case of AdS5|8 using the following ansatz:
44
S = S 0 + a1 S 1 + a2 S 2

= d5 x e du (D )4 + a1 J (D )2 + a2 (J )2 L(+4) ,
(4.16)
L(+4) = 0,
where L(+4) is now covariantly analytic, D+
1
(4.17)
D D D D ,
96
and a1 , a2 are two constants to be determined. It is assumed that the above action is evaluated
in WessZumino gauge (2.31), using the bar projection (2.32), and as usual e stands for the
determinant of the vielbein, e = det(em a ), with em a ea n = m n .
In accordance with the definition of S, there are several rules for integration by parts which
one can use in practice:

d5 x e du Da Qa = 0,
(4.18)

d5 x e du D ++ Q | = d5 x e du D Q++ | = 0,
(4.19)

d5 x e du J Q(0) = 0.
(4.20)
(D )2 = D D
,
(D )4 =
Here Q(0) is a covariantly analytic superfield of harmonic charge 0.

Our aim is to find the constants a1 , a2 for which S is invariant under the isometry transformations of AdS5|8 . Let us first compute the variation of S0 under infinitesimal isometry
transformations. Due to (3.1), we have

(D )4 L(+4) = (D )4 L(+4) = (D )4 ( + iJ )L(+4)

= + iJ + M (D )4 L(+4) = ( + iJ )(D )4 L(+4) . (4.21)

Since L(+4) is covariantly analytic, we obtain

5
S0 = d x e du ( + iJ )(D )4 L(+4)

(D )4 D+
, (D )4 L(+4) .
= d5 x e du + D
(4.22)
Here we have also used Eqs. (4.18) and (4.20).

(D )4 in (4.22), we observe that D[ D D D
D = 0, and then
To compute D
D
0 = 5 D[ D D D
]
[D
D D D D

+ D D
D D D

+ D D
D D D
+ D
D D D D

+ D D D D D ].
(4.23)
Moving D
in each term to the left gives
)4 = J [4M D D D + 3D M D D
(
D
D

120

+ 2D D M D + D D D M ].
(4.24)
45
This can be further transformed by moving all the Lorentz generators to the right and factors of
D
to the left using iteratively the algebra of covariant derivatives. We end up with

)4 = D 5 J (D )2 + 3(J )2
D
(
D
12

1
1 2 8
2
2
J D (D ) (J ) D + (J ) D M
M .
8
3
2
(4.25)
The expression in the second line does not contribute when acting on a Lorentz scalar such as
L(+4) .
To compute [D+
, (D )4 ] in (4.22), we should iteratively use the algebra of covariant deriva
tives. This is an obvious but tedious procedure. The result is:
1
3
1
, (D )4 = iD D (D )2 + J D
(D )2 J iD D
D+
4
8
3
13
1
2 J J D
+ J + D
(D )2 52 J J + D
2
4
1
3
3
J (D )2 D+
+ J D
D D+ J D D
D+
8
8
8
55
3
+ (J )2 D+
+ J iD D M 2 J J D M
12
2
1
+ 22 J J + D M J + D (D )2 M
4
2 J J + D M M .
(4.26)
Using the relations (4.25) and (4.26), and also the integration by parts identities (4.18) and (4.20),
variation (4.22) turns into

5
)2 + 3(J )2 D L(+4)
(
D
S0 = d5 x e du + J D

12

1
3

(D )2
+ d5 x e du iD D (D )2 + J D
4
8

1
13

+ J iD D 2 J J D
3
2

1
)2 + 52 J J + D L(+4) .
J + D
(
D
(4.27)

4
Finally, it remains to note J = J + J + , and also make use of Eq. (3.14) projected
to the minus-harmonics
5
iD = (J + J + ).
2
As a result, the variation of S0 under the isometry transformations takes the final form:
(4.28)
46

2
8
S0 = d x e du J + D
(D )2 (J )2 + D
3
3

32
+ 2 J J + D
L(+4) .
(4.29)
The next step is to compute the variation of the functional S1 appearing in our action (4.16).
Here the procedure is the same as for S0 . Varying

(D )2 L(+4) = (D )2 ( + iJ )L(+4) = ( + iJ )(D )2 L(+4) ,
(4.30)
we get
S1 =

d5 x e

du J + D
(D )2 J + D+
, (D )2 L(+4) .
(4.31)
Using the algebra of covariant derivatives gives

+
D , (D )2 = 4iD D 6J D
+ 12J + D
2J D+
+ 8J + D M .
As a result, the variation of S1 is

S1 = d5 x e du J + D
(D )2 + 4(J )2 + D

L(+4) .
162 J J + D
(4.32)
(4.33)
It is seen that (4.33) is proportional to (4.29). Therefore, our ansatz (4.16) leads to the unique
supersymmetric action: a1 = 2/3 and a2 = 0.
The supersymmetric action is

2 2 (+4)
5
4
D+
L(+4) = 0.
S = d x e du (D ) + J (D ) L
(4.34)
,
3
This is the main result of this section.
By construction, the Lagrangian in (4.34) is a covariantly analytic superfield of harmonic
charge +4. It should be also chosen to be real with respect to analyticity preserving conjugation
[1] (see also Section 5.1), and then action (4.34) can be seen to be real. Otherwise, L(+4) is
completely arbitrary. Therefore, a great many flat superspace actions [2] can be lifted to the AdS
superspace. For instance, an off-shell hypermultiplet can be realized in terms of a covariantly
analytic superfield q + (z, u) and its conjugate q + (z, u), with respect to the analyticity preserving
conjugation. To describe its dynamics, one can choose
L(+4) = q + D ++ q + + (q + q + )2 ,
(4.35)
with a coupling constant.

5. Projective superspace approach
In the projective superspace approach to d-dimensional theories with eight supercharges, one
deals with superfields that live in Md|8 S 2 , where Md|8 denotes the conventional superspace,
d 6, and S 2 the two-sphere. Such superfields are required to (i) be Grassmann analytic, i.e. to
be annihilated by one half of the supercharges; (ii) be holomorphic on an open domain of S 2 .
47
The latter requirement is equivalently achieved by considering superfields (n) (z, u+ ) which are
holomorphic functions of a single isotwsitor u+i C2 {0}, and have definite degree of homogeneity with respect to u+ , (n) (z, cu+ ) = cn (n) (z, u+ ). The variables u+i can be viewed
as homogeneous coordinates for CP 1 . A second linearly independent isotwistor, ui , is only
required (as a purely auxiliary means, without any intrinsic significance) for constructing a supersymmetric action which was proposed originally in four dimensions in [3] and then reformulated
in [4] in terms of the projective isotwistor u+i . The terminology isotwistor is due to [35,36].
In the flat global case, the 5D N = 1 extension of the 4D N = 2 supersymmetric action [4] is
as follows5 [25]:
+ +i

ui du
1
5
)4 L++ (z, u+ ),
d
x
(
D
(5.1)
2
(u+ u )4
where
D+
L++ (z, u+ ) = 0,
L++ (z, cu+ ) = c2 L++ (z, u+ ),
c C .
The action is invariant under arbitrary projective transformations of the form

a 0
GL(2, C).
(ui , ui + ) (ui , ui + )R, R =
b c
(5.2)
(5.3)
This gauge-like symmetry implies that the action is actually independent of u

i . It can be fixed
by imposing, for instance, the gauge
u+i (1, ) = i u+
i (, 1) = i ,
ui (0, 1) u
i (1, 0),
(5.4)
in which the action (5.1) reduces to the standard 5D N = 1 projective superspace action [10,25].
5.1. Projective multiplets
Here we introduce several off-shell projective multiplets that are most interesting from the
point of view of model building. By definition, a projective superfield Q(n) (z, u+ ) lives on the
anti-de Sitter superspace and depends parametrically on a non-vanishing isotwistor u+i = 0. It is
defined to be analytic,
D+
Q(n) = 0,
(5.5)
and transform by the rule

Q(n) = ( + iJ )Q(n)
under the isometry group. We specify J to act on
J Q(n) =
(5.6)
Q(n)
as follows
J ++ D Q(n) nJ + Q(n)
.
(u+ u )
(5.7)
This definition involves an external isotwistor u

i subject to the only requirement
(u+ u ) = 0.
5 Note that the action given in Eq. (B.1) of [25] contains a wrong overal factor of
(5.8)
1.
48
Since Q(n) is independent of u , it is natural to require J Q(n) to be independent of u as well,

that is
J ++
(n)
Q(n) nJij u+j Q(n) = u+
i JQ .
u+i
(5.9)
Contracting this with u+i gives

J ++ u+i
Q(n) = nJ ++ Q(n) .
u+i
(5.10)
Therefore, Q(n) is a homogeneous function of u+ of degree n,

Q(n) (z, cu+ ) = cn Q(n) (z, u+ ),
c C .
(5.11)
The Q(n) will be called a projective superfield of weight n.

As is obvious, the complex conjugate of an analytic superfield is not analytic. However, one

+i , which is
can introduce a generalized, analyticity-preserving conjugation [1,3,7], u+i u
+i
obtained by composing the complex conjugation, u u+i , with the antipodal map u+i
u+
i . In what follows, it is called smile-conjugation. It is thus defined to act on the isotwistor
+
u = (u+i ) by the rule6
u+ u + = i2 u+ ,

u+i = u+i ,
(5.12)
with 2 the second Pauli matrix. Its action on the projective superfields is defined to be
(n) (u+ ) Q
(n) (u + ),
Q(n) (u+ ) Q
(n) = (1)n Q(n) .

Q
(5.13)
(n) (u+ ) is a homogeneous function of u+ of degree n, that is Q

(n) (cu+ ) =
It is clear that Q
(n) (u+ ), with c C . Due to the identity
cn Q
(n)
+ (n)
(n) ,
D
Q = (1) (Q ) D+ Q
(5.14)
the smile-conjugation indeed preserves analyticity.

It is important to note that, in accordance with (5.13), for an even integer weight, n = 2p, one
can consistently define real projective superfields R (2p) with respect to the smile-conjugation:
R (2p) = R (2p) .
Now, let us show that the smile-conjugation is compatible with the superfield transformation
law (5.6). To evaluate the smile-conjugate of J Q(n) , Eq. (5.7), we conventionally define the
operation of smile-conjugate for u = (ui ) to be identical to that we have already chosen for
the isotwistor u+ , that is
u u = i2 u ,

ui = ui .
(5.15)
We should emphasize that such a definition is completely conventional in the sense that the
projective superfields are independent of the isotwistor u . Then it holds
6 Due to projective invariance, u+i cu+i , the smile-conjugation could be also defined as u+ u + = i u+ ,
2
instead of (5.12).

= J ,
J
= D ,

D
+ = J + ,
J
+ u ) = (u+ u ).

(u
49
(5.16)
This implies
(n) ,
J
Q(n) = J Q
(5.17)
and the smile-conjugate of the transformation law (5.6) is

(n) = ( + iJ )Q
(n) = Q (n) .
Q
(5.18)
Therefore, the smile-conjugation preserves the superfield transformation laws under the isometry
group.
As is known, the space CP 1 can be covered by two charts that are defined in terms of u+ =
+1
(u , u+2 ) as follows: (i) the north chart on which u+1 = 0; (ii) the south chart on which u+2 = 0.
As will be described below, the projective action involves the line integral over a closed contour
in CP 1 , and this contour can be chosen to lie inside one of the coordinate charts. The latter can
be chosen to be the north chart, and that is why our local considerations will be mainly restricted
to that chart. In the north chart, we can introduce a projective invariant complex coordinate
defined as u+i = u+1 (1, ), with = u+2 /u+1 . Since u +i = (u+2 , u+1 ), the smile-conjugation
acts as follows:
1
.
(5.19)
The simplest solution to Eq. (5.11) is the O(n) multiplet defined by Eqs. (4.2) and (4.3).
This multiplet is globally defined on CP 1 . Allowing for singularities at some points in CP 1
offers the possibility to generate many more interesting supermultiplets. For example, a charged
hypermultiplet is described by a weight-one projective superfield + (u+ ) being holomorphic on
CP 1 {N }, where the North pole is identified with u+i (0, 1). We can represent + (u+ ) as

+ (u+ ) = u+1 + u+i /u+1 u+1 ( ),
(z, ) =
k (z) k .
(5.20)
k=0
Its smile-conjugate + (u+ ) is holomorphic on CP 1 {S}, where the South pole is identified
with u+i (1, 0). We can represent + (u+ ) as
+ (u+ ) = u+2 + (u+i /u+2 ) u+2 ( ),
(z, ) =
k=0
k (z)
(1)k
,
k
(5.21)
with k (z) the complex conjugate of k (z). To describe an off-shell vector multiplet, one should
use a real weight-zero projective superfield V (u+ ) being holomorphic on CP 1 {N S}. It can
be represented as
V (z, ) =
k=
Vk (z) k ,
Vk = (1)k Vk .
(5.22)
50
5.2. Projective action principle

Our aim here is to find a generalization of the flat superspace action (5.1) to the case of AdS5|8
superspace. We start with the following ansatz7
S = S0 + 1 S1 + 2 S2
+ +i

4
ui du
1
5
) + 1 J (D )2 + 2 (J )2 L++ (z, u+ ).
d
x
e
(
D
=
2
(u+ u )4
(5.23)
Here L++ (z, u+ ) is a covariantly analytic superfield, D+
L++ = 0, which is homogeneous in
+
ui of degree +2. The line integral in (5.23) is carried out over a closed contour, = {u+
i (t)},
in the space of u+ variables. The integrand in (5.23) involves a constant (i.e. time-independent)
+
isotwistor u
i subject to the only condition that u (t) and u form a linearly independent basis
+
at each point of the contour , that is (u u ) = 0.

Our first requirement is that the action (5.23) be invariant under the projective gauge transformations (5.3). First of all, it is obvious that (5.23) is invariant under arbitrary scale trans+
formations u+
i (t) c(t)ui (t), with c(t) = 0. It is thus only necessary to analyse projective
transformations of u of the form
+
u
i u
i = a(t)ui + b(t)ui (t),
a(t) = 0.
(5.24)
and
should be time independent, the coefficients should obey the equations
Since both
(using the notation f df (t)/dt , for a function f (t)):
a=b
(u+ u+ )
,
(u+ u )
b = b
(u+ u )
.
(u+ u )
(5.25)
As is obvious, the action (5.23) is invariant under arbitrary scale transformations u

i a(t)ui ,
with a = 0. Therefore, it only remains to analyse infinitesimal transformations of the form
+
u
i = b(t)ui , with b(t) obeying the differential equation (5.25). This transformation induces
the following variations:
D
= bD+
,
J = 2bJ + .
(5.26)
Let us start by evaluating the variation of S0 . Using the condensed notation

d++
we obtain
S0 =
+i
1 u+
1 (u+ u+ )
i du
=
dt,
2 (u+ u )4
2 (u+ u )4
++
(5.27)

d5 x e (D )4 L++

d++ b d5 x e 3D
D {D+ , D
} + 2D
{D+ , D }D

96

L++ .
+ {D+
, D }D D
(5.28)
7 An alternative approach to introduce the projective action consists in using a proper generalization of the procedure
given in [37]. The latter allows one to derive the projective action as a singular limit of the harmonic action.
51
Now, making use of the covariant derivatives algebra (2.28a)(2.29b) and the identities
1
J ++ D
),
(J + D+
(u+ u )
1
[J, D
] = + (J D+
J + D
),
(u u )
]=
[J, D+
(5.29a)
(5.29b)
we can systematically move in (5.28) all spacetime derivatives to the left (neglecting total
spacetime derivatives) and the J operator to the right. This gives

7
3
++
5
S0 = d b d x e J + (D )2 (u+ u )(D )2 J
12
8

11
+ (u+ u )2 J J L++ .
(5.30)
2
To transform the second and third terms in the square brackets, we should first recall how J acts
on the Lagrangian,
J L++ =
1
(J ++ D L++ + 2J + L++ ).
(u+ u )
(5.31)
Since L++ is a homogeneous function of degree two, we also have
d ++ (u+ u ) +i
(u+ u+ ) ++
++
L
L = + u
D L
dt
(u u )
u+i
(u+ u )
=2
(u+ u ) ++ (u+ u+ ) ++
L + D L .
(u+ u )
(u u )
The latter results leads to
(u+ u+ )J L++ = J ++
d ++
(u+ u )
(u+ u+ )
L 2 + J ++ L++ + 2 + J + L++ .
dt
(u u )
(u u )
One more technical observation,
(5.32)
(u+ u )
(u+ u+ )
d ++
= 2 + J ++ 2 + J + ,
J
dt
(u u )
(u u )
(5.34)
allows us to obtain the following identity:

(u+ u+ )
d
bJ ++ ++
(u+ u+ ) + ++
++
b + 3JL =
L
J L .
+
4b
dt (u+ u )3
(u u )
(u+ u )4
(5.33)
Then (5.30) becomes

25
S0 = d++ b d5 x e J + (D )2 + 222 J J + L++ .
12
Using the same procedure, for S1 and S2 we find

S1 = d++ b d5 x e 2J + (D )2 + 482 J J + L++ ,

++
S2 = d b d5 x e42 J J + L++ .
(5.35)
(5.36)
(5.37a)
(5.37b)
52
The relations obtained show that the requirement of projective invariance, S = S0 + 1 S1 +

2 S2 = 0, uniquely fixes the coefficients in (5.23) as follows: 1 = 25/24 and 2 = 18. We
end up with the projective-invariant action

+ +i

ui du
1
5
)4 + 25 J (D )2 18(J )2 L++ . (5.38)
d
x
e
(
D
S=

2
24
(u+ u )4
Now, we are going to demonstrate that (5.38) is supersymmetric, that is this action is invariant
under the isometry group of AdS5|8 . This requires us to carry out calculations that are very similar
to those presented in Section 4 for the harmonic case. But there are two technical features being
specific for the projective case: (i) unlike the harmonic case, we have (u+ u ) = 1 in general,
and therefore it is necessary to keep track ofthe factors of (u+ u ); (ii) unlike the harmonic
superspace identity (4.20), in general we have d++ J Q = 0. In all variations involving the
U(1) generator J , we will systematically move J s to the right to hit the Lagrangian L++ , so that
Eqs. (5.31) and (5.33) can be applied.
We start by computing the variation of S0 under the infinitesimal isometry transformation.
Making use of

(D )4 L++ = (D )4 ( + iJ )L++ = ( + iJ )(D )4 L++
(5.39)
gives

S 0 =

1 + 4
, (D )4
D (D ) D+
(u+ u )

++
4
4
i J, (D ) + (D ) J L .
d
++
d xe
(5.40)
(D )4 L++ , we note that Eq. (4.25) holds even if (u+ u ) = 1, since in the derivaTo evaluate D
tion of (4.25) we only used Eq. (2.28c) and the commutation relations of the Lorentz generator
M with the covariant derivatives, and both results are clearly not affected by the normalization
of (u+ u ). Therefore, for the first term on the right of (5.40) we have

5
4 ++
2
2
D (D ) L = J D (D ) + 3(J ) D L++ .
(5.41)
12
, (D )4 ], which appears in (5.40), we have derived Eq. (4.26) in the
For the operator [D+
harmonic case. Now, in evaluating the second term on the right of (5.40), we should take care of
the factors of (u+ u ), as well as to move the U(1) generator J to the right. This gives
1
3
7
, (D )4 = (u+ u )iD D (D )2 + (u+ u )D

(D )2 J J + D
(D )2
D+
4
8
8
11
17
(u+ u )J iD D (u+ u )2 J D
J
6
4
33
+ 2 J J + D
(5.42)
+ ,
where the dots denote those terms which do not contribute when acting on Lorentz scalar and
analytic superfields such as the Lagrangian L++ . Inserting (5.42) into S0 , one can get read of
the terms with vector covariant derivatives by taking into account the integration by parts rule
53
(4.18) and
iD =
5
(J + J + ).
2(u+ u )
(5.43)
To evaluate the contributions to S0 which contain J L++ , we note that Eqs. (5.33) and (5.34)
imply

d J ++ L++
(u+ u+ )
(u+ u+ ) + ++
++
J
L
=
J L .
(5.44)
+
4
dt (u+ u )4
(u+ u )4
(u+ u )5
The latter observation tells us

+

++
++
++ J
++
= 4 d
O(u )L ,
d O(u )J L
+
(u u )
for any operator O(u ) independent of u+ . It follows that

++
d5 x e D+
d
, (D )4 L++

5
J
1
++
5
d x e + + D
=
(D )2
d
2
(u u )
4

11 + 5 + ++
+ J D + J D L .
6
3
Completely similar considerations, using also D = 0 (3.13), give

++
d5 x e J, (D )4 L++
d

J +
++
5
d x e + 4(D )4
= d
(u u )

25 2
2
++
+ J (D ) 22(J ) L .
12
As a result, the variation S0 can be represented in the form

25 J + 2 91 (J )2 +
D (D ) +
D
S0 = d++ d5 x e
24 (u+ u )
12 (u+ u )
(5.45)
(5.46)
(5.47)

(J )2 J +
25 2 J J + 25 J J + 2
++
D
+
)
22
(
D
L
.
6 (u u )
12 (u u )
(u u )
(5.48)
The variations S1 and S2 can be computed by similar means. The results are:

J + 2
(J )2 +
D
(
D
)
+
10
D
S1 = d++ d5 x e
(u+ u )
(u+ u )

2 J J +
J J + 2
(J )2 J +
++
4
D 2
(D ) 48
L ,
+
(u u )
(u u )
(u u )
(5.49a)

)2
)2 J +

(J
(J
4
+ D
L++ .
S2 = d++ d5 x e
(5.49b)
(u+ u )
(u+ u )
+
54
Collecting all the results obtained, we conclude

S = 0,
(5.50)
and therefore the action (5.38) is supersymmetric. Actually, it proves to be the only supersymmetric action in the family (5.23). It is quite remarkable that projective invariance implies
supersymmetry and vice versa.
6. Dynamical systems in projective superspace
In this section we study in more detail the projective multiplets and then consider several
important supersymmetric theories. To simplify the analysis, it is useful to choose the projective
gauge u
2 = 0. Without loss of generality, one can also work in a representation of the algebra in
which J 11 = J 22 = 0, and hence J = 0.
6.1. Projective multiplets revisited
In each of the two coordinate charts for CP 1 , one can describe the projective multiplets by
superfields invariant under the projective transformation (5.11). Let us restrict our consideration to the north chart. Given a complex projective multiplet of weight n, Q(n) (u+ ), it can be
equivalently described by a holomorphic function Q[n] ( ) defined as follows:

n
Q[n] ( ) Q(n) (1, ).
Q(n) u+i = u+1 Q[n] ( ),
(6.1)
Here Q[n] ( ) is clearly invariant under (5.11). For the smile-conjugate of Q(n) (u+ ), we get

n [n]
[n] (1/ ).
( ), Q [n] ( ) = Q
Q (n) u+i = u+2 Q
(6.2)
Given a real projective multiplet R (2p) (u+ ), with respect to the smile-coinjugation, it can be
represented
p

R (2p) (u+ ) = iu+1 u+2 R [2p] ( ), R [2p] ( ) R [2p] (1/ ) = R [2p] ( ).
(6.3)
The most general form for Q[n] (z, ) is
Q[n] (z, ) =
k
Q[n]
k (z) .
(6.4)
k=
11 = J 22 = 0), the action of the operator J on our
In the projective gauge chosen (u
2 = 0, J
superfield becomes

1
[n] +
++
+
[n] +
12
+2
J Q (u ) = + (J D nJ )Q (u ) = J
n 2u
Q[n] (u+ ).
(u u )
u+2
Then, since the isotwistor u+i is neutral under the action of J , it holds

+1 n
n
n

u
J Q[n] ( ) = J Q(n) (u+ ) = J 12 n u+1 Q[n] ( ) 2u+2 +2 u+1 Q[n] ( )
u

+1 n 12
J
= u
nQ[n] ( ) 2 Q[n] ( ) ,
and therefore
[n]
J Q (z, ) = J
55
n Q[n] (z, ),
2
[n]
1
J Q[n]
k (z) = (2k n)J 1 Qk (z).
(6.5)
In the case of a real superfield R (2p) (z, u+ ) = (iu+1 u+2 )p R [2p] (z, ), we have for R [2p]
+
R [2p] (z, ) =
[2p]
Rk
(z) k ,
[2p]
[2p]
R k = (1)k Rk .
(6.6)
k=
The operator J is represented as follows:

J R [2p] (z, ) = 2J 1 1
[2p]
R (z, ),
[2p]
J Rk
[2p]
(z) = 2kJ 1 1 Rk
(6.7)
(z).
Let us analyse the implications of the analyticity condition, D+

Q(n) (u+ ) = 0. It is useful
to change the representation for the projective superfields, Q(n) (u+ ) Q[n] ( ). We then have
1
2
Q[n] ( ) = u+1 ( D D )Q[n] ( ), and therefore the analyticity condition is equivalent to
D+
D Q[n] ( ) = D Q[n] ( ).
2
(6.8)
For the component superfields Q[n]

k (z), this implies
[n]
D Q[n]
k = D Qk1 .
2
(6.9)
1
It is natural to think of D and D as the covariant derivatives associated with two 5D Dirac
spinor coordinates, 1 and their conjugates 2 . It then follows from (6.8) that the dependence of
Q[n] ( ) on 2 is completely determined by the dependence of Q[n] ( ) on 1 .
Suppose that the expansion of Q[n] ( ) in powers of terminates from below
[n]
Q (z, ) =
k
Q[n]
k (z) .
(6.10)
k=L
Then, Eq. (6.9) tells us that the two lowest components of Q[n] are constrained as follows:
D Q[n]
L = 0,
2 2 [n]
D QL+1 = 12J Q[n]
L ,
2
where
D i
2
= Di Di ,
i = 1, 2.
(6.11)
(6.12)
[n]
Therefore, Q[n]
L is a five-dimensional chiral superfield, while QL+1 a complex linear superfield.
[n]
[n]
The union of QL and QL+1 forms a 5D analogue of the famous chiral-nonminimal doublet in
4D supersymmetry [38].
Given a real O(2) multiplet H (2) (z, u+ ), we can represent H [2] (z, ) in the form
H [2] (z, ) =
(z) + G(z) (z),
(6.13)
56
where is a five-dimensional chiral superfield, and G a real linear superfield,

2 2
2
= G.
D = 0,
D G = 0, G
(6.14)
If the expansion of Q[n] ( ) in powers of terminates from above,

L
Q[n] (z, ) =
k
Q[n]
k (z) ,
(6.15)
k=
then Eq. (6.9) implies that the two highest components of Q[n] are constrained as follows:
1 2 [n]
1
D QL1 = 12J Q[n]
D Q[n]
(6.16)
L = 0,
L .
[n]
Therefore, Q[n]
L is a five-dimensional antichiral superfield, while QL1 a complex antilinear
superfield.
For further analysis, it is useful to switch from the 5D four-component spinor notation to the
4D two-component one by representing
i

D
Di
Di =
=
.
D i
ij D
j
In such a notation, the algebra of covariant derivatives (2.21a)(2.21c) takes the form
i
j
D , D = 2 ij D5 3 ij J 4J ij M ,
i

D , D j = 2iji a Da 4J i j M ,

D i , D j = 2ij D5 3ij J 4Jij M ,

i

i
j
Da , Di = J ij (a ) D j ,
Da , D i = Jij ( a )
D ,
2
2
1
1

D5 , D i = Ji j D j ,
D5 , Di = J i j Dj ,
2
2
[Da , D5 ] = 2 J 2 Ma5 .
[Da , Db ] = 2 J 2 Mab ,
In the two-component spinor notation, the analyticity condition
to
D Q[n] ( ) = D Q[n] ( ),
2
D+
Q[n] ( ) = 0
1
D 2 Q[n] ( ) = D 1 Q[n] ( ).
(6.17a)
(6.17b)
(6.17c)
(6.17d)
(6.17e)
(6.17f)
is equivalent
(6.18)
For the component superfields Q[n]

k (z), this implies
[n]
D Q[n]
k = D Qk1 ,
2
[n]
D 2 Q[n]
k = D1 Qk+1 .
(6.19)
By analogy with the flat case, these constraints indicate an interesting interpretation. Let us intro1
2
duce two sets of spinor derivatives, (D , D 1 ) and (D , D 2 ) which can be viewed as the covariant
derivatives corresponding to two different sets of Grassmann variables 1 and 2 . Then, the
above constraints imply that the dependence of the projective superfields on 2 is uniquely
determined in terms of their dependence on 1 . Unlike the flat case, such an interpretation is
somewhat limited in the sense that one cannot consistently switch off the variables 2 (what
would be necessary for reducing the multiplets to 4D N = 1 superfields). It follows from the al1
gebra of covariant derivatives, specifically from Eq. (6.17d), that [Da , D ] = 2i J 12 (a ) D 2
57
D , and therefore the commutation relations mix all the spinor

and [Da , D 1 ] = 2i J 12 ( a )
derivatives. This is an important difference between the flat and curved cases.
The constraints (6.19) simplify if the series in (6.4) or (6.6) is bounded from below (above).
Consider a real O(2n) multiplet H (2n) (z, u+ ). In accordance with the above general consideration, it can be described by the superfield H [2n] (z, ) which is defined by H (2n) (z, u+ ) =
(iu+1 u+2 )n H [2n] (z, ), and can be represented in the form
2
H [2n] (z, ) =
+n
Hk[2n] (z) k ,
[2n]
H k[2n] = (1)k Hk
.
(6.20)
k=n
The analyticity constraints (6.19) imply that the two lowest component superfields are constrained by
[2n]
= 0,
D 1 Hn
[2n]

[2n]
[2n]
= (4D5 + 6J )Hn
= 4D5 12nJ 1 1 Hn
,
(D 1 )2 Hn+1
where we have defined
i 2
D Di Di ,
(D i )2 D i D i .
(6.21)
(6.22)
Consider an arctic multiplet of weight n 0, (n) (u+ ), defined to be holomorphic on CP 1

{N}. It can be represented as

n
(n) (z, u+ ) = u+1 [n] (z, ),
[n] (z, ) =
k[n] (z) k .
(6.23)
k=0
Then the constraints on the two lowest components superfields are

D 1 0[n] = 0,

(D 1 )2 1[n] = (4D5 + 6J )0[n] = 4D5 6nJ 1 1 0[n] .
(6.24)
In the flat superspace limit, 0, the constraints (6.21) and (6.24) reduce to those given in
[10].
6.2. Projective action
Here we turn to a more detailed analysis of the projective action (5.38). In the projective
11 = J 22 = 0) used throughout this section, we have J = 0, and therefore
gauge (u
2 = 0, J
the projective action simplifies
+ +i

ui du
1
d5 x e(D )4 L++ .
S=
(6.25)
+
4
2
(u u )
Of course, the Lagrangian L++ should be real with respect to the smile conjugation, and can be
represented as
L++ (z, u+ ) = iu+1 u+2 L(z, ).
Then, the action turns into

2 2
d
1
d5 x e D 1 D 1 L(z, ),
S=
32
2i
(6.26)
(6.27)
58

1
= ( + + ).
where we have taken into account the fact that {D , D } = 0 in the projective gauge, and also
made use of the identity
2
2 [n]
D2 Q = 2 D1 Q[n] + 12 J Q[n] ,
Using the relation

(6.28)
we can express action (6.27) in the equivalent forms

2 2 2

d
1
d5 x e D 1
S=
D
12J L(z, ),
32
2i
and
1
S=
32
d
2i

2 2
d xe D
5
(6.29a)

1 2 1
+ 12J L(z, ),
D
(6.29b)
where we have used the identities

1 2 2
2
2
= 18J 1 1 D + 4iD D2 + 6D J 8J 1 1 D2 M ,
D , D
1 2 2 2
D , D
1 2

1 2
1
2
= 4iD D , D + 6 D , D + 962 J 1 1 J 8J 1 1 D , D M .
(6.30)
(6.31)
Then we can represent the action in the form

1 2 2 2
2 1 2
d
1
d5 x e
S=
+ 24J 1 1 D 1 + D 2
D , D
L(z, )
32
2i
(6.32)
which makes manifest the reality of S with respect to the smile-conjugation.
It can be seen from the above relations that there exists a natural gauge freedom in the
choice of L++ . It occurs in the three incarnations:
L++ L++ + ++ + ++ ,
L++ L++ + iJ ++ ( + ),
L
++
++
+H
++
(6.33)
(6.34)
(6.35)
++
H ++
and arctic multiplets (6.23) of weight +2 and 0, respectively, and

a real O(2)
with
multiplet.
It is also instructive to express the action in a 4D N = 1 form by switching to the twocomponent spinor notation
1 1
2 2
D
D
D
D
1
2
=
,
D =
=
.
D =
(6.36)
D
D2
D
D 1
Using the analyticity conditions (6.18) we can express D 2 via D 1 . As a result our action (6.25)
becomes

1 2
d
1 1 2 2
5
1
d xe
S=
(6.37)
D (D1 ) J 1 D
L(z, ).
2i
16
Using the identities
1 2

1
D 1 , D
= 8J 1 1 D 2 4i a Da D + 8J 1 1 D1 M ,
(6.38)
2

2
D1 , (D 1 )2 = 16J 1 1 D 1 D 2 + 16J 1 1 D1 D + 962 J 1 1 J

a 1
+ 4i Da D , D 1 + 8J 1 1 D1 , D 1 M ,
the action can also be rewritten in the following form

d
1 2 1 2 1 2
d5 x e
S=
(D1 ) D + J 1 (D1 ) L(z, ),
2i
16
59
(6.39)
(6.40)
or in the manifestly real form

1 2
1 1 2 2 1
1 2
d
5
1
d xe
D , (D1 ) J 1 D + J 1 (D1 ) L(z, ).
S=
2i
32
2
2
(6.41)
As compared with the flat superspace action [10], the second and third terms on the right of
(6.41) are due to the non-vanishing curvature.
6.3. Nonlinear sigma-models
We consider a system of interacting artic weight-one multiplets + (z, u+ ) and their smileconjugates + described by the Lagrangian
L++ = iK( + , + ),
K( I , J )
(6.42)
L++
L++ (z, u+ )
a real analytic function. Since

=
is required to be a weightwith
two projective superfield, the potential K has to respect the following homogeneity condition

I
I
= 2K(, ).
(6.43)
+
K(, )
I
I
For L++ to be real, it is sufficient to require a stronger condition
= K(, ).
K(, )
(6.44)
I
Such a Lagrangian corresponds to the superconformal sigma-model introduced in [25]. Then,
representing + (z, u+ ) = u+1 (z, ) and + (z, u+ ) = u+2 (z, ), we can rewrite the Lagrangian in the form
I
L++ (z, u+ ) = iu+1 u+2 L(z, ),
L = K(, ).
(6.45)
Because of freedom (6.33) in the choice of Lagrangian, we can generalize the above construc
tion by replacing K( I , J ) in (6.42) with

() = 2(),
K
I , J = K I , J + I J , I
(6.46)
I
with () a holomorphic homogeneous function of degree +2. Then, the homogeneity condition (6.44) turns into
= K
(, )
+ () + (
).
(6.47)
K
(, )
I
We can also consider a system of interacting arctic weight-zero multiplets (z, u+ ) and their
smile-conjugates described by the Lagrangian
I
i
L++ = J ++ K( , ),
2
(6.48)
60
with K( I , J ) a real function which is not required to obey any homogeneity condition. Due
to the gauge freedom (6.34), the action is invariant under Khler transformations of the form
),
K( , ) K( , ) + ( ) + (
(6.49)
with a holomorphic function. Such dynamical systems generalize the hyper-Khler sigmamodels on cotangent bundles of Khler manifolds [3941].
6.4. Vector multiplet and ChernSimons couplings
An Abelian vector multiplet can be described by a weight-zero real projective superfield
V (z, u+ ) which is required to be holomorphic on CP 1 {N S}.
V (z, u+ ) = 0,
D+
V (z, cu+ ) = V (z, u+ ),
c C .
(6.50)
In the North chart, it is characterized by the series (5.22). It is defined to possess the gauge
freedom
V V + + ,
(z, ) =
k (z) k
(6.51)
k=0
(z, u+ )
with
an arctic multiplet of weight 0. Using considerations similar to those given in
Section 5.2, the field strength (compare with the flat superspace expression [25])
+ +i

ui du 2
1
W (z) =
(6.52)
(D ) 12J V (z, u+ )
16i
(u+ u )2
can be shown to be invariant under the projective transformations (5.3). The field strength turns
out to be invariant under the gauge transformations (6.51). In the projective gauge (u
2 = 0,
11
22
J = J = 0), the field strength takes the form

2
1
W (z) =
(6.53)
d D 1 V (z, ),
16i
compare with the flat superspace result [10].
The AdS transformation law of V ,
V = ( + iJ )V ,
(6.54)
can be shown to imply that W transforms as

W = W
(6.55)
under the isometry group.

The field strength can be shown to obey the Bianchi identity
1
j)
j)
D(i D W = D (i D W,
4
and therefore
(i
k)
k)
D D D W = 2 J (ij D W,
(6.56)
(6.57)
compare with the flat superspace case [9,10]. The Bianchi identity implies that

1
+ )2 W 2J ++ W 2
G++ (z, u+ ) = i D+ W D+
W
+
W
(
D
61
(6.58)
is a composite O(2) multiplet,

D+
G++ = 0,
+
G++ (z, u+ ) = Gij (z)u+
i uj .
(6.59)
H ++ (z, u+ )
be a real O(2) multiplet. Then, similarly to the flat superspace case [10,25],
Let
the supersymmetric action associated with the Lagrangian
L++ = V (z, u+ )H ++ (z, u+ )
(6.60)
can be shown to be invariant under the gauge transformations (6.51).

Given several Abelian vector multiplets VI (z, u+ ), where I = 1, . . . , n, the composite superfield (6.58) is generalised to the form:

1
++
+
+
+ 2
++
G++
=
G
=
i
D
W
D
W
+
(
D
)
W
2J
W
W
W
I
J
I J ,
(I
J)
IJ
(I J )
2
D+
G++ = 0,
IJ
+ +
+
G++
I J (z, u ) = GI J (z)ui uj .
ij
(6.61)
We then can construct a supersymmetric ChernSimons action associated with the Lagrangian
1
+
(6.62)
cI,J K VI (z, u+ )G++
J K (z, u ), cI,J K = cI,KJ ,
12
for some constant parameters cI,J K (compare with the flat superspace case [10,25]). In accordance with the above result, the ChernSimons action is gauge invariant.
L++
CS =
6.5. Tensor multiplet and vector-tensor couplings

Given several O(2) (or, equivalently, tensor) multiplets HI++ (z, u+ ), a supersymmetric action
is generated by the Lagrangian
L++ = F(HI++ ),
I = 1, . . . , n
(6.63)
where F (H ) is a weakly homogeneous function of first degree in the variables H ,

HI
F (H )
F(H ) = I HI ,
HI
(6.64)
for some constants s.8 Such a Lagrangian occurs in the models for superconformal tensor
multiplets in four [42] and five dimensions [25].
One can also consider systems of coupled vector and tensor multiplets described by a Lagrangian of the form

1
L++ = F(HI++ ) + VI I HI++ + cI,J K G++
(6.65)
JK ,
12
for some coupling constants I and cI,J K .
8 The projective action principle formulated in Section 5.2 requires the Lagrangian to be a projective weight-two
multiplet. With I = 0 in (6.64), the Lagrangian (6.63) does not have any definite weight, and hence the results of
Section 5.2 are not applicable directly. We plan to discuss the case with I = 0 in more detail somewhere else.
62
7. Coset space realization

In this section we would like to give an explicit realization for the N = 1 AdS5 supergeometry which we have studied in Section 2 using the representation-independent approach. From
the group-theoretical point of view, it is known that the N = 1 AdS5 superspace (or simply
AdS5|8 ) can be identified with the coset space SU(2, 2|1)/SO(4, 1) U(1). Using the formalism of nonlinear realizations9 [44] (or Cartans coset construction), here we introduce a suitable
coset representative that makes possible to realize one half of AdS5|8 as a trivial fiber bundle
with fibers isomorphic to four-dimensional Minkowski superspace. This realization should be
useful if one is interested in having the 4D N = 1 super-Poincar symmetry manifest. However,
since it corresponds to one half of AdS5|8 (known as the Poincar patch [45]), it is not suitable
to describe the supersymmetric actions.
The analysis of this section builds on the construction given in [46], see also [47] for related
issues. Note that we use the superform conventions of [19].
7.1. Coset representative
As is well known, the supergroup SU(2, 2|1) is the four-dimensional N = 1 superconformal
group. It is generated by Lie-algebra elements of the form (parametrization (7.1) was used in
[48,49])

w
ib
2

X = ia
(7.1)
w +
2
,
)
2(
2
which satisfy the conditions

str X = 0,
BX B = X,
B=

0 1 0
1 0 0 .
0 0 1
(7.2)
The matrix elements in (7.1) correspond to a 4D Lorentz transformation (w , w ), a transla , a special conformal transformation b , a Q-supersymmetry ( , ), an S-supersymtion a

metry ( , ), and a combined scale and U(1)-chiral transformation = 12 + 3i .
The explicit parametrization for the algebra su(2, 2|1), which is given in (7.1), is ideally suited
to describe the compactified Minkowski space SU(2, 2|1)/(P C ), where P denotes the N = 1
super Poincar group (generated by the parameters (w, w,
b, , )
in (7.1)), and C denotes the
in (7.1). In
group of scale and chiral transformations generated by the parameters and
the case of the coset space SU(2, 2|1)/SO(4, 1) U(1), however, this parametrization should
be slightly modified. In addition, a re-scaling of some matrix elements is needed in order to
incorporate the AdS curvature 2 into the formalism.
As is known, a key role in the coset construction for M = G/H is played by a coset representative S(p) defined to be a smooth mapping S : U G, for some open domain U M, such
that S(p)p0 = p, for any point p U , where p0 U is a fixed point having H as its isotropy
group. On topological grounds, it is not always possible to extend U to the whole coset space M.
9 Many years ago, this formalism was also applied to introduce the 4D N = 1 AdS superspace [27,43].
63
As a coset representative, S(z), for AdS5|8 = SU(2, 2|1)/SO(4, 1) U(1), following mainly
[46] we choose
S(z) = g(z) gS gD
1
1
1
0
0
e 2 y 1
0
0
1 2 2 2
1
1
= ix+ 1 2 2 0
1
0 0
e 2 y 1 0
1
1
1
0
0
1
0 2 2
1
2 2 0
1
1
1 y
y
2
2
e
2e
2 2
1
1
1
= ie 2 y x+ e 2 y ( 2i2 x+ + 4 ) 2 2 ( ix+ ) ,
1
2 2 e 2 y
2 2 e 2 y ( + 2 )
(1 + 4 )
(7.3)
i a
where
=
denote ordinary 4D N = 1 (anti)chiral bosonic variables. It is worth

pointing out that the coset representative g(z) corresponds to the coset P/SO(3, 1) and provides
a matrix realization10 for 4D N = 1 Minkowski superspace, with coordinates z = (x a , , ).
Note that the isotropy group at z = 0 is H = SO(4, 1) U(1) SU(2, 2|1) = G, and it is generated by matrices of the form
i
0
0
31
w 2i b 0
H = 2i b w
0 ,
3i 1
0 + 0
0
0
0
0
0
4i3
a
x
xa
= w ,
w
tr w = 0,
(7.4)
= .
b = b,
Setting = 1 in (7.3) gives the parametrization used in [46].

Once the coset representative S(p) is chosen, the next step in the coset construction for
M = G/H is to compute the MaurerCartan one-form S 1 dS which proves to encode all the information about the geometry of M. Let G and H be the Lie algebras of G and H , respectively,
and G H be a complement of H in G such that [G H, H] G H. Then, the Maurer
Cartan one-form can be uniquely decomposed as S 1 dS = S 1 dS|G H + S 1 dS|H , where
S 1 dS|G H is identified with the vielbein, and S 1 dS|H with the connection.
In our case, the vielbein E = S 1 dS|G H and the connection = S 1 dS|H are:
S 1 dS = E + ,
1
2 Ey

E = i E
2
1
2
2 (E )
2i E
1
2 Ey
2 (E )
(7.5)
1
2
3i U(1)
i
=
2 2 (E )
1
)
2 2 (E
,
2i
3i U(1)
0
.
0
4i
3 U(1)
(7.6)
10 It is a curious historic fact that the above matrix realization for 4D N = 1 Minkowski superspace was introduced by
Akulov and Volkov [50] a year before the official discovery of superspace.
64
The components of the vielbein are given by the one-forms

E = e ey 1 e2y 2 2 2 + 2iey d + 2iey d
+ 4iey d 2 + 4iey d 2 ,

Ey = dy + d (2 ) + d 2 ,

1
1

(E ) = d e 2 y + em ie 2 y m ,

1
1
1
(E ) = d e 2 y + d 2e 2 y 2 + d 2e 2 y

1

+ em i2 e 2 y 2 m .
(7.7a)
(7.7b)
(7.7c)
(7.7d)
The components of the SO(4, 1) U(1) connection read

= d 4 2 + em 2i2 m i2 m ,

= e ey 1 + 2 e2y 2 2 + d 2iey + d 2iey

+ d 4i2 ey 2 + d 4i2 ey 2 ,

U(1) = d (3i ) + d 3i + em 32 m
,
(7.8a)
(7.8b)
(7.8c)
where
em = dx m i d a + i a d ,
(7.9)
is the spacetime component of the N = 1 flat superspace vielbein [19].

Note that under a group transformation g SU(2, 2|1)
g) S
h,
gS(z) = S(g z)h(z;
g) H,
h(z;
(7.10)
the vielbein E and the connection transform as follows:

h 1 ,
E
= hE
h 1 (dh)
h 1 .

= h
M
(7.11)
A
It is useful to introduce the inverse EA of the vielbein supermatrix EM implicitly used in

the previous equations (EA M EM B = A B , EM A EA N = M N ). With the definitions

M = em , dy, d , d , d , d = EA EA M ,
(7.12)

a
A
A

M
E = E , Ey , (E ) , (E ) , (E ) , (E ) = EM ,
(7.13)
E , we find
and Ea = 12 ( a )

3
3
e = E ey 1 + 2 e2y 2 2 + (E ) (2ie 2 y ) + (E ) 2ie 2 y

5
5
+ (E ) 2ie 2 y 2 + (E ) 2ie 2 y 2 ,
1

1

dy = Ey + (E ) 2e 2 y + (E ) 2e 2 y ,

1
3
d = (E ) e 2 y 1 22 e2y 2 2 + (E ) 2e 2 y 2

3
+ (E ) 2e 2 y + Ea iey a ,

1
1
d = (E ) e 2 y 1 42 e2y 2 2 + (E ) 2e 2 y 2

1
+ (E ) 2e 2 y + Ea 2i2 ey a 2 .
where
em
e
= 12 ( m )

(7.14a)
(7.14b)
(7.14c)
(7.14d)
It is also useful to decompose the connection with respect to the curved basis {EA }

1
3
= (E ) e 2 y 4 2 + (E ) e 2 y 4 2 2 2

+ Ea 2 ey 2i a + i a ,

1
1
= E 1 22 e2y 2 2 + (E ) 4ie 2 y + (E ) 4ie 2 y

3
3
+ (E ) 4i2 e 2 y 2 + (E ) 4i2 e 2 y 2 ,

1
1
U(1) = Ea 32 ey a + (E ) 3ie 2 y + (E ) 3ie 2 y

3
) 6i2 e 32 y 2 .
+ (E ) 6i2 e 2 y 2 + (E
65
(7.15a)
(7.15b)
(7.15c)
7.2. SO(4, 1) U(1) covariance

To better understand the relation between the above coset construction and the AdS5|8 supergeometry of Section 2, it is necessary to figure out the precise meaning of the SO(4, 1) U(1)
covariance of the vielbein and the connection. We will use several results which are collected in
Appendix A and concern the reduction of 5D spinors into 4D ones.
First of all, let us recall that choosing g = h H in relations (7.10), (7.11) gives h = h =
const, and the group transformations (7.11) reduce to
E
= hEh1 ,

= hh1 ,
h SO(4, 1) U(1).
(7.16)
In particular, a 5D Lorentz transformation acts as follows:

E
= E1 ,
where
=
0
0
0 0 1

= 1 ,

,

1 cd
= exp (cd )
.
2
(7.17)
(7.18)
This transformation law allows us to combine components of the connection into fivedimensional vector and spinor. Explicitly, we can write

1
2i Ea (a ) 2 2 E
E=
(7.19)
,
1
2 2 E
0

1 0
3
1 a b (a b ) 0
0
,
=
(7.20)
0 + i U(1)
2
0 0 0
0 0 43
where

Ea = Ea , E5 = Ea , Ey ,

(E )
= (E ) , (E ) ,
E =
,
E
(E )

a b = ab , a5 ,

1
,
a5 = a .
ab = ab + ab
(7.21a)
(7.21b)
(7.21c)
(7.21d)
66
, and U(1)
Note that Ea , a b = b a and U(1) are real. It follows that Ea , E , E
a b
transform under the 5D Lorentz group SO(4, 1) respectively as a vector, a Dirac spinor, its Dirac
conjugate spinor, an antisymmetric two-tensor and a scalar. Due to (7.21a) we can identify
x 5 y.
(7.22)
Note also that we can combine the two spinors E and

defined as follows:

Ei = Ei , E i ,
E1
= (E ) ,
E2
1
E
= (E ) ,
= (E ) ,
into a 5D pseudo-Majorana spinor

(7.23)
2
E
= (E ) .
(7.24)
It remains to consider the transformation properties of the vielbein and the connection under
the U(1) part of the isotropy group. In accordance with (7.16), they transform as

= 1 ,
E
= E 1 ,

0
[exp( 13 i)]
0
.
=
43 i
e
0 0
Clearly is invariant under the U(1) transformation, while E transforms as
i

a ( ) 2 12 (ei E )
E
2
E
=
,
1
)
2 2 (ei E
0
(7.25)
(7.26)
and hence Ea is invariant. Note also that (7.25) induces the following transformation of Ei :

1 0
Ji j = (3 )i j =
.
E
i = exp(iJ ) i j Ej ,
(7.27)
0 1
7.3. Representation of covariant derivatives
With the vielbein and the connection having been introduced, we can now construct the covariant derivatives
1
DA = EA + iA J + A bc Mb c
2
1
= EA + iA J + A bc Mbc + A b5 Mb5
2

1
2
= Da , Di = Da , D5 , D , D 1 , D , D 2 .
The vector fields EA are defined by

EA = Ea , Ei = EA M DM ,

= EM A EA .
DM = m , , D , D , ,
y

(7.28)
(7.29)
Here the supermatrices EA M and EM A have been defined in Section 7.1. It should be pointed out
that (m , D , D ) are the 4D N = 1 flat superspace covariant derivatives, D = + i
and D =
67
. Furthermore, the connection superfields in D are defined as

+ i
A
U(1) = EA A ,
a b = EA A a b .
(7.30)
It can be shown that the explicit expressions for the covariant derivatives are as follows:

Da = ey 1 + 2 e2y 2 2 a iey ( a ) D iey (a ) D
+ 2i2 ey 2 ( a ) + 2i2 ey 2 (a )
3i2 ey (a )J

+ 2 ey ab bcde (c )M
de 1 + 22 e2y 2 2 Ma5 ,
D5 =
,
y
(7.31a)
(7.31b)

1
1
1
1
D = e 2 y 1 22 e2y 2 2 D 2e 2 y 2 2e 2 y

1
5
1
1
ie 2 y 2 m m 3e 2 y J + 2e 2 y ab Mab
+ 2 e 2 y
y

3
(7.31c)
+ 2i2 e 2 y 2 a Ma5 ,

1
3
3
2
D = e 2 y 1 42 e2y 2 2
+ 2e 2 y 2 D 2e 2 y D

3
3
3
+ ie 2 y m m 62 e 2 y 2 J + 22 e 2 y 2 ab Mab

1
2ie 2 y a Ma5 ,
(7.31d)

1
1
1
D 1 = e 2 y 1 22 e2y 2 2 D 2e 2 y 2
2e 2 y

1
5
1
1
ie 2 y 2 m m + 3e 2 y J + 2e 2 y ab Mab
+ 2 e 2 y
y

2 32 y 2 a
(7.31e)
+ 2i e Ma5 ,

1
3
3
3
D 2 = e 2 y 1 42 e2y 2 2
+ 2e 2 y 2 D 2e 2 y D + ie 2 y m m

1
2 32 y 2
2 32 y
(7.31f)
+ 6 e J + 2 e 2 ab Mab 2ie 2 y a Ma5 .
It is interesting to consider a flat superspace limit, 0, for the covariant derivatives. In this
limit, one finds
DA |0 = eU DA eU ,
U = + ,
(7.32)
where DA = (a , Di ) are 5D flat global covariant derivatives,

Di =

i b i b ,

with i = (i , i
) and i = ( , ).
(7.33)
68
7.4. Torsion and curvature

Now, we are prepared to demonstrate that the geometry described in the present section reproduces the geometry of AdS5|8 constructed in Section 2.
We proceed by recalling that, in accordance with the coset construction, the torsion T and
curvature R two-forms are defined as follows:
T = dE E E ,
R = d .
(7.34)
Under group transformations (7.10) they transform covariantly

h 1 ,
T
= hT
h 1 .
R
= hR
(7.35)
Keeping in mind the definition E + = G1 dG, we get

dE + d = G1 dG G1 dG
= E E + E + E + ,
(7.36)
from which we obtain

dE = (E E)|G H + E + E,
d = (E E)|H + ,
(7.37)
since (E + E) G H and H. Using the previous formulae we are able to

see that the torsion and curvature two-forms are given by simple expressions
T = (E E)|G H ,
R = (E E)|H .
Therefore, it remains to compute E E.

Direct calculations give

1 2 a
3
2 a
2 E E (a b ) + 4E2 E1 i E E2 (a )
,
EE=
3
i 2 E1 Eb (b )
4E1 E2
and this we should represent as E E = (E E)|G H + (E E)|H . We end up with
i

1
2 Ta (a ) 2 2 T2
T=
,
1
2 2 T1
0

1 0
3
1 a b (a b ) 0
0
R= R
,
0 + iRU(1)
2
0 0 0
0 0 43
where

1

Ta = Ek Ej 2i j k a ,
2

1 c
i
i
j
k
b
Ti = E Ej
(3 )i (c ) + Ek E (3 )i (b )
,
2
2
2

1

Ra b = Ed Ec 2 ca b + cb a + El Ek 4 ki (3 )li a b ,
d
d
2

1

kl
RU(1) = El Ek 3i .
2
(7.38)
(7.39)
(7.40)
(7.41)
(7.42)
(7.43)
(7.44)
(7.45)
69
Using standard superform definitions [19], we define the components of the torsion and curvature
as follows:
1
TA = EC EB TB C A ,
(7.46)
2
1
1
RU(1) = ED EC (RU(1) )C D .
Ra b = ED EC RC D a b ,
(7.47)
2
2
Now, let us return to the covariant derivatives described in Section 2. Their algebra given by
Eqs. (2.21a)(2.21c) can be represented concisely as
1
[DA , DB } = TA B C DC + i(RU(1) )A B J + RA B cd Mcd .

(7.48)
2
Comparing (2.21a)(2.21c) with Eqs. (7.42)(7.45), we find that all the components of the torsion and curvature coincide provided
J i j = (3 )i j .
(7.49)
This completes our analysis of the coset construction.

Acknowledgements
This work is supported in part by the Australian Research Council and by a UWA research
grant.
Appendix A. 5D conventions
Our 5D notation and conventions correspond to [10]. The 5D gamma-matrices m =
(m , 5 ), with m = 0, 1, 2, 3, are defined by
{m , n } = 2m n 1,
(m ) = 0 m 0
are chosen in accordance with

0
(m )
,
(m ) =
( m )
0
(A.1)

(5 ) =
i
0
0
i
(A.2)
such that 0 1 2 3 5 = 1. The charge conjugation matrix, C = ( ), and its inverse, C 1 =

C = ( ) are defined by

0
1
T

Cm C = (m ) ,
(A.3)
=
,
=
.
0
0
The antisymmetric matrices and are used to raise and lower the four-component spinor
indices.
A Dirac spinor, = ( ), and its Dirac conjugate, = ( ) = 0 , look like

= , .
= ,
(A.4)
One can now combine = ( , ) and = = ( , ) into a SU(2) doublet,

i = i , i
(A.5)
= i
, i = 1, 2,
i
,
70
with 1 = and 2 = . It is understood that the SU(2) indices are raised and lowered by
= ij . The Dirac spinor i = ( i )
ij and ij , 12 = 21 = 1, in the standard fashion: i
j
satisfies the pseudo-Majorana condition i T = Ci . This will be concisely represented as

i
= i .
(A.6)
With the definition m n = n m = 14 [m , n ], the matrices {1, m , m n } form a basis in
the space of 4 4 matrices. The matrices and (m ) are antisymmetric, (m ) = 0,

while the matrices (m n ) are symmetric.
It is useful to write explicitly the 4D reduction of these matrices

0
i
0
(m )
(m ) =
(A.7)
,
(5 ) =
,
(m )
0
0
i

0
2i (m )
0
(mn )
,
(
, (A.8)
(mn ) =
)
=
m5
0
( mn )
m )
0
2 (

i

0
(mn )
0
2 (m )
,
(
(mn ) =
, (A.9)
)
=
m5
0
0
( mn )
2 (m )
where (mn ) = 14 (m n n m ) and ( mn ) = 14 ( m n n m ) .
Given a 5-vector V m and an antisymmetric tensor F m n = F n m , we can equivalently represent them as the bi-spinors V = V m m and F = 12 F m n m n with the following symmetry
properties
V = 0,
V = V ,
F = F .
(A.10)
The two equivalent descriptions Vm V and Fm n F are explicitly described as follows:

1
Vm = (m ) V ,
4
V = V m (m ) ,
(A.11)
F = F m n (m n ) ,
Fm n = (m n ) F .
2
These results can be easily checked using the identities
(A.12)
= + + ,
=
1 m
1
(m ) + ,
2
2
(A.13)
and therefore
1 m
1
(m ) + ,
2
2
with the completely antisymmetric fourth-rank tensor.
Complex conjugation gives
=
( ) = ,
(V ) = V ,
provided V m and F m n are real.
(A.14)
(F ) = F ,
(A.15)
71
References
[1] A.S. Galperin, E.A. Ivanov, S.N. Kalitsyn, V. Ogievetsky, E. Sokatchev, Unconstrained N = 2 matter, YangMills
and supergravity theories in harmonic superspace, Class. Quantum Grav. 1 (1984) 469.
[2] A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E.S. Sokatchev, Harmonic Superspace, Cambridge Univ. Press, Cambridge, UK, 2001, 306p.
[3] A. Karlhede, U. Lindstrm, M. Rocek, Self-interacting tensor multiplets in N = 2 superspace, Phys. Lett. B 147
(1984) 297.
[4] W. Siegel, Chiral actions for N = 2 supersymmetric tensor multiplets, Phys. Lett. B 153 (1985) 51.
[5] U. Lindstrm, M. Rocek, New hyper-Khler metrics and new supermultiplets, Commun. Math. Phys. 115 (1988)
21.
[6] U. Lindstrm, M. Rocek, N = 2 super-YangMills theory in projective superspace, Commun. Math. Phys. 128
(1990) 191.
[7] A.A. Rosly, Super-YangMills constraints as integrability conditions, in: M.A. Markov (Ed.), Proceedings of the
International Seminar on Group Theoretical Methods in Physics, vol. 1, Zvenigorod, USSR, 1982, Nauka, Moscow,
1983, p. 263 (in Russian); English translation: in: M.A. Markov, V.I. Manko, A.E. Shabad (Eds.), Group Theoretical
Methods in Physics, vol. 3, Harwood Academic Publishers, London, 1987, p. 587.
[8] E. Witten, An interpretation of classical YangMills theory, Phys. Lett. B 77 (1978) 394.
[9] B. Zupnik, Harmonic superpotentials and symmetries in gauge theories with eight supercharges, Nucl. Phys. B 554
(1999) 365, hep-th/9902038;
B. Zupnik, Nucl. Phys. B 644 (2002) 405, Erratum.
[10] S.M. Kuzenko, W.D. Linch III, On five-dimensional superspaces, JHEP 0602 (2006) 038, hep-th/0507176.
[11] P.S. Howe, K.S. Stelle, P.C. West, N = 1 D = 6 harmonic superspace, Class. Quantum Grav. 2 (1985) 815;
B.M. Zupnik, Six-dimensional supergauge theories in the harmonic superspace, Sov. J. Nucl. Phys. 44 (1986) 512.
[12] J. Grundberg, U. Lindstrm, Actions for linear multiplets in six dimensions, Class. Quantum Grav. 2 (1985) L33.
[13] S.J. Gates Jr., S. Penati, G. Tartaglino-Mazzucchelli, 6D supersymmetry, projective superspace and 4D, N = 1
superfields, JHEP 0605 (2006) 051, hep-th/0508187;
S.J. Gates Jr., S. Penati, G. Tartaglino-Mazzucchelli, 6D supersymmetric nonlinear sigma-models in 4D, N = 1
superspace, JHEP 0609 (2006) 006, hep-th/0604042.
[14] W. Siegel, Curved extended superspace from YangMills theory a la strings, Phys. Rev. D 53 (1996) 3324, hepth/9510150.
[15] W.D. Linch III, B.C. Vallilo, Covariant N = 2 heterotic string in four dimensions, hep-th/0611105.
[16] N. Berkovits, Covariant quantization of the GreenSchwarz superstring in a CalabiYau background, Nucl. Phys.
B 431 (1994) 258, hep-th/9404162;
N. Berkovits, C. Vafa, N = 4 topological strings, Nucl. Phys. B 433 (1995) 123, hep-th/9407190;
N. Berkovits, W. Siegel, Superspace effective actions for 4D compactifications of heterotic and type II superstrings,
Nucl. Phys. B 462 (1996) 213, hep-th/9510106.
[17] A.S. Galperin, N.A. Ky, E. Sokatchev, N = 2 supergravity in superspace: Solution to the constraints, Class. Quantum Grav. 4 (1987) 1235;
A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E. Sokatchev, N = 2 supergravity in superspace: Different versions
and matter couplings, Class. Quantum Grav. 4 (1987) 1255.
[18] E. Sokatchev, Off-shell six-dimensional supergravity in harmonic superspace, Class. Quantum Grav. 5 (1988) 1459.
[19] J. Wess, J. Bagger, Supersymmetry and supergravity, Princeton Univ. Press, Princeton, USA, 1992, 259p.
[20] W. Siegel, S.J. Gates Jr., Superfield supergravity, Nucl. Phys. B 147 (1979) 77;
S.J. Gates Jr., M.T. Grisaru, M. Rocek, W. Siegel, Superspace, or One Thousand and One Lessons in Supersymmetry, Benjamin/Cummings, 1983, hep-th/0108200.
[21] I.L. Buchbinder, S.M. Kuzenko, Ideas and Methods of Supersymmetry and Supergravity or a Walk Through Superspace, IOP, Bristol, 1998.
[22] A.Y. Segal, A.G. Sibiryakov, Explicit N = 2 supersymmetry for higher-spin massless fields in D = 4 AdS superspace, Int. J. Mod. Phys. A 17 (2002) 1207, hep-th/9903122.
[23] S.J. Gates Jr., S.M. Kuzenko, A.G. Sibiryakov, N = 2 supersymmetry of higher superspin massless theories, Phys.
Lett. B 412 (1997) 59, hep-th/9609141;
S.J. Gates Jr., S.M. Kuzenko, A.G. Sibiryakov, Towards a unified theory of massless superfields of all superspins,
Phys. Lett. B 394 (1997) 343, hep-th/9611193.
[24] I.A. Bandos, E. Ivanov, J. Lukierski, D. Sorokin, On the superconformal flatness of AdS superspaces, JHEP 0206
(2002) 040, hep-th/0205104.
72
[25] S.M. Kuzenko, On compactified harmonic/projective superspace, 5D superconformal theories, and all that, Nucl.
Phys. B 745 (2006) 176, hep-th/0601177.
[26] N. Dragon, Torsion and curvature in extended supergravity, Z. Phys. C 2 (1979) 29.
[27] E.A. Ivanov, A.S. Sorin, Superfield formulation of Osp(1, 4) supersymmetry, J. Phys. A 13 (1980) 1159.
[28] M.F. Sohnius, Supersymmetry and central charges, Nucl. Phys. B 138 (1978) 109.
[29] P. Breitenlohner, M.F. Sohnius, Superfields, auxiliary fields, and tensor calculus for N = 2 extended supergravity,
Nucl. Phys. B 165 (1980) 483.
[30] J. Wess, Supersymmetry and internal symmetry, Acta Phys. Austriaca 41 (1975) 409;
W. Siegel, Off-shell central charges, Nucl. Phys. B 173 (1980) 51.
[31] M.F. Sohnius, K.S. Stelle, P.C. West, Representations of extended supersymmetry, in: S.W. Hawking, M. Rocek
(Eds.), Superspace and Supergravity, Cambridge Univ. Press, Cambridge, 1981, p. 283.
[32] S.V. Ketov, New self-interaction for N = 2 multiplets in 4d and ultraviolet finiteness of two-dimensional N = 4
sigma-models, in: Proceedings of the International Seminar Group Theory Methods in Physics, vol. 1, Urmala,
USSR, May 1985, Nauka, Moscow, 1985, p. 87;
S.V. Ketov, B.B. Lokhvitsky, Some generalizations of N = 2 YangMills matter couplings, Class. Quantum Grav. 4
(1987) L137;
S.V. Ketov, B.B. Lokhvitsky, I.V. Tyutin, Hyperkahler sigma models in extended superspace, Theor. Math. Phys. 71
(1987) 496.
[33] F. Gonzalez-Rey, U. Lindstrm, M. Rocek, R. von Unge, S. Wiles, Feynman rules in N = 2 projective superspace.
I: Massless hypermultiplets, Nucl. Phys. B 516 (1998) 426, hep-th/9710250.
[34] A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, Duality transformations and most general matter self-coupling in N =
2 supersymmetry, Nucl. Phys. B 282 (1987) 74.
[35] A.A. Rosly, Gauge fields in superspace and twistors, Class. Quantum Grav. 2 (1985) 693.
[36] A.A. Rosly, A.S. Schwarz, Supersymmetry in a space with auxiliary dimensions, Commun. Math. Phys. 105 (1986)
645.
[37] S.M. Kuzenko, Projective superspace as a double-punctured harmonic superspace, Int. J. Mod. Phys. A 14 (1999)
1737, hep-th/9806147.
[38] B.B. Deo, S.J. Gates Jr., Comments on nonminimal N = 1 scalar multiplets, Nucl. Phys. B 254 (1985) 187.
[39] S.J. Gates Jr., S.M. Kuzenko, The CNM-hypermultiplet nexus, Nucl. Phys. B 543 (1999) 122, hep-th/9810137;
S.J. Gates Jr., S.M. Kuzenko, 4D N = 2 supersymmetric off-shell sigma models on the cotangent bundles of Khler
manifolds, Fortschr. Phys. 48 (2000) 115, hep-th/9903013.
[40] M. Arai, M. Nitta, Hyper-Khler sigma models on (co)tangent bundles with SO(n) isometry, Nucl. Phys. B 745
(2006) 208, hep-th/0602277.
[41] M. Arai, S.M. Kuzenko, U. Lindstrm, Hyperkhler sigma models on cotangent bundles of Hermitian symmetric
spaces using projective superspace, JHEP 0702 (2007) 100, hep-th/0612174.
[42] B. de Wit, M. Rocek, S. Vandoren, Hypermultiplets, hyper-Kaehler cones and quaternion-Kaehler geometry,
JHEP 0102 (2001) 039, hep-th/0101161.
[43] B.W. Keck, An alternative class of supersymmetries, J. Phys. A 8 (1975) 1819;
B. Zumino, Nonlinear realization of supersymmetry in anti-de Sitter space, Nucl. Phys. B 127 (1977) 189.
[44] S. Coleman, J. Wess, B. Zumino, Structure of phenomenological Lagrangians. 1, Phys. Rev. 177 (1969) 2239;
C.G. Callan, S. Coleman, J. Wess, B. Zumino, Structure of phenomenological Lagrangians. 2, Phys. Rev. 177 (1969)
2247;
C.J. Isham, A group-theoretic approach to chiral transformations, Nuovo Cimento 59A (1969) 356;
C.J. Isham, Metric structures and chiral symmetries, Nuovo Cimento 61A (1969) 188;
A. Salam, J. Strathdee, Nonlinear realizations. 1: The role of Goldstone bosons, Phys. Rev. 184 (1969) 1750;
D.V. Volkov, Phenomenological Lagrangians, Sov. J. Part. Nucl. 4 (1973) 3;
V.I. Ogievetsky, Nonlinear realizations of internal and spacetime symmetries, in: Proceedings of the Xth Winter
School of Theoretical Physics in Karpacz, vol. 1, Wroslaw, 1974, p. 227.
[45] O. Aharony, S.S. Gubser, J.M. Maldacena, H. Ooguri, Y. Oz, Large N field theories, string theory and gravity, Phys.
Rep. 323 (2000) 183, hep-th/9905111.
[46] S.M. Kuzenko, I.N. McArthur, Goldstone multiplet for partially broken superconformal symmetry, Phys. Lett. B 522
(2001) 320, hep-th/0109183.
[47] S. Bellucci, E. Ivanov, S. Krivonos, Goldstone superfield actions in AdS(5) backgrounds, Nucl. Phys. B 672 (2003)
123, hep-th/0212295.
[48] H. Osborn, N = 1 superconformal symmetry in four-dimensional quantum field theory, Ann. Phys. 272 (1999) 243,
hep-th/9808041.
73
[49] S.M. Kuzenko, S. Theisen, Correlation functions of conserved currents in N = 2 superconformal theory, Class.
Quantum Grav. 17 (2000) 665, hep-th/9907107.
[50] V.P. Akulov, D.V. Volkov, Goldstone fields with spin 1/2, Theor. Math. Phys. 18 (1974) 28, Teor. Mat. Fiz. 18
(1974) 39.
Resolved CalabiYau cones and flows

from Labc superconformal field theories
W. Chen a,1 , M. Cvetic b,2 , H. L a,,1 , C.N. Pope a,1 , J.F. Vzquez-Poritz a,1
a George P. & Cynthia W. Mitchell Institute for Fundamental Physics, Texas A&M University,
College Station, TX 77843-4242, USA

b Department of Physics & Astronomy, University of Pennsylvania, Philadelphia, PA 19104-6396, USA

Abstract
We discuss D3-branes on cohomogeneity-three resolved CalabiYau cones over Labc spaces, for which
a 2-cycle or 4-cycle has been blown up. In terms of the dual quiver gauge theory, this corresponds to motion
along the non-mesonic, or baryonic, directions in the moduli space of vacua. In particular, a dimension-two
and/or dimension-six scalar operator gets a vacuum expectation value. These resolved cones support various
harmonic (2, 1)-forms which reduce the ranks of some of the gauge groups either by a Seiberg duality
cascade or by Higgsing. We also discuss higher-dimensional resolved CalabiYau cones. In particular, we
obtain square-integrable (2, 2)-forms for eight-dimensional cohomogeneity-four CalabiYau metrics.
Published by Elsevier B.V.
1. Introduction
The AdS/CFT correspondence relates type IIB string theory on AdS5 S 5 to four-dimensional
N = 4 U (N ) superconformal YangMills theory [13]. More generally, type IIB string theory
on AdS5 X 5 , where X 5 is an EinsteinSasaki space such as T 1,1 , Y pq [4,5] or Labc [6,7],
corresponds to an N = 1 superconformal quiver gauge theory. The dual gauge theories have
been identified in [8] for T 1,1 , in [9,10] for Y pq and in [1113] for Labc .
E-mail address: honglu@physics.tamu.edu (H. L).

1 Research supported in part by DOE grant DE-FG03-95ER40917.
2 Research supported in part by DOE grant DE-FG02-95ER40893, NSF grant INTO3-24081, and the Fay R. and
Eugene L. Langberg Chair.

0550-3213/$ see front matter Published by Elsevier B.V.
doi:10.1016/j.nuclphysb.2007.06.018
W. Chen et al. / Nuclear Physics B 785 (2007) 7497
75
There is a prescription for mapping perturbations of the supergravity background to operators

in the dual gauge theory [2,3]. In particular, motion in the Khler moduli space of the Calabi
Yau cone over the EinsteinSasaki space corresponds to giving vacuum expectation values (vevs)
to the fundamental fields, such that only non-mesonic operators get vevs. This is because the
mesonic directions of the full moduli space correspond to the motion of the D3-branes in the
CalabiYau space whereas the non-mesonic, or baryonic, directions are associated with either
deformations of the geometry or turning on B-fields. This has been studied for a blown-up 2cycle in the resolved conifold in [14], as well as for a blown-up 4-cycle in the resolved conifold
[15], Y pq cones [16], Labc cones [17] and general CalabiYau cones [18]. All of these resolved
CalabiYau cones with blown-up 4-cycles follow the general construction given in [19,20].
In this paper, we shall apply the state/operator correspondence to a general class of resolved
CalabiYau cones over Labc with a blown-up 2-cycle or 4-cycle. These metrics can be obtained
from the Euclideanization of the BPS limit of the six-dimensional KerrNUTAdS solutions
[21,22].3 In particular, blowing up a 2-cycle or 4-cycle corresponds to giving a vev to a real
dimension-two and/or six scalar operator. Although cycles are being blown up, in all but two
cases there remain singularities [24,25]. However, there is a countably infinite subset of cases
where there is an ALE singularity, on which perturbative string dynamics is well-defined. Some
of these cases were studied in [18]. While adding a large number of D3-branes that are uniformly
distributed, or smeared, on the blown-up cycle ends up inducing a power-law singularity at
short distance,4 the resulting backgrounds can nevertheless be reliably used to describe perturbations around the UV conformal fixed point of the quiver gauge theories. Close to the UV fixed
point, blowing up a 2-cycle on the Labc cone corresponds to giving a vev to an operator that
is analogous to the case of the resolved conifold. Therefore, we shall refer to these spaces as
resolved cones, though it should be understood that there are still orbifold-type singularities.
The supergravity background can also be perturbed by adding a harmonic 3-form which lives
on the CalabiYau metrics. If this is a pure (2, 1)-form then supersymmetry will be preserved.
Furthermore, if this form carries nontrivial flux then it corresponds to D5-branes wrapped on a
2-cycle in the CalabiYau space. The introduction of these fractional D3-branes eliminates the
conformal fixed point in the UV limit of the quiver gauge theory. The theory undergoes a Seiberg
duality cascade and the ranks of some of the gauge groups are reduced with decreasing energy
scale. The supergravity solutions corresponding to fractional branes have been constructed for
the cones over T 11 [27,28], Y pq [29] and Labc spaces [30,31]. Fractional branes have also been
considered for CalabiYau spaces with blown-up cycles, such as the deformed conifold [32],
resolved conifold [33] and regularized conifold [15], as well as the resolved Y pq cones with
blown-up 4-cycles [18]. We shall also consider continuous families of 3-forms that do not have
nontrivial flux. In this case, there remains a conformal fixed point in the UV limit of the field theory. It has been proposed that the ranks of some of the gauge groups are reduced with decreasing
energy scale via the Higgs mechanism [34].
Since the Labc spaces have cohomogeneity two, the form fields constructed on the corresponding CalabiYau spaces will generally have nontrivial dependence on the radial direction as
well as the two non-azimuthal coordinates of Labc . In addition, these forms generally break the
U (1)R U (1) U (1) global symmetry group of the theory down to a U (1) U (1) symmetry
3 This is the even-dimensional analog of the relation between the EinsteinSasaki spaces constructed in [23] and odd-
dimensional BPS KerrNUTAdS solutions.

4 This singularity is due to the smearing of the D3-brane charge on the blown-up cycle. A completely non-singular
solution with D3-branes stacked at a single point on the resolved conifold has been constructed [26].
76
Fig. 1. RG flows from the superconformal fixed point of the Labc quiver gauge theory correspond to various deformations
of the supergravity background.
group which, in particular, breaks the R-symmetry. However, this is done in such a way that the
theory preserves N = 1 supersymmetry.
The various perturbations of the AdS5 Labc supergravity background that will be discussed
are shown in Fig. 1. These perturbations, which can be superimposed with one another, correspond to continuous families of Renormalization Group (RG) flows from the UV superconformal
fixed point of the quiver gauge theory.
The paper is organized as follows. In Section 2, we discuss the geometry of the resolved
CalabiYau cones over the Labc spaces. A subset of these are the resolved cones over Y pq and
their various limits. We find various harmonic (2, 1)-forms on these metrics, some of which
carry nontrivial flux and some of which do not. In Section 3, we apply some of our results
to the AdS/CFT correspondence. In particular, we relate the perturbations of the AdS5 Labc
background to various flows from the UV conformal fixed point of the dual quiver gauge theory.
In Section 4, we consider eight-dimensional resolved cones over Lpqrs and the various harmonic
forms that live on them. In Section 5, we carry out the corresponding analysis for the higherdimensional resolved cones. Lastly, conclusions are presented in Section 6.
2. Six-dimensional resolved CalabiYau cones
Although the Labc spaces themselves are non-singular for appropriately chosen integers p, q,
r [6,7], the cones over these spaces have a power-law singularity at their apex. In the case of the
cone over T 1,1 , this singularity can be smoothed out in two different ways [35]. Firstly, one can
blow up a 3-cycle, which corresponds to a complex deformation. The resulting deformed conifold
has been crucial for the construction of a well-behaved supergravity dual of the IR region of the
gauge theory, providing a geometrical description of confinement [32].
One might hope that a similar resolution procedure could be performed on other Labc cones.
Although a first-order deformation of the complex structure of Y pq cones has been found in [36],
there exists an obstruction to finding the complex deformations beyond first order [37,38]. There
is also evidence from the field theory side that such deformations will break supersymmetry for
the Y pq cones [3942] as well as for a large class of Labc cones [42]. Nevertheless, there are
Labc cones which allow for complex structure deformations [37,38], which can be understood
77
Fig. 2. A 4-cycle within the base space of a cone over Labc can be blown up. Within this 4-cycle lies a 2-cycle. The
volumes of these two cycles correspond to two independent Khler moduli.
from the corresponding toric diagrams [12].5 However, the explicit metrics for these deformed
Labc cones, let alone the solutions for D3-branes on these cones, are not known.
The second way in which the T 1,1 cone can be rendered regular is by blowing up a 2-cycle
[35]. Also, for the case of a cone over T 1,1 /Z2 , the singularity can be resolved by blowing up
a 4-cycle. Both of these resolutions are examples of Khler deformations which, as we shall
see shortly, can also be performed on the Labc cones C(Labc ). Moreover, the 2-cycle actually
lives within the 4-cycle, as illustrated in Fig. 2. This means that there are two Khler moduli
associated with the 4-cycle. For certain parameter choices, we can have the 4-cycle corresponds
to the EinsteinKhler base space of Labc , whose metric can be obtained by taking a certain
scaling limit of a Euclideanized form of the PlebanskiDemianski metric [30]. It is also possible
to have the volume of the 4-cycle vanishes, whilst keeping a 2-cycle blown up.
It has been found that the cone over Y 2,1 can be rendered completely regular by blowing
up an appropriate 4-cycle [24]. However this, together with the resolved cones over T 1,1 and
T 1,1 /Z2 , constitute the only examples of non-singular resolved cones over the Labc spaces [25].
Although we shall refer to these spaces as resolved Labc cones, there are generally orbifoldtype singularities remaining. In the limit of a vanishing 2-cycle, this can be seen simply because at
short distance the geometry becomes a direct product of R2 and the four-dimensional Einstein
Khler base space of Labc , which is itself an orbifold. Nevertheless, the resolved cones over
Labc can be embedded in ten dimensions to give Ricci-flat backgrounds Mink4 C(Labc ), on
which perturbative string dynamics is well-defined. However, as we shall see in Section 3, the
back-reaction of D3-branes leads to a power-law singularity at short distance. This singularity is
due to the fact that we are smearing the D3-branes on the blown-up cycle. For the case of the
resolved conifold, it has been shown that if the D3-branes are stacked at a single point then the
supergravity solution is completely regular [26].
2.1. Resolved cones over Y pq
Before turning to resolved cones over the general cohomogeneity-two Labc spaces, it is instructive first to consider the subset involving the cohomogeneity-one Y pq spaces. The resolved
5 We thank Angel Uranga for discussions on this point.
78
cone over Y pq has the metric [21]

2
2
x+y 2
X
x +y 2
Y
y
x
2
dx +
d +
3 +
dy +
d
3
ds6 =
4X
x+y
2
4Y
x+y
2

xy 2
+
1 + 22 ,
4
where
2
2
X = x(x + )
,
Y = y( y) + ,
x
y
(2.1)
(2.2)
and that
3 = d + cos d,
12 + 22 = d 2 + sin2 d 2 .
(2.3)
It has been shown that the only completely regular examples are the resolved cones over T 1,1 ,
T 1,1 /Z2 and Y 2,1 [24,25]. We shall now consider various limits of the metric (2.1).
Resolved conifold
In order to reduce to a resolved cone over T 11 (or T 11 /Z2 ), we need to select such that Y (y)
2 3
. Making the coordinate redefinition
has a double root. This happens when = 27
2
1
2
y = + cos ,
= 3 + 2 ,
3
27
2
2
2
= ,
3 3 +
d,
9
3
and setting the parameter to zero, we find that the metric becomes
ds62 =
x + 23 2
X
2
dx +
(3 + cos d )
4X
9(x + 23 )

1

1
2 2
+
x + d + sin2 d 2 + x 12 + 22 .
6
3
6
(2.4)
(2.5)
If = 0, there is a blown-up S 2 and the solution describes the resolved conifold [35]. If, on the
other hand, = 0, then there is a blown-up S 2 S 2 and the solution describes the regularized
conifold [15]. In fact, it has been shown that one can always blow up a 4-cycle on any cone
over an EinsteinSasaki space [19,20]. We shall now take a look at the analogous limits for the
resolved cones over the Y pq spaces.
The = 0 limit
If we let y y, 3 and then take 0, we obtain the limit

2

2

x
X
dy
1
1
dx 2 +
d + y3 + x
+ Y 32 + y 12 + 22 ,
ds 2 =
4X
x
2
4Y
4
(2.6)
where
X = x2
2
,
x
Y = y(1 y) +
2
.
y
(2.7)
There is a single Khler modulus, which corresponds to a blown-up 4-cycle with a volume parameterized by . This is the analog of the resolved cone for general Y pq spaces. However, unlike
79
the T 1,1 /Z2 case, this metric has an orbifold-type singularity at its apex, since the geometry
reduces to the direct product of R2 and an EinsteinKhler orbifold.
The = 0 limit: Blowing up 2-cycles
One can also consider the limit in which vanishes, in which case x runs from 0 to asymptotic
. Near x = 0, we can express the metric as

2

2
dy 2
1 2
2
1 2 2
1 2
2
2
2
+ Y d r 3 , (2.8)
ds = y dr + r 3 + d + r 1 + 2 +
4
y
4
4Y
2
where x = r 2 . At r = 0 there is a collapsing 3-sphere, instead of a circle as in the previous limit.
There is a single Khler modulus corresponding to the volume of a blown-up 2-cycle, which is
parameterized by . However, unlike the analogous resolved conifold for which there is a smooth
2-sphere, in general this 2-cycle is a tear-drop with a conical singularity.
CalabiYau structure
The CalabiYau structure on the metric (2.1) is given by a Khler form J and a holomorphic
(3, 0)-form G(3) . These can be expressed in the complex vielbein basis
1 = e1 + ie2 ,
2 = e3 + ie4 ,
3 = e5 + ie6 ,
where the vielbein is conveniently chosen to be

x
+
y
X
y
1
2
e =
dx,
e =
d +
3
4X
x +y
2

Y
xy
x
4
5
d
3 ,
1 ,
e =
e =
x +y
2
4
(2.9)

e =
3

e =
6
x +y
dy, ,
4Y
xy
2 .
4
(2.10)
The Khler 2-form is then given by

i
J = i i ,
2
and the complex self-dual harmonic (3, 0)-form is given by
G(3) = e3i 1 2 3 W(3) + iW(3) .
(2.11)
(2.12)
Harmonic (2, 1)-forms

We are interested in harmonic (2, 1)-forms that live on the resolved Y pq cones, since their
presence preserves the minimal supersymmetry of the theory. We find there exist the following
five such (2, 1)-forms:
1 =
e3i
1 2 3 ,
xX
2 =
e3i
2 1 3 ,
yY
e3i
3 1 2 ,
xyXY

1
1
1
4 =
2 ( 3 3 1 1 ) 1 ( 3 3 2 2 ) ,
xy x + y x Y
y X
3 =
80

1
1
1
5 =
2 ( 3 3 1 1 ) + 1 ( 3 3 2 2 ) .
x + y x2 Y
y2 X
(2.13)
All of these forms have singularities at all distances x, for certain values of y, except for 1 ,
which has a singularity only at small distance. 1 has a rapid fall off at large distance, such that
it does not support nontrivial flux. On the other hand, in the large-x limit the last harmonic form
behaves like

1
2
1
1
5 = 1 2 3 + d + 2 3 d dy +
21 2 d
4
y
4x
2y

1
1
y
1
+i
(2.14)
1 2 2 3 dy dx + 1 2 dy + O 2 .
y
Y
y
x
This indicates that this form does support nontrivial flux. In the = 0 limit, in which we have
first rescaled y y, 4 and 5 reduce to the same form. This form has a singularity that is
confined to small distance.
2.2. Resolved cones over Labc
We now turn to the resolved cones over the general cohomogeneity-two Labc spaces. The
metric is given by [22]

2
1
1 2 2
u dx + v 2 dy 2 + w 2 dz2 + 2 d + (y + z) d + yz d
4
u
2
2
1
1
+ 2 d + (x + z) d + xz d + 2 d + (x + y) d + xy d ,
v
w
where the functions u, v, w are given by
ds 2 =
(2.15)
(y x)(z x)
(x y)(z y)
(x z)(y z)
,
v2 =
,
w2 =
,
X
Y
Z
X = x( x)( x) 2M,
Y = y( y)( y) 2L1 ,
u2 =
Z = z( z)( z) 2L2 .
(2.16)
Notice that the coordinates x, y and z appear in the metric on a symmetrical footing. We shall
choose x to be the radial direction, and y and z to be the non-azimuthal coordinates on the Labc
level sets. This reduces to the Y pq subset when a = p q, b = p + q and c = d = p.
CalabiYau structure
The complex vielbein can be written as
1 = e1 + ie2 ,
in the vielbein basis
1
e1 = u dx,
2
1
e3 = v dy,
2
1
5
e = w dz,
2
2 = e3 + ie4 ,
3 = e5 + ie6 ,

1
d + (y + z) d + yz d ,
u

1
e4 = d + (x + z) d + xz d ,
v

1
6
e =
d + (x + y) d + xy d .
w
(2.17)
e2 =
(2.18)
81
Then the Khler 2-form and complex self-dual harmonic (3, 0)-form are given by
i
J = i i ,
2
where
G(3) = ei 1 2 3 ,
= 3 + 2( + ) + .
(2.19)
(2.20)

There is a harmonic (2, 1)-form given by
ei 1
2 3 .
X
Using this, one can then construct a general class of harmonic (2, 1)-forms
1 =
(2.21)
1 = f ( )1 ,
(2.22)
for any function f so long as d 1 = 0. This orthogonality condition is obeyed by

Y Z i2
e ,
X
as can be seen by calculating its exterior derivative:
=
d =

2
u(y z)X 1 v(z x)Y 2 w(x y)Z 3 .
(x y)(y z)(z x)
(2.23)
(2.24)
We can consider the special case for which

1 =
(Y Z) i(2+1) 1
e
2 3 ,
X +1
(2.25)
where is a continuous parameter. Due to the dependence, this field only preserves U (1)2
of the U (1)3 isometry of the six-dimensional space. Although the full U (1)3 is preserved for
= 1/2, the form field would blow up at the degeneracies of X, Y and Z, which would lead
to a singular surface in the ten-dimensional geometry. In order for the singularity to be confined
to X = 0, so that we have a reasonable gravity description near the UV region of the dual field
theory, we require that 0.
We find there exist the following (2, 1)-forms:

Y Z i2 ei 1
1 = f
e
2 3 ,
X
X

XZ i2 ei 1
e
2 3 ,
2 = f
Y
Y

XY i2 ei 1
3 = f
e
2 3 ,
Z
Z

4 = a1 A 1 2 2 3 3 + a2 B 2 3 3 1 1

+ a3 C 3 1 1 2 2 ,
82

5 = b1 Ax 1 2 2 3 3 + b2 By 2 3 3 1 1

+ b3 Cz 3 1 1 2 2 ,
(2.26)
where
A1 = (y z)2 (y x)(z x)X,
B 1 = (x z)2 (x y)(z y)Y ,
C 1 = (x y)2 (x z)(y z)Z,
(2.27)
and ai and bi are constants which satisfy the conditions a1 + a2 + a3 = 0 and b1 + b2 + b3 = 0.

Notice that the first three forms in (2.27) are related to each other by interchanging the x, y and
z coordinates, while the last two forms remain invariant. This reflects the fact that the x, y and
z coordinates appear in a completely symmetric manner in the metric of the resolved cone over
Labc . 1 has a singularity that is confined to small distance, as do 4 and 5 if one performs
the rescaling y y, z z and then takes the limit 0. 4 and 5 have nontrivial flux,
while 1 does not.
In the cohomogeneity-two limit, the resolved Labc cones reduce to the resolved Y pq cones.
In this limit, 4 and 5 reduce to the corresponding forms given in (2.13), while the first three
forms generalize those in (2.13) to include an arbitrary function f . In particular, taking f = 1
reproduces the 1 and 2 in (2.13), whilst taking f to be the inverse of its argument reproduces 3 .
3. D3-branes and the AdS/CFT correspondence
A supersymmetric D3-brane solution of the type IIB theory with six-dimensional CalabiYau
transverse space is given by

1
1
ds = H 2 dt 2 + dx12 + dx22 + dx32 + H 2 ds62 ,
F5 = G(5) + G(5) ,
G(5) = dt dx1 dx2 dx3 dH 1 ,
RR
NS
F(3) = F(3)
+ iF(3)
= m(3) ,
(3.1)
with
6 H = m2 |(3) |2 .
(3.2)
and (3) is a harmonic (2, 1)-form

Here the 6 is a Laplacian of the CalabiYau metric
in ds62 . We shall refer to this as a modified D3-brane solution, owing to the inclusion of the
additional 3-form. If this 3-form carries nontrivial flux, then it corresponds to fractional a D3brane.
We shall take the six-dimensional metric ds62 of the transverse space to be the resolved cone
over Labc . We first consider the case of vanishing m. It was shown in [4345] that the Klein
Gordon equation for the general AdS-Kerr-NUT solutions constructed in [22] is separable. Since
our metrics arise as the Euclideanization of the supersymmetric limit of AdS-Kerr-NUT solutions, the corresponding equation for H is hence also separable. To see this, we consider a real
superposition of the ansatz
ds62
H = H1 (x)H2 (y)H3 (z)e2i(a0 a1 +a2 ) .

In general, this ansatz breaks the U (1)3 global symmetry.
(3.3)
The Laplace equation is then given by

(XH1 ) (a0 + a1 x + a2 x 2 )2
1
0=
(y x)(z x)
H1
X

(Y H2 ) (a0 + a1 y + a2 y 2 )2
1
+
(x y)(z y)
H2
Y

(ZH3 )
1
(a0 + a1 z + a2 z2 )2
+
,
(x z)(y z)
H3
Z
83
(3.4)
where a prime denotes a derivative with respect to the separated variable associated with the
function Hi . This equation can be expressed as three separate equations in x, y and z:

(a0 + a1 x + a2 x 2 )2

(XH1 )
+ b0 + b1 x H1 = 0,
X

(a0 + a1 y + a2 y 2 )2
(Y H2 )
+ b0 + b1 y H2 = 0,
Y

(a0 + a1 z + a2 z2 )2
(ZH3 )
(3.5)
+ b0 + b1 z H3 = 0,
Z
where b0 and b1 are separation constants. These equations do not have explicit closed-form
solutions for general ai and bi . We shall consider the simplest solution obtained by setting all of
the ai and bi to zero and letting H depend on x only. The solution is given by
H = c0
c1 log(x x1 )
c1 log(x x2 )
c1 log(x x3 )
+
+
.
(x1 x2 )(x1 x3 ) (x2 x1 )(x2 x3 ) (x3 x1 )(x3 x2 )
(3.6)
where x1 , x2 and x3 are the three roots of X, satisfying

x1 + x2 + x3 = + ,
x1 x2 + x1 x3 + x2 x3 = ,
x1 x2 x3 = 2M.
(3.7)
Consider the radial coordinate x with x1 x . Then the function H has a logarithmic divergence at x1 .
We now consider solutions for which the 3-form (3) is turned on. A simple solution can
be obtained by rescaling y y, z z and then taking the limit = = 0. The general
construction of [19,20] is recovered for the case of this class of resolved Labc cones [17]. We
can then take (3) to be the harmonic (2, 1)-form 1 given by (2.21). Then, for a certain choice
of integration constants, the resulting H is given by
x
,
H=
(3.8)
18M(x 3 2M)
which diverges at x 3 = 2M.
The divergence of the H function in both (3.6) and (3.8) corresponds to a naked singularity in
the short-distance region of the geometry. This singularity of the D3-brane solution arises even in
the case of the resolved cone over Y 2,1 , which itself is completely regular [24]. This singularity
is due to the fact that the D3-branes have been smeared over the blown-up 4-cycle. A shell
of uniformly distributed branes tends to be singular at its surface. For the case of the resolved
conifold, in which there is a blown-up 2-cycle, a completely regular solution has been found for
which the D3-branes are stacked at a single point [26]. This involves solving the equations (3.5)
for the case of T 1,1 for which there is a delta function source. The solution is expressed as an
84
expansion in terms of the angular harmonics. It would be interesting to explore than analogous
construction for the resolved Labc cones. All of these other examples, with the sole exception of
Y 2,1 , will still have orbifold singularities.
Another possible way in which regular solutions can be obtained is to blow up a 3-cycle
instead of a 4-cycle. Then an appropriate 3-form would prevent the 3-cycle from collapsing, as
in the case of the deformed conifold [32]. As already discussed in the previous section, while
there exists an obstruction to complex deformations of Y pq cones there are other subsets of the
Labc cones which do allow for complex structure deformations [12,37,38]. However, the explicit
metrics for these deformed Labc cones are not known.
Although the solution describing D3-branes on a resolved Labc cone becomes singular at short
distance, we can still use this background at large distance to study various flows of the quiver
gauge theory in the region of the UV conformal fixed point. At large x, (3.2) becomes
4
(3.9)
x (Xx H ) = m2 |(3) |2 ,
x2
where X is given by (2.16). Note that this equation applies for arbitrary and , since for large
x we can consistently neglect the dependence of H on the non-azimuthal angular coordinates
y and z. Again considering the case of the self-dual harmonic (2, 1)-form 1 given by (2.21), the
resulting asymptotic expansion of H is

c6
Q
c4
c2
H = 2 1+
(3.10)
+ 2 + 3 + ,
x
x
x
x
where
2
c2 = ( + ),
3

1 2
c4 = + + 2 ,
2

1 m2
+ 12 2 + 2 ( + ) + 2M .
c6 =
30 Q
(3.11)
We have set an additive constant to zero so that the geometry is asymptotically AdS5 Labc .
This can be seen from the leading x 2 r 4 term in H (since x has dimension two, we can take
x r 2 for large x). The transformation properties and dimensions of the operators being turned
on in the dual field theory can be read off from the linearized form of the supergravity solution
(3.1). The metric perturbations due to H have the same form as those within the metric ds62 itself.
Therefore, from the asymptotic expansion of H given in (3.10), we can read off that there are
scalar operators of dimension two, four and six with expectation values that go as c2 , c4 and c6 ,
respectively. This is consistent with the perturbations of the 2-form and 4-form potentials. We
shall now discuss the gauge theory interpretation of the blown-up 2-cycles, as well as the 3-form,
in more detail.
Blown-up 2-cycle
First, we consider the case with vanishing M, for which the six-dimensional space is the Labc
analog of the resolved conifold, in the sense that there is a blown-up 2-cycle. The volume of
the 2-cycle is characterized by the parameters and . This is a global deformation, in that it
changes the position of the branes at infinity [18].
85
The parameters and specify the expectation values of dimension n non-mesonic scalar
operators in the dual gauge theory. For the case = , c2 and c6 vanish, while c4 can only
vanish for = = 0. To identify the specific dimension-two operator whose expectation value
goes as c2 , it is helpful to consider the description of the resolved cone over Labc in terms of four
complex numbers zi which satisfy the constraint
4

Qi |zi |2 = t,
(3.12)
i=1
where one then takes the quotient by a U (1) action [9]. The parameter t is the area of the blownup CP 1 and corresponds to the coefficient of the FayetIliopoulos term in the Lagrangian of
the field theory. The zi correspond to the lowest components of chiral superfields. This can be
described as a gauged linear sigma model with a U (1) gauge group and 4 fields with charges
Qi . Then the above constraint corresponds to setting the D-terms of the gauged linear sigma
model to zero to give the vacuum. For the Labc
spaces, the Qi are given by Qi = (a, c, b, d)
where d = a + b c [11]. The requirement 4i=1 Qi = 0 guarantees that the 1-loop -function
vanishes, so that the sigma model is CalabiYau.
Since t acts as a natural order-parameter in the gauge theory, from (3.12) it is reasonable to
suppose that blowing up the 2-cycle corresponds to giving an expectation value that goes as c2
to the dimension-two scalar operator6
K = aA A cB B + bC C dD D .
(3.13)
This operator lies within the U (1) baryonic current multiplet. Since this conserved current has
no anomalous dimension, the dimension of K is protected. K reduces to the operator discussed
in [18] for the case of a resolved cone over T 11 /Z2 , for which a = b = c = d = 1.
Blown-up 4-cycle
For nonvanishing M in the function X, one generically blows up a 4-cycle. Unlike the case
of a blown-up 2-cycle, this is a local deformation since it does not change the position of the
branes at infinity [18]. In the limit of vanishing and , one recovers the general construction
obtained in [19,20] that has been recently discussed in [1618]. Also note that c6 vanishes for
the appropriate values of M, and .
It has been shown that the number of formal FayetIliopoulos parameters can be matched
with the possible deformations, which is suggestive that the dimension-six operator that is turned
on is associated with the gauge groups in the quiver. Although the specific operator has not been
identified, it has been proposed that they are of the schematic form [18]
Oi =
g,
ci,g Wg W
(3.14)
where the gauge groups in the quiver have been summed over, Wg is an operator associated with
the field strength for the gauge group g, and ci,g are constants. The dimension-six operator might
also have contributions from the bifundamental fields of the form
6 We thank Amihay Hanany and Igor Klebanov for correspondence on this point.
86
a1 A A B B C C + a2 A A B B D D + a3 A A C C D D
+ a4 B B C C D D ,
(3.15)
where the ai are constants. It is proposed that a particular combination of all of these terms in
(3.14) and (3.15) correspond to the blown-up 4-cycle.7 One possibility is that the contributions
from the bifundamental fields in (3.15) are present only when and are nonvanishing.
Turning on the 3-form
Turning on a 3-form results in the ranks of some of the gauge groups of the dual quiver
gauge theory being reduced with decreasing energy scale. For the case in which the 3-form has
nontrivial flux, the theory undergoes a Seiberg duality cascade [27,28,32]. On the other hand,
the 3-form 1 given by (2.21) does not have nontrivial flux. For a case such as this, it has been
proposed that the reduction in ranks of gauge groups is due to Higgsing [34]. In particular, from
(3.10), we see that the parameter m associated with the 3-form also contributes to the expectation
value c6 of a dimension-six scalar operator. An additional effect of this 3-form is that the U (1)
R-symmetry is broken. The theory still preserves N = 1 supersymmetry.
4. Eight-dimensional resolved CalabiYau cones
4.1. Cohomogeneity-two metrics
We now turn to eight-dimensional CalabiYau spaces, which can be used to construct M2brane solutions of eleven-dimensional supergravity. Before considering the general cohomogeneity-four resolved cones over Lpqrs , we shall first look at the cohomogeneity-two metrics, which
can be built over an S 2 S 2 base space. These metrics are given by [21,22]
2
2

1 2 2 1 2 2
1
1
y
x
2
ds8 = u dx + v dy + 2 d +
(3 + 3 ) + 2 d
((3) + 3 )
4
4
3
3
u
v

+ c2 12 + 22 + 12 + 22 ,
x +y
x+y
xy
u2 =
,
v2 =
,
c2 =
,
X
Y
6
2
2
X = x(x + ) 2 ,
(4.1)
Y = y( y) + 2 ,
x
y
Completely regular examples were discussed in [25].
CalabiYau structure
We can define the vielbein basis

1
1
1
y
1
2
e = u dx,
e3 = v dy,
e = d +
(3 + 3 ) ,
2
u
3
2

1
x
e4 =
e5 = c1 ,
e6 = c2 ,
d
(3 + 3 ) ,
v
3
e7 = c 1 ,
e8 = c 2 ,
7 We thank Sergio Benvenuti for correspondence on this point.
(4.2)
87
and then the complex vielbein

1 = e1 + ie2 ,
2 = e3 + ie4 ,
3 = e5 + ie6 ,
4 = e7 + ie8 .
(4.3)
The Khler 2-form and holomorphic (4, 0)-form are given by

i
J = i i ,
2
(4.4)
G(4) = e4i 1 2 3 4 .
(4.5)
and

We find four self-dual (2, 2)-forms; they are given by
( 1 1 + 2 2 ) ( 3 3 + 4 4 ) 2( 1 1 2 2 + 3 3 4 4 )
,
x3y3
( 1 1 2 2 ) ( 3 3 4 4 )
2 =
,
xy(x + y)2
1 =
e4i ( 1 2 3 4 + 1 2 3 4 )
,
x 2 y 2 XY
( 1 2 1 2 ) ( 3 3 + 4 4 )
4 =
.
xy XY
3 =
(4.6)
Notice that 1 and 2 are square integrable, in that they are well behaved at both small and large
asymptotic distance. For the cases in which the eight-dimensional CalabiYau spaces are regular
[25], these harmonic forms can be used to construct completely non-singular M2-brane solutions
to eleven-dimensional supergravity.
4.2. Cohomogeneity-four metrics on resolved cones over Lpqrs
We now turn to the general cohomogeneity-four metrics on resolved CalabiYau cones over
the seven-dimensional EinsteinSasaki spaces Lpqrs , which can be written as [22]
ds82 =
where

1 2 2
u1 dx1 + u22 dx22 + u23 dx32 + u24 dx42
4
2
1
+ 2 d + (x2 + x3 + x4 ) d + (x2 x3 + x2 x4 + x3 x4 ) d + x2 x3 x4 d
u1
2
1
u2
2
1
u3
2
1
+ 2 d + (x1 + x2 + x3 ) d + (x1 x2 + x1 x3 + x2 x3 ) d + x1 x2 x3 d ,
u4
(4.7)
88
(x2 x1 )(x3 x1 )(x4 x1 )

(x1 x2 )(x3 x2 )(x4 x2 )
,
u22 =
,
X1
X2
(x1 x3 )(x2 x3 )(x4 x3 )
(x1 x4 )(x2 x4 )(x3 x4 )
,
u24 =
,
u23 =
X3
X4
X1 = x1 (a x1 )(b x1 )(c x1 ) 2M1 ,
u21 =
X2 = x2 (a x2 )(b x2 )(c x2 ) 2M2 ,

X3 = x3 (a x3 )(b x3 )(c x3 ) 2M3 ,
X4 = x4 (a x4 )(b x4 )(c x4 ) 2M4 .
(4.8)
CalabiYau structure
We shall choose the vielbein basis
1
1
1
1
e5 = u3 dx3 ,
e7 = u4 dx4 ,
e1 = u1 dx1 , u3 = u2 dx2 ,
2
2
2
2

1
d + (x2 + x3 + x4 ) d + (x2 x3 + x2 x4 + x3 x4 ) d + x2 x3 x4 d ,
e2 =
u1

1
e4 =
u2

1
e6 =
u3

1
d + (x1 + x2 + x3 ) d + (x1 x2 + x1 x3 + x2 x3 ) d + x1 x2 x3 d .
e8 =
u4
(4.9)
The holomorphic vielbein are then given by

1 = e1 + ie2 ,
2 = e3 + ie4 ,
3 = e5 + ie6 ,
4 = e7 + ie8 .
(4.10)
Defining

i 1
1 + 2 2 + 3 3 + 4 4 ,
2
= ei 1 2 3 4 ,
J=
(4.11)
where
= 4 + 3(a + b + c) + 2(ab + bc + ca) + abc,
(4.12)
it is straightforward to verify that

dJ = 0,
d = 0,
(4.13)
and hence that the metric is indeed Ricci-flat Khler, with J being the Khler form and the
holomorphic (4, 0)-form.
We find that harmonic (3, 1)-forms can be constructed as follows. First, it can be verified that
G(3,1) =
1 i 1
e 2 3 4
X1
(4.14)
is closed, and hence harmonic. Next, we define the function

X2 X3 X4 i
e ,
=
X1
89
(4.15)
which can be shown to satisfy the relation

d =

u1 ei
u1 (x2 x3 )(x2 x4 )(x4 x3 )X1 1
u2 u3 u4 (x1 x2 )(x1 x3 )(x1 x4 )
u2 (x3 x1 )(x3 x4 )(x4 x1 )X2 2 + u3 (x1 x2 )(x4 x1 )(x4 x2 )X3 3

+ u4 (x1 x2 )(x3 x1 )(x2 x3 )X4 4 ,
(4.16)
where Xi denotes the derivative of Xi with respect to its argument xi . It therefore follows that
d G(3,1) = 0, and so
(3,1) = f ( )G(3,1)
(4.17)
is a harmonic (3, 1)-form for any function f . In particular, we have a family of harmonic (3, 1)forms given by
(3,1) =
X2 X3 X4
X1+1
e(2+1)i 1 2 3 4 ,
(4.18)
for any constant . For nonzero , these forms preserve only a U (1)3 subgroup of the U (1)4
isometry of the eight-dimensional space. Note that (3,1) has a singularity only at short distance
if 0, where we have taken x1 to be the radial direction. Additional harmonic (3, 1)-forms can
be constructed by permuting the xi directions, but these forms have singularities for all x1 . They
are analogous to the (2, 1)-forms 1 , 2 and 3 in (2.27) for a six-dimensional space, and they
do not support nontrivial flux.
We can also construct harmonic (2, 2)-forms as follows. We define (2, 2)-forms

G(2,2) = f 1 1 2 2 + 3 3 4 4

+ g 1 1 3 3 + 2 2 4 4

+ h 1 1 4 4 + 2 2 3 3 ,
(4.19)
where f , g and h are functions of (x1 , x2 , x3 , x4 ). Imposing the closure of G(2,2) leads to three
independent solutions for f , g and h, namely
f = g = h = 1,
1
,
f=
2
(x1 x2 ) (x1 x3 )(x2 x4 )(x3 x4 )2
x1 (2x4 x2 x3 ) + x2 (2x3 x4 ) x3 x4
g=
,
(x1 x2 )2 (x1 x3 )2 (x2 x4 )2 (x3 x4 )2
1
,
h=
(x1 x2 )(x1 x3 )2 (x2 x4 )2 (x3 x4 )
(4.20)
(4.21)
90
and
1
,
(x1 x3 )(x2 x3 )2 (x1 x4 )2 (x2 x4 )
1
g=
,
(x1 x3 )2 (x2 x3 )(x1 x4 )(x2 x4 )2
x1 (x3 + x4 2x2 ) + x2 (x3 + x4 ) 2x3 x4
.
h=
(x1 x3 )2 (x2 x3 )2 (x1 x4 )2 (x2 x4 )2
f=
(4.22)
These forms are somewhat analogous to the (2, 1)-forms 4 and 5 given in (2.27) for a sixdimensional space. The first solution, (4.20), is just the harmonic (2, 2)-form J J . It follows
from (4.19) that J G(2,2) is proportional to (f + g + h), and so J G(2,2) is non-zero for
(4.20). However, each of the solutions (4.21) and (4.22) satisfies f + g + h = 0, and so these two
harmonic (2, 2)-forms satisfy the supersymmetric condition
J G(2,2) = 0.
(4.23)
Notice also that these harmonic (2, 2)-forms are square integrable. These can be used to construct
modified M2-brane solutions, which have only orbifold-type singularities. Note that none of
these cohomogeneity-four CalabiYau spaces are completely regular [25].
4.3. M2-brane solutions
We can use these eight-dimensional spaces, and the harmonic 4-forms which they support, to
construct a modified M2-brane solution to eleven-dimensional supergravity, given by

2
= H 2/3 dt 2 + dx12 + dx22 + H 1/3 ds82 ,
ds11
F(4) = dt dx1 dx2 dH 1 + mL(4) ,
(4.24)
where
1 2 2
(4.25)
m L(4) ,
48
is an (anti)self-dual harmonic 4-form on the eight-dimensional space with the metric
H =
and L(4)
ds82 .
Let us first consider the case with m = 0, for which the Laplace equation on the CalabiYau
metric is separable. The solution for general dimensionality is presented in Appendix B. Here we
just give a solution for the eight-dimensional case that depends only on the radial variable x1 ; it
is given by
x1
H=
3Q
dx .
X(x1 ) 1
Thus in the asymptotic region at large x1 , the function H has the behavior

Q
3
c2
H = 3 1+
+ , where c2 = ( + + ).
x
4
x1
1
(4.26)
(4.27)
We have taken an arbitrary additive constant to zero, so that the geometry is asymptotically
AdS4 Lpqrs . Since x1 has dimension two, we see that there is a non-mesonic dimension-two
scalar operator being turned on with expectation value c2 .
91
It is especially interesting to construct M2-brane solutions using one of the square-integrable

harmonic (2, 2)-forms that we found previously, since this guarantees that with the appropriate integration constants the only singularities are of orbifold type. This is because the
4-form prevents the blown-up 4-cycle from collapsing. Moreover, examples of regular eightdimensional CalabiYau spaces that have been discussed in [25] can be used to construct
completely non-singular M2-brane solutions. The resulting geometry smoothly interpolates between AdS4 Lpqrs asymptotically, and a direct product of Minkowski3 and a compact space
at short distance. Many examples of cohomogeneity-one solutions of this type were constructed
in [4951]. Although not much is known even about the UV conformal fixed point of the dual
three-dimensional N = 2 super-YangMills field theory, based on the geometrical properties of
the supergravity background it flows to a confining phase in the IR region.
5. Harmonic forms on higher-dimensional resolved cones
In this section, we extend some of the constructions of harmonic middle-dimension forms to
the case of higher-dimensional metrics on the resolutions of cones over EinsteinSasaki spaces.
We take as our starting point the local Ricci-flat Khler metrics in dimension D = 2n + 4 that
were considered in [25]:

x +y 2 x +y 2
X
Y
y 2
x 2 xy
2
d s =
dx +
dy +
d + +
d +
dn2 ,
4X
4Y
x +y
x+y
2
2
= d + A,
X = x(x + ) n ,
(5.1)
Y = y( y) + n ,
x
y
where dn2 is a metric on a 2n-dimensional EinsteinKhler space Z, satisfying Rab = 2(n +
1)gab , with Khler form J = 12 dA. (We have made some minor changes of coordinates compared
to the metric presented in [25].) For convenience, we shall set the constant to unity. This can
always be done, when
= 0, by means of coordinate scalings together with an overall rescaling
of the Ricci-flat metric. The special case = 0 can be recovered via a limiting procedure.
Next, we define the 2-forms
1
1
x = dx (d + y ),
(5.2)
y = dy (d x ),
= xyJ.
2
2
It can easily be verified that J x y + is closed and, in fact, this is the Khler form of the
Ricci-flat Khler metric (5.1). In the case that n is even (n = 2m), we find that the middle-degree
form

1
1
1
m1
m
m+1
G(2m+2) =
x
y
x
y
m+1
m(m + 1)
(xy)2m+1
(5.3)
is closed. Since it is also self-dual, it follows that it is a harmonic form. This generalises the
harmonic (2, 2)-form 1 in eight dimensions given in (4.6) and is somewhat analogous to the
(2, 1)-forms 4 and 5 given in (2.27) for a six dimensions.
Further harmonic forms can be obtained if one takes the EinsteinKhler base metric dn2
to be a product of EinsteinKhler metrics. For example, if we choose it to be the product of
metrics on two copies of CPm (recall that we are considering the case where n = 2m is even),
with Khler forms J1 and J2 respectively (so J = J1 + J2 ), then defining
1 = xyJ1 ,
2 = xyJ2 ,
(5.4)
92
we find that
(2m+2) =
G

1
mp
p
(x + y )
(1)p 1
2
2
m
(x + y) (xy)
m
(5.5)
p=0
is closed and self-dual, and therefore it is harmonic.

6. Conclusions
We have investigated the Khler moduli associated with blowing up a 2-cycle or 4-cycle on
CalabiYau cones over the Labc spaces. This yields a countably infinite number of backgrounds
with ALE singularities on which perturbative string dynamics is well defined. Although adding
D3-branes induces a power-law type singularity at short distance, one can still use the AdS/CFT
dictionary to relate the blown-up cycles to deformations of the dual quiver gauge theory close
to the UV conformal fixed point. In particular, we identify the non-mesonic dimension-two real
scalar operator that acquires a vev, thereby generalizing the state/operator correspondence for the
resolved conifold over T 11 [14] and T 11 /Z2 [18] to resolved cones over the Labc spaces. On the
other hand, blowing up a 4-cycle corresponds to a dimension-six non-mesonic scalar operator
getting a vev.
The resolved cones over the cohomogeneity-two Labc spaces support various harmonic (2, 1)forms, some of which depend nontrivially on three non-azimuthal coordinate directions. These
forms can be further generalized by a multiplicative function, so long as the exterior derivative of
this function satisfies a certain orthogonality condition. In particular, there are harmonic (2, 1)forms which depend on continuous parameters. 3-forms carrying nontrivial flux correspond to
fractional D3-branes, while those which do not correspond to giving a vev to a dimension-six
operator.
For the D3-brane solutions constructed with resolved cones over Labc , we have restricted
ourselves to the case in which the D3-branes are smeared over the blown-up cycle. As we already
mentioned, this yields to a power-law singularity at short distance. For solutions involving a 3form field, one may be able to smooth out this singularity by a complex deformation of the
CalabiYau space that results in a blown-up 3-cycle. Although it has been shown that there are
obstructions to the existence of complex deformations of cones over Y pq spaces, there are other
subsets of the Labc cones which do allow for complex structure deformations [12,37,38]. It would
be useful to construct the explicit metrics describing these deformed Labc cones, as well as the
non-singular supergravity solutions that describe fractional D3-branes on these spaces.
Alternatively, one can consider stacking the D3-branes at a single point. For the case of the
resolved conifold, this has been shown to yield a completely regular solution [26]. Perhaps there
are analogous constructions with the resolved cones over Labc . With the exceptions of T 1,1 ,
T 1,1 /Z2 and Y 2,1 , the resolved Labc cones have orbifold singularities. Although these singularities will remain there when D3-branes are stacked at a single point, perturbative string dynamics
is well-defined on such backgrounds.
One can also consider fibering a D3-brane worldvolume direction (which need not be compact) over a resolved Labc cone in such a way that the resulting geometry only has orbifold-type
singularities. For the case of the resolved conifold, such a D3-brane solution has already been
constructed and is completely regular, and it is also supersymmetric [46]. The corresponding
D3-brane solutions for the resolved Labc cones are currently being investigated [47].
We also discussed the geometry of higher-dimensional CalabiYau spaces with blown-up cycles, as well as the various harmonic forms which live on them. In particular, we have found
93
that eight-dimensional resolved cones over the Lpqrs spaces support harmonic 4-forms that are
square integrable. They can be used to construct M2-brane solutions of eleven-dimensional supergravity which have only orbifold-type singularities. Unfortunately, not much is known about
the dual three-dimensional N = 2 gauge theories, other than that they flow from a UV conformal
fixed point to a confining phase in the IR region.
Lastly, the type IIB supergravity backgrounds dual to certain marginal deformations ( deformations) of the conformal fixed point of the Y pq and Labc quiver gauge theories were obtained
in [48,52]. The solution-generating method works for any gravity solution with U (1) U (1)
global symmetry. It might be interesting to see if these deformations can be applied to the gravity solutions discussed in this paper, since they possess the necessary global symmetry.
Acknowledgements
We should like to thank Philip Argyres, Sergio Benvenuti, Aaron Bergman, Davide Forcella,
Amihay Hanany, Igor Klebanov, Jason Kumar, Louis Leblond, Leopoldo Pando Zayas and Angel
Uranga for helpful conversations and correspondence. M.C. thanks the George P. & Cynthia
W. Mitchell Institute for Fundamental Physics for hospitality during the course of this work.
Appendix A. Complex structure and first-order equations
In this appendix, we construct Ricci-flat Khler spaces in dimension D = 2n + 4, built over
an EinsteinKhler base space of real dimension 2n with metric dn2 . We normalise this metric
so that it satisfies Rij = 2(n + 1)gij . Its Khler form will be written as J = 12 dA. We may also
assume that it admits a holomorphic (n, 0)-form , satisfying (see, for example, Section 4 of
[23])
d = i(n + 1)A .
(A.1)
The ansatz for the (3n + 4)-dimensional Ricci-flat Khler metrics will be
d s 2 = u2 dx 2 + v 2 dy 2 + a 2 (d + f1 )2 + b2 (d + f2 )2 + c2 dn2 ,
(A.2)
where a, b, c, u, v, f1 and f2 are functions of x and y, and

= d + A.
(A.3)
We define the vielbein

e1 = u dx,
e2 = a(d + f1 ),
e4 = b(d + f2 ),
e3 = v dy,
ei = cei ,
(A.4)
where ei is a vielbein for the EinsteinKhler base metric dn2 .

We make the ansatz
J = e1 e2 + e3 e4 + c2 J
(A.5)
for the Khler form. It is then natural to define a complex vielbein by

1 = e1 + ie2 ,
where
2 = e3 + ie4 ,
i = c i ,
is a complex vielbein for the base metric
= ei +i cn 1 2
dn2 .
(A.6)
We also make the ansatz
(A.7)
94
for the holomorphic (n + 2, 0)-form. The conditions for d s 2 to be Ricci flat and Khler are then
given by
d J = 0,
d = 0.
(A.8)
One immediately finds that the constant should be chosen to be

= n + 1.
(A.9)
However, the constant can be left arbitrary.

We now obtain the first-order equations:
2
2 .
d J = 0: (bv) (au). = 0,
c 2auf1 = 0,
c 2bvf2 = 0,

.

n
n
n
d = 0: uvc avc buc = 0,

bucn f2 (n + 1)bucn + abcn (f1 f2 ) = 0,

.
avcn f1 (n + 1)avcn abcn (f1 f2 ) = 0.
(A.10)
The constant appearing in the first-order equations (A.10) is always trivial, in the sense that it
can be set to any chosen non-zero value without loss of generality. To see this, we perform the
following rescaling of coordinates and functions:
x x,
y y,
c c,
f1 f1
f2 f2 ,
(A.11)
whilst leaving the functions a, b, u and v unscaled. It can be seen that the effect of these rescalings is to scale the metric d s 2 in (A.2) according to
d s 2 2 d s 2 .
(A.12)
The rescalings have the effect of replacing by in the first-order equations (A.10), thus giving
2
2 .
d J = 0: (bv) (au). = 0,
c 2auf1 = 0,
c 2bvf2 = 0,

.

n
n
n
d = 0: uvc avc buc = 0,

bucn f2 (n + 1)bucn + abcn (f1 f2 ) = 0,

.
avcn f1 (n + 1)avcn abcn (f1 f2 ) = 0.
(A.13)
Since a rescaling of a Ricci-flat metric by a non-zero constant leaves it Ricci-flat, it follows that
the constant can be chosen at will, and so no generality is lost by setting to any desired finite
and non-zero value.
Appendix B. Separability of Laplacian on CalabiYau metrics
We consider the CalabiYau metrics obtained in [21,22]. The metric can be expressed as

n1
2
n

U dx2
X
2
+
Wi di
,
ds =
4X
U
=1
X = x
n1

i=1
i=0
(i x ) 2 ,
U =
n

=1
(x x ),
(B.1)
95
where Wi is defined by
n

(1 + qx )
=1
n1

Wi q i+1 .
(B.2)
i=0
It turns out that the equation H = 0 is separable in the x coordinates, where is the Laplacian
taken on the above metric. (The separability for the more general non-extremal KerrNUTAdS
metrics was shown explicitly in [4345].) Making the ansatz

n1

n

H (x ) exp 2i
(1)i ai n1i ,
H=
(B.3)
=1
i=0
for the harmonic function, we find that the H (x ) satisfy

n1

n2
( i=0 ai xi )2

i
(X H )
+
bi x H = 0,
X
(B.4)
i=1
where a prime on H or X denotes a derivative with respect to its argument x . The system
thus has 2n 1 independent separation constants a0 , a1 , . . . , an1 and b0 , b1 , . . . , bn2 .
References
[1] J.M. Maldacena, The large N limit of superconformal field theories and supergravity, Adv. Theor. Math. Phys. 2
(1998) 231;
J.M. Maldacena, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200.
[2] S.S. Gubser, I.R. Klebanov, A.M. Polyakov, Gauge theory correlators from non-critical string theory, Phys. Lett.
B 428 (1998) 105, hep-th/9802109.
[3] E. Witten, Anti-de Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150.
[4] J.P. Gauntlett, D. Martelli, J. Sparks, D. Waldram, Supersymmetric AdS5 solutions of M-theory, Class. Quantum
Grav. 21 (2004) 4335, hep-th/0402153.
[5] J.P. Gauntlett, D. Martelli, J. Sparks, D. Waldram, SasakiEinstein metrics on S 2 S 3 , Adv. Theor. Math. Phys. 8
(2004) 711, hep-th/0403002.
[6] M. Cvetic, H. L, D.N. Page, C.N. Pope, New EinsteinSasaki spaces in five and higher dimensions, Phys. Rev.
Lett. 95 (2005) 071101, hep-th/0504225.
[7] M. Cvetic, H. L, D.N. Page, C.N. Pope, New EinsteinSasaki and Einstein spaces from Kerrde Sitter, hep-th/
0505223.
[8] I.R. Klebanov, E. Witten, Superconformal field theory on threebranes at a CalabiYau singularity, Nucl. Phys. B 536
(1998) 199, hep-th/9807080.
[9] D. Martelli, J. Sparks, Toric geometry, SasakiEinstein manifolds and a new infinite class of AdS/CFT duals, Commun. Math. Phys. 262 (2006) 51, hep-th/0411238.
[10] S. Benvenuti, S. Franco, A. Hanany, D. Martelli, J. Sparks, An infinite family of superconformal quiver gauge
theories with SasakiEinstein duals, JHEP 0506 (2005) 064, hep-th/0411264.
[11] S. Benvenuti, M. Kruczenski, From SasakiEinstein spaces to quivers via BPS geodesics: Lpqr, JHEP 0604 (2006)
033, hep-th/0505206.
[12] S. Franco, A. Hanany, D. Martelli, J. Sparks, D. Vegh, B. Wecht, Gauge theories from toric geometry and brane
tilings, JHEP 0601 (2006) 128, hep-th/0505211.
[13] A. Butti, D. Forcella, A. Zaffaroni, The dual superconformal theory for Lpqr manifolds, JHEP 0509 (2005) 018,
hep-th/0505220.
[14] I.R. Klebanov, E. Witten, AdS/CFT correspondence and symmetry breaking, Nucl. Phys. B 556 (1999) 89, hepth/9905104.
[15] L.A. Pando Zayas, A.A. Tseytlin, 3-branes on spaces with R S 2 S 3 topology, Phys. Rev. D 63 (2001) 086006,
hep-th/0101043.
[16] S.S. Pal, A new Ricci flat geometry, Phys. Lett. B 614 (2005) 201, hep-th/0501012.
96
[17] K. Sfetsos, D. Zoakos, Supersymmetric solutions based on Y pq and Lpqr , Phys. Lett. B 625 (2005) 135, hep-th/
0507169.
[18] S. Benvenuti, M. Mahato, L.A. Pando Zayas, Y. Tachikawa, The gauge/gravity theory of blown up four cycles,
hep-th/0512061.
[19] L. Berard-Bergery, Quelques exemples de varietes riemanniennes completes non compactes a courbure de Ricci
positive, C. R. Acad. Sci. Paris, Ser. I 302 (1986) 159.
[20] D.N. Page, C.N. Pope, Inhomogeneous Einstein metrics on complex line bundles, Class. Quantum Grav. 4 (1987)
213.
[21] W. Chen, H. L, C.N. Pope, Kerrde Sitter black holes with NUT charges, Nucl. Phys. B 762 (2007) 38, hep-th/
0601002.
[22] W. Chen, H. L, C.N. Pope, General KerrNUTAdS metrics in all dimensions, Class. Quantum Grav. 23 (2006)
5323, hep-th/0604125.
[23] H. L, C.N. Pope, J.F. Vazquez-Poritz, A new construction of EinsteinSasaki metrics in D 7, Phys. Rev. D 75
(2007) 026005, hep-th/0512306.
[24] T. Oota, Y. Yasui, Explicit toric metric on resolved CalabiYau cone, Phys. Lett. B 639 (2006) 54, hep-th/0605129.
[25] H. L, C.N. Pope, Resolutions of cones over EinsteinSasaki spaces, hep-th/0605222.
[26] I.R. Klebanov, A. Murugan, Gauge/gravity duality and warped resolved conifold, hep-th/0701064.
[27] I.R. Klebanov, N.A. Nekrasov, Gravity duals of fractional branes and logarithmic RG flow, Nucl. Phys. B 574
(2000) 263, hep-th/9911096.
[28] I.R. Klebanov, A.A. Tseytlin, Gravity duals of supersymmetric SU(N ) SU(N + M) gauge theories, Nucl. Phys.
B 578 (2000) 123, hep-th/0002159.
[29] C.P. Herzog, Q.J. Ejaz, I.R. Klebanov, Cascading RG flows from new SasakiEinstein manifolds, JHEP 0502 (2005)
009, hep-th/0412193.
[30] D. Martelli, J. Sparks, Toric SasakiEinstein metrics on S 2 S 3 , Phys. Lett. B 621 (2005) 208, hep-th/0505027.
[31] D. Gepner, S.S. Pal, Branes in Lpqr , Phys. Lett. B 622 (2005) 136, hep-th/0505039.
[32] I.R. Klebanov, M.J. Strassler, Supergravity and a confining gauge theory: Duality cascades and SB-resolution of
naked singularities, JHEP 0008 (2000) 052, hep-th/0007191.
[33] L.A. Pando Zayas, A.A. Tseytlin, 3-branes on resolved conifold, JHEP 0011 (2000) 028, hep-th/0010088.
[34] O. Aharony, A note on the holographic interpretation of string theory backgrounds with varying flux, JHEP 0103
(2001) 012, hep-th/0101013.
[35] P. Candelas, X.C. de la Ossa, Comments on conifolds, Nucl. Phys. B 342 (1990) 246.
[36] B.A. Burrington, J.T. Liu, M. Mahato, L.A. Pando Zayas, Towards supergravity duals of chiral symmetry breaking
in SasakiEinstein cascading quiver theories, JHEP 0507 (2005) 019, hep-th/0504155.
[37] K. Altmann, The versal deformation of an isolated toric Gorenstein singularity, alg-geom/9403004.
[38] K. Altmann, Infinitesimal deformations and obstructions for toric singularities, alg-geom/9405008.
[39] D. Berenstein, C.P. Herzog, P. Ouyang, S. Pinansky, Supersymmetry breaking from a CalabiYau singularity,
JHEP 0509 (2005) 084, hep-th/0505029.
[40] S. Franco, A. Hanany, F. Saad, A.M. Uranga, Fractional branes and dynamical supersymmetry breaking, JHEP 0601
(2006) 011, hep-th/0505040.
[41] M. Bertolini, F. Bigazzi, A.L. Cotrone, Supersymmetry breaking at the end of a cascade of Seiberg dualities, Phys.
Rev. D 72 (2005) 061902, hep-th/0505055.
[42] A. Brini, D. Forcella, Comments on the non-conformal gauge theories dual to Y pq manifolds, JHEP 0606 (2006)
050, hep-th/0603245.
[43] D.N. Page, D. Kubiznk, M. Vasudevan, P. Krtou, Complete integrability of geodesic motion in general Kerr
NUTAdS spacetimes, Phys. Rev. Lett. 98 (2007) 061102, hep-th/0611083.
[44] V.P. Frolov, P. Krtou, D. Kubiznk, Separability of HamiltonJacobi and KleinGordon equations in general Kerr
NUTAdS spacetimes, JHEP 0702 (2007) 005, hep-th/0611245.
[45] P. Krtou, D. Kubiznk, D.N. Page, V.P. Frolov, KillingYano tensors, rank-2 Killing tensors and conserved quantities in higher dimensions, JHEP 0702 (2007) 004, hep-th/0612029.
[46] H. L, J.F. Vzquez-Poritz, S 1 -wrapped D3-branes on conifolds, Nucl. Phys. B 633 (2002) 114, hep-th/0202175.
[47] P. Argyres, M. Edalati, J.F. Vzquez-Poritz, in preparation.
[48] O. Lunin, L. Maldacena, Deforming field theories with U (1) U (1) global symmetry and their gravity duals,
JHEP 0505 (2005) 033, hep-th/0502086.
[49] M. Cvetic, H. L, C.N. Pope, Brane resolution through transgression, Nucl. Phys. B 600 (2001) 103, hepth/0011023.
97
[50] M. Cvetic, G.W. Gibbons, H. L, C.N. Pope, Ricci-flat metrics, harmonic forms and brane resolutions, Commun.
Math. Phys. 232 (2003) 457, hep-th/0012011.
[51] M. Cvetic, G.W. Gibbons, H. L, C.N. Pope, Hyper-Khler Calabi metrics, L2 harmonic forms, resolved M2-branes,
and AdS4 /CFT3 correspondence, Nucl. Phys. B 617 (2001) 151, hep-th/0102185.
[52] C. Ahn, J.F. Vzquez-Poritz, Marginal deformations with U (1)3 global symmetry, JHEP 0507 (2005) 032, hepth/0505168.
Massive N = 1 supermultiplets with arbitrary

superspins
Yu.M. Zinoviev
Institute for High Energy Physics, Protvino, Moscow Region, 142280, Russia
Abstract
In this paper we give explicit construction of massive N = 1 supermultiplets in flat d = 4 Minkowski
spacetime. We work in a component on-shell formalism based on gauge invariant description of massive
integer and half-integer spin particles where massive supermultiplets are constructed out of appropriate set
of massless ones.
0. Introduction
In a flat spacetime massive spin s particles in a massless limit decompose into massless spin
s, s 1, . . . ones. This, in particular, leads to the possibility of gauge invariant description of
massive spin s particles, e.g. [113]. In this, two different approaches could be used. From one
hand, one can start with usual non-gauge invariant description of massive particle and achieve
gauge invariance through the introduction of additional fields (thus promoting second class constraints into the first class ones). From the other hand, one can start with the appropriate set of
massless particles having gauge invariance from the very beginning and obtain massive particle
description as a deformation of massless theory. This last approach closely mimic situation in
spontaneous gauge symmetry breaking where gauge field has to eat some Goldstone field(s) to
become massive.
In the supersymmetric theories all particles must belong to some supermultiplet, massive or
massless. Till now most of investigations in supersymmetric theories where bounded to massless
supermultiplets. Quite a few results on massive supermultiplets mainly devoted to superspins 1
E-mail address: yurii.zinoviev@ihep.ru.
doi:10.1016/j.nuclphysb.2007.06.008
99
Yu.M. Zinoviev / Nuclear Physics B 785 (2007) 98114
and 3/2 exist [1421]. The aim of this paper is to extend these results to include massive N = 1
supermultiplets with arbitrary superspins. Certainly, it would be nice to have superfield off-shell
description of such supermultiplets, but as previous results clearly show it is a highly non-trivial
task. So in this paper we restrict ourselves with component on-shell formalism in terms of physical fields. The same reasoning on the massless limit means that massive supermultiplets could
(should) be constructed out of the massless ones in the same way as massive particles out of the
massless ones. So our approach will be supersymmetric generalization of the second approach
to massive particle description mentioned above. Namely, we will start with appropriate set of
massless supermultiplets and obtain massive one as a smooth deformation.
The paper is organized as follows. Though our previous examples on massive superspin 1
[21] and superspin 3/2 [15] supermultiplets already give important hints on how general case of
arbitrary superspin could looks like, due to peculiarities of lower spin fields they are not enough
to achieve such generalization. Thus, in the first two sections we give two more concrete examples, namely massive supermultiplets with superspins 2 and 5/2, correspondingly. All these and
subsequent results heavily depend on the gauge invariant description of massive particles with
integer [2,3] and half-integer [10] spins as well as on the known form of massless supermultiplets [22]. For reader convenience and to make paper self-contained, in the next two sections we
give all necessary formulas in compact condensed notations. One of the lessons from previous
investigations is that the structures of massive supermultiplets with integer and half-integer superspins are different, so in the last two sections we consider these two cases separately. We will
see that, in spite of large number of fields, all calculations are pretty straightforward and mainly
combinatorial.
1. Superspin 2
Massive superspin 2 supermultiplet contains four massive particles with spins 5/2, 2, 2
and 3/2, correspondingly. In the massless limit massive supermultiplets must decompose into
the appropriate set of massless ones in the same way as massive spin s particlesinto massless
spin s, s 1, . . . ones. Simple counting of physical degrees of freedom immediately gives:
5/2
2

2
3/2
5/2
2
2
3/2
3/2
1
1
1/2
1/2
0, 0

.
So we will start with five massless supermultiplets ( , h ), (f , ), ( , A ), (B , )

and (, z). From our previous experience with massive superspin 1 and superspin 3/2 supermultiplets we know that it is crucial for the construction of massive supermultiplets to make dual
rotation of vector A and axial-vector B fields mixing massless supermultiplets containing
these fields. But now we have two tensor fields h and f as well, moreover they necessarily
must be tensor and pseudo-tensor ones. Thus we have to consider the possibility to mix massless
supermultiplets with these fields as well and the real structure of massless supermultiplets we are
going to work with looks like:

.
z
Then, introducing a sum of the massless Lagrangians for bosonic fields:
100
1
1
L0 = h h (h) (h) + (h) h h h
2
2
1
1
+ f f (f ) (f ) + (f ) f f f
2
2
1 2
1 2
1
1
2
2
A B + ( ) + ( )
4
4
2
2
as well as sum of the massless Lagrangians for fermionic fields:
(1)
i
i
2i(
) + i(
) () + i(
) (
)
L0 =
2
4
i
i
i
) +
+ i( )( ) ( )(
2
2
2
i
i
) + i
+ i(
)(
)() (

(2)
2
2
2
it is not hard to check that the most general supertransformations leaving sum of massless Lagrangians invariant have the form (round brackets denote symmetrization):

i
= cos(2 )h( ) sin(2 )5 f( ) ,
2
h = 2 cos(2 )( ) + i sin(2 )( ( ) ),
f = 2 sin(2 )( 5 ) + i cos(2 )( ( ) 5 ),

= sin(2 )h + cos(2 )f 5 ,
(3)
for the supermultiplets containing spin 2 fields and

i
= cos(1 )A sin(1 )5 B ,
2 2
),
A = 2 cos(1 )( ) + i sin(1 )(
5 ),
B = 2 sin(1 )( 5 ) + i cos(1 )(

1
= sin(1 )A + cos(1 )5 B
2
for those with (axial-)vector ones. The last supermultiplets is simple:
= i ( + 5 ),
= ( ),
(4)
= ( 5 ).
To construct massive supermultiplet we have to add mass terms for all fields as well as appropriate corrections to fermionic supertransformations. In this, the most important question is
which lower spin fields play the roles of Goldstone ones and have to be eaten to make main
gauge fields massive. For the bosonic fields (taking into account parity conservation) the choice
is unambiguous: vector A and scalar fields for tensor field h and axial-vector B and
pseudo-scalar for pseudo-tensor f . Thus bosonic mass terms will be:

1
L1 = 2 h A h(A) 3A + 2 f B f (B) 3B ,
m

1
3
3
1
1
2
2
2
h + f h f
f + 2.
(5)
L2 = h h h
2
2
2
2
2
m
101
But for fermions we have two spin 3/2 and two spin 1/2 fields and there is no evident choice.
Thus we introduce the most general mass terms for the fermions:
1
1
1
) ( ) +
Lm = + (
m
2
4

1
1
i1 ( ) i2 ( )
2
2

)( )
+ a1 ( )( ) + a2 (

)
+ a3 ( )( ) + ia4 ( ) + ia5 ( ) + ia6 (
) + a8
+ a9
+ a10
+ ia7 (
(6)
and proceed with calculations. Cancellation of variations with one derivative gives:
1
1
1 = ,
sin(2 ) = cos(2 ) = ,
2
2
1
sin(1 ) = cos(1 ) = ,
2 = 2,
2
1
a3 = 1,
a1 = ,
a4 = 2,
a5 = 0,
4
a6 = a2 2,
a7 = 3,
a8 = 0,
a9 = 6,
while variations without derivatives give:
1
1
a10 = .
a2 = ,
2
2
Resulting fermionic mass terms:
1
1
1
) ( ) +
Lm = + (
m
2
4
i
i
i
) i 2 + (
)
+ (
2
2 2
2
1
1
1
1
)( ) + ( )( )
+ ( )( ) + (
4
4
2
2
i
1
) + i 3(
) 6

+ i 2( ) + (
(7)
2
2
correspond to invariance of the Lagrangian (besides global supertransformations) under three
local spinor gauge transformations:
im
m
m
( ) + g 1 + g 2 ,
2
4 2
2 2
m
im
im
1 +
2 ,
= 1 +
4
2
2
im
im
1 +
2 ,
= 2 + m 2 +
2
2
m
= m 32 .
= m 21 + 2 ,
2
= ( ) +
(8)
102
From these formulas one can easily determine which combination of spin 3/2 fields and
plays the role of Goldstone field for spin 5/2 field . Indeed, if one introduces two orthogonal
combinations:
1
2
2
1
= +
= + ,
5
5
5
5
then after diagonalization of mass terms one finds that is a Goldstone field, while
physical field with the same mass as .
Similar to the case of massive supermultiplet with superspin 1, both mixing angles have been
fixed: 1 = 2 = /4 and this, in turn, means that all bosonic fields enter through the complex
combinations only:
H = h + 5 f ,
C = A + 5 B ,
z = + 5 .
Introducing gauge covariant derivatives:
m
z = z m 3C
H = H C g ,
2
we can write final form of fermionic supertransformations as:

3
i
m
m
= H ( ) mH + ( ( H )) +
g z ,
2
4
4 2

1
im
1
im
= H ( H ) + z + z ,
3
4 3
2
2 2

i
im
1
im
= C + ( H ) + z z ,
4
3
2 3
2
1
= C ,
(9)
= i z.
2 2
Note also that due to complexification of bosonic fields the Lagrangian and supertransformations are invariant under global axial U (1)A symmetry, axial charges for all fields being:
Field
, , , ,
H , C , z
qA
+1
2. Superspin 5/2
Our next examplemassive supermultiplet with superspin 5/2. It also contains four massive
fields: with spins 3, 5/2, 5/2 and 2 and in the massless limit it should reduce to six massless
supermultiplets:

3
3
5/2
2
3/2
1
1/2
.
5/2
5/2
5/2
2
3/2
1
1/2
0, 0
2
By analogy with all previous cases we will take into account possible mixing for bosonic tensor
and vector fields, so we will start with the following structure of massless supermultiplets:

h
.
f A
B
z
103
So we introduce sum of the massless Lagrangians for bosonic fields:

1
3
3
L0 = + () () 3() +
2
2
2
3
1
1
1
+ ()2 + h h (h) (h) + (h) h h h A2
4
2
2
4
1
1
1
+ ( )2 + f f (f ) (f ) + (f ) f f f
2
2
2
1 2
1
B
(10)
+ ( )2
4
2
as well as sum of the massless Lagrangians for fermionic fields:
i
2i( ) ( ) + i( ) (
) + i( ) i
L0 =
2
4
i
i
2i(
) + i(
) () + i(
) (
)
+
2
4
i
i
i
) +
+ i( )( ) ( )(
2
2
2
i
i
) + i
+ i(
)(
)() (

(11)
2
2
2
and start with the following global supertransformations:

1
= i(( ) ),
(12)
= + ( ) ( )
4
for the supermultiplet (3, 5/2),

i
= cos(2 )h( ) sin(2 )5 f( ) ,
2
h = 2 cos(2 )( ) + i sin(2 )( ( ) ),
f = 2 sin(2 )( 5 ) + i cos(2 )( ( ) 5 ),

= sin(2 )h + cos(2 )f 5
(13)
for the mixed (5/2, 2) and (2, 3/2) supermultiplets,

i
= cos(1 )A sin(1 )5 B ,
2 2
),
A = 2 cos(1 )( ) + i sin(1 )(
5 ),
B = 2 sin(1 )( 5 ) + i cos(1 )(

1
= sin(1 )A + cos(1 )5 B
2
for the mixed (3/2, 1) and (1, 1/2) supermultiplets and
= i ( + 5 ),
= ( ),
(14)
= ( 5 )
for the last one.

By analogy with the superspin 3/2 case, we will assume that fermionic mass terms are Dirac
ones:
104
1
1
Lm = + 2( ) ( ) +
m
2

5
1
1
+ ( ) + ( )
+i
2
2
2
3
3
) 3
+ ( )( ) + 2i( ) + 2i(
(15)
2
2
where all coefficients are completely fixed by the requirement that the Lagrangian has to be
invariant not only under the global supertransformations, but under four (by the number of fermionic gauge fields) spinor gauge transformations:
im
m 5
= ( ) +
( ) +
g 1 ,
2
4 2
im
m 5
( ) +
g 2 ,
= ( ) +
2
4 2
5
5
3
3
+ im 2 ,
+ im 1 ,
= 1 + m
= 2 + m
2
4
2
4
= 2m2 .
= 2m1 ,
As for the bosonic fields, here the roles of the fields are evident (again taking into account
parity conservation): we need tensor h , vector A and scalar fields to make spin 3 field
massive, while pseudo-tensor f field needs to eat axial-vector B and pseudo-scalar
fields. So the bosonic mass terms are also completely fixed:

1
1
L1 = 3 h + 2 (h) h + 5 h A h(A)
m
2

6A + 2 f B f (B) 3B ,
(16)
1
5
15
15
1
3
3
3
A A2
h + 2
L2 = + h2 +
2
2
4
2
4
2
2
m2

3
1
f + 2.
f f f 2
(17)
2
2
Now we require that the whole Lagrangian be invariant under global supertransformations
with appropriate corrections to fermionic transformations. This fixes both mixing angles:
5
2
1
1
,
cos(2 ) = ,
,
cos(1 ) =
sin(2 ) =
sin(1 ) =
6
3
6
3
and gives the following form of additional terms for fermionic supertransformations:

1
3
5
1 5
=
( ( h))
f 5 +
( (f )) 5 ,
m
4
3
4 3

1
5
1
1 5
1 5
1
= i ( ) + g ( )
( A) +
g A
( B) 5
m
8
4 3
4 3
2 6

1 5
g B5 ,
+
2 6
105

1
1 5
1
5
1
5
=
( ) A A B 5 B5 ,
m
8 2
2 6
2 6
2 3
2 3

3 5
1 3
1
= i
( h) +
(f ) 5 5 ,
m
2 6
2 2
1
1
5 ].
(18)
= [ 25 ],
= i[ 6A + 3B
m
m
The complete supertransformations for fermionic fields could be simplified by introduction of
gauge invariant derivatives;
m 5
= m 6A ,
A g ,
h = h
2
m
= m 3B .
f = f B g ,
2
This time bosonic fields do not combine into complex combinations, but due to the fact that
fermionic mass terms are Dirac ones the Lagrangian and supertransformations are invariant under
global axial U (1)A transformations, provided axial charges of all fields are assigned as follows:
Field
, , ,
h , f , A , B , ,
, ,
qA
+1
3. Massive particles
All our previous and subsequent calculations heavily depend on the gauge invariant description of massive high spin particles. For reader convenience and to make paper self-contained we
will give here gauge invariant formulations for massive particles with arbitrary integer [2,3] and
half-integer [10] spins. We restrict ourselves to flat d = 4 Minkowski space but all results could
be easily generalized to the case of (A)dS space with arbitrary dimension d.
3.1. Integer spin
The simplest way to describe massless bosonic field with arbitrary spin s is to use completely symmetric rank s tensor (1 2 ...s ) which is double traceless. In what follows we will
use condensed notations where index denotes just number of free indices and not the indices
themselves. For example, the tensor field itself will be denoted as s , its contraction with derivative as ()s1 , its trace as s2 and so on. As we will see this does not lead to any ambiguities
then working with free Lagrangians quadratic in fields. In these notations the Lagrangian for
massless particles of arbitrary spin s could be written as:

s
s(s 1) s2
1
L0 = (1)s s s ()s1 ()s1
s2

2
2
4

s(s 1)
s(s 1)(s 2)
s3
s3 ( )
+
(19)
()s1 (1 s2)
( )
2
8
where = 0. This Lagrangian is invariant under the following gauge transformations:
0 s = (1 s1) ,
s3 = 0
106
where parameter s1 is completely symmetric traceless tensor of rank s 1.

To construct gauge invariant Lagrangian for massive particle which has correct (i.e. with right
number of physical degrees of freedom) massless limit, we start with the sum of massless Lagrangians with 0 k s:
L0 =
1 k
k
k(k 1) k2
k ()k1 ()k1

k2
2
2
4
k=0

k(k 1)
k(k 1)(k 2)
k1
k3
k3
.
() (1 k2)
( ) ( )
+
2
8
(1)k
(20)
Then we add the following cross terms with one derivative as well as mass terms without derivatives:

s

1
k
(1) ak ()k1 k1 + (k 1) k2 ()k2
L1 =
m
k=1

(k 1)(k 2)
k3 k3 ,
( )
+
4

1
L2 =
(1)k dk k k + ek k2 k2 + fk k2 k2
2
m
s
(21)
k=0
and try to achieve gauge invariance with the help of appropriate corrections to gauge transformations:
1
k = k k + k g(2 k2) .
m
Straightforward but lengthy calculations give a number of algebraic equations on the unknown
coefficients which could be solved (and this is non-trivial because we obtain overdetermined
system of equations) and give us:
(s k)(s + k + 1)
k+1
k , 0 k s 1,
,
k+1 =
2k
2(k + 1)2
(s k + 1)(s + k)
(s k 1)(s + k + 2)
,
dk =
,
ak =
2
4(k + 1)

k(k 1)
ek =
(s k + 2)(s + k 1) + 6 ,
16(k + 1)
1
fk =
(s k + 2)(s + k 1)(s k + 1)(s + k).
4
k2 =
3.2. Half-integer spin

For the description of massless spin s + 1/2 particles we will use completely symmetric
rank s tensor-spinor s such that ( )s3 = 0 (in the same condensed notations as before).
Then Lagrangian for such field could be written as:
107
s
1 s s
)s1
s( )s1 ( )s1 + ( )s1 (
2
2

s(s 1)
s(s 1) s2 s2
+
( )s2 s2
2
8
L0 = i(1)
(22)
and is invariant under the following gauge transformations:

0 s = (1 s1) ,
( ) = 0,
where gauge parameter s1 is a -traceless tensor-spinor of rank s 1.

Once again we start with the sum of massless Lagrangians with 0 k s:
L0 =
k
1 k k
)k1
k( )k1 ( )k1 + ( )k1 (
2
2
k=0

k(k 1)
k(k 1) k2 k2
+
( )k2 k2
.

2
8
i(1)
(23)
To combine all these massless fields into one massive particle we have to add the following
mass terms:

s

k(k 1) k2 k2
s +1
k
k k
k1
k1
Lm =
(1)
k( ) ( )
2(k + 1)
4
k=0

k 1 k2
ick ( )k1 k1
(24)
( )k2
2
and corresponding corrections to gauge transformations:
k = k k + ik (1 k1) + k g(2 k2) .
Then total Lagrangian will be gauge invariant provided:

(s + 1)2 k 2
ck+1
s+1
ck =
,
k =
,
k =
,
2
k+1
2k(k + 1)
k =
ck
.
2k
4. Massless supermultiplets
It is not easy to find in the recent literature the explicit component form of massless supermultiplets with arbitrary superspin [22], so for completeness we will give their short description
here. Off-shell superfield formulation could be found in [2325]. As we have already seen on
the lower superspin cases, supermultiplets with integer and half-integer superspins have different
structure and have to be considered separately.
(s, s +1/2). Supermultiplet with integer superspin s contains bosonic spin s field and fermionic spin s + 1/2 one. In this and in two subsequent sections we will use the same condensed
notations as in the previous one. By analogy with superspins 1 and 2 supermultiplets, we start
with the following ansatz for the supertransformations:
s = i1 (s1 1) ,
s = ( s ).
108
Indeed, calculating variations of the sum of two massless Lagrangians one can see that most of
variations cancel, provided one set 1 = 2 . The residue:
(s 1)(s 2)
2( )s2 s2 s2 2 s2
4
s3 2( )s2 s2 + 2( )s3 ( )
s3
(s 2)( )s3 ( )

)
s3
( )s3 (
L = (1)s
contains terms with s2 only, so we proceed by adding to fermionic supertransformations one

more term:
s = i2 (1 1 s2) .
Then the choice 2 =
(s1)(s2)
4s
leaves us with:

(s 1)(s 2)
)
s3 ( )s3 (
s3
2( )s3 ( )
4
s3 . So we make one more (last) correction to superwhere the only terms are whose with ( )
transformations:
L = (1)s
s3)
s = i3 g(2 1 ( )
and obtain full invariance with 3 = (s1)(s2)
. To fix concrete normalization we will use
4s
closure of the superalgebra. Calculating the commutator of two supertransformations we obtain:
[1 , 2 ]s = i 2 ( 2 1 ) s +
where dots mean up to gauge transformation. So we set =
2 and our final result looks like:

i
i(s 1)(s 2)
s3) ,
(1 1 s2) g(2 1 ( )
s = (s1 1) +
2
2 2s
s = 2(s ).
(25)
(s + 1/2, s + 1). Half-integer superspin multiplet contains fermionic spin s + 1/2 fields and
bosonic spin s + 1 one. Again by analogy with lower superspin case we will make the following
ansatz for supertransformations:
s = 1 (s) ,
s+1 = i( (s 1) ).
This time most of the variations cancel if one set 1 = leaving us with:
s(s 1)
s2 s2 2 ( )
s2
2( )s2 ( )
4

)
s3 .
s2 (s 2)( )s3 ( )
2( )s2 (
L = i(1)s
Then the full invariance could be achieved with the following correction to supertransformations:
s2)
s = 2 (1 1 ( )
provided 2 = s1
4 . To check the closure of superalgebra and to choose normalization we calculate commutator of two supertransformations:
[1 , 2 ]s+1 = 2i 2 ( 2 1 ) s+1 + .
109
Then our choice will be = 1 and our final form:

s 1
s2) ,
s+1 = i( (s 1) ).
(1 1 ( )
4
Note that starting with superspin 2 the structure of supertransformations are defined up to possible field dependent gauge transformations and our choice differs from that of [22]. It makes no
difference for massless theories but for massive case the structure of corrections for fermionic
supertransformations depends on the choice made.
s = (s) +
5. Integer superspin
Now, having in our disposal gauge invariant description of massive particles with arbitrary
(half-)integer spins, known form of supertransformations for massless arbitrary superspin supermultiplets and concrete examples of massive supermultiplets with lower superspins, we are
ready to construct massive arbitrary superspin supermultiplets. As we have seen, integer and
half-integer cases have different structures and have to be considered separately.
In this section we consider massive supermultiplet with integer superspin. Such supermultiplet
also contains four massive fields: two bosonic spin s fields (with opposite parity) and fermionic
spin s +1/2 and s 1/2 ones. Calculating total number of physical degrees of freedom and taking
into account possible mixing of supermultiplets containing bosonic fields with equal spins and
opposite parity, we start with the following structure of massless supermultiplets:

s
k
s

0
.
As
Bs
Ak
Bk
z
s1
k1
k=1
By analogy with superspin 1 and 2 cases, we will assume that all bosonic fields enter through
the complex combinations Ck = Ak + iBk only (so that all possible mixing angles are fixed and
equal /4). Thus we choose the following form of supertransformations for massless supermultiplets with 1 k s:

i
i(k 1)(k 2)
k3) ,
(1 1 C k2) g(2 1 ( C)
k = C (k1 1) +
2
4k
C k = 2( k ) + i 2( (k1 1) ),
1
k2
k3)
k1 = C(k1) + (1 1 ( C)
2
4 2
and also
0 = i z,
(26)
z = 2( 0 ).
As a result of our assumption mass terms for bosonic fields are completely fixed:

s

(1)k ck C k (1 Ck1) (k 1)C k2 (C)k2
L1 =
k=0

(k 1)(k 2) k2
C (1 Ck3) + h.c. ,
4
s

(1)k dk C k Ck + ek C k2 C k2 + fk C k2 Ck2 + h.c.
L2 =
+
k=0
(27)
110
where
1 (s + k)(s k + 1)
(s k 1)(s + k + 2)
,
dk =
,
ck =
2
2
4(k + 1)

k(k 1)
(s k + 2)(s + k 1) + 6 ,
ek =
16(k + 1)
1
fk =
(s k + 2)(s + k 1)(s k + 1)(s + k).
8
As for the fermionic mass terms, a priori we do not have any restrictions on them so we have to
consider the most general possible form:

s

1
k(k 1) k2
)k1 ( )k1
k2
Lm =
(1)k a1k k k k(
m
4
k=0

k(k 1) k2
)k1 ( )k1
+ a2k k k k(
k2
4

k(k 1) k2
+ a3k k k k( )k1 ( )k1
k2
4

k 1 k2
)k1 k1
+ ib1k (
( )k2
2

k 1 k2
k1
( )k2
+ ib2k ( ) k1
2

k 1 k2
k1
+ ib3k ( ) k1
( )k2
2

k 1 k2
k1
+ ib4k ( ) k1
(28)
( )k2
2
where:
1
a2s = a3s = b3s = b4s = 0.
a1s = ,
2
The requirement that total Lagrangian be invariant under (appropriately corrected) supertransformations gives:
1
2 2
k
a1k = ,
a2k =
ck+1 ,
,
a3k =
2
k+1
2(k + 1)
1
b1k = 2ck ,
b3k = 0,
b2k = ,
b4k = 2ck+1 .
2
In this, additional terms for fermionic supertransformations look like:

2ick+1
k(k 1)
(k 1)2 (2k + 1)
1
k2)
k =
( C)k +
(1 C k1)
g(2 ( C)
m
k+1
4(k + 1)
8k(k + 1)
k1
(k 1)(k 2)
(1 ( C)k1
g(2 C k2)
Ck +
2k
8k 2

ick
(1 Ck1) g(2 ( C)k2) ,
(29)
111
1
kck+2
k1)
(1 ( C)
k =
m
2 2(k + 1)

ik
k(k 1)
(k 1)(3k + 1)
k2)
(1 C k1) +
g(2 ( C)
( C)k
4(k + 1)
8k(k + 1)
2(k + 1)

2ck+1
1
(k 1)(k 2)
(30)
Ck + (1 ( C)k1) +
g(2 Ck2) .
k+1
k
4k 2
Here the supertransformations for k field contain terms with Ck+1 , Ck and Ck1 fields in the
first, second and third lines correspondingly, while that of k contain terms with Ck+2 , Ck+1 and
Ck fields.
6. Half-integer superspin
Next we turn to the half-integer superspin case. This time we have two fermionic spin s + 1/2
fields and bosonic ones with spins s + 1 and s. Usual reasoning on physical degrees of freedom
and possible mixings leads us to the following structure of massless supermultiplets we will start
with:

s
As+1
k
0
As+1
.
s
s
Ak
Bk
s
z
Bs
k1
k=1
We see that this structure is rather similar to that of integer superspin case. The main difference
(besides the presence of As+1 , s supermultiplet) comes from the mixing of bosonic fields. We
have no reasons to suggest that all mixing angles could be fixed from the very beginning so we
have to consider the most general possibility here. Let us denote:
Ck = cos(k )Ak + 5 sin(k )Bk ,
Dk = sin(k )Ak + 5 cos(k )Bk .
In these notations supertransformations for massless supermultiplets could be written as follows.

Highest supermultiplet:
As+1 = i( (s 1) ),
s = A(s) +
s 1
s2) .
(1 1 ( A)
4
Main set (1 k s):

i
i(k 1)(k 2)
k3) ,
k = C (k1 1) +
(1 1 C k2) g(2 1 ( C)
2
2 2k
Ak = 2 cos(k )(k ) + i sin(k )((k1 1) ),
Bk = 2 sin(k )( k 5 ) + i cos(k )( (k1 1) 5 ),

k2
k3
(1 1 ( D)
k1 = D(k1) +
4
and the last supermultiplet:
0 = i z,
(31)
z = 2( 0 ).
By analogy with superspin 3/2 and 5/2 cases we will assume that fermionic mass terms are
Dirac ones. This immediately gives:
112

s

s +1
k(k 1) k2
k
k
k1
k k( ) ( )k1
k2
Lf =
(1)
k+1
4
k=0

k 1 k2
k1
( )k2 + ( )
ick ( ) k1
2
where
(32)
(s + k + 1)(s k + 1)
.
2
The choice for the bosonic mass terms (taking into account parity) is also unambiguous:
ck =
L1 =

(k 1)(k 2) k2
(1)k ak Ak (1 Ak1) (k 1)A k2 (A)k2 +
A (1 A k3)
4
k=0

s

(1)k bk B k (1 Bk1) (k 1)B k2 (B)k2
+
s+1

k=0
(k 1)(k 2) k2
+
B (1 B k3)
4

(33)
for the terms with one derivative, where:
(s + k + 1)(s k + 2)
(s + k)(s k + 1)
ak =
,
bk =
2
2
and the following terms without derivatives:

1
L2 =
(1)k dk Ak Ak + ek A k2 A k2 + fk A k2 Ak2
2
m
s+1
k=0

(1)k dk B k Bk + ek B k2 B k2 + fk B k2 Bk2 .
(34)
k=0
Here:

(s k)(s + k + 3)
k(k 1)
,
ek =
(s k + 3)(s + k) + 6 ,
dk =
4(k + 1)
16(k + 1)

1
fk =
(s k + 3)(s + k)(s k + 2)(s + k + 1),
4

(s k 1)(s + k + 2)
k(k 1)
dk =
,
ek =
(s k + 2)(s + k 1) + 6 ,
4(k + 1)
16(k + 1)
1
fk =
(s k + 2)(s + k 1)(s k + 1)(s + k).
4
Note that hatted coefficients differ from the unhatted ones by replacement s s + 1.
Now we require that total Lagrangian be invariant under the supertransformations. First of all
this fixes all mixing angles:

s +k+1
s k+1
,
cos(k ) =
sin(k ) =
2(s + 1)
2(s + 1)
113
and gives us additional terms for fermionic supertransformations:

1
k = 1 Ak + 2 (1 ( A)k1) + 3 g(2 A k2)
m
+ 1 Bk + 2 (1 ( B)k1) + 3 g(2 B k2)

kck+1
k1) + cos(k+1 )(1 ( B)
k1) ,
sin(k+2 )(1 ( A)
+
4(k + 1)
1
k2)
k = 4 ( A)k + 5 (1 A k1) + 6 g(2 ( A)
m
k2)
+ 4 ( B)k + 5 (1 B k1) + 6 g(2 ( B)

ak
cos(k ) (1 Ak1) g(2 ( A)k2
k 2

bk
sin(k ) (1 Bk1) g(2 ( B)k2)
k 2
where
1 =
s k
cos(k ),
k2 + s + k + 1
cos(k ),
2 =
2k(k + 1)
2(k + 1)
(s + 1)(k 1)(k 2)
cos(k ),
3 =
4 2k 2 (k + 1)
s +k+2
k2 s + k 1
sin(k ),
sin(k ),
1 =
2 =
2(k + 1)
2k(k + 1)
(s + 1)(k 1)(k 2)
sin(k ),
3 =
4 2k 2 (k + 1)
s +1
k(k 1)(s k)
sin(k+1 ),
4 =
5 =
sin(k+1 ),
k+1
4(k + 1)2
(k 1)[(k + 1)(s + 1) 2k 2 (s k)]
6 =
sin(k+1 ),
8k(k + 1)2
s+1
k(k 1)(s + k + 2)
cos(k+1 ),
4 =
5 =
cos(k+1 ),
k+1
4(k + 1)2
(k 1)[(k + 1)(s + 1) 2k 2 (s + k + 2)]
6 =
cos(k+1 ).
8k(k + 1)2
We have explicitly checked that (rather complicated) formulas from this and previous sections
correctly reproduce all lower superspins results.
7. Conclusion
Thus, using supersymmetric generalization of gauge invariant description for massive particles, we managed to show that all massive N = 1 supermultiplets could be constructed out of
appropriate set of massless ones. In this, in spite of large number of fields involved, all calculations are pretty straightforward and mainly combinatorial. Certainly, using gauge invariance
one can fix the gauge where all but four physical massive fields are equal to zero. But in this
case all supertransformations must be supplemented with field dependent gauge transformations
114
restoring the gauge. So the structure of resulting supertransformation becomes very complicated
and will contain higher derivative terms.
References
[1] Yu.M. Zinoviev, Gauge invariant description of massive high spin particles, Preprint 83-91, IHEP, Protvino, 1983.
[2] S.M. Klishevich, Yu.M. Zinoviev, On electromagnetic interaction of massive spin-2 particle, Phys. At. Nucl. 61
(1998) 1527, hep-th/9708150.
[3] Yu.M. Zinoviev, On massive high spin particles in (A)dS, hep-th/0108192.
[4] N. Arkani-Hamed, H. Georgi, M.D. Schwartz, Effective field theory for massive gravitons and gravity in theory
space, Ann. Phys. 305 (2003) 96, hep-th/0210184.
[5] S. Hamamoto, Possible nonlinear completion of massive gravity, Prog. Theor. Phys. 114 (2006) 1261, hep-th/
0505194.
[6] M. Bianchi, P.J. Heslop, F. Riccioni, More on La Grande Bouffe, JHEP 0508 (2005) 088, hep-th/0504156.
[7] K. Hallowell, A. Waldron, Constant curvature algebras and higher spin action generating functions, Nucl. Phys.
B 724 (2005) 453, hep-th/0505255.
[8] I.L. Buchbinder, V.A. Krykhtin, Gauge invariant Lagrangian construction for massive bosonic higher spin fields in
D dimensions, Nucl. Phys. B 727 (2005) 537, hep-th/0505092.
[9] I.L. Buchbinder, V.A. Krykhtin, P.M. Lavrov, Gauge invariant Lagrangian formulation of higher spin massive
bosonic field theory in AdS space, Nucl. Phys. B 762 (2007) 344, hep-th/0608005.
[10] R.R. Metsaev, Gauge invariant formulation of massive totally symmetric fermionic fields in (A)dS space, Phys.
Lett. B 643 (2006) 205212, hep-th/0609029.
[11] Yu.M. Zinoviev, On massive spin 2 interactions, Nucl. Phys. B 770 (2007) 83, hep-th/0609170.
[12] R.R. Metsaev, Gravitational and higher-derivative interactions of massive spin 5/2 field in (A)dS space, hep-th/
0612279.
[13] I.L. Buchbinder, V.A. Krykhtin, A.A. Reshetnyak, BRST approach to Lagrangian construction for fermionic higher
spin fields in (A)dS space, hep-th/0703049.
[14] I.L. Buchbinder, S.J. Gates Jr., J. Phillips, W.D. Linch, New 4D, N = 1 superfield theory: Model of free massive
superspin-3/2 multiplet, Phys. Lett. B 535 (2002) 280288, hep-th/0201096.
[15] Yu.M. Zinoviev, Massive spin-2 supermultiplets, hep-th/0206209.
[16] I.L. Buchbinder, S.J. Gates Jr., W.D. Linch III, J. Phillips, Dynamical superfield theory of free massive superspin-1
multiplet, Phys. Lett. B 549 (2002) 229236, hep-th/0207243.
[17] T. Gregoire, M.D. Schwartz, Y. Shadmi, Massive supergravity and deconstruction, JHEP 0407 (2004) 029, hep-th/
0403224.
[18] S.J. Gates, S.M. Kuzenko, 4D N = 1 higher spin gauge superfields and quantized twistors, JHEP 0510 (2005) 008,
hep-th/0506255.
[19] I.L. Buchbinder, S.J. Gates Jr., S.M. Kuzenko, J. Phillips, Massive 4D, N = 1 superspin 1 & 3/2 multiplets and
dualities, JHEP 0502 (2005) 056, hep-th/0501199.
[20] S.J. Gates Jr., S.M. Kuzenko, G. Tartaglino-Mazzucchelli, New massive supergravity multiplets, hep-th/0610333.
[21] Yu.M. Zinoviev, Massive supermultiplets with spin 3/2, hep-th/0703118.
[22] T. Curtright, Massless field supermultiplets with arbitrary spin, Phys. Lett. B 85 (1979) 219.
[23] S.M. Kuzenko, A.G. Sibiryakov, V.V. Postnikov, Massless gauge superfields of higher half integer superspins, JETP
Lett. 57 (1993) 534.
[24] S.M. Kuzenko, A.G. Sibiryakov, Massless gauge superfields of higher integer superspins, JETP Lett. 57 (1993) 539.
[25] S.M. Kuzenko, A.G. Sibiryakov, Free massless higher superspin superfields on the anti-de Sitter superspace, Phys.
At. Nucl. 57 (1994) 1257.
Axionic symmetry gaugings in N = 4 supergravities

and their higher-dimensional origin
Jean-Pierre Derendinger a , P. Marios Petropoulos b,a , Nikolaos Prezas c,
a Institut de Physique, Universit de Neuchtel, Breguet 1, 2000 Neuchtel, Switzerland
b Centre de Physique Thorique, Ecole Polytechnique, CNRS 1 , 91128 Palaiseau, France
c CERN PH-TH, 1211 Genve, Switzerland

Abstract
We study the class of four-dimensional N = 4 supergravities obtained by gauging the axionic shift and
axionic rescaling symmetries. We formulate these theories using the machinery of embedding tensors,
characterize the full gauge algebras and discuss several specific features of this family of gauged supergravities. We exhibit in particular a generalized duality between massive vectors and massive two-forms
in four dimensions, inherited from the gauging of the shift symmetry. We show that these theories can
be deduced from higher dimensions by a ScherkSchwarz reduction, where a twist with respect to a noncompact symmetry is required. The four-dimensional generalized duality plays a crucial role in identifying
the higher-dimensional ascendent.
1. Introduction
The effective field theories describing the low-energy dynamics of typical string and M-theory
compactifications are plagued with massless scalar fields that hamper any attempt to contact fourdimensional phenomenology. One way of eliminating some of these unwanted massless scalars
is the introduction of fluxes in the internal compactification space (see [1] for a review). From
the effective field theory perspective, turning on fluxes corresponds to a gauging of the original
E-mail addresses: jean-pierre.derendinger@unine.ch (J.-P. Derendinger), marios@cpht.polytechnique.fr

(P.M. Petropoulos), nikolaos.prezas@cern.ch (N. Prezas).
1 Unit mixte UMR 7644.
doi:10.1016/j.nuclphysb.2007.06.021
116
J.-P. Derendinger et al. / Nuclear Physics B 785 (2007) 115134
theory obtained without fluxes. The term gauging refers to the fact that a subgroup of the global
duality symmetry of the original theory is promoted to a local gauge symmetry. Simultaneously,
part of the original Abelian gauge symmetry is promoted to a non-Abelian one and various fields,
including the scalars, acquire minimal couplings to the gauge fields. The connection with the
moduli stabilization issue stems from the fact that for extended supergravity theories, the only
way to generate a potential is through the gauging. The resulting theories are known as gauged
supergravities (see [2] for a concise review).
One can envisage a bottomup approach to the problem of moduli stabilization where instead
of looking for a specific higher-dimensional background whose corresponding low-energy effective theory is phenomenologically viable, one first constructs a gauged supergravity theory with
the required phenomenological properties and then attempts to engineer it from higher dimensions. Such a programme was initiated in [3] and models were proposed there exhibiting full
moduli stabilization.
Evidently such a programme depends crucially on having a good picture of the landscape
of possible gauged supergravity theories. Therefore, a problem of paramount importance is to
classify and describe all possible gaugings of supergravity in diverse spacetime dimensions and
with various amounts of supersymmetry. This problem was tackled in a series of publications
[48] where the general method of embedding tensors was developed.
Ultimately, however, one would like to know that the effective theory constructed in the
bottomup approach is consistent, i.e. can be embedded in a certain string or M-theory setup. Although for many classes of gauged supergravities their higher-dimensional realization is known,
for the generic gauged supergravity there is no recipenot even guaranteed existence of a higherdimensional origin. Presumably such an endeavor would require a better understanding of the
classes of string backgrounds dubbed non-geometric, whose significance has recently been investigated in the framework of flux compactifications and supergravity theories.
Here, our aim is to contribute in this direction by analyzing a family of four-dimensional
N = 4 gauged supergravities and explain how they can be obtained from string theory or higherdimensional supergravity theories. Gauged supergravity with N = 4 supersymmetry was the
arena of [3] since string theories with 16 supercharges are the starting point of a variety of realistic
string constructions. Recently the most general theory of this type was explicitly constructed in
[9] using the formalism of embedding tensors. We will use this formalism to study in detail the
specific class of N = 4 theories obtained by gauging the axionic symmetries, namely the axionic
shifts and rescalings. Consistency requires that several SO(6, 6) directions are also gauged. The
general structure of the gauge algebra is systematically worked out and it exhibits the following
characteristic property: it is non-flat contrary to what happens in more conventional gaugings.
In order to uncover the higher-dimensional origin of these gaugings one needs to perform
a ScherkSchwarz reduction of heterotic supergravity (or of the common sector in general)
on a torus. The crucial ingredient here is a twist by a non-compact duality symmetry of tendimensional supergravity.
The identification of the theory obtained from the reduction with the gauged supergravity
under consideration is not at all straightforward. It relies heavily on a duality between massive
vectors and massive two-forms in four dimensions. This duality is a necessary extension of the
more standard duality between a massless two-form and an axion scalar field in four dimensions:
it incorporates the Stckelberg-like terms that are generated by the axionic gauging.
The organization of this paper is as follows. Section 2 is devoted to a reminder of gauged
[1012] N = 4 supergravity in D = 4 [1315]. We emphasize the approach of the embedding
tensor following Refs. [8,9,16]. In Section 3 we specialize to the so-called electric gaugings and
117
we analyze a particular class of algebras for which we elaborate on the corresponding gauged
supergravity. These theories admit two equivalent formulations related by a duality between massive vectors and massive two-forms in four dimensions. In Section 4 we move on to higher
dimensions. Our aim is to analyze the ten-dimensional origin of the four-dimensional theory
constructed in Section 3. We show that it can be obtained using a generalized ScherkSchwarz
reduction of heterotic N = 1 D = 10 supergravity. Furthermore, we comment on the higherdimensional origin of other classes of N = 4 gaugings. In Section 5 we present our conclusions
and discuss some open problems.
2. Reminder on N = 4 gauged supergravities in D = 4
In this section we review N = 4 gauged supergravity following Ref. [9]. After some general
remarks on the possible gauge algebras and the constraints on the gauging parameters we present
for reference the most general bosonic Lagrangian.
2.1. Gauge algebras and the embedding tensor
Four-dimensional N = 4 supergravity has a very restricted structure. It contains generically
the gravity multiplet and n vector multiplets. The bosonic sector of the theory consists of the
graviton, n + 6 vectors and 2 + 6n scalars. In the ungauged version, the gauge group is Abelian,
U (1)6+n , and there is no potential for the scalars.
Interactions are induced upon elimination of the auxiliary fields. They affect the scalars whose
non-linearities are captured by a universal coset manifold [17,18]
M=
SO(6, n)
SL(2, R)
.
U (1)
SO(6) SO(n)
(2.1)
All fields are non-minimally coupled to the Abelian vectors. The gauge kinetic terms have scalarfield-dependent coefficients, whereas the action is at most quadratic in the gauge-field strengths
with no explicit dependence on the gauge potentials. For this reason, the SL(2, R) SO(6, n)
Sp(12 + 2n, R) symmetry of the scalar manifold is globally realized as a U-duality symmetry.
Although the scalar manifold survives any deformation of the plain theory triggered by gaugings,
the U-duality is broken as a consequence of the introduction of non-Abelian field strengths and
minimal couplings, all of which depend explicitly on the gauge potentials.
The duality group acts as a symmetry of the field equations and the Bianchi identities of the
gauge fields. In the standard formulation of supergravity only a subgroup of it is realized off-shell
as a genuine symmetry of the Lagrangian. This includes the SO(6, n) plus a two-dimensional
non-semi-simple subalgebra of SL(2, R) generated by axionic shifts and axionic rescalings. The
third transformation in SL(2, R), corresponding to the truly electricmagnetic duality, is an onshell symmetry which relates different Lagrangians associated to different choices of symplectic
frames.
The only known deformation of N = 4 supergravity compatible with supersymmetry is the
gauging. This consists in transmuting part of the U (1)6+n local symmetry into an non-Abelian
gauge symmetry, or equivalently in promoting part of the global U-duality symmetry to a local
symmetry, using some of the available vectors. This operation should not alter the total number
of propagating degrees of freedom, as required e.g. by supersymmetry.
It is possible to parameterize all gaugings of N = 4 supergravity by using the so-called embedding tensor. The latter describes how the gauge algebra is realized in terms of the U-duality
118
generators. It captures all possible situations, including those where some of the gauge fields
are the magnetic duals of the vectors originally present in the Lagrangian, as well as the option of gauging the duality rotation between electric and magnetic vectors, which, as already
stressed, appears only as an on-shell symmetry. In this procedure, one doubles the number of
vector degrees of freedom, keeping however unchanged the number of propagating ones thanks
to appropriate auxiliary fields. One therefore avoids any choice of symplectic frame until a specific gauging is performed. We will not elaborate on these general properties but summarize the
structure of the embedding tensor that we will use in the ensuing. We refer the reader to [8,9,16]
for further information on this subject.
The n + 6 electric vector fields AM+ , M = 1, . . . , 6 + n belong to the fundamental vector
representation of SO(6, n). Their magnetic duals AM form also a vector of SO(6, n), but carry
opposite charge with respect to the SO(1, 1) SL(2, R) that generates the axionic rescalings.
The SL(2, R) algebra is generated by S = S , , {+, }, which obey the following
commutation relations:
[S , S ] = S S S S
with + = 1 = + . In
read2

0
S++ = S =
2

0
S = S ++ =
0
(2.2)
the vector representation (S ) = + and they explicitly

0
,
0

2
.
0
S+ = S + =

1 0
,
0 1
(2.3)
The axion a and the dilaton form a complex scalar = a + ie2 , which parameterizes the
SL(2, R)/U (1) coset. We define
2

1
| |
Re
M =
(2.4)
1
Im Re
and we denote by M its inverse. The action of

a b
g=
SL(2, R)
c d
(2.5)
is linear on M: M gMg T while it acts on as a Mbius transformation: a + b/c + d.

Therefore, S ++ generates the axionic shifts, S + the axionic rescalings, whereas the electric
magnetic duality is generated by S . In this basis, ({AM+ }, {AM }) form a doublet of SL(2, R)
with diagonal S + and correspondingly the rescaling charges are +1 and 1.
The SO(6, n) is generated by TMN = TN M , M, N {1, . . . , 6 + n} obeying
[TKL , TJ M ] = LJ TKM + KM TLJ KJ TLM LM TKJ ,
with LJ being the SO(6, n)
J.
(TKL )I J = KI LJ LI K
metric.3
(2.6)
In the fundamental representation the generators read:
2 Since with the present conventions for , = , we can raise and lower -indices unambiguously as

follows: A = A and B = B . This leads to A+ = A and A = A+ . In particular, S++ = S , S+ =
S + and S = S ++ .
3 Indices M, N, . . . are lowered and raised with
KM (inverse matrix).
LJ and
119
The 6n scalars coming from the n vector multiplets live on the SO(6, n)/(SO(6) SO(n))
coset and can be parameterized by a symmetric matrix M of elements MMN . Introducing vielm , V a ) with m = 1, . . . , 6 and a = 1, . . . , n we can write M = VVT and define the
beins V = (VM
M
fully anti-symmetric tensor
p
m n
VN VP VQ VRr VSs
MMNP QRS = mnpqrs VM
(2.7)
that appears in the scalar potential. We will denote by M MN the inverse of MMN .
The gauging of N = 4 supergravity proceeds by selecting a subalgebra of SL(2, R) SO(6, n)
generated by some linear combination of {S , TKL }. The coefficients of this combination
are the components of the embedding tensor and are subject to various constraints discussed
in detail in the aforementioned references. In summary, this tensor belongs a priori to the
(2 3, Vec) + (2, Vec Adj) of SL(2, R) SO(6, n). However, there are linear constraints
resulting from the requirement that the commutator provides an adjoint action and that supersymmetry is preserved.4 These reduce the representation content of the embedding tensor to
(2, Vec) + (2, Ant[3] ). Furthermore, there are quadratic constraints guaranteeing the closure of
the gauge algebraJacobi identity. Putting everything together, one finds that the admissible
generators of the gauge algebra are of the form

1
fLMN T MN P Q P TQL + L S
2
where fLMN are fully anti-symmetric in L, M, N and L satisfy
L =
(i)
MN M N = 0
, .
(2.8)
(2.9)
These parameters characterize completely the gauging.

The set of consistency conditions is completed as follows:
(ii)
(iii)
MN (M f)N I J = 0,

I J + MN M fN I J = 0,
(2.10)
(2.11)
and
MN fMI [J fKL]N
1
1
1
= [J fKL]I + I fJ KL MN M fN[J K L]I
2
6
2
1
fJ KL I ,
(2.12)
6
where [ ] and ( ) stand for anti-symmetrization and symmetrization with respect to different indices belonging to the same family (e.g. [LN ] = 1/2(LN NL)).
Several comments are in order here. A general gauging is manifestly expressed in terms of
2 (6 + n) + 2 (6 + n)(5 + n)(4 + n)/6 parameters subject to four conditions (i)(iv). The
gauge algebra, as defined by this set of parameters, is characterized by the commutation relations of the subset of independent L s. These generators are indeed constrained (they satisfy
e.g. (L )M LM = 0 as a consequence of their definition (2.8) and Eqs. (2.9), (2.10), (2.11))
(iv)
4 This is actually the minimal set of constraints that one can consistently impose and they also guarantee that a Lagrangian exists propagating the correct number of degrees of freedom, as one learns from general studies on gaugings of
maximal supergravities [48].
120
as they should since no more that 6 + n vectors can propagate. The structure constants of the
gauge algebra are not directly read off from the fLMN s, which are not necessarily structure
constants of some algebra. They can, however, be expressed in terms of the fLMN s and M s
and describe a variety of situations which capture simple, semi-simple or even non-semi-simple
examples. For all of these, MN always provides an invariant metric although the CartanKilling
metric of the corresponding gauge algebra can be degenerate.
The duality phases of de Roo and Wagemans [1921] are also captured by the present formalism when the M s are absent,5 as relative orientations of f+LMN with respect to fLMN
for each simple component [9]. We will not elaborate any longer on the general aspects of the
embedding tensor and the variety of physical possibilities and for further details we will refer the
reader to the already quoted literature.
2.2. Lagrangian formulation
The bosonic Lagrangian corresponding to the most general N = 4 gauging was presented
in [9]. For the sake of completeness we reproduce this result here and provide several comments.
The Lagrangian consists of a kinetic term, a topological term, and a potential for the scalars:
L = Lkin + Ltop + Lpot .
(2.13)
The kinetic term reads

1
1
1
e1 Lkin = R + D MMN D M MN
D D
2
16
4(Im )2
1
1
M N
M N
H
+ aMN H
H
e2 MMN H
4
8
with the covariant derivatives defined as
D MMN = MMN + 2gAP P (M Q MN )Q ,

M+
M
M+
2
D = + gAM
+M + g A +M A M igA M
(2.14)
(2.15)
(2.16)
and the generalized gauge-field strengths being

g
M N P +
M
M
NP
H
= 2[ AM+
] g fNP A[ A] + N P C
2
g
++
+
+ + M C
+ M C
.
(2.17)
2
The combinations MNP and fMNP are defined in terms of the gauging parameters fMNP
and N as
MNP = fMNP [N P ]M ,
(2.18)
3
fMNP = fMNP [M P ]N N MP .
2
(2.19)
MN = C [MN ] and C
The tensor gauge fields C
= C are auxiliary and their elimination
ensures that the correct number of gauge field degrees of freedom are propagated. For that purpose one needs to introduce a topological term in the Lagrangian
()
5 For vanishing
M s, Eq. (2.12) is the only constraint; it is an ordinary Jacobi identity when = .
121
e1 Ltop

g
P+
P
N+
M N+
=
+M N P AM
A A (fMNP + 2N MP )A A A
2
g
g
N+ P Q
N P QR
fMNR fP Q R AM
C
+MNP M QR C
A A A +
4
16

1
R
NP
+
++
QR M AQ
MNP C
2 AM
.
+ M C
+ +M C
g
f
A
4
(2.20)
Finally, there is a potential for the scalar fields that takes the form

g2
2 MQ
1
1 MQ N R P S
MQ
NR P S
M M +
M
fMNP fQRS M
M

e Lpot =
16
3
3

4
fMNP fQRS M MNP QRS + 3M N M MMN .
(2.21)
9
The basic feature of this Lagrangian is that it depends explicitly on both the electric gauge
M
potentials AM+
and their magnetic duals A . Therefore it allows the gauging of any subgroup
of the full duality group SL(2, R) SO(6, n), where in principle both electric and magnetic
potentials can participate. The field equations derived from this Lagrangian and the gauge transformations can be found in [9].
3. Gauging the N = 4 axionic symmetries
Here we present a class of N = 4 gauged supergravities obtained by making local the axionic
shifts and axionic rescalings. This is a subclass of the electric gaugings and we will refer to
them as non-unimodular gaugings. Their existence was pointed out in [9] but here we analyze
in detail and full generality the properties of the corresponding gauge algebra and discuss some
interesting features of their Lagrangian description, such as a duality between massive vectors
and massive two-forms.
3.1. Electricmagnetic duality and electric gaugings
The most general N = 4 gauging is described in terms of two tensors fLMN and I ,
, {+, } and I, L, M, N {1, . . . , 6 + n}, satisfying four quadratic conditions (i)(iv) displayed in Eqs. (2.9)(2.12). This general formalism, defined in an arbitrary symplectic frame,
captures in particular the gauging of the electricmagnetic duality symmetry generated by S++
(see Section 2.1).
Gaugings with pure fLMN s have been studied extensively in the literature. They correspond
to switching on gauge algebras entirely embedded in SO(6, n), as shown by (2.8). On the other
hand, turning on the I s allows one to gauge both SL(2, R) and SO(6, n). This situation has not
attracted much attention and only a few examples of the corresponding gauge algebras have been
analyzed (see e.g. [9,35]). Our aim is to study systematically a class of such gaugings and show
that they correspond to a specific pattern of higher-dimensional reduction which generalizes the
ScherkSchwarz mechanism.
We will focus here on electric gaugings, namely gaugings that do not involve the S++ generator of SL(2, R). Axionic shifts S or axionic rescalings S+ will however be gauged,
accompanied by the appropriate SO(6, n) generators. Hence, this class of gaugings is defined
122
by setting
I = 0.
(3.1)
We will also set

fLMN = 0,
(3.2)
while keeping +I I = 0 and

although this is not compulsory for electric
f+LMN fLMN = 0. Furthermore, we will focus on the case n = 6, which is related to pure
gravity in ten dimensions, and adopt the off-block-diagonal 6 + 6 metric:

0 I6
=
(3.3)
.
I6 0
gaugings,6
The quadratic constraint (iii), Eq. (2.11), is now automatically satisfied while the constraints
(i), (ii), (iv)Eqs. (2.9), (2.10), (2.12)reduce to
MN M N = 0,
MN fMI [J fKL]N
MN M fN I J = 0,
2
= f[I J K L] .
3
(3.4)
(3.5)
3.2. Non-unimodular gaugings

The fundamental representation 12 of SO(6, 6) decomposes into 6+1 + 61 under the diagonal
GL(6) = U (1) SL(6) subgroup and correspondingly the I -indices decompose into (i, i ) with
both i and i ranging from 1 to 6. Then, the 6 + 6 metric can be written in the following way:
ij = i j = 0, whereas ii = i i = ii . In this basis the SO(6, 6)-invariant inner product takes

the form AM B M = Am B m + Am B m and we can write Am Am , Am Am .
A specific solution with f s and s
Now, a non-trivial solution to Eqs. (3.4) and (3.5) is
i = i ,
i = 0 for all i ,
fi ij = fij i = fj i i = [i j ]i
(3.6)
all others vanishing,
(3.7)
with i arbitrary real numbers. The existence of the gauging described in Eqs. (3.6) and (3.7)
was pointed out in Ref. [9]. Actually, there exist a whole class of gaugings of this type with more
components of the tensor fI J K turned on. As discussed in [9], besides Eqs. (3.6) and (3.7) one
can turn on fij k . Then, the quadratic constraints reduce to a single equation
f[ij k l] = 0.
(3.8)
Since the novel feature of this class of gaugings is the presence of a non-zero parameter, we
will restrict ourselves to the simplest example, namely the gauging with fij k = 0.
The gauging under consideration will be called non-unimodular for reasons that will become clear at the end of Section 4.2, or traceful since
5
fij j = i .
2
(3.9)
6 What is meant by electric gauging is not universally set in the literature. In [9], for example, electric gaugings are
defined as those with I = fLMN = 0, which is somewhat too restrictive.
123
This is slightly misleading, however, since the gauge algebra is traceless as a consequence of
the full anti-symmetry of its genuine structure constants. The latter are not fij k but specific
combinations of fI J K and I read off from the commutation relations of generators (2.8).
The gauge algebra
In the rest of this section, we will characterize the gauge algebra, which will be further studied from a higher-dimensional perspective in the next chapters. Using Eqs. (3.6) and (3.7) in
expression (2.8), we obtain:
i
S i ,
2
i = 0,

i j
+i =
T j + S+ i ,
2
+i = j T j i i .
i =
(3.10)
(3.11)
(3.12)
(3.13)
The gauge algebra at hand has at most 18 non-vanishing generators but only 7 are independent:

, , plus 5 of the i due to the constraint i i = 0.
Their commutation relations follow from Eqs. (2.2) and (2.6):
[i , j ] = 0,
(3.14)
[, j ] = 0,
(3.15)
[i , ] = i ,
(3.16)
[, ] = .
(3.17)
The set {, i } spans an Abelian Lie subalgebra. Furthermore, the algebra generated by {, }
is a non-compact A2,2 subalgebra of SL(2, R) SO(6, 6).
The above algebra is non-flat,7 in contrast to the algebras obtained by standard Scherk
Schwarz reductions. As we will see in later, the gaugings at hand are related to twisted versions
of these reductions, which relax therefore the flatness of the gauge algebras. The particular case
i = 1i appears when compactifying from five to four dimensions (see [35]) with a noncompact twist generated by the five-dimensional rescaling. The full algebra (3.10)(3.17) will
also emerge (see Section 4) as a ten-dimensional heterotic reduction with twist. Richer nonAbelian extensions are possible in this case, that eventually lead to non-vanishing fij k as already
advertised, and which are not possible when compactifying from five dimensions. These issues
will be extensively analyzed in Section 4.2.
3.3. Lagrangian description
It is straightforward to derive the bosonic Lagrangian for the gaugings we have just described
by using the general formulas of the previous section. As a first step, we will implement (3.1) and
(3.2), and later set (3.6) and (3.7). Finally, we will dualize a vector, which acquires a Stckelberglike mass via the gauging at hand, into a massive two-form field potential.
7 An algebra is called flat when it is generated by a set {Q, X } satisfying the commutation relations [Q, X ] = M j X ,
i
i
i j
j
[Xi , Xj ] = 0 with Mi = Mji . Then the Levi-Civita connection on the corresponding group manifold has zero curva-
ture, therefore justifying the name flat.
124
Electric gaugings I = fKLM = 0

The kinetic terms read
1
1
1
e1 Lkin = R + D MMN D M MN e4 D aD a D D
2
16
4
1
1 2
M N
M N
+ aMN H
H ,
e MMN H H
4
8
where now
(3.18)
D MMN = MMN + 2gAP P (M Q MN )Q
(3.19)
g M
M
M N P
H
= 2[ AM
] g fN P A[ A] + C
2
(3.20)
and
++ .
with C := C
Since fMNP and M are zero and hence MNP and fMNP are zero as well, we have
M+
M
M
omitted the + index from all coefficients. We will use the notation AM
A , A X
for the gauge potentials in order to avoid cluttering of the formulas with indices. Furthermore,
we define the linear combinations
X M XM ,
A M AM
.
(3.21)
The gauge covariant derivatives of the axion and dilaton take the form8
D a = a + X + A a,
(3.22)
1
D = A .
(3.23)
2
By turning on the parameters we have gauged a non-Abelian two-dimensional subgroup of
the SL(2, R) global axion-dilaton symmetry. The magnetic potential X corresponds to the
gauging of the shift symmetry of the axion a a + c and acts as a Stckelberg field, while A
gauges the dilatation symmetry a e2 a, + . In terms of the notation of the previous
subsection, the corresponding gauge algebra is the one spanned by and with commutation
relation (3.17).
The topological term in the Lagrangian for the class of gaugings under consideration becomes

1
1
P
M
e1 Ltop = M N P XM AN
A M C G
2
4

1
N P Q
fMNR fP Q R AM
A
A
X
(3.24)
.
4
We have defined the following gauge field strengths
M
M N P
F
= 2[ AM
] fN P A[ A] ,
(3.25)
M
M Q R
GM
= 2[ X] fQR A[ X] ,
(3.26)
M = FM + 1MC .
in terms of which one can write H
2
8 From now on we set the coupling constant g equal to 1 for simplicity.
Finally, the potential terms are

1 ++
1 MQ N R P S
1
MN
+ fMNP fQRS
M M
M
e Lpot = M
3M N M
16
3

2 MQ
+
M MQ N R P S
125
(3.27)
with M ++ = e2 .
Non-unimodular gaugings
We will now specify the gauge parameters by setting (3.6) and (3.7). With this gauging, the
indices i and i corresponding to the 6+1 and 61 of U (1) SL(6) SO(6, 6) are treated differ
ently. We first notice that the magnetic field strengths Gm
do not appear in the Lagrangian. The
gauge field strengths that appear are
m
= 2[ Am
F
] ,
(3.28)
m
F
(3.29)

j
j
m
+ 2j Am
[ A] Aj [ A] ,

m
i
m
m i
Gm
= 2[ X] + i A[ X] + i A[ X] ,

= 2[ Am
]
(3.30)
where now
X M XM = m Xm ,
m
A M AM
= m A .
(3.31)
One further notices that the Lagrangian actually depends only on the linear combination m Gm
.
This combination can be written in terms of A and X as
G = 2[ X] + 2A[ X] .
(3.32)
m = 2 A .
Similarly we introduce F = m F
[ ]
The natural prescription is to integrate out the auxiliary two-form C in order to obtain the
final Lagrangian. If we do so, starting from (3.18), (3.24) and (3.27), we obtain
1
1
1
e1 L = R + D MMN D M MN e4 D aD a D D
2
16
4
1
1
1 2
M N
M N
F
+ aMN F
2 e2 Z Z
F
e MMN F
4
4
4
1
X + e1 Lpot .
12
(3.33)

In this expression Z = aF + G + e2 m Mm N F N , where F N is the Hodge
N
2
Poincar dual of F ; we have also defined = m M mn n and we have introduced the Chern
Simons form:
1
1
n
m
= An Fn + An F
A Am
Am + A A Am + cyclic.
2
2
(3.34)
The local gauge invariance under axionic shifts enables us to gauge away the axion a = 0.
Then, the magnetic potential X acquires a mass through its Stckelberg coupling to the axion.
The final expression for the gauge-fixed Lagrangian is therefore
126
1
1
1
M N
e1 L = R + D MMN D M MN D D e2 MMN F
F
2
16
4

1

N
2 e2 G + e2 m Mm N F N G + e2 m Mm N F
4
1
X + e1 Lpot .
12
It captures all the relevant information carried by the axionic-symmetry gauging.
(3.35)
Stckelberg mass and dualization

There is a different formulation of the theory, which is actually more suggestive of a higherdimensional origin. Instead of integrating out the auxiliary anti-symmetric tensor C , one can
promote C to a propagating field and integrate out X . Using as previously (3.18), (3.24) and
(3.27), this procedure yields the following Lagrangian:
1
1
e1 L = R + D MMN D M MN D D
2
16

1
1
1 2
M
+ M C F N + N C
e MMN F
4
2
2

2
3
1
e4 [ C] A[ C] + e1 Lpot ,
8
3
(3.36)
where the two-form C acquires now a scalar-field-dependent mass.

The key observation is that the theory described by (3.36) is related to that described by
(3.35) by an interesting generalized duality. Recall that a massless two-form potential in four
dimensions carries one propagating degree of freedom and can be dualized into a scalar. In the
case where the two-form comes from the reduction of the ten-dimensional NSNS two-form in
the gravity multiplet, the dual scalar is the axion. If instead, as it happens here, the two-form
is massive, it carries three degrees of freedom and the dual potential is a massive vector. This
duality is a particular instance of a generalized duality between massive p-forms and massive
(D p 1)-forms in D dimensions [22,23].
For the benefit of the reader we will present schematically how this generalized duality works,
leaving as an exercise its precise implementation between (3.35) and (3.36). Start from a massive
gauge field X with Lagrangian
L=
1

1
( X X ) X X + m2 X X .
2
2
4g
(3.37)
This Lagrangian can be obtained from

1
L = 2C X + g 2 C C + m2 X X
(3.38)
2
by integrating C , which is an auxiliary non-propagating antisymmetric tensor. Instead, we
can integrate by parts L so that X becomes non-dynamical, while C acquires a dynamics.
Integrating out finally X yields an action for a massive two-form,
2
L = 2 C C + g 2 C C ,
(3.39)
m
which can be brought to a more familiar form for anti-symmetric tensors, by HodgePoincardualizing C .
127
Following a similar pattern, one can replace the massive vector X in L (Eq. (3.35)) by a
two-form. This yields precisely L (Eq. (3.36)), therefore demonstrating that the Lagrangians L
and L describe equivalent physics.
4. Higher-dimensional origin
In this section we perform a generalized dimensional reduction of heterotic supergravity to
four dimensions and show that the resulting effective theory belongs to the class of N = 4 gauged
supergravities studied in the previous section. Recall that the usual dimensional reduction of
heterotic supergravity, which can be thought of as compactification on a six-torus keeping only
the massless modes, results in a four-dimensional theory with 16 supercharges, Abelian gauge
group U (1)12+p and a global off-shell symmetry SO(6, 6 + p) [24]. Six of the Abelian vectors
are graviphotons while another six of them come from reducing the NSNS two-form on the
1-cycles of the torus.
One can also reduce some of the vectors present already in ten dimensions and obtain p
additional Abelian gauge fields. Usually this is done for the vectors lying in the Cartan torus
of the ten-dimensional gauge group, therefore yielding a theory with p = 16. Since our goal
here is to make contact with a gauged supergravity with p = 0 we will ignore this possibility
and consider the reduction of the gravity-multiplet fields only. It is straightforward to extend the
reduction to the YangMills sector and obtain generalizations of the gaugings we discussed so
far.
4.1. Heterotic reduction with duality twist
Our starting point is the bosonic action of heterotic supergravity in the string frame9

S = dx dy Ge
R + GMN M N
M4
K6

1 MM N N KK
G
G
HMNK HM N K .
G
12
(4.1)
We assume a decomposition of the ten-dimensional spacetime in a four-dimensional noncompact part M4 parameterized by coordinates x and a six-dimensional internal manifold K6
parameterized by y i . The spacetime indices will be decomposed as M = (, i). As usual is
the dilaton and H = dB is the three-form field strength of the NSNS anti-symmetric tensor B.
As said, we neglect the ten-dimensional YangMills fields.
Taking K6 to be a flat six-torus and keeping only the y-independent modes yields a theory
with 12 Abelian vectors and 38 massless scalars, two of which are the dilaton and the axion. The
latter is the dual of the two-form B obtained by reducing BMN . One way to obtain a more
interesting theory is by introducing fluxes in the torus. The most general reduction of this type,
where both NSNS and geometric fluxes were present, was studied in [25].
The introduction of geometric fluxes has an alternative interpretation as a reduction with a
twist for the spacetime fields [26]. This is actually a particular case of a generalized reduction
9 The ten-dimensional spacetime indices M, N, . . . of this section should not be confused with the fundamental
SO(6, 6) indices of the previous sections.
128
scheme that usually is referred to as ScherkSchwarz reduction [27]. The characteristic property of reductions of this type is that the reduction ansatz can incorporate a dependance on the
coordinates of the internal torus. This dependance is not arbitrary however; on a technical level
it is dictated by the requirement that the Lagrangian should be independent of the internal coordinates. This is implemented by selecting a profile for the fields whose consistency is guaranteed
by some symmetry of the original theory. Such reductions in the context of supergravity have
been studied in [2836] while the reader is referred to [37] for a general discussion on reductions
with duality twists.
A subtle point that is not usually emphasized is the following. These reduction schemes yield
an effective theory for a finite set of modes selected out of the infinitude of higher-dimensional
modes according to some symmetry principle. Hence, it is not necessarily true that they encompass all low-energy modes and further analysis is required in order to establish that the effective
theory obtained through a ScherkSchwarz reduction is actually a low-energy effective theory. It
would be interesting to perform such an analysis for the reduction scheme presented below but
this issue lies beyond the scope of this paper.
The symmetry we will employ in the present paper is the SO(1, 1) scaling symmetry of (4.1),
under which the fields transform as
GMN e GMN (x),
+ 4,
BMN e BMN (x).
(4.2)
This enables us to trade the usual periodic ansatz, which assumes no dependance on the torus
coordinates, with the following one
(x, y) = (x) + 4i y i ,
GMN (x, y) = ei y GMN (x),
BMN (x, y) = ei y BMN (x).
(4.3)
The parameters i are arbitrary real numbers that dictate the twisting of the fields along the six
one-cycles of the torus.
The decomposition of the ten-dimensional metric tensor in terms of four-dimensional fields is
the usual one
G = g + Ai Aj hij ,
Gi = Aj hij ,
Gij = hij .
(4.4)
Here g is the metric on M4 , hij are the 21 metric moduli of the six-torus and Ai are the
KaluzaKlein gauge fields. Similarly, the anti-symmetric tensor is decomposed as
B = b ,
Bi = bi ,
Bij = bij .
(4.5)
Bi
We obtain a two-form b , which usually is dualized to an axion, six vectors

and 15 scalar
moduli bij .
The above decompositions hold for the full ten-dimensional fields that are assumed to have
a dependance on y i of the type dictated by the SO(1, 1) scaling symmetry. Consistency implies
that
k
g (x, y) = ek y g (x),
b (x, y) = ek y b (x),
hij (x, y) = ek y hij (x),

k
bij (x, y) = ek y bij (x),

(x, y) = (x) + k y k ,
(4.6)
Ai (x, y) = Ai (x),
bi (x, y) = ek y bi (x),
(4.7)
(4.8)
129
where the four-dimensional dilaton is defined as = 12 log det h. The y-independent modes
on the right-hand sides are the four-dimensional fields for which we would like to derive the
effective action. The ansatz is consistent in the sense that the y-dependance is totally eliminated
from the action and the integration over y i yields an overall multiplicative factor. Notice that the
volume of the internal six-torus is encoded in the metric moduli hij .
Let us first reduce the EinsteinHilbert part of the action along with the dilaton kinetic term.
The most efficient way of performing this reduction is the following. We start from the metric
G MN (x, y) = 2 (y)GMN (x),
(4.9)
where (y) = exp( 12 y i i ), use the relation between Ricci scalars for conformally related metrics, and finally apply the usual reduction formulas
for GMN (x). After the above redefinition of
the dilaton, necessary for absorbing the factor of h, the conformal rescaling g exp2 g ,
which brings us to the Einstein frame in four dimensions, and a final rescaling 2, we obtain

1
1
Sgravity = dx g R4 + hij Ak k hij hij + A hij
2
8
M4

1 i
1 j
g
A i A j
2
2

1
1 2
i
j
2
ij
e i h j ,
e hij F F
4
2
(4.10)
m = Am Am and all fields are exclusively y-dependent.

where F

There are several observations in order. First, the metric moduli hij become charged under the
KaluzaKlein gauge fields Ai . The charges are given by the vector of twisting coefficients i .
The dilaton is also coupled in a Stckelberg fashion to Ai . This signals the gauging of a shift
symmetry as expected from a reduction where the ansatz was twisted by employing such a symmetry.10 Notice, furthermore, that the twisting does not result in a non-Abelian gauge symmetry
for the KaluzaKlein gauge fields. It does however lead to a potential for the dilaton and the
metric moduli.
The reduction of the NSNS part of the Lagrangian is more easily performed using the
tangent-space components of the anti-symmetric three-form [24]. Furthermore, some field redefinitions are necessary in order to bring the resulting Lagrangian to a more standard form. One
defines vector fields Yn and a two-form B as
Yn = bn + bnm Am
,
B = b + Am
[ Y]m
We get

SNSNS =
(4.11)
n
Am
A bmn .
(4.12)

1

1
1
hmn hk D bm D bnk e4 3 [ B] k Ak[ B]
8
6
2
M4

1
[
3 [ B ] A B ]
2
10 Note that the dilaton is shifted under the SO(1, 1) scaling.
130

1

Yn + n B bnk F k
e2 hmn Ym + m B bm F
4

1
g dx,
+ e2 i hij hk bm hmn bnk 2hik bkm hmn bnr hrj j
16
(4.13)
where we have defined

D bn = bn + Yn n Y m Am
bn ,

1
Y = Y Y +
mn m n (Am Yn Yn Am ),
2
(4.14)
(4.15)
while the ChernSimons three-form is

1
1
m
= Y A + A Y A Am
(4.16)
Ym + A A Ym + cyclic.
2
2
We observe that the NSNS moduli are charged under the KaluzaKlein gauge fields but have
also Stckelberg couplings to the NSNS gauge potentials. The latter couplings are due to the
gauging of the shift symmetries of those moduli induced by the duality twist. A crucial difference
with the case of the ordinary dimensional reduction is that the four-dimensional two-form B
acquires a mass. This prohibits the standard dual formulation in terms of an axion but, according
to the discussion of the previous section on dualities between massive fields, suggests that a dual
formulation in terms of a massive vector is possible. Let us finally stress that the reduction of the
NSNS sector also contributes to the potential for the hij and bij moduli (last line of (4.13)).
4.2. Contact with N = 4 gauged supergravity
We will now show that the effective theory described by the sum of actions (4.10) and (4.13)
is nothing but the N = 4 gauged supergravity worked out in Section 3. Using the standard parameterization of the moduli matrix M MN
mn

h
hmk bkn
M MN =
(4.17)
,
bmk hkn hmn bmk hk bn
the N = 4 potential (3.27) obtained for the non-unimodular gauging reads:

1 2 ij
e i 8h hij hk bm hmn bnk + 2hik bkm hmn bnr hrj j .
(4.18)
16
This is precisely the potential in the effective theory (4.10) plus (4.13). Notice that this identification clarifies the higher-dimensional interpretation of the gauging parameters : they correspond
to the parameters used to twist the boundary conditions by SO(1, 1) scalings along the six onecycles of the torus.
It is straightforward to check that the rest of the terms in (4.10) plus (4.13) match exactly those
m
of (3.36) provided we identify the gauge fields Am
, A in the gauged-supergravity Lagrangian
with the KaluzaKlein and NSNS gauge fields Am
, Ym in the heterotic reduction and the antisymmetric tensors as C 2B . This elucidates the higher-dimensional origin of the gauged
supergravity of Section 3 and confirms the prominent role of the generalized duality performed
in four dimensions for reaching (3.36). It is amusing that the four-dimensional two-form B that
comes from the NSNS anti-symmetric tensor in ten dimensions is actually the auxiliary tensor
gauge field required for consistency of the gauging in the formalism of [8].
V=
131
Let us mention at this point that ordinary reductions of the heterotic theory with NSNS
fluxes and geometric fluxes also yield N = 4 gauged supergravities [25]. The correspondence
with the embedding-tensor language is as follows: there are no s turned on and the only nonvanishing parameters are the f+I J K fI J K . Under the decomposition of indices I = (i, i ), the
background NSNS fluxes ij k and geometric fluxes11 ij k are identified with the components
of fI J K as
fij k = 3ij k ,
fij k = 2ij k ,
(4.19)
all other components being zero. The remaining non-trivial quadratic constraint is (iv) (Eq. (2.12))
and it corresponds to the Bianchi identities for the NSNS fluxes and the Jacobi identity for the
geometric fluxes. From this we conclude that the more general class of gaugings we mentioned
in Section 3.2 with non-zero fij k originates from a ten-dimensional reduction with an SO(1, 1)
duality twist combined with background NSNS fluxes. The condition (3.8) found then is a
consequence of the Bianchi identity resulting from the ansatz (4.3).
An interesting observation is in order here. The correspondence between the components
fij k and the geometric fluxes ij k provides an alternative perspective on the gauging we have
performed in Section 3.2 and the subsequent heterotic reduction of Section 4.1. Indeed, if we
interpret the y-dependance of the internal metric (cf. Eqs. (4.6)) as inducing a geometric flux, this
flux is automatically non-unimodular since ij i = 0. In ordinary reductions, the unimodularity
condition ensures consistency of the truncation of the higher-dimensional Lagrangian [26,38].
This well-known obstruction is circumvented in our approach thanks to the compensating duality
twist.12
5. Conclusions and open problems
In this paper we studied in detail the class of N = 4 axionic-symmetry gaugings and established that they can be embedded in heterotic theory. More specifically, they arise through a
reduction where the boundary conditions for the fields are twisted by an SO(1, 1) scaling symmetry. Similar reductions are possible for type II strings yielding N = 8 gauged supergravities
or N = 4 upon appropriate orbifolding/orientifolding. For M-theory, instead, there is no scaling
symmetry of the action and that implies that the Lagrangian cannot be consistently truncated for
fields with boundary conditions of this type. However, one can still perform such reductions at
the level of the equations of motion and it is expected that the reduced equations of motion correspond to gaugings of the type we studied here. In passing, we also note that since the dilaton
becomes a component of the internal geometry when a type IIA background is lifted to M-theory,
reductions with dilaton twists should lift to M-theory reductions with purely geometric twists of
the type studied in [39,40].
Some obvious extensions of the current work include twisted reductions of the heterotic theory
taking into account the ten-dimensional gauge fields or similar type II reductions in the presence
of branes and orientifolds. This should yield electric N = 4 gaugings where the gauge algebra is
a subgroup of SL(2, R) SO(6, n) for n 6.
The fact that the formulation of gauged supergravity through the embedding tensor is dualitycovariant implies that these theories capture the effective dynamics of backgrounds related by
11 We use the notation of [25].
12 This statement refers to a reduction performed in the string frame. The metric in the Einstein frame is not affected by
the duality twist and the corresponding geometric flux must always be unimodular.
132
duality transformations. Recently it has become increasingly clear that the majority of these
backgrounds are non-geometric and cannot be described using the familiar notions of geometry
and ordinary fluxes. From one point of view this demonstrates the power of the effective bottom
up approach, since four-dimensional physics can be derived without the need to delve into the
microscopic details of a higher-dimensional setup. On the other hand, one could argue that a better understanding of non-geometric backgrounds may still be obtained through analyzing gauged
supergravity.
For instance, the non-geometric fluxes Q and R proposed in [41] as T-dual of the familiar
NSNS and geometric fluxes, are automatically captured for heterotic compactifications on a
six-torus by the formulation of N = 4 gauged supergravity we have being discussing. Using the
notation of [41] our gauging parameters f+I J K describe all possible situations through
f+ij k Hij k ,
f+ij k fij k ,
f+i j k Qij k ,
f+i j k R ij k .
(5.1)
Besides this set of SO(6, 6)-dual fluxes, the most general N = 4 gauging comprises of another
set of S-dual fluxes fI J K . It would be extremely interesting to understand the microscopic
origin of all those non-geometric fluxes directly in ten dimensions and derive the corresponding gauged supergravities using an appropriate reduction scheme (see [42] for some recent
ideas in this direction. Also, the non-geometric fluxes can be interpreted as geometric fluxes
in an appropriate generalized geometry [43]). Among others, this should shed some light on the
open problem of lifting gauged supergravities with non-trivial duality phases in heterotic string
theory.
A related question concerns the higher-dimensional origin of the gauging constraints (i)(iv).
For example, although some of these constraints have a clear origin as Bianchi identities in the
internal space, this is not so for the null condition (i) M M = 0. The reduction we performed
depends naturally on six parameters that fill up M in such a way that it is automatically null. It
would be interesting to understand how N = 4 gaugings with more general parameters M can
be obtained from higher dimensions and where the null condition comes from.
We conclude by emphasizing that formulations of string and M-theory of the type presented
in [4446] as well as the mathematical framework of generalized complex geometry [47] may
provide the appropriate tools for resolving the above issues.
Acknowledgements
The authors wish to thank L. Carlevaro, G. DallAgata, S. Ferrara and H. Samtleben for
stimulating scientific discussions and J. Schn, M. Trigiante and M. Weidner for very helpful
correspondence. This work was supported in part by the EU under the contracts MEXT-CT2003-509661, MRTN-CT-2004-005104, MRTN-CT-2004-503369 and by the Agence Nationale
pour la Recherche, France. Marios Petropoulos acknowledges financial support by the Swiss
National Science Foundation.
References
[1] M. Graa, Flux compactifications in string theory: A comprehensive review, Phys. Rep. 423 (2006) 91, hepth/0509003.
[2] M. Trigiante, Dual gauged supergravities, hep-th/0701218.
[3] J.-P. Derendinger, C. Kounnas, P.M. Petropoulos, F. Zwirner, Superpotentials in IIA compactifications with general
fluxes, Nucl. Phys. B 715 (2005) 211, hep-th/0411276.
133
[4] B. de Wit, H. Samtleben, M. Trigiante, On Lagrangians and gaugings of maximal supergravities, Nucl. Phys. B 655
(2003) 93, hep-th/0212239.
[5] B. de Wit, H. Samtleben, M. Trigiante, Maximal supergravity from IIB flux compactifications, Phys. Lett. B 583
(2004) 338, hep-th/0311224.
[6] B. de Wit, H. Samtleben, M. Trigiante, Gauging maximal supergravities, Fortschr. Phys. 52 (2004) 489, hepth/0311225.
[7] B. de Wit, H. Samtleben, M. Trigiante, The maximal D = 5 supergravities, Nucl. Phys. B 716 (2005) 215, hepth/0412173.
[8] B. de Wit, H. Samtleben, M. Trigiante, Magnetic charges in local field theory, JHEP 0509 (2005) 016, hepth/0507289.
[9] J. Schn, M. Weidner, Gauged N = 4 supergravities, JHEP 0605 (2006) 034, hep-th/0602024.
[10] D.Z. Freedman, J.H. Schwarz, N = 4 supergravity theory with local SU(2) SU(2) invariance, Nucl. Phys. B 137
(1978) 333.
[11] S.J.J. Gates, B. Zwiebach, Gauged N = 4 supergravity theory with a new scalar potential, Phys. Lett. B 123 (1983)
200.
[12] S.J.J. Gates, B. Zwiebach, Searching for all N = 4 supergravities with global SO(4), Nucl. Phys. B 238 (1984) 99.
[13] A. Das, SO(4) invariant extended supergravity, Phys. Rev. D 15 (1977) 2805.
[14] E. Cremmer, J. Scherk, Algebraic simplifications in supergravity theories, Nucl. Phys. B 127 (1977) 259.
[15] E. Cremmer, J. Scherk, S. Ferrara, SU(4) invariant supergravity theory, Phys. Lett. B 74 (1978) 61.
[16] J.-P. Derendinger, Supergravity gaugings and moduli superpotentials, Fortschr. Phys. 54 (2006) 366.
[17] A.H. Chamseddine, N = 4 supergravity coupled to N = 4 matter, Nucl. Phys. B 185 (1981) 403.
[18] J.P. Derendinger, S. Ferrara, N = 1 and N = 2 supergravities coupled to matter: Superhiggs effect and geometrical
structure, in: B. de Wit, P. Fayet, P. van Nieuwenhuizen (Eds.), Supersymmetry and Supergravity 84, Proceedings
of the Trieste Spring School.
[19] M. de Roo, P. Wagemans, Gauge matter coupling in N = 4 supergravity, Nucl. Phys. B 262 (1985) 644.
[20] P. Wagemans, Breaking of N = 4 supegravity to N = 1, N = 2 at = 0, Phys. Lett. B 206 (1988) 241.
[21] P. Wagemans, Aspects of N = 4 supergravity, Ph.D. Thesis, Groningen University report RX-1299, 1990.
[22] P.K. Townsend, K. Pilch, P. van Nieuwenhuizen, Selfduality in odd dimensions, Phys. Lett. B 136 (1984) 38;
P.K. Townsend, K. Pilch, P. van Nieuwenhuizen, Phys. Lett. B 137 (1984) 443, Addendum.
[23] F. Quevedo, Duality beyond global symmetries: The fate of the B field, hep-th/9506081.
[24] J. Maharana, J.H. Schwarz, Noncompact symmetries in string theory, Nucl. Phys. B 390 (1993) 3, hep-th/9207016.
[25] N. Kaloper, R.C. Myers, The O(d, d) story of massive supergravity, JHEP 9905 (1999) 010, hep-th/9901045.
[26] J. Scherk, J.H. Schwarz, How to get masses from extra dimensions, Nucl. Phys. B 153 (1979) 61.
[27] J. Scherk, J.H. Schwarz, Spontaneous breaking of supersymmetry through dimensional reduction, Phys. Lett. B 82
(1979) 60.
[28] E. Bergshoeff, M. de Roo, M.B. Green, G. Papadopoulos, P.K. Townsend, Duality of type II 7-branes and 8-branes,
Nucl. Phys. B 470 (1996) 113, hep-th/9601150.
[29] P.M. Cowdall, H. Lu, C.N. Pope, K.S. Stelle, P.K. Townsend, Domain walls in massive supergravities, Nucl. Phys.
B 486 (1997) 49, hep-th/9608173.
[30] E. Bergshoeff, M. de Roo, E. Eyras, Gauged supergravity from dimensional reduction, Phys. Lett. B 413 (1997) 70,
hep-th/9707130.
[31] I.V. Lavrinenko, H. Lu, C.N. Pope, Fibre bundles and generalised dimensional reductions, Class. Quantum Grav. 15
(1998) 2239, hep-th/9710243.
[32] N. Kaloper, R.R. Khuri, R.C. Myers, On generalized axion reductions, Phys. Lett. B 428 (1998) 297, hepth/9803066.
[33] C.M. Hull, Massive string theories from M-theory and F-theory, JHEP 9811 (1998) 027, hep-th/9811021.
[34] C.M. Hull, A. Catal-Ozer, Compactifications with S-duality twists, JHEP 0310 (2003) 034, hep-th/0308133.
[35] G. Villadoro, F. Zwirner, The minimal N = 4 no-scale model from generalized dimensional reduction, JHEP 0407
(2004) 055, hep-th/0406185.
[36] R.A. Reid-Edwards, Geometric and non-geometric compactifications of IIB supergravity, hep-th/0610263.
[37] A. Dabholkar, C. Hull, Duality twists, orbifolds, and fluxes, JHEP 0309 (2003) 054, hep-th/0210209.
[38] C.M. Hull, R.A. Reid-Edwards, Flux compactifications of string theory on twisted tori, hep-th/0503114.
[39] G. DallAgata, N. Prezas, ScherkSchwarz reduction of M-theory on G2 -manifolds with fluxes, JHEP 0510 (2005)
103, hep-th/0509052.
[40] C.M. Hull, R.A. Reid-Edwards, Flux compactifications of M-theory on twisted tori, JHEP 0610 (2006) 086, hepth/0603094.
134
[41] J. Shelton, W. Taylor, B. Wecht, Nongeometric flux compactifications, JHEP 0510 (2005) 085, hep-th/0508133.
[42] A. Dabholkar, C. Hull, Generalised T-duality and non-geometric backgrounds, JHEP 0605 (2006) 009, hepth/0512005.
[43] G. DallAgata, N. Prezas, H. Samtleben, M. Trigiante, in preparation.
[44] C.M. Hull, A geometry for non-geometric string backgrounds, JHEP 0510 (2005) 065, hep-th/0406102.
[45] C.M. Hull, Doubled geometry and T-folds, hep-th/0605149.
[46] C.M. Hull, Generalised geometry for M-theory, hep-th/0701203.
[47] M. Gualtieri, Generalized complex geometry, math.DG/0401221.
Thermodynamics of apparent horizon

in brane world scenario
Rong-Gen Cai a, , Li-Ming Cao a,b
a Institute of Theoretical Physics, Chinese Academy of Sciences, PO Box 2735, Beijing 100080, China
b Graduate School of the Chinese Academy of Sciences, Beijing 100039, China
Received 5 January 2007; received in revised form 12 June 2007; accepted 21 June 2007
Abstract
In this paper we discuss thermodynamics of apparent horizon of an n-dimensional Friedmann
RobertsonWalker (FRW) universe embedded in an (n + 1)-dimensional AdS spacetime. By using the
method of unified first law, we give the explicit entropy expression of the apparent horizon of the FRW universe. In the large horizon radius limit, this entropy reduces to the n-dimensional area formula, while in the
small horizon radius limit, it obeys the (n + 1)-dimensional area formula. We also discuss the corresponding bulk geometry and study the apparent horizon extended into the bulk. We calculate the entropy of this
apparent horizon by using the area formula of the (n + 1)-dimensional bulk. It turns out that both methods
give the same result for the apparent horizon entropy. In addition, we show that the Friedmann equation on
the brane can be rewritten to a form of the first law, dE = T dS + W dV , at the apparent horizon.
1. Introduction
Thermodynamics of black hole has been studied for a long time. However, most discussions of black hole thermodynamics have been focused on the stationary case. For dynamical (i.e., non-stationary) spherically symmetric black holes, Hayward has proposed a method
to deal with thermodynamics associated with trapping horizon of a dynamical black hole in
4-dimensional Einstein theory [14]. In this method, for spherical symmetric spacetimes,
ds 2 = h dx dx + r 2 d22 , Einstein equations can be rewritten in a form called unified first
E-mail addresses: cairg@itp.ac.cn (R.-G. Cai), caolm@itp.ac.c (L.-M. Cao).

doi:10.1016/j.nuclphysb.2007.06.016
136
R.-G. Cai, L.-M. Cao / Nuclear Physics B 785 (2007) 135148
law
dE = A + W dV ,
(1)
h r r); A =
where E is the so-called MisnerSharp energy [5], defined by E =
2
3
4r is the sphere area with radius r and V = 4r /3 is the volume; and the work density
W 12 T h and the energy supply vector T r + W r with T being the energy
momentum tensor of matter in the spacetime. Projecting this unified first law along a trapping
horizon, one gets the first law of thermodynamics for dynamical black hole
dE, =
(2)
dA, + W dV , ,
8G
where is surface gravity, defined by = 21h ( hh r), on the apparent horizon, is a

projecting vector. In a recent paper [6], we have applied this theory to study the thermodynamics
of apparent horizon of a FRW universe in higher dimensional Einstein gravity and some nonEinstein theories, such as Lovelock gravity and scalar-tensor gravity by rewriting gravity field
equations to standard Einstein equations with a total energymomentum tensor. The total energy momentum tensor consists of two parts: one is just the ordinary matter energymomentum
tensor; and the other is an effective one coming from the contribution of the higher derivative
terms (for example, in Lovelock gravity) and scalar field (for example, in scalar-tensor gravity
theory). The total energymomentum tensor will enter into the energy-supply and work term
W in (1). Therefore, the work density and energy-supply vector in (1) also can be decomposed
r
2G (1
into ordinary matter part and effective part: = + , W =W + W . That is, and W are the
e
energy-supply vector and work density from the ordinary matter contribution, while and W
come from the effective energymomentum tensor. The matter part of energy-supply is the energy flux defined by pure ordinary matter, so its integration on the sphere (after projecting along
the apparent horizon) naturally defines the heat flow Q in the Clausius relation Q = T dS. On
the other hand, the unified first law tells us
m
e
Q = A , =
(3)
dA, A , .
8G
The right-hand side of the above equation should be written in the form of T S by considering
spacetime horizon with T = /2 as an equilibrium thermodynamic system. Thus we find a
method to get the entropy S of the apparent horizon because the right-hand side of the above
equation is easy to calculate. What one needs to do is to put the right-hand side of the above
equation into a form of total differential projecting along . This total differential gives the variation of the horizon entropy S. By using this method we have indeed given the entropy expression
not only in the Einstein gravity, but also in the Lovelock gravity. The resulting expression of
the apparent horizon entropy is the same as the one for black hole horizon in each theory. For
related discussions on the thermodynamics of apparent horizon in the FRW universe, see [714].
In the setup of static, spherically symmetric black hole spacetimes, there are also some discussions on the relation between the field equations at the black hole horizon and the first law of
thermodynamics [1416]. For the Rindler causal horizon, related discussions see [17,18].
It is interesting to apply this method developed in [6] to study the entropy of the apparent
horizon of a FRW universe in the brane world scenario. This is partially because the gravity on
the brane is not the Einstein theory, the well-known area formula for black hole horizon entropy
must not hold in this case, and partially because exact analytic black hole solutions on the brane
have not been found so far, and then it is not known how the horizon entropy of black hole on the
137
brane is determined by the horizon geometry. On the other hand, the exact Friedmann equations
of the FRW universe on the brane have been derived for the RSII model some years ago [19].
Therefore with the entropy expression of the apparent horizon by using the Friedmann equations,
our method can give some clues to study the thermodynamics of the black holes on the brane. In
this paper, we are indeed able to give the explicit entropy expression of the apparent horizon of
FRW universe in the RSII brane world scenario.
Brane world scenario, based on the assumption that our universe is a 3-brane embedded in
a higher dimensional bulk spacetime, has been intensively studied over past years [2022]. In
the scenario, the standard model fields are confined on the brane, while gravity can propagate
in the whole spacetime. The effective gravity on the brane is different from the standard Einstein gravity due to the existence of extra dimension. In this paper we focus on the so-called
RSII model [22]. The effective equations of motion on the 3-brane living in 5-dimensional bulk
with Z2 symmetry have been given in [23]
(4)
G = 4 q + 8G4 + 54 E ,
(4)
where
1
(5)
4 ,
48 5

1
1
4 = 52 5 + 52 2 ,
(6)
2
6
1
1
1
1
= + + q q 2 ,
(7)
4
12
8
24
and E is the electric part of the 5-dimensional Weyl tensor. Here and are the vacuum
energy and energymomentum tensor of matter on the brane, while 5 and 5 are 5-dimensional
gravity coupling constant and cosmological constant, respectively. 4 is the effective cosmological constant on the brane. Therefore it can be seen from the right-hand side of (4) that except for
the energymomentum tensor of matter , there exist two additional terms and E . This
implies that these two terms can be regarded as effective energymomentum tensors, which cannot enter the definition of Q in the Clausius relation as mentioned above. If the bulk is a pure
AdS spacetime, then E vanishes, and is the only effective energymomentum tensor.
Thus, we can use our method to obtain the entropy expression of apparent horizon in the FRW
universe on the brane. We expect that this entropy expression should reveal some properties of
the horizon entropy of brane world black hole. Indeed our result is consistent with the one in
reference [24], where the authors give a horizon entropy of an n-dimensional black hole on a
brane, which is embedded in an (n + 1)-dimensional bulk, in the limit of large horizon radius.
The apparent horizon of FRW universe on the brane will extend into the bulk (AdS space
time) with a finite distance. The total area of this apparent horizon which extends into the bulk
can be directly calculated from the bulk geometry. Since the gravity in the bulk is the Einsteins
general relativity, the well-known area formula of horizon entropy holds. Therefore, from the
higher dimensional area formula of entropy we can get an entropy of this apparent horizon. We
find that the horizon entropy, according to the area formula in the bulk, is completely the same
as the one obtained by using the unified first law on the brane.
The paper is organized as follows. In Section 2, we give the effective equations of motion on
the (n 1)-brane embedded in an (n + 1)-dimensional bulk with Z2 symmetry by generalizing
the result of [23]. In Section 3, we consider a FRW universe on the brane, and calculate the
entropy of apparent horizon by use of the unified first law. In Section 4, we study the bulk
G4 =
138
geometry and the apparent horizon extended into the bulk. By calculating the area of this apparent
horizon extended into the bulk, we get the same entropy from the (n + 1)-dimensional area
formula as the one obtained in Section 3. We end this paper with conclusion in Section 5.
2. Effective equations of motion on the brane
In the brane world scenario, the n-dimensional world is described by an (n 1)-brane
(M, q ) in an (n + 1)-dimensional spacetime (V, g ). We denote the vector unit normal
to M by n and the induced metric on M by q = g n n (we use the notations in [23]).
Then from the Gauss equation
(n)
R = (n+1) R q q q q + K K K K ,
(8)
and (n + 1)-dimensional Einstein equations

1
2
R g (n+1) R = n+1
T ,
(9)
2
where T is the (n + 1)-dimensional energymomentum tensor in the bulk, together with the
relation of Weyl tensor and Riemann tensor, we obtain the n-dimensional equations on the brane

n2 2
1
(n)
G =
n+1 T q q + T n n T q
n1
n

1
+ KK K K q K 2 K K E ,
(10)
2
where
(n+1)
E (n+1) C n n q q
(11)
is the electric part of bulk Weyl tensor. We choose a local Gauss normal coordinate such that
the hypersurface = 0 coincides with the brane world and n dx = d . Assume that the bulk
has only a cosmological constant n+1 , the (n + 1)-dimensional energymomentum tensor then
has the form
T = n+1 g + S (),
(12)
where
S = q + ,
(13)
with = 0, and are the brane tension and the energymomentum tensor of matter on
the (n 1)-brane. The singular behavior in the energymomentum tensor can be attributed to
extrinsic curvature from the so-called Israels junction condition
n
[q ] = 0,
2
[K ] = n+1

1
q S ,
S
n1
(14)
(15)
where [X] := lim+0 X lim0 X = X + X .

After imposing the Z2 -symmetry on the bulk spacetime, the extrinsic curvature of the brane
can be expressed in terms of the energymomentum tensor on the brane

1 2
1
+
K = K = n+1 S
(16)
q S .
2
n1
139
Substituting this equation (the is not relevant because of the Z2 symmetry) into Eq. (10), we
obtain the gravitational field equations on the (n 1)-brane
(n)
4
G = n q + 8Gn + n+1
E ,
(17)
where
n2
4 ,
32(n 1) n+1

n2 2
n2
2
2
n+1 +
,
n = n+1
n
8(n 1) n+1
1
1
1
1
= +
+ q
q 2 ,
4
4(n 1)
8
8(n 1)
Gn =
(18)
(19)
(20)
and E is the electric part of the (n + 1)-dimensional Weyl tensor defined in Eq. (11). The
above equations reduce to the ones in [23] as n = 4, as expected. Eq. (17) is not closed since
the bulk geometry cannot be determined by using only these equations without solving the Einstein equations in the bulk. The standard Einstein equations can be recovered by taking the limit
n+1 0 while keeping Gn finite. Hereafter, we specialize to the RS II model with vanishing
cosmological constant on the brane, which gives
n+1 =
n(n 1)
,
2 2
2n+1
2(n 1)
.
2
n+1
(21)
This also implies

2
2
Gn+1
= n+1 =
.
Gn
8Gn n 2
(22)
Now we consider a FRW universe with perfect fluid confined on the (n 1)-brane. The perfect
fluid has the energymomentum tensor
= ( + p)t t + pq ,
(23)
where t satisfies t t = 1. By using the expression of , we find

=

n2
2( + p)t t + ( + 2p)q .
8(n 1)
The equations of motion on the brane then become

4
n 2 n+1
(n)
G = 8Gn + p +
( + p) t t
n 1 32Gn

4
n 2 n+1
+ 8Gn p +
( + 2p) q E .
n 1 64Gn
(24)
(25)
For a locally conformal flat bulk spacetime (for example, a pure AdS spacetime), the bulk Weyl
tensor vanishes, so we can omit the E term in that case. Choosing q to be n-dimensional
FRW metric (FRW universe can indeed be embedded in a pure AdS spacetime), i.e.,
a(t)2
2
dr 2 + a(t)2 r 2 dn2
1 kr 2
2
= h ab dx a dx b + r 2 dn2
,
ds 2 = q dx dx = dt 2 +
(26)
140
where r = a(t)r, x 0 = t, x 1 = r, from (25) we obtain the Friedmann equations on the brane
without dark radiation term (E = 0)

4
k
n 2 n+1
2 ,
(n 1)(n 2) H 2 + 2 = 16Gn +
(27)
n 1 64Gn
a

4
n 2 n+1
k
( + p) .
(n 2) H 2 = 8Gn + p +
(28)
n 1 32Gn
a
The matter density on the brane satisfies the continuity equation
+ (n 1)H ( + p) = 0.
(29)
This equation and the Friedmann equations will be used in the next sections.
3. Unified first law and thermodynamics of apparent horizon
In this section we will give the explicit entropy expression of apparent horizon in a FRW
universe on the brane by using the method developed in [6]. Introducing the effective energy
momentum tensor
4
n+1
8Gn
(30)
we can rewrite the gravitational field equations on the brane as

e
(n)
G = 8Gn + ,
(31)
which is in the form of the standard Einstein field equations. Therefore the unified first law (1) is
applicable to Eq. (31). It is easy to find
e
t t =
4
n 2 n+1
2,
n 1 64Gn
r r =
4
n 2 n+1
( + 2p).
n 1 64Gn
(32)
The work density term has the form

m
W =W + W ,
(33)
with
m
1
W = ( p),
2
W=
4
n 2 n+1
p.
n 1 64Gn
(34)
Note that quantities with over m are calculated through the matter energymomentum tensor ,
e
while quantities with over e are calculated through the effective energymomentum tensor .
m
e
e
Namely, W = 12 h , and W = 12 h . Similarly, the energy supply vector can be decomposed as
m
= + ,
(35)
m
1
1
= ( + p)H r dt + ( + p)a dr,
2
2
(36)
with
141
4
4
n 2 n+1
n 2 n+1
( + p)H r dt +
( + p)a dr,
n 1 64Gn
n 1 64Gn
(37)
where they are defined by = r + W r and = r + W r , respectively. Using these quantities and the MisnerSharp energy in n dimensions inside the apparent horizon [6,25]
E=
1
(n 2)n2 rAn3 ,
16Gn
(38)
and An2 = n2 rAn2 and Vn2 = An2 rA /(n 1) are the area and volume of the (n 2)
sphere with the apparent horizon radius rA = 1/ H 2 + k/a 2 of the FRW universe, respectively,
we can put the (00) component of equations of motion (31) into the form of the unified first law
dE = An2 + W dVn2 .
(39)
After projecting along a vector = t (1 2)H rr with = r A /2H rA , we get the first law
of thermodynamics of the apparent horizon [6],
dAn2 , + W dVn2 , ,
dE, =
(40)
8Gn
where = (1 r A /(2H rA ))/rA is the surface gravity of the apparent horizon. The pure matter
m
energy-supply An2 (after projecting along the apparent horizon) gives the heat flow Q in
the Clausius relation Q = T dS. By using the unified first law on the apparent horizon, we have

m
e
Q An2 , =
dAn2 , An2 , ,
8Gn
4
n 2 n+1
=
(41)
dAn2 , +
(1 )( + p)An2 H rA .
8Gn
n 1 32Gn
From the Friedmann equations given in the previous section, one finds

8(n 1) 1
1
( + p) =
(42)

.
4
n+1
rA2
r 4 + 2 r 2
A
Therefore, we arrive at

(1 )
1
Q =
(n 2)An2 H rA
4Gn
rA4 + 2 rA2

n 2 n2 rAn2

=T
d rA , = T dS, = T dS,
4Gn r 2 + 2
(43)
where
(n 2)n2
S=
4Gn
rA
0
r n2
A
d rA .
rA2 + 2
Integrating (44), we obtain the entropy expression associated with the apparent horizon

2
An2
n 2 rA
rA
n1 1 n+1
S=
, ,
,
,
2 F1
4Gn
n1
2
2
2

(44)
(45)
142
where 2 F1 [, , , z] is Gaussian hypergeometric function. When n = 4, we have

A2
2
2
rA
S=
1 + 2 2 arcsinh
.
4G4

rA rA
(46)
Here some remarks are in order.

(i) Taking the limit n+1 0 while keeping Gn finite, we have 0. In this limit, the ndimensional Einstein gravity is recovered on the brane. This limit can also be understood as the
one of the large apparent horizon radius, namely, rA . In this limit, we see from (45) that the
entropy expression of the apparent horizon reduces to the well-known area formula of horizon
entropy in n dimensions. This is an expected result since the gravity is an Einstein one in this
limit, where the area formula holds. The entropy can also be expressed by

An2
2
An2
=
.
S=
(47)
4Gn
4Gn+1 n 2
This indicates that in the limit of large horizon radius, the entropy is proportional to the area
of the cylinder with length 2/(n 2) if we look at it from the viewpoint of the bulk. This
coincides with the argument in [24].
(ii) For small rA , we can expand the hypergeometric function, and reach

n 2 rA
An2
+ O rA2 .
S=
(48)
4Gn
n1
Using the relation (22), we have, up to the first order,
S=
An2 n 2 rA
1 2An2 rA
=
.
4Gn n 1
4Gn+1 n 1
(49)
Note that the volume of the (n 2)-dimensional sphere with horizon radius is Vn2 =
An2 rA /(n 1), we find the entropy in the small apparent horizon limit
S=
2Vn2
.
4Gn+1
(50)
Namely, this entropy produces the (n + 1)-dimensional properties of theory, and obeys the area
formula of horizon in (n + 1) dimensions. The factor 2 is due to the Z2 symmetry of the bulk.
Therefore, in the small apparent horizon limit, the (n + 1)-dimensional effect will be remarkable.
The Vn2 is just the volume of the disk inside the sphere which has area An2 . The areas of
these two disks are negligible under the large horizon radius limit [24], while they are important
in the small radius limit.
(iii) The unified first law can be rewritten to be
e
dE An2 W dVn2 = An2 + W dVn2 .
(51)
By using the first, second Friedmann equations, (27) and (28), and the continuity equation (29),
we find
e
dE An2 W dVn2 = d(Vn2 ).
(52)
Here Vn2 is nothing, but the total energy of matter inside the apparent horizon. Thus we can
rewrite the unified first law as
m
d E = An2 + W dVn2 ,
(53)
143
where E = Vn2 . After projecting along the horizon using , Eq. (53) gives us with
m
m

d E , = T dS, + W dVn2 , .
(54)
This relation is nothing but the first law of thermodynamics associated with the apparent horizon [13], dE = T dS + W dVn2 , with identifying the inner energy E to be Vn2 , temperature
T to /2 , entropy S to the form given in (45), and the work density W = ( p)/2.
(iv) If the bulk Weyl tensor does not vanish, the thing becomes complicated. Without the
knowledge of bulk geometry, we cannot obtain an entropy expression of apparent horizon in
terms of the horizon radius.
4. Bulk geometry and apparent horizon extended into bulk
From the global point of view, the apparent horizon on the brane will extend into the bulk.
And then the entropy of the apparent horizon can be determined by the area formula in the bulk
(here we have assumed that the gravity is the Einstein one in the bulk). In this section, we will
directly calculate the area of the apparent horizon which extends into the bulk, and give the
entropy expression of apparent horizon from the bulk geometry. For this aim, we have to first
find out the bulk geometry. Assume that the metric in the bulk has the form
d s 2 = N 2 (t, z) dt 2 +
A2 (t, z) 2
2
dr + A2 (t, z)r 2 dn2
+ dz2 ,
1 kr 2
(55)
where A and N satisfy (the brane is supposed to locate at z = 0 with Z2 symmetry)

N(t, 0) = 1,
A(t, 0) = a(t),
N(t, z) = N (t, z),
A(t, z) = A(t, z).
Then we have the nonvanishing components of Einstein tensor

N A A
(n+1)
G0z = (n 1)
,
N A
A
2

2

1
2 A
A
N2
(n+1)
2 A
G00 = (n 1)(n 2) 2 N
+
+k 2 ,
2
n2 A
A
A2
A

2
1
2 A
N
A
(n+1)
Gzz = (n 1)(n 2)
+
2
2
n2 A N
A

2

2 AN
2 A
k
1 A
+
2 ,
2
n2A
N A2 n 2 A N
A

2
1
2 A
N
2 A
A
(n+1)
Gij = (n 2)(n 3)A2 ij
+
+
2
2
n3 A N
n3 A
A

2

2 N A
N
1
2 A
k
2
A
+ 2 2+
2 ,
+
(n 2)(n 3) N
n3N A n3A
N
A
A
(56)
(57)
(58)
(59)
(60)
where prime denotes the derivative with respect to z and overdot stands for the derivative with
respect to t . In the bulk, the equations of motion is just the Einstein equations with a cosmological
constant
(n+1)
2
G = n+1
n+1 g .
(61)
144
From (n+1) G0z = 0, we obtain

N
A
A
= .
N A
A
Defining
(62)

A 2
F (t, z) = An2 A
2 2 k ,
N
(63)
we have

2

A
An1 A 2
2 A
N2
2 A
N
+
+k 2 ,
F = (n 2)
n2 A
N2
A2
A2
A
2
A
A
n1
+
F = (n 2)AA
n2 A N
A2

2

2 A N
2 A
k
1 A
.
2
n2A
N A2 n 2 A N
A2
(64)
(65)
In terms of F , we can express the (00) and (zz) components of equations of motion as
n1 2
2A
An1 2
2AA
F =
n+1 n+1 ,
n+1 .
n1
n 1 n+1
Integrating the first one in (66), we get
F
=
F+
2
2n+1
n+1
n(n 1)
An + C = 0,
(66)
(67)
where C is a function of t. Integrating the second equation leads to the conclusion that C is a
constant. Thus we have
2

2
2 2 n+1
A
k
C
A
(68)
+ 2 = n+1
+ n.
+
NA
n(n 1)
A
A
A
We can calculate the Weyl tensor for the metric (55) and find that the nonvanishing components
of Weyl tensor have the form

2

1 A 2
k
A
A
A
N
1 A N
A
+ 2
+
+
+ 2
.
C 2
(69)
A
A
N
A N
A
N A
A
N AN
If we impose the constraint

2 2 n+1
A
A
N
1 A N
A
1
=
=
+ 2
= n+1
= 2,
A
N
A N
A
n(n 1)
N AN

(70)
we find that the condition with vanishing Weyl tensor exactly leads to C = 0 in Eq. (68). Substituting this constraint into the Einstein equations, we can see that the second-order differential
equation for the Einstein equations reduce to a first-order one in the case, which is just Eq. (68)
with vanishing C. Therefore to solve Eq. (68) becomes simple when the Weyl tensor vanishes.
Otherwise, one has to solve a second-order differential equation (nonlinear term will appear in
this second-order differential equation if n > 4).
Considering the Israel junction condition on the brane, one can give the equations of motion
on the brane (Friedmann equation) from (68). The equation (n+1) G0z = 0 also implies that we
145
have
A
= (t).
N
(71)
Substituting this equation into the (00) equation or (68) with vanishing C, we have
2 + k (A
)2
2n+1 2
A = 0.
n(n 1)
(72)
0) = a(t)
Since A(t,
and N (t, 0) = 1, we have (t) = a(t).
Eq. (72) has the solution

1

1 2
1 2
2
2|z| 2
2z
A(t, z) = a(t) 2 + 1 +
.
1 + 2 sinh
cosh
2 rA
2 rA2

rA
(73)
One can check that this solution satisfies the constraint (70). For n = 4, this solution is just the
one found in [19] with vanishing C.
The apparent horizon for fixed z has the form

rA |z =
2

1 A 2
k 2
a
k 2
+
=
+
,
N 2 A2 A2
A2 A2
1
(74)
namely, we have
rA |z = rA f (z, rA ),
(75)
with rA = rA |0 and

1

1 2
1 2
2
2|z| 2
2z
f (z, rA ) = 2 + 1 +
.
1 + 2 sinh
cosh
2 rA
2 rA2

rA
(76)
We show this function (the red line for = 1, rA = 0.5, green one for = 1, rA = 2, and blue one
for = 1, rA = 3) in Fig. 1. For any fixed rA , we find that this function has a zero at z = zmax .
The value zmax is determined by solving the equation f (z) = 0, which gives

rA
zmax (rA ) = arcsinh
(77)
.

This indicates that the apparent horizon extends into the bulk with a finite distance, zmax .
Once the bulk geometry is known, we can calculate the area of the apparent horizon using the
bulk geometry. Let us first consider the case with n = 4. Considering the Z2 symmetry, we have
the horizon area
A = 2 4 rA2
zmax

f 2 (z, rA ) dz.
(78)
Carrying out the integration, we arrive at

2
2
rA
2
.
A = 4 rA 1 + 2 2 arcsinh

rA rA
(79)
146
Fig. 1. The function f (z) has a zero at zmax . The red line is for the case with = 1, rA = 0.5, the green one for = 1,
rA = 2, and the blue one for = 1, rA = 3, respectively. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
According to the 5-dimensional area formula, we obtain the entropy associated with this 5dimensional apparent horizon

2
2
rA
A2
1 + 2 2 arcsinh
.
S=
(80)
4G5

rA rA
Using the relation (22), we can rewrite this entropy as

A2
2
2
rA
S=
1 + 2 2 arcsinh
.
4G4

rA rA
(81)
This is exactly the form (46) we have given in the previous section by using the method developed
in [6].
For the case with an arbitrary n (n > 4), the area of the apparent horizon in the (n + 1)dimensional bulk can be expressed as
zmax
(rA )
A = 2 n2 rAn2
f n2 (z, rA ) dz.
(82)
After some calculations, one can obtain

2
rA
2
n1 1 n+1
n2 rAn1 2 F1
, ,
,
A=
.
n1
2
2
2

By using the (n + 1)-dimensional area formula of horizon entropy, we have

2
An2 2
A
rA
rA
n1 1 n+1
=
S=
, ,
,
.
2 F1
4Gn+1 4Gn+1 n 1
2
2
2

Using the relation (22), once again, we arrive at

2

An2 n 2 rA
rA
n1 1 n+1
F1
S=
, ,
,
.
4Gn n 1 2
2
2
2

(83)
(84)
(85)
147
This is completely the same as the entropy expression (45) associated with the apparent horizon
on the brane.
Clearly, when the bulk Weyl tensor is nonvanished, one has to solve a nonlinear second-order
differential equation for n > 4, in order to give the bulk geometry. Solving this nonlinear equation
is not an easy matter, but it is worth trying. Once the bulk geometry is given, one can obtain the
entropy of apparent horizon by using the area formula in the bulk.
5. Conclusion and discussion
In this paper we have discussed thermodynamics of the apparent horizon of an n-dimensional
FRW universe in the RSII brane world scenario. The gravity theory on the brane is not the Einstein one due to the existence of extra dimension, therefore the well-known area formula for
horizon entropy is not applicable on the brane. By using the method we have developed in [6],
we have obtained an explicit entropy expression of the apparent horizon in the n-dimensional
FRW universe embedded in an (n + 1)-dimensional pure AdS spacetime. In this model, the
effective energymomentum tensor coming from the non-Einstein part is just the in the effective field equations on the brane. From this effective energymomentum tensor, we define an
effective energy-supply. By using the unified first law, together with the Clausius relation Q =
m
dA, A , = T dS (this also suggests that the associated thermodynamA , = 8G

ics with the apparent horizon is an equilibrium one), we have obtained an analytic expression (45)
of the entropy for the apparent horizon. The entropy expression has expected properties. In the
limit of large horizon radius, this entropy reduces to the n-dimensional area formula, where the
Einstein gravity on the brane is recovered. In this case, this entropy can be interpreted as the area
of the cylinder as described in [24]. On the other hand, in the limit of small horizon radius,
the entropy reduces to (n + 1)-dimensional area formula, as expected, and can be interpreted as
the area of two disks, which is negligible in the large radius limit. In addition, we have shown
that the Friedmann equation on the brane can be rewritten as a universal form like the first law,
dE = T dS + W dV .
Our entropy expression for the apparent horizon in FRW universe is useful in the study of
thermodynamics of black holes on the brane since we expect that the entropy associated with
apparent horizon and black hole horizon has the same expression.
We have also discussed the bulk geometry for an arbitrary dimension n 4, and given a
solution with vanishing Weyl tensor. We have found that the apparent horizon extends into the
bulk with a finite distance which are labelled by zmax (rA ) = arcsinh(rA /). Using the bulk
geometry and the area formula in the bulk, we have calculated the entropy associated with the
apparent horizon extended into the bulk. We have found that both methods give the completely
same expression for the apparent horizon entropy. It is of great interest to extend our method to
other brane world scenarios such as DGP model, models with a bulk GaussBonnet term, etc.
Acknowledgements
This research was initiated during R.G. Cais visit to the department of physics, Fudan university, the warm hospitality extended to him is appreciated. R.G. Cai thanks B. Wang and
R.K. Su for helpful discussions. This work was supported partially by grants from NSFC, China
(Nos. 10325525 and 90403029), and a grant from the Chinese Academy of Sciences.
148
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
S.A. Hayward, Phys. Rev. D 49 (1994) 6467.

S.A. Hayward, Phys. Rev. D 53 (1996) 1938, gr-qc/9408002.
S.A. Hayward, Class. Quantum Grav. 15 (1998) 3147, gr-qc/9710089.
S.A. Hayward, S. Mukohyama, M.C. Ashworth, Phys. Lett. A 256 (1999) 347, gr-qc/9810006.
C.W. Misner, D.H. Sharp, Phys. Rev. 136 (1964) B571.
R.G. Cai, L.M. Cao, Phys. Rev. D 75 (2007) 064008, gr-qc/0611071.
R.G. Cai, S.P. Kim, JHEP 0502 (2005) 050, hep-th/0501055.
M. Akbar, R.G. Cai, Phys. Lett. B 635 (2006) 7, hep-th/0602156.
A.V. Frolov, L. Kofman, JCAP 0305 (2003) 009, hep-th/0212327.
U.K. Danielsson, Phys. Rev. D 71 (2005) 023516, hep-th/0411172.
R. Bousso, Phys. Rev. D 71 (2005) 064024, hep-th/0412197.
G. Calcagni, JHEP 0509 (2005) 060, hep-th/0507125.
M. Akbar, R.G. Cai, Phys. Rev. D 75 (2007) 084003, hep-th/0609128.
M. Akbar, R.G. Cai, Phys. Lett. B 648 (2007) 243, gr-qc/0612089.
T. Padmanabhan, Class. Quantum Grav. 19 (2002) 5387, gr-qc/0204019;
T. Padmanabhan, Phys. Rep. 406 (2005) 49, gr-qc/0311036;
T. Padmanabhan, gr-qc/0606061.
A. Paranjape, S. Sarkar, T. Padmanabhan, Phys. Rev. D 74 (2006) 104015, hep-th/0607240.
T. Jacobson, Phys. Rev. Lett. 75 (1995) 1260.
C. Eling, R. Guedens, T. Jacobson, Phys. Rev. Lett. 96 (2006) 121301, gr-qc/0602001.
P. Binetruy, C. Deffayet, U. Ellwanger, D. Langlois, hep-th/9910219.
N. Arkani-Hamed, S. Dimopoulos, G.R. Dvali, Phys. Lett. B 429 (1998) 263, hep-ph/9803315;
I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, G.R. Dvali, Phys. Lett. B 436 (1998) 257, hep-ph/9804398.
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hep-ph/9905221.
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690, hep-th/9906064.
T. Shiromizu, K.i. Maeda, M. Sasaki, Phys. Rev. D 62 (2000) 024012, gr-qc/9910076;
A.N. Aliev, A.E. Gumrukcuoglu, Class. Quantum Grav. 21 (2004) 5081, hep-th/0407095.
R. Emparan, G.T. Horowitz, R.C. Myers, JHEP 0001 (2000) 007, hep-th/9911043.
D. Bak, S.J. Rey, Class. Quantum Grav. 17 (2000) L83, hep-th/9902173.
Supersymmetric Standard Model

from the heterotic string (II)
Wilfried Buchmller a , Koichi Hamaguchi a,b, , Oleg Lebedev c ,
Michael Ratz c
a Deutsches Elektronen-Synchrotron DESY, 22603 Hamburg, Germany
b Department of Physics, University of Tokyo, Tokyo 113-0033, Japan
c Physikalisches Institut der Universitt Bonn, Nussallee 12, 53115 Bonn, Germany
Received 13 September 2006; received in revised form 19 June 2007; accepted 22 June 2007
Available online 17 July 2007
Abstract
We describe in detail a Z6 orbifold compactification of the heterotic E8 E8 string which leads to the
(supersymmetric) Standard Model gauge group and matter content. The quarks and leptons appear as three
16-plets of SO(10), two of which are localized at fixed points with local SO(10) symmetry. The model has
supersymmetric vacua without exotics at low energies and is consistent with gauge coupling unification.
Supersymmetry can be broken via gaugino condensation in the hidden sector. The model has large vacuum
degeneracy. Certain vacua with approximate BL symmetry have attractive phenomenological features.
The top quark Yukawa coupling arises from gauge interactions and is of the order of the gauge couplings.
The other Yukawa couplings are suppressed by powers of Standard Model singlet fields, similarly to the
FroggattNielsen mechanism.
1. Introduction and summary

The Standard Model is a remarkably successful theory of the structure of matter. It is a chiral
gauge theory with the gauge group GSM = SU(3)c SU(2)L U(1)Y and three generations
of quarks and leptons. All masses are generated by the Higgs mechanism which involves an
SU(2) doublet of scalar fields. Its unequivocal prediction is the existence of the Higgs boson
* Corresponding author at: Deutsches Elektronen-Synchrotron DESY, 22603 Hamburg, Germany.
E-mail address: koichi.hamaguchi@desy.de (K. Hamaguchi).

doi:10.1016/j.nuclphysb.2007.06.028
150
W. Buchmller et al. / Nuclear Physics B 785 (2007) 149209
which still remains to be discovered. From a theoretical perspective, the minimal supersymmetric
extension of the Standard Model, the MSSM, is particularly attractive. Apart from stabilizing the
hierarchy between the electroweak and Planck scales and providing a natural explanation of
the observed dark matter, it predicts unification of the gauge couplings at the unification scale
MGUT 2 1016 GeV.
Even more than the unification of gauge couplings, the symmetries and the particle content
of the Standard Model point towards grand unified theories (GUTs) [1,2]. Remarkably, one generation of matter, including the right-handed neutrino, forms a single spinor representation of
SO(10) [3,4]. It therefore appears natural to assume an underlying SO(10) structure of the theory.
The route of unification, continuing via exceptional groups, terminates at E8 , which is beautifully
realized in the heterotic string [5,6].
An obstacle on the path towards unification are the Higgs fields, which are SU(2)L doublets,
while the smallest SO(10) representation containing the Higgs doublets, the 10-plet, predicts
additional SU(3)c triplets. The fact that Higgs fields form incomplete split GUT representations
is particularly puzzling in supersymmetric theories where both matter and Higgs fields are chiral
multiplets. The triplets cannot have masses below MGUT since otherwise proton decay would be
too rapid. This then raises the question why SU(2)L doublets are so much lighter than SU(3)c
triplets. This is the notorious doublettriplet splitting problem of ordinary 4D GUTs.
Higher-dimensional theories offer new possibilities for gauge symmetry breaking connected
with compactification to four dimensions. A simple and elegant scheme, leading to chiral fermions in four dimensions, is the compactification on orbifolds, first considered for the heterotic
string [713], and more recently applied to GUT field theories [1419]. Such orbifold GUTs
appear as intermediate effective field theories in compactifications of the heterotic string when
some of the compact dimensions are of order 1/MGUT and therefore large compared to the string
length [2023].
In orbifold compactifications, gauge symmetry of the 4D effective theory is an intersection
of larger symmetries at orbifold fixed points. Massless modes located at these fixed points all
appear in the 4D theory and form representations of the larger local symmetry groups. Zero
modes of bulk fields, on the contrary, are only representations of the smaller 4D gauge symmetry
and form in general split multiplets. When the local symmetry at some orbifold fixed points is
a GUT symmetry, one obtains the picture of local grand unification. The SM gauge group can
be thought of as an intersection of different local GUT groups. Matter fields appear as complete
GUT representations localized at the fixed points, whereas the Higgs doublets are associated
with bulk fields, and therefore split multiplets. In this way the structure of the Standard Model is
naturally reproduced [2325].
Recently, we have obtained the gauge group and matter content of the supersymmetric Standard Model from the heterotic string by using the picture of local grand unification as the guiding
principle [26]. Quarks and leptons appear as three 16-plets of SO(10), two of which are localized
at orbifold fixed points with local SO(10) symmetry. For generic vacua, no exotic states appear
at low energies and the model is consistent with gauge coupling unification. In this paper we
describe our construction in detail.
It is well known that the number of possible string vacua is huge. Early estimates of the total
number of different vacua of the heterotic string gave numbers like 101500 [27], which came as
a complete surprise. More recent studies, based on flux compactifications, give similarly large
numbers [28]. Searches for Standard Model-like vacua have been based on orbifold compactifications [29,30], the free fermionic formulation [3133], intersecting D-brane models [34] and
Gepner orientifolds [35]. Despite the huge number of vacua, it turned out to be extremely dif-
151
ficult to construct a consistent ultraviolet completion of the (supersymmetric) Standard Model,

and only recently several examples have been obtained [26,36,37].1 This suggests that not all
field theories can be embedded into string theory and that a consistent ultraviolet completion of
the Standard Model may eventually lead to some testable low energy predictions.
In this paper, the model presented in [26] is described in detail. We hope that this will be useful
for further phenomenological studies of the model and also for the search for other embeddings
of the Standard Model into the heterotic string. In order to keep the paper self-contained, we
recall the basics of strings on orbifolds in Sections 24. In Section 2, the boundary conditions
for untwisted and twisted strings, the mode expansion and the massless spectrum are discussed;
furthermore, a simple derivation of the projection conditions for physical states is given. Our
orbifold model is based on the 6D torus defined by the G2 SU(3) SO(4) root lattice, which
has a Z6-II = Z3 Z2 discrete symmetry. The geometry is described in Section 3 with emphasis
on the localization of twisted states. In Section 4, the string selection rules for superpotential
couplings of the Z6-II orbifold are reviewed and somewhat extended.
The main results of this paper are contained in Sections 58 and in Appendices AD. After
describing our search strategy for compactifications with local SO(10) symmetry, we study the
unbroken gauge group G and the massless spectrum of the model in Section 5. We also list the
GUT representations at various fixed points and the 6D orbifold GUTs which one obtains for two
compact dimensions of size 1/MGUT . The FayetIliopoulos (FI) D-term of an anomalous U(1)
triggers further symmetry breaking [39]. In particular,
G SU(3)c SU(2)L U(1)Y Ghidden ,
(1.1)
with Ghidden = SU(4) SU(2) is possible, in which case the model has a truly hidden sector
admitting spontaneous SUSY breaking. We further show that, for generic vacua, unwanted exotic
states attain large masses and decouple. This is one of the central results of our paper.
The decoupling of exotic states can be achieved without breaking supersymmetry. In Section 6, we discuss D- and F -flat directions in the field space as well as general supersymmetric
field configurations, neglecting supergravity corrections. The model naturally accommodates
spontaneous supersymmetry breaking via hidden sector gaugino condensation, which is described in Section 7.
In generic F - and D-flat configurations, our model yields the MSSM spectrum, however the
analysis of proton decay and flavour becomes intractable. In Section 8, we therefore identify a
simple and phenomenologically attractive D-flat field configuration, without proving F -flatness,
which preserves

GSM U(1)BL SU(4) .
(1.2)
Here we keep the hidden sector SU(4) unbroken which is needed for gaugino condensation. We
show that unwanted exotics can be decoupled in this case as well. Further, we identify two Higgs
doublets and discuss the pattern of Yukawa couplings. The top quark Yukawa coupling arises
from gauge interactions and is of the order of the gauge couplings. Other Yukawa couplings
are suppressed by powers of Standard Model singlet fields, similarly to the FroggattNielsen
mechanism [40].
Finally, in Section 9, we conclude with a brief outlook on open questions and further challenges for realistic compactifications of the heterotic string.
1 In intersecting brane constructions, a model containing the spectrum of the MSSM plus vector-like exotics and
additional U(1) factors was obtained in [38].
152
2. Strings on orbifolds
In the following subsections we collect the basic notions and formulae which are needed to
describe propagation of the E8 E8 heterotic string on orbifolds T6 /ZN [7,8]. We follow the
definitions of Katsuki et al. [41].
2.1. Lattices and twists
The torus is obtained as the quotient T6 = R6 /2, where is the lattice of a semi-simple
Lie algebra of rank 6 with a ZN discrete symmetry. The 6 compact coordinates of the torus x i ,
i = 4, . . . , 9, are conveniently combined into 3 complex coordinates zi = 1 (x 2i+2 + ix 2i+3 ),
i = 1, . . . , 3. Points in R6 differing by a lattice vector,
z z + 2,
(2.1)
with = ma ea , ma Z (a = 1, . . . , 6), are identified. Here ea denote the basis vectors in the
three planes of the lattice.
The lattice has a ZN discrete symmetry which acts crystallographically, i.e., it maps the lattice
onto itself,
i
z z,
ji = e2ivN ji ,
N = 1,
i
NvN
= 0 mod 1.
i, j = 1, . . . , 3,
(2.2)
with
(2.3)
Here we assume the factorization T6 = T2 T2 T2 . N = 1 supersymmetry in 4D requires that

the ZN twist be contained in the SU(3) subgroup of SO(6), i.e.,

i
(2.4)
vN
= 0 mod 1.
i
Lattice translations and twists k (k = 0, . . . , N 1) form the space group S whose elements
are denoted by ( k , ). The orbifold T6 /ZN can also be defined as the quotient R6 /S, where

z k , z k z + 2.
(2.5)
The multiplication rule in the space group is given by
k

1 , 1 k2 , 2 = k1 k2 , k1 2 + 1 .
(2.6)
An orbifold has fixed points f , which are invariant under the action of a space group element
( k , ),

f = k , f = k f + 2, = ma ea , ma Z.
(2.7)
Here k and depend on the fixed point f . Since the position of the fixed point is defined only up
to a lattice vector, is defined up to a translation in the sublattice

k 1 k = | = 1 k , .
(2.8)
Each fixed point ( k , ) is associated with a sublattice = + k , and there are as many sublattices as fixed points. The dimension of a sublattice k can be smaller than dim = 6 if (1 k )
has eigenvectors with eigenvalue 0. In this case the element ( k , ) describes fixed planes.
153
2.2. Untwisted and twisted strings

In the light-cone gauge the heterotic string can be described by the following world-sheet
fields [42]: 8 string coordinates and 8 right-moving NeveuSchwarzRamond fermions ( =
),
X i (, ) = XLi (+ ) + XRi ( ),
i ( ),
i = 2, . . . , 9,
(2.9)
and 32 left-moving fermions I ,

I (+ ),
I = 1, . . . , 32.
(2.10)
Here i is the spacetime index, while index I is associated with E8 E8 gauge degrees of
freedom. It is convenient to combine the string coordinates in the compact dimensions into 3
complex variables Z i and, similarly, the right moving fermions into 3 complex NSR fermions
i ,

1
Z i = X 2i+2 + iX 2i+3 ,
2

1
i = 2i+2 + i 2i+3 ,
2
(2.11)
where i = 1, . . . , 3. The ZN twist acts on these fields as
Z Z,
(2.12)
Closed strings on ZN orbifolds can be untwisted or twisted. In the former case the string is closed
already on the torus and has the boundary conditions,
Z( + 2) = Z( ) + 2ma ea ,
+ 2) = (
),
(
ma Z,
(2.13)
(2.14)
whereas in the latter case the string is closed on the orbifold but not on the torus and has the
boundary conditions (k = 1, . . . , N 1),
Z( + 2) = k Z( ) + 2ma ea ,
+ 2) = k (
),
(
(2.15)
(2.16)
where k and ma depend on the fixed point f . The lattice translation in Eq. (2.15) enters the
space group element associated with the fixed point, Eq. (2.7). The plus and minus signs in
Eqs. (2.14) and (2.16) correspond to the Ramond and the NeveuSchwarz sectors, respectively.
Twisted strings are localized at the orbifold fixed points, whereas untwisted strings can propagate
freely on the orbifold (Fig. 1).
Modular invariance usually requires that the ZN SO(6) twist of the spacetime degrees of
freedom be accompanied by a ZN E8 E8 twist of the fermions I , representing the internal
symmetry group. On the complex fermions

1
I = 2I 1 + i2I ,
2
I = 1, . . . , 16,
(2.17)
the ZN twist acts as

,
JI = e2iVN JI ,
(2.18)
154
Fig. 1. Twisted and untwisted strings. The dots denote orbifold fixed points.
where
N = 1,
8
VNI = N
I =1
16
VNI = 0 mod 2,
(2.19)
I =9
with integer N VNI . The fermions I can have untwisted (k = 0) or twisted (k = 1, . . . , N 1)

boundary conditions,
).
+ 2) = k (
(
(2.20)
This makes the parallel between and transparent. Extending vN by a zero entry acting on
1 , v 2 , v 3 ; 0), we note that vectors N v and N V lie
the uncompactified dimensions, vN (vN
N
N
N N
on the root lattices SO(8) and SO(16)SO(16) , respectively. In an orthonormal basis, SO(2N ) is

defined by vectors (n1 , . . . , nN ) with integer ni and N
i=1 ni = 0 mod 2. One can show that the
gauge symmetry of this theory is E8 E8 which contains SO(16) SO(16) as a subgroup [6].
A convenient formulation of the heterotic string is obtained by representing fermionic degrees
of freedom in terms of bosons. In this case one replaces the 8 right-moving and 32 left-moving
fermions with 4 right-moving and 16 left-moving bosons,
i
i ( ) = e2iH ( ) ,
I (+ ) = e
2iX I (
+)
i = 1, . . . , 4,
(2.21)
I = 1, . . . , 16.
(2.22)
are compactified on a 16-dimensional torus represented by the E8 E8 root lattice,

1
1
n 1 + , . . . , n8 +
,
E8 : p = (n1 , . . . , n8 ) or
(2.23)
2
2

where ni integer with 8i=1 ni = 0 mod 2, and similarly for the second E8 . This gives rise to
gauge multiplets of the E8 E8 group in 10 dimensions, coupled to supergravity.
Compactifying the extra 6 dimensions on an orbifold amounts to modding the string coordinates by the space group and its gauge counterpart. The latter is obtained by embedding the
twists and lattice shifts into gauge degrees of freedom X I as

k
I
,
, ma ea 1, kVNI + ma Wna
(2.24)
The fields X I
I denotes a Wilson line of order n. Here N V and nW (n N ) are required to lie on

where Wna
N
n
the E8 E8 root lattice.2 Thus, a twist of the spacetime degrees of freedom is accompanied by
2 This generalizes V of Eq. (2.18) in which case N V lies on the SO(16) SO(16) root lattice.
N
N
155
a shift kVN of the gauge coordinates, while a torus lattice translation is accompanied by a gauge
coordinate shift ma Wna . This corresponds to generalizing the boundary condition (2.20) for the
left-moving fermions to
I
I
I ( + 2) = e2i(kVN +ma Wna ) I ( ).
(2.25)
The bosonic field boundary conditions then read (k = 0, . . . , N 1)

i
mod SO(8) ,
H i ( + 2) = H i ( ) kvN

I
X I ( + 2) = X I ( ) + kVNI + ma Wna
mod E8 E8 .
(2.26a)
(2.26b)
Here SO(8) denotes the weight lattice of SO(8) given in the orthonormal basis by
SO(8) :
q = (n1 , n2 , n3 , n4 ),
(2.27)

where ni integer with i ni odd or ni half-integer with i ni even.
To summarize, the heterotic string can be described by the left moving bosonic fields ZLi (+ ),
i
ZL (+ ), X I (+ ) and the right moving bosonic fields ZRi ( ), ZRi ( ), H i ( ). They fall into
untwisted or twisted categories depending on whether they represent strings closed on a torus or
on an orbifold only.
2.3. Modular invariance and local twists
The gauge shift VN and the Wilson lines Wn are subject to consistency conditions. First of all,
NVN and nWn are vectors of the E8 E8 root lattice,
NVN E8 E8 ,
nWn E8 E8 .
(2.28)
Second, modular invariance of the theory requires that they satisfy additional constraints (see
e.g., [21]):

2
= 0 mod 2,
N VN2 vN
(2.29a)
NVN Wn = 0 mod 1,
(2.29b)
NWn Wm = 0 mod 1 (Wn = Wm ),
(2.29c)
NWn2
= 0 mod 2.
(2.29d)
By adding E8 E8 root lattice vectors to VN and Wn satisfying these conditions, one can bring
VN ,Wn to the form which obeys a stronger constraint,3

1 2
2
= 0 mod 1,
VN v N
2
VN Wn = 0 mod 1,
Wn Wm = 0 mod 1 (Wn = Wm ),
1 2
W = 0 mod 1.
2 n
3 There are exceptions to this statement, for instance, when V = 0.
(2.30a)
(2.30b)
(2.30c)
(2.30d)
156
This form has the advantage that the analysis of physical states of the theory simplifies significantly. These equations can also be written as

1
2
(2.31)
= 0 mod 1, r = 0, 1,
(rVN + ma Wna )2 r 2 vN
2
where 0 ma n 1 for a Wilson line Wn of order n.
The twist can be thought of as a local quantity, that is, depending on the fixed point and the
twisted sector. Indeed, Eqs. (2.25) and (2.26b) show that what matters at a particular fixed point
f is the combination
I
VfI = kVNI + ma Wna
,
(2.32)
which plays the role of the local gauge twist, as well as its right-moving counterpart kvN . Each
local twist Vf can be expressed as the sum of the twist kVN for vanishing Wilson lines and a
linear combination of Wilson lines determined by the location of the fixed point f . The local
twists satisfy modular invariance conditions (2.31) and can be treated on the same footing as VN .
This observation will be important for the concept of local GUTs.
2.4. Mode expansion and massless spectrum
The boundary conditions discussed in Section 2.2 lead to the following mode expansion for
the untwisted string (i = 1, . . . , 3),
1
i 1 i in i 1 i in+
Z i (, ) = zi + p i + ma eai +
+
,
e
e
2
2 n n n
2 n n n
1
i 1 i in i 1 i in+
Z i (, ) = zi + p i + ma eai +
+
.
e
e
2
2 n n n
2 n n n
(2.33a)
(2.33b)
Twisted strings have the expansion (cf. Eq. (2.15))

Z i (, ) = f i +
i
2
Z i (, ) = f i +

i
nZ+kvN
i
2
1 i in i
+
e
n fn
2
i
nZkvN

i
nZkvN
1 i in i
+
e
n fn
2
1 i in+
,
e
n fn
i
nZ+kvN
1 i in+
.
e
n fn
(2.34)
(2.35)
In this case, there is no center-of-mass string motion, i.e., p i = ma = 0. If there is a fixed plane,
the boundary conditions for strings in the fixed plane are untwisted and the expansion is given
by Eqs. (2.33).
The bosonized NSR fermions have the expansion (i = 1, . . . , 4)
H i (, ) = hi +

1 i
i 1 i in
i
+
,
q + kvN
e
2
2
n n
(2.36)
n=0
while the gauge coordinates are given by (I = 1, . . . , 16)

X I (, ) = x I +

1 I I
i 1 I in+
I
.
+ +
p + kVN + ma Wna
e
2
2
n fn
n=0
(2.37)
157
The momentum vectors q i and p I specify the Lorentz and gauge quantum numbers of the string
states. Note that the creation and annihilation operators of the twisted string (2.34), (2.35) and
the left-moving string (2.37) depend on the fixed point f .
States of the heterotic string are given by a direct product of the right- and left-moving parts.
A basis in the Hilbert space of the quantised string is obtained by acting with the creation operators fi n , fi n , fi n , fI n (n < 0) on the ground states of the untwisted sector U (k = 0) and the
twisted sectors Tk (k = 1, . . . , N 1). Massless states in the untwisted sector as well as twisted
states living on fixed planes have pi = ma = 0. The ground states of the different sectors depend on the momentum vectors q i , p I and, for the twisted sectors, also on the fixed point f
(cf. (2.32)),
|q, p |q |p ,
|f ; q, p |q + kvN |p + Vf .
(2.38)
It turns out that for the model discussed below only oscillator modes of the left-moving
strings ZLi (+ ), ZLi (+ ) and X I (+ ) are relevant. The corresponding twisted sector states are
(ni , mi < 0)
j
fi1n1 fi2n2 f m1 1 f m2 2 fIl11 fIl22 |f ; q, p .
(2.39)
Massless states of the untwisted sector satisfy the following mass equations:
1 2 1 2 1
m = q + N + N = 0,
8 R 2
2
1 2 1 2
m = p 1 + N + N = 0,
8 L 2
(2.40a)
(2.40b)
where N, N , N , N are the integer oscillator numbers. Twisted massless states obey
1
1 2 1
(k)
(k)
mR = (q + kvN )2 + c(k) + i Nf i + i Nf i = 0,
8
2
2
1 2 1
(k)
(k)
m = (p + Vf )2 1 + c(k) + i N f i + i N f i = 0,
8 L 2
(2.41a)
(2.41b)
where
c(k) =
1 (k)
(k)
i 1 i ,
2
(2.42)
(k)
(k)
(k)
(k)
with i = (kvN )i mod 1, so that 0 < i 1, and i = (kvN )i mod 1 so that 0 < i 1.
(k)
(k)
This implies that i = i = 1 for (kvN )i integer. In Eq. (2.41), Nf i , Nf i , N f i and N f i N
represent the oscillator numbers of the right- and left-movers in zi and z i directions, respectively.
Note that Nf i and Nf i , as well as N f i and N f i , denote independent quantities. They are the
eigenvalues of the corresponding number operators N f i ,
1 i
f n fi n ,
N f i = (k)
i n>0
and analogously for Nf i , N f i , N f i . The sum
in the literature.
(2.43)

(k)
i (ki Nf i
(k)
+ ki
Nf i ) is often referred to as N
158
2.5. Projection conditions for physical states

As discussed in Section 2.2, an orbifold is obtained by identifying points in flat space which
transform into each other under the action of the space group,
x gx,
x R6 , g S.
(2.44)
Quantized strings whose boundary conditions are related by a symmetry transformation must
lead to the same Hilbert space of physical states. In particular, strings with the boundary conditions
( + 2) g( ) and
( + 2) hgh1 ( )
(2.45)
produce the same Hilbert space for any h S [8]. Here stands for Z i , Z i , H i and X I . For
each conjugacy class consisting of elements hgh1 one therefore has a separate Hilbert space.
h 1 = g, leave the string boundary
Space group elements h which commute with g, i.e. hg
conditions invariant. Hence, their representation in the Hilbert space must act as the identity on
physical states,
h|phys
= |phys .
(2.46)
This is the invariance or projection condition for physical states.

A space group element h = ( k , )
acts as a translation on the center-of-mass coordinates of
the bosonic fields H i and X I (cf. (2.26)),

I
I
N +m
.
x I x I + kV
a Wna
i ,
hi hi kv
N
(2.47)
Hence, the momentum eigenstates in twisted sectors transform as
|f ; q, p e2i(kvN (q+kvN )+(kVN +m a Wna )(p+Vf )) |f ; q, p ,
(2.48)
and similarly for untwisted states. From Eqs. (2.34) and (2.35) one reads off the transformation
properties of the creation operators,
fi n e2ikvN fi n ,
fin e2ikvN fin .

i
(2.49)
A state with non-vanishing oscillator numbers then transforms as

fi n fim |f ; q, p
)kv
N (q+kvN )+(kV
N +m
a Wna )(p+Vf )) i
f n fim |f ; q, p .
e2i(kvN (NN
(2.50)
Physical states have to satisfy Eq. (2.46), which yields the projection conditions

N (q + kvN ) + (kV
N +m
N N f N f kv
a Wna ) (p + Vf ) = 0 mod 1,
kv
(2.51)
for values of k and m

a which depend on the conjugacy class. As we will discuss in Section 2.5.2,
in non-prime orbifolds Eq. (2.51) gets modified for higher twisted sector states. Below we analyze in detail the projection conditions for the untwisted and twisted sectors.
159
2.5.1. Untwisted sector

The untwisted sector (k = 0) is associated with the space group element g = (1, 0), and
Eq. (2.51) has to be satisfied for the full space group, i.e., for all values k and m
a . This yields the
projection conditions
vN q VN p = 0 mod 1,
Wn p = 0 mod 1,
(2.52)
= 2) and q is the SO(8) weight lattice mowhere p is the E8 E8 root lattice momentum
mentum (q 2 = 1). The E8 momenta lie on the same lattice as the E8 coordinates because of
self-duality.
The untwisted sector contains gauge and matter supermultiplets of the 4D effective theory.
For the former vN q = 0 mod 1 yielding gauge bosons with q = (03 ; 1) and gauginos with
q = ( 12 , 12 , 12 ; 12 ). For the matter multiplets, vN q = n/N mod 1 with n = 1, . . . , N 1 leading
to the bosonic SO(8) momenta ( 1, 0, 0; 0) where the underline denotes permutations, and their
fermionic partners.
Since gauge multiplets satisfy vN q = 0 mod 1, the conditions
(p 2
VN p = 0 mod 1,
Wn p = 0 mod 1,
(2.53)
determine the roots p of the unbroken 4D gauge group. It is instructive to rewrite this set of
equations as
Vf p = 0 mod 1,
for all fixed points f,
(2.54)
where Vf is the local shift (2.32) associated with the fixed point. At each fixed point the gauge
group is broken locally to a subgroup of E8 E8 . The states surviving all local projection conditions, i.e., those corresponding to the intersection of all local gauge groups, yield the gauge fields
of the low-energy gauge group.
Matter multiplets (vN q = 0 mod 1) originate from the 10D gauge fields polarized in the
compact directions and their fermionic partners. They form chiral superfields transforming as
the coset of E8 E8 and the unbroken 4D gauge group. All untwisted states are bulk fields in the
compactified dimensions.
2.5.2. Twisted sectors
For the twisted sectors Tk (k = 1, . . . , N 1), the projection conditions depend on k. Consider
k = 1 and a fixed point f with the space group element g = (, ). The space group elements

commuting with g are h = ( k , )
= (, )n , n N. The resulting projection condition is

vN N f N f vN (q + vN ) + Vf (p + Vf ) = 0 mod 1,
(2.55)
where Vf = VN + ma Wna . Using strong modular invariance (2.30), one can show that all massless states (cf. (2.41)) satisfy this condition. Therefore all massless modes in the first twisted
sector correspond to physical states. In the case of prime orbifolds, Eq. (2.55) also holds for
higher twisted sectors with Vf = kVN + ma Wna .
For non-prime orbifolds the situation is more complicated. Some of the higher twisted sectors
Tk , k > 1, are related to lower order twists ZN/k which leave one of the T2 tori invariant. This
results in additional projection conditions. Furthermore, fixed points of the lower order twists are
not necessarily fixed points of the original twist ZN . The ZN twist transforms these fixed points
into each other such that they are mapped into the same singular point in the fundamental domain
of the orbifold. Physical states correspond to linear combinations of the states appearing at the
fixed points of the ZN/k twist.
160
The conjugacy classes of higher twisted sectors Tk are given by hgh1 where both g and h
have the form ( k , ). The number of the conjugacy classes is the number of the fixed points of
the lower order twist ZN/k . In general, twists of other orders ZN/k transform these classes into
each other. In particular, the ZN twist acts on the ZN/k conjugacy classes gi as
1 h 1 = g2 ,
hg
2 h 1 = g3 ,
hg
...,
n h 1 = g1 ,
hg
(2.56)
with h of the form (, ) and n 1. In this case, the higher twisted states transform as
= |2 ,
h|1
= |3 ,
h|2
= |1 .
h|n
...,
(2.57)
From linear combinations of these localized states one obtains a basis of physical states which
[8,43,44],
are ZN twist and h-eigenstates
1 2isq
|phys, q =
e
|s ,
n
n
(2.58)
s=1
where q = 0, 1/n, 2/n, . . . , 1. As a consequence,
h|phys,
q = e2iq e2i(kvN (Nf Nf )kvN (q+kvN )+(kVN +m a Wa )(p+Vf )) |phys, q , (2.59)
where we have used Eq. (2.50) and h = ( k , m

a ea ) is assumed to mix the conjugacy classes of
Tk as above. This leads to the modified projection conditions for the superpositions (2.58):

N (q + kvN ) + (kV
N +m
N N f N f kv
a Wna ) (p + Vf ) + q = 0 mod 1.
kv
(2.60)
In this paper we are especially interested in a Z6-II orbifold which has Z3 and Z2 subtwists
with invariant tori. The corresponding twist vector is v6 = (1/6, 1/3, 1/2). As we shall discuss in detail in Section 3, two different fixed points in the T2,4 twisted sectors are related by
Eq. (2.56) with h = ( 3 , 0). The eigenstates of ( 3 , 0) are

1
|phys, = |1 |2 ,
2
(2.61)
where the states |1 , |2 correspond to the two fixed points of 2 away from the origin. The
projection condition (2.60) becomes

3v6 N f N f 3v6 (q + kv6 ) + 3V6 (p + Vf ) + q = 0 mod 1,
(2.62)
with q = 1/2, 1 for k = 2, 4. Here Vf = kV6 + m3 W3 is the local Z3 gauge shift. Physical states
of T2,4 must also satisfy additional projection conditions which stem from invariance of the third
T2 torus (the SO(4) torus) under 2 . Clearly, translations 3 in this torus commute with 2 .
Thus invariance under space group transformations (1, 3 ) requires
W2 (p + Vf ) = 0 mod 1,
(2.63a)
W2
(2.63b)
(p + Vf ) = 0 mod 1,
W2
are two discrete Wilson lines in the SO(4) torus.

where W2 and
The Z6-II orbifold also has a Z2 subtwist. The fixed points of the T3 twisted sector are mapped
into each other by the space group element h = ( 2 , 0). Invariance under 2 leads to the projection condition

2v6 N f N f 2v6 (q + 3v6 ) + 2V6 (p + Vf ) + q = 0 mod 1,
(2.64)
161
with q = 1/3, 2/3, 1 and the local Z2 gauge shift Vf = 3V6 +n2 W2 +n2 W2 . The Z2 twist leaves
the second torus (the SU(3) torus) invariant. Invariance of the T3 states under translations in
this torus requires
W3 (p + Vf ) = 0 mod 1.
(2.65)
Here W3 is a discrete Wilson line in the SU(3) torus.

T5 twisted sector contains anti-particles of the T1 sector, and will not be treated separately in
the following.
The above projection conditions are relevant to our model. A sample calculation of the physical spectrum is given in Appendix A.
2.6. Local GUTs
I .
Consider a fixed point f which is associated with the local gauge shift VfI = kVNI + ma Wna
A local GUT can be defined by the E8 E8 roots p (p 2 = 2) satisfying
p Vf = 0 mod 1.
(2.66)
These roots represent a local gauge symmetry supported at the fixed point. Twisted matter appears in a representation of the local GUT. Each representation is characterized by the square of
the shifted momentum, (p + Vf )2 , which is the same for members of the same multiplet.
The concept of local GUTs is important for construction of realistic models. In particular, all
massless states of the T1 sector survive the ZN projection and represent physical states. They
form complete multiplets of the corresponding local GUT, although this GUT does not appear in
4D. As discussed in Section 1, this may naturally explain why the SM gauge (and Higgs) bosons
do not form complete GUT multiplets, while the matter fields do.
Let us illustrate how a local SO(10) structure arises. Consider a Z6-II heterotic orbifold based
on the Lie lattice G2 SU(3) SO(4) with v6 = (1/6, 1/3, 1/2; 0), the gauge shift

1 1 1
1
V6 =
(2.67)
, , , 0, 0, 0, 0, 0
, 0, 0, 0, 0, 0, 0, 0 ,
2 2 3
3
and arbitrary Wilson lines.
The local gauge shift at the origin in the T1 sector is Vf = V6 . The local GUT roots are found
from
p V6 = 0 mod 1.
(2.68)
This corresponds to SO(10) SU(2)2 symmetry in the observable sector. The SO(10) roots are
given by
p = (0, 0, 0, 1, 1, 0, 0, 0 ),
(2.69)
where the underline denotes all possible permutations of the corresponding entries.
Relevant twisted matter fields of the T1 sector satisfy the masslessness condition4
(p + V6 )2obs =
23
18
(2.70)
4 The invariance conditions (2.55) are satisfied automatically once V is brought to the strong modular invariant form
6
(5.3).
162
for the p + V6 components in the first E8 . The solution is

1
1 5
(p + V6 )obs = 0, 0, , odd
,
6
2
(2.71)
where odd (1/2)5 denotes all combinations containing an odd number of minus signs. This
is a 16-plet of SO(10). The Z6 invariant states have the right-mover shifted momentum q + v6 =
(1/6, 1/3, 1/2; 0) for spacetime bosons and analogously for spacetime fermions. All of
these states appear in the physical spectrum of the model.
The Wilson lines can be chosen such that the gauge group in 4D is that of the Standard Model
(times extra factors). This does not affect the above considerations and the local SO(10) GUT
structure remains intact.
3. Geometry of the Z 6-II orbifold
In this section we describe geometrical features of the Z6-II = Z3 Z2 orbifold based on the
G2 SU(3) SO(4) Lie algebra lattice, which is required for construction of our model.
3.1. Fixed points and fundamental region
The Z6-II orbifold with the G2 SU(3) SO(4) lattice is based on the twist vector5
1
v6 = (1, 2, 3; 0).
(3.1)
6
This orbifold allows for one discrete Wilson line of degree 3 in the SU(3) plane and two Wilson
lines of degree 2 in the SO(4) plane. The Z6 action on the torus coordinates zi ,
i
zi e2iv6 zi ,
(3.2)
is illustrated in Fig. 2. This orbifold has Z6 , Z3 and Z2 fixed points defined by

6
f i e2i v6 f i G2 SU(3)SO(4) ,
= 6, 3, 2,
(3.3)
where G2 SU(3)SO(4) is the torus lattice. The 12 Z6 fixed points are shown in Fig. 2, the 9 Z3
fixed pointsin Fig. 3 and the 16 Z2 fixed pointsin Fig. 4. It is a characteristic feature of nonprime orbifolds that the Z3 and Z2 fixed points are generally different from the Z6 fixed points.
The Z3 subtwist leaves the SO(4) plane invariant, whereas under the Z2 subtwist the SU(3) plane
is fixed.
The orbifold is flat apart from the singular points (conical singularities) corresponding to the
Z6 , Z3 and Z2 fixed points. Twisted states are localized at these singularities. In what follows,
we detail their localization properties in each T2 torus.
3.2. Twisted states location
3.2.1. G2 plane
In the G2 plane, there is one point fixed under Z6 located at the origin, 3 points x, y, z fixed
under Z3 , and 4 points a, b, c, d fixed under Z2 (Fig. 5). Some of them transform into each other
5 The overall sign of v is chosen such that one obtains left-chiral states (q = 1/2 for fermions) in the first twisted
6
4
sector. This convention differs from that of our earlier work [26].
163
Fig. 2. G2 SU(3) SO(4) torus lattice of the Z6-II orbifold. Possible Wilson lines are denoted by W3 , W2 and W2 .
Fig. 3. Z3 fixed points.
Fig. 4. Z2 fixed points.
(a) Modding out to the Z6 pillow.
(b) Z6 pillow.
Fig. 5. The G2 plane. The two simple roots of G2 are given by the arrows in (a) with the shaded area spanned by them
being the fundamental region of the torus. The fundamental region of the orbifold is one sixth of this region (darker area)
and can be represented by the pillow in (b). The latter corresponds to folding the fundamental region along the dashed
edge and gluing the other edges together (cf. [45,46]).
under Z6 twisting and correspond to the same points in the fundamental domain of the orbifold.
For the three Z3 fixed points
x ,
y ,
z,
(3.4)
one has
x x,
y z,
z y,
(3.5)
164
under the Z2 twist 3 , and the four Z2 fixed points

a ,
b,
c ,
(3.6)
d,
transform under the Z3 twist 2 as

a a,
b c,
c d,
d b.
(3.7)
Thus we have the following mapping from the fundamental domain of the torus to the fundamental domain of the orbifold:

b
x
y
c .
,
(3.8)
,
a
z
d
Consequently, T2,4 twisted matter lives at or points of the orbifold pillow, whereas T3
twisted matter lives at or .
As explained in Section 2.5.2, the fact that the Z3 and Z2 fixed points are not fixed under
Z6 introduces a new quantum number for physical states, a phase = e2iq with fractional q .
Consider the T2,4 twisted sectors. Among the states localized at , there are two linear combinations

1
1
|; 1 = |y |z ,
|; +1 = |y + |z ,
(3.9)
2
2
which are Z2 (and Z6 ) eigenstates with eigenvalues = 1,
3 |; +1 = |; +1 ,
3 |; 1 = |; 1 .
(3.10)
These eigenstates can be labelled by the order of the twist k = 2, 4 and the parameter q ,
|; +1 = |k = 2, 4; q = 1 ,
|; 1 = |k = 2, 4; q = 1/2 .
(3.11)
The state at the origin has = 1 and can be labelled as

|x = |; +1 = |k = 2, 4; q = 0 .
(3.12)
To distinguish = 1 states at from those at , we assign q = 0 to the former and q = 1 to

the latter.
The T3 states are treated analogously. There are three linear combinations of states located at
, with Z3 eigenvalues 1, e2i/3 , and 1 ,

1
|; 1 = |b + |c + |d ,
(3.13a)
3

1
|; = |b + 1 |c + 2 |d ,
(3.13b)
3

; 1 = 1 |b + |c + 2 |d .
(3.13c)
3
The Z3 (and Z6 ) eigenstates can again be characterized by the order of the twist and q ,

; 1 = |k = 3; q = 1/3 .
|; 1 = |k = 3; q = 1 ,
(3.14)
The state at the origin is now labelled as
|a = |; 1 = |k = 3; q = 0 .
(3.15)
165
Fig. 6. Location of the twisted states in the G2 plane.

Table 3.1
Localization quantum numbers and space group elements
n3
0
1
2
Location (in 2 units)
Space group element
0
1e +
3 3
2e +
3 3
2e
3 4
1e
3 4
k=1
m3
k=2
m3
k=4
m3
(, 0)
(, e3 )
0
1
( 2 , 0)
( 2 , e3 )
0
2
( 4 , 0)
( 4 , e3 )
0
1
(, e3 )
( 2 , e3 )
( 4 , e3 )
The T1,5 twisted sector states are localized at the origin, which corresponds to a Z6 eigenstate
with eigenvalue = 1, i.e.,
|; 1 = |k = 1, 5; q = 0 .
(3.16)
The location of all Tk twisted states is illustrated in Fig. 6.

The effect of the quantum number q on the projection conditions for physical states has been
discussed in Section 2.
3.2.2. SU(3) plane
States twisted by k with k = 1, 2, 4, 5 are localized at the three fixed points in the SU(3)
plane, whereas T3 and untwisted states live in the bulk. The localization is specified by the quantum number n3 (cf. Fig. 7(a)). Table 3.1 lists the coordinates of the fixed points in the SU(3) torus
as well as the corresponding space group elements. The coordinates are defined up to translations
in the sublattice 2[ne3 + (n 3m)e4 ] with n, m Z.
As discussed in Section 2, a fixed point or plane with the space group element ( k , ae3 + be4 )
corresponds to the local gauge shift
Vf = kV6 + m3 W3 ,
m3 = a + b mod 3,
(3.17)
up to terms involving W2 and W2 . Note that m3 depends not only on the location (n3 ) but also
on the order of the twist k (Table 3.1). The above local shift is equivalent to
Vf = k(V6 + n3 W3 ).
(3.18)
3.2.3. SO(4) plane

Twisted states from T1,5 and T3 are localized at the four fixed points in the SO(4) plane
whereas T2 , T4 and untwisted states correspond to bulk fields. The fixed points are labelled by n2
and n2 (Fig. 7(b)). Table 3.2 lists the coordinates of the fixed points and the corresponding space
group elements. The coordinates are defined up to translations in the sublattice 2[2ne5 + 2me6 ]
166
(a) SU(3) plane.
(b) SO(4) plane.
Fig. 7. Localization quantum numbers n3 , n2 and n2 .

Table 3.2
Localization quantum numbers and space group elements
(n2 , n2 )
Location (in 2 units)
(0, 0)
(0, 1)
Space group element
0
1e
2 6
1e
2 5
1 (e + e )
6
2 5
(1, 0)
(1, 1)
k=1
k=3
(, 0)
(, e6 )
( 3 , 0)
( 3 , e6 )
(, e5 )
( 3 , e5 )
(, e5 + e6 )
( 3 , e5 + e6 )
where n, m Z. The local shift for the k sectors (k = 1, 3, 5) reads

Vf = k(V6 + n2 W2 + n2 W2 )
(3.19)
up to terms involving W3 .
4. Superpotential
In this section, we discuss the superpotential couplings in heterotic orbifolds. Interactions
on orbifolds are calculated using superconformal field theories [43,47]. This leads to a set of
selection rules dictating which couplings are allowed. For our purposes, it suffices to identify the
allowed couplings without knowing their precise strength. The following discussion is closely
related to the analysis of Kobayashi et al. [22], with some extensions.6
4.1. Vertex operators and correlation functions
In orbifold conformal field theory, couplings are obtained from correlation functions of vertex
operators for the corresponding physical states. The vertex operators for bosons in the (1)-ghost
picture read (cf. [22]):
V 1 = e e2i(q+kvN )H e2i(p+Vf )X
(f )
3

i
Z
N f i
i
Z
N
fi
f .
i=1
6 Certain corrections to the selection rules of [22] will be discussed in detail in Ref. [48].
(4.1)
167
Here q, k, p, f and N f i , N f i are the quantum numbers described in Section 2, and f is the
twist field which creates the vacuum of the twisted sector at the fixed point f from the untwisted
vacuum (cf. [43,47,4951]); is the bosonized superconformal ghost (cf. [52]). Vertex operators
for untwisted states correspond to k = Vf = 0, f = 1.
In the 0-ghost picture, (4.1) is replaced with
(f )
V 0 = e2i(q+kvN )H e2i(p+Vf )X
3

i
Z
N f i
i
Z
3
N

j
j
e2iH Z j + e2iH Z j f .
fi
j =1
i=1
(4.2)
The vertex operator for fermions is given by
V 1/2 = e 2 e2i(q+kvN )H e2i(p+Vf )X

(f )
3

i
Z
N f i
i
Z
N
fi
f .
(4.3)
i=1
In what follows, we will mainly be interested in the superpotential couplings. These are extracted
from couplings between 2 fermions and n 2 bosons given by the correlation functions
(f1 ) (f2 ) (f3 ) (f4 )
(f )
V 1/2 V 1/2 V 1 V 0 V 0 n .
(4.4)
The correlation function (4.4) factorizes into correlators involving separately the fields , H , X I ,
Z i and the twist fields [43,47,4951]. Z6 invariance of each correlator leads to various selection
rules which we discuss in the following.
4.2. Selection rules
4.2.1. Gauge invariance
Consider a coupling of n massless physical states labelled by index r. As expected, the coupling has to obey gauge invariance. The gauge quantum numbers are specified by the shifted
momenta p + Vf which play the role of the weight vectors w.r.t. the unbroken subgroup of
E8 E8 . For the correlation function to be non-zero, the states have to form a gauge singlet,
n

(p + Vf )(r) = 0.
(4.5)
r=1
It is instructive to interpret a coupling among twisted fields in terms of local gauge groups.
Suppose that the twisted states form representations R, R , etc., under the local non-Abelian
gauge groups Glocal , Glocal , etc. Then the coupling among these states is invariant under the
intersection of these groups,
Gintersection = Glocal Glocal Glocal E8 E8 ,
(4.6)
which is given by the E8 E8 roots common to all of the local groups. The remaining gauge
invariance conditions concern U(1) charges. R, R , etc. can be decomposed into representations
of Gintersection such that the invariant couplings involve the latter. This implies, for instance, that
a coupling between localized 16-plets of SO(10) and other twisted states need not be invariant
under the full SO(10). As a result, a mass term for the SM singlet in the 16-plet can be written
without invoking large SO(10) representations such as 126-plets, which are necessary in 4D
GUTs.
168
Table 4.1
Z6-II orbifold: H -momenta for bosons containing no oscillators
k
H -momentum
1 (1, 2, 3)
6
1 (1, 2, 0)
3
1 (3, 0, 3)
2
1 (2, 1, 0)
3
1 (5, 4, 3)
6
2
3
4
5
4.2.2. H -momentum rules

Twist invariance of the compact 6D space requires that the superpotential be a scalar with
respect to discrete rotations in the compact space. In other words, the H -momenta must add up
to zero (up to a discrete ambiguity). The H -momenta invariant under the ghost picture changing
are defined by [22]

i
= q i + kv6i (r) N f i N f i (r)
R(r)
(4.7)
and can be thought of as discrete R-charges [49,53]. They lie on the SO(8) weight lattice.
For an allowed coupling between 2 fermions and n 2 bosons, the sum of the H -momenta
must vanish. This rule can be reformulated in terms of bosonic H -momenta only. Specifically,
n

r=1
n

r=1
n
1
R(r)
= 1 mod 6,
(4.8a)
2
R(r)
= 1 mod 3,
(4.8b)
3
R(r)
= 1 mod 2,
(4.8c)
r=1
i are the H -momenta of the bosonic components of chiral superfields. For the Z
where R(r)
6-II
orbifold these are listed in Table 4.1.7 We note that gauge invariance requires strict vanishing
of the sum of E8 E8 momenta, whereas the sum of H -momenta must vanish up to a discrete
shift as given above. The difference between the two rules stems from the fact that the gauge
16D torus possesses continuous symmetries, while in the case of the 6D orbifold they are only
discrete.
4.2.3. Space group selection rules

The space group selection rule [43,47] states that the string boundary conditions have to match
in order for the coupling to be allowed. Consider twisted states living at the fixed points f1 ,
f2 , . . . , fn corresponding to the space group elements ( k(1) , (1) ), ( k(2) , (2) ), . . . , ( k(n) , (n) ).
A coupling of these states is allowed if (cf. [49])
k

(1) , (1) k(2) , (2) k(n) , (n) = (1, 0)
(4.9)
7 Our sign convention is opposite to that of [22].
(a)
(b)
169
(c)
Fig. 8. Allowed (a), (b) and forbidden (c) 3-point couplings between localized states |A , |B , |C in the SU(3) plane.

up to a torus lattice vector nr=1 k(r) , where k(r) = (1 k(r) ). The untwisted sector corresponds to the space group element (1, 0). The above condition is equivalent to (cf. Appendix B)
n

r=1
n
k(r) = 0 mod 6,
(r) = 0 mod
r=1
(4.10a)
n
k(r) .
(4.10b)
r=1
The first equation restricts the twisted sectors that can couple and states that the total twist of the
coupling must be 0 mod 6. The second condition puts a restriction on the fixed points. In terms
of the localization quantum numbers, it reads
SU(3) plane:
SO(4) plane:
n

r=1
n

r=1
n
k(r) n3(r) = 0 mod 3,
(4.11a)
n2(r) = 0 mod 2,
(4.11b)
n2(r) = 0 mod 2,
(4.11c)
r=1
plus an additional condition to be discussed below. The quantum numbers n3(r) , n2(r) and n2(r)
have been defined in Section 3.8 The space group selection rule for the SU(3) plane is illustrated
in Fig. 8.
For the G2 plane, there is a non-trivial selection rule if only T2 and T4 , or only T3 states are
involved in the coupling. As we show in Appendix B, the coupling must satisfy
{q (1) , . . . , q (n) }
/ permutations{x, 0, . . . , 0}
(4.12)
for x = 0.
To summarize, we have presented the string selection rules which determine whether a given
superpotential coupling is allowed. Apart from gauge invariance, such couplings enjoy certain
discrete symmetries related to the localization properties of the states involved.
8 Note that the sum rule for the SU(3) plane differs from the corresponding rule in [22] by the factors k .
(r)
170
(a)
(b)
Fig. 9. 3 vs. 2 equivalent families.
5. The MSSM from the heterotic string

In this section, we present an orbifold compactification of the E8 E8 heterotic string which
yields the MSSM spectrum and gauge group at low energies. Apart from the MSSM sector, the
model contains a hidden sector which can account for low-energy supersymmetry breakdown.
In this section we present basic features of the model, whereas other important aspects such as
vacuum configurations, SUSY breaking, and phenomenology will be discussed in Sections 68.
5.1. Search strategy
It is well known that with an appropriate choice of the gauge shift V and Wilson lines, it is
not difficult to get the Standard Model gauge group times extra group factors. The real challenge
however is to get three generations of the SM matter.
We base our search on the concept of local GUTs. Since one complete matter generation
(plus a right-handed neutrino) is a 16-plet of SO(10), we use the gauge shifts which admit local
SO(10) symmetry and 16-plets at the fixed points. There are only two such shifts in a Z6-II
orbifold [54,55],

1 1 1
1
V6 =
, , , 0, 0, 0, 0, 0
, 0, 0, 0, 0, 0, 0, 0 ,
2 2 3
3

1
1
1
1 1
, , , 0, 0, 0, 0, 0
, , 0, 0, 0, 0, 0, 0 .
V6 =
(5.1)
3 3 3
6 6
Each of them ensures that there are 16-plets in the T1 sector, which remain in the massless
spectrum regardless of the Wilson lines. Further, one adjusts the Wilson lines such that the gauge
group in 4D is that of the Standard Model times additional factors.
To obtain three matter generations, the simplest option is to use three equivalent fixed points
with local SO(10) symmetry [23], Fig. 9(a). This would provide an intuitive explanation for
triplication of fermion families. However, our scan over such models shows that in this case
there are always chiral exotic states in the spectrum (cf. Appendix C).9 Such states get masses
due to electroweak symmetry breaking and generally are inconsistent with experiment. A similar
statement applies to other Zn orbifolds with n 6.
This result implies that the three families of 16-plets are not all equivalent, at least in the
context of Z6 orbifold models. We are thus led to consider the next-to-simplest possibility:
2 equivalent families and one different family, Fig. 9(b). The equivalent 16-plets can appear due
9 We find that some models have exotic matter which is vector-like with respect to SU(3) SU(2) but chiral with
c
L
respect to correctly normalized U(1)Y . In particular, our earlier model [23] suffers from this problem.
171
to 2 equivalent fixed points in the SO(4) plane with one Wilson line W2 . The remaining family
then has to come from other sectors of the model. We find that this procedure is successful and,
in many cases, the exotic matter is vector-like with respect to the Standard Model. Furthermore,
we find that the vector-like matter can be consistently decoupled at least in one case.
5.2. The model
Our model is a Z6-II heterotic orbifold based on the Lie lattice G2 SU(3) SO(4). It involves
two Wilson lines: one of order 2, W2 , and another of order 3, W3 , and has the gauge shift V6
consistent with the local SO(10) structure. Specifically, the gauge shift and the Wilson lines are
given by [26]

1 1 1
1
, , , 0, 0, 0, 0, 0
, 0, 0, 0, 0, 0, 0, 0 ,
V6 =
2 2 3
3

1
1 1 1
3 1 1 1 1 1 1 1
W2 =
, 0, , , , 0, 0, 0 , , , , , , , ,
2
2 2 2
4 4 4 4 4 4 4 4

1
1 1 1 1 1
1 1 1
, 0, 0, , , , ,
1, , , , 0, 0, 0, 0 .
W3 =
(5.2)
3
3 3 3 3 3
3 3 3
By adding elements of the root lattice E8 E8 to the shift and Wilson lines, one can transform
this set to

1 1 1
17 5 5 5 5 5 5 5
V6 = , , , 0, 0, 0, 0, 0
, , , , , , ,
,
2 2 3
6
2 2 2 2 2 2 2

1
1 1 1
23 25 21 19 25 21 17 17
, , , , , , ,
,
W2 = , 0, , , , 0, 0, 0
2
2 2 2
4
4
4
4
4
4
4 4

1 1 1 1 1 1 1 1
2 1 4
0, , , , 1, 0, 0, 0 ,
W3 = , , , , , , ,
(5.3)
6 2 2 6 6 6 6 6
3 3 3
which fulfills the strong modular invariance conditions (2.30).
The gauge group after compactification is

G = SU(3) SU(2) SU(4) SU(2) U(1)9 .
(5.4)
Here the brackets [ ] indicate a subgroup of the second E8 factor. The generators of the U(1)
factors can be chosen as

1 1 1 1 1
t1 = tY = 0, 0, 0, , , , ,
(0, 0, 0, 0, 0, 0, 0, 0),
2 2 3 3 3
t2 = (1, 0, 0, 0, 0, 0, 0, 0)(0, 0, 0, 0, 0, 0, 0, 0),
t3 = (0, 1, 0, 0, 0, 0, 0, 0)(0, 0, 0, 0, 0, 0, 0, 0),
t4 = (0, 0, 1, 0, 0, 0, 0, 0)(0, 0, 0, 0, 0, 0, 0, 0),
t5 = (0, 0, 0, 1, 1, 1, 1, 1)(0, 0, 0, 0, 0, 0, 0, 0),
t6 = (0, 0, 0, 0, 0, 0, 0, 0)(1, 0, 0, 0, 0, 0, 0, 0),
t7 = (0, 0, 0, 0, 0, 0, 0, 0)(0, 1, 1, 0, 0, 0, 0, 0),
t8 = (0, 0, 0, 0, 0, 0, 0, 0)(0, 0, 0, 1, 0, 0, 0, 0),
t9 = (0, 0, 0, 0, 0, 0, 0, 0)(0, 0, 0, 0, 1, 1, 1, 1).
(5.5)
172
Table 5.1
Quantum numbers of the massless states w.r.t. GSM [SU(4) SU(2)] and a field naming convention
Name
Irrep
Count
Name
qi
di
i
mi
si
si
fi
wi
(3, 2; 1, 1)1/6
1; 1, 1)1/3
(3,
(1, 2; 1, 1)1/2
(1, 2; 1, 1)0
(1, 1; 1, 1)1/2
(1, 1; 1, 1)0
(1, 1; 4, 1)0
(1, 1; 6, 1)0
3
7
5
8
16
69
4
5
u i
di
i
ei
si+
hi
fi
Irrep
1; 1, 1)2/3
(3,
(3, 1; 1, 1)1/3
(1, 2; 1, 1)1/2
(1, 1; 1, 1)1
(1, 1; 1, 1)1/2
(1, 1; 1, 2)0
1)0
(1, 1; 4,
Count
3
4
8
3
16
14
4
One of the U(1) factors is anomalous. It is generated by

tanom =
8
ci ti
i=1

11 1 3 1
1 1
with ci = 0, , , , , 1, , , 0 .
6 2 2 6
3 3
(5.6)
The sum of the anomalous U(1) charges is

tr tanom = 88,
(5.7)
which is relevant to the calculation of the FayetIliopoulos term.

The factors SU(3) and SU(2) in G are identified with the color SU(3)c and the weak SU(2)L
of the Standard Model. The hypercharge generator is given by tY . It is embedded in SO(10) just
like in usual 4D GUTs,
SU(3)c SU(2)L U(1)Y SO(10).
(5.8)
Thus it automatically has the correct normalization and is consistent with gauge coupling unification. It is also important that this hypercharge is non-anomalous, tY tanom = 0.
The massless matter states are listed in Table 5.1. They appear in both the untwisted and
twisted sectors, apart from T5 which has no left-chiral superfields. The spectrum can be summarized as follows:
matter:
3 16 + vector-like.
(5.9)
Two generations are localized in the compactified space and come from the first twisted sector
T1 , whereas the third generation is partially twisted and partially untwisted:
2 16 T1 ,
16 U, T2 , T4 .
(5.10)
In particular, the up-quark and the quark doublet of the third generation are untwisted, which
results in a large Yukawa coupling, whereas the down-quark is twisted and its Yukawa coupling
is suppressed. The 16-plet quantum numbers of the third generation are not enforced by local
GUTs, but are related to the Standard Model anomaly cancellation.
Apart from the 3 matter families, the model contains extra states which are vector-like with
respect to the Standard Model gauge group. These include a pair of Higgs doublets and additional exotic matter which, as we show in the subsequent sections, can be consistently decoupled.
A complete list of quantum numbers of the massless states is given in Tables D.2 and D.3.
173
5.3. Local GUT representations

The matter states of the model can be viewed as originating from representations of local
GUTs supported at certain fixed points or planes. States from the first twisted sector correspond
to brane fields living at the orbifold fixed points. As discussed in Section 2, such states are
invariant under the orbifold action. Thus they all survive in 4D and furnish complete representations of the local GUTs. On the other hand, states from higher twisted (as well as untwisted)
sectors are not automatically invariant under the orbifold action. Part of the GUT multiplet is
projected out such that the surviving states produce incomplete (split) multiplets in 4D. In
particular, the gauge multiplets of E8 reduce to those of the Standard Model (and extra group
factors). The latter can be viewed as an intersection of local GUTs at various orbifold fixed
points (see e.g. [25]). We survey the local GUTs and their representations in Table D.1.
5.4. Spontaneous gauge symmetry breaking
The effective low energy theory of our orbifold model has, in general, smaller gauge symmetry
and fewer massless states than those in Eq. (5.4) and Table 5.1. One of the reasons is that there
is an anomalous U(1) which induces a FI D-term,
Danom =
(i)
qanom
|i |2 +
gMP2
tr tanom ,
192 2
(5.11)
(i)
where the sum runs over all scalars i with anomalous charges qanom . This D-term must be zero
in a supersymmetric vacuum, so at least some of the scalars are forced to attain large vacuum
expectation values, typically not far below the string scale. As a result, the anomalous U(1)
gets broken. Generically, this also triggers breakdown of other gauge symmetries, under which
the above mentioned scalars are charged. The resulting gauge group and matter fields at low
energies are therefore a subset of those in Eq. (5.4) and Table 5.1. A list of anomalous charges
of SM singlets is given in Table D.4.
More generally, some of the scalars can attain VEVs as long as it is consistent with supersymmetry, Fi = Da = 0. In the simplest case, such scalars are associated with flat directions in
the field space. In general, supersymmetric configurations are described by non-trivial solutions
of Fi = Da = 0, which correspond to points or low-dimensional manifolds in the field space. In
either case, this breaks part of the gauge symmetry,
VEVs
GGlow-energy .
(5.12)
Furthermore, such VEVs provide mass terms for some of the matter states. In particular, if the
superpotential coupling
W = xi xj s1 sn
(5.13)
exists, with xi and xj being vector-like states w.r.t. Glow-energy and sk being the scalars attaining
VEVs, then xi and xj become massive and decouple from the low energy theory.
It is common that orbifold models contain states which are charged under both GSM and other
gauge factors originating from the second E8 . As long as such gauge factors are unbroken, there
is no hidden sector in the model, which is usually required for spontaneous SUSY breaking. The
separation between the visible and the hidden comes about when some of the scalars attain
VEVs thereby breaking the unwanted gauge factors. In our model, this occurs, in particular, when
some of the 69 si states break U(1)8 .
174
An interesting property of the model is that none of the oscillator states is charged under
GSM (cf. Tables D.2, D.3). If all the oscillators develop VEVs, the unbroken gauge group is
GSM [SU(4) U(1)], while the SM matter is neutral under the additional U(1). This might be
important as it has been argued that giving VEVs to oscillator modes corresponds to resolving
the conical singularities associated with the fixed points [49]. This means that the phenomenologically relevant gauge group survives the naive blowing-up procedure.
Orbifold models with the same gauge shifts and Wilson lines but different scalar VEVs lead
to distinct low-energy theories. For example, in some of them, the Standard Model gauge group
is broken. To obtain realistic models, one has to make sure that, first of all,
GSM is unbroken,
exotic matter is heavy.
There are also further phenomenological constraints which we discuss in the subsequent sections.
5.5. Decoupling the exotic states
A necessary condition for the decoupling of vector-like exotic states, without breaking the
Standard Model gauge group, is the existence of the superpotential couplings
xi xj (SM singlets).
(5.14)
Furthermore, the rank of the xi , xj mass matrix must be maximal such that no massless vectorlike states survive. We find that in our model the required mass terms are allowed and the exotic
states can be decoupled.
The exotic states charged under GSM are pairs of di and di , i and i , si and si+ , and mi . The
mass terms for these states have the form
ij
ij
ij
ij
Wmass = di Md (s)dj + i M (s)j + mi Mm (s)mj + si+ Ms (s)sj ,
where s denotes some SM singlets. Taking s = {si }, we find

5 5 5 5 5 3 3
s s s s s s s
s1 s1 s3 s3 s3 s3 s3
ij
Md (s) = 1 1 3 3 3 3 3 ,
s s s s s s s
s6 s6 s6 s3 s3 s6 s6
3 4 4 1 1 1 1 1
s s s s s s s s
s1 s2 s2 s5 s5 s3 s3 s3
ij
M (s) = s 1 s 2 s 2 s 5 s 5 s 3 s 3 s 3 .
1 2 2 5 5 6 3 3
s s s s s s s s
s1
ij
s6
s6
s3
s3
s6
s3
(5.15)
(5.16)
(5.17)
s3
ij
Mm (s) and Ms (s) are given in Eqs. (D.1) and (D.2) in Appendix D, respectively. Here, an
entry s N indicates the existence of a coupling which involves N singlets. For instance, the (1, 1)
entry of the dd mass term includes
Wd1 d1 = d1 d1 (s3 s20 s39 s44 s65 + s7 s34 s35 s40 s41 + ),
(5.18)
175
where the coefficients are omitted. Different entries generally involve different combinations of
the singlets as well as different couplings, such that the rank of each mass matrix is maximal.
We note that higher N does not necessarily imply significant suppression of the coupling [56]:
s can be close to the string scale and, furthermore, the coefficient in front of the coupling grows
with N . We find that all mass matrices have maximal rank at order 8. A zero in the mass matrices
(D.1)(D.2) of Appendix D.4 indicates that up to order 8 no coupling appears.
This result implies that all of the exotic states can be decoupled below the GUT scale or
so. In particular, the rank of Md is 4 such that only 3 down-type quarks survive. M has, in
general, rank 5 resulting in 3 massless doublets of hypercharge 1/2. In order to get an extra
pair of (Higgs) doublets with hypercharge 1/2 and 1/2, one has to adjust the singlet VEVs
such that the rank reduces to 4. This large finetuning constitutes the well known supersymmetric
-problem and will be discussed in subsequent sections. A further constraint on the above texture
comes from the top Yukawa coupling: it is order one if the up-type Higgs doublet has a significant
component of 1 .
In the above mass matrices, s are chosen to be singlets under SU(4) SU(2) such that their
VEVs break
G SU(3)c SU(2)L U(1)Y Ghidden ,
(5.19)
with Ghidden = SU(4) SU(2) . Now the model has a truly hidden sector which can be responsible for spontaneous SUSY breaking.
In the next section we show that the required configurations of the singlet VEVs are in general
consistent with supersymmetry, e.g. Fi = Da = 0. The D-flatness is ensured by constructing
gauge invariant monomials out of the singlets [57,58] involved in the mass terms for the exotic
states. We further show that generally there exist non-trivial solutions to Fi = Da = 0 in the form
of low-dimensional manifolds in the field space.
Not all vacuum configurations consistent with supersymmetry and the decoupling are phenomenologically viable. Further important constraints are due to
absence of rapid proton decay,
realistic flavour structures,
small -term.
This strongly restricts allowed VEVs for the singlets. As we show in Section 8, these constraints
motivate certain patterns of the VEVs, in particular those which preserve BL symmetry at the
GUT scale.
Finally, let us remark on gauge invariance of the couplings in the framework of local GUTs.
As stated in Eq. (4.6), a coupling among twisted states is invariant under the intersection of
local gauge groups supported at the corresponding fixed points, but not necessarily under each
of the groups. To give an example, consider an allowed coupling s4 s26 s57 . Each of these singlets
originates from a larger representation of the local group. The above coupling arises from the
of
coupling of states contained in (16, 1, 1; 1) of SO(10) SU(2) SU(2) SO(14), (1; 1, 4)
SU(7) [SO(8) SU(4)], and (14; 1) of SO(14) [SO(14)]. Clearly, it is not SO(10) invariant.
This is a special feature of local GUTs.
To summarize, we have shown that our model reproduces the exact MSSM spectrum and the
gauge group at low energies. The matter multiplets appear as 3 16-plets of SO(10). Since Md is
a 4 7 matrix and M is a 5 8 matrix, there exists one pair of SU(2) Higgs doublets which
do not form complete GUT representations. The model also has a hidden sector.
176
(a) SO(4) plane.
(b) SU(3) plane.
(c) G2 plane.
Fig. 10. Orbifold GUT limits. In each case, a plane with a large compactification radius is displayed. Only subgroups of
the first E8 are shown and U(1) factors are omitted.
5.6. Orbifold GUT limits

One of the motivations for revisiting orbifold compactifications of the heterotic string is the
phenomenological success of orbifold GUTs [1419]. In our model, the hypercharge is correctly
normalized and the spectrum is that of the MSSM, which leads to gauge coupling unification
at about 2 1016 GeV. It is therefore interesting to study orbifold GUT limits of the model,
which correspond to anisotropic compactifications where some radii are significantly larger than
the others. Such anisotropy may mitigate the discrepancy between the GUT and the string scales
and can be consistent with perturbativity for one or two large radii of order (2 1016 GeV)1
[59,60]. In the energy range between the compactification scale and the string scale one obtains
an effective higher-dimensional field theory.
In Z6-II orbifolds, there are four independent radii: two are associated with the G2 and SU(3)
planes, respectively, and the other two are associated with the two independent directions in the
SO(4)-plane. Any of these radii can in principle be large leading to a distinct GUT model.
The bulk gauge group and the amount of supersymmetry are found via a subset of the invariance conditions (2.53) with N < 6. Consider a subspace S of the 6D compact space with large
compactification radii. This subspace is left invariant under the action of some elements of the
orbifold space group, i.e., a subset of twists and translations G. The bulk gauge multiplet in S is
part of the N = 4 E8 E8 gauge multiplet which is invariant under the action of G, i.e. a subset
of conditions (2.46) restricted to G [25].
In our model, the intermediate orbifold picture can have any dimensionality between 5 and
10. For example, the 6D orbifold GUT limits are (up to U(1) factors):
177
SO(4) plane: bulk GUT = SU(6), N = 2,

SU(3) plane: bulk GUT = SU(8), N = 2,
G2 plane:
bulk GUT = SU(6) SO(4), N = 4,
where the plane with a large compactification radius is indicated and N denotes the amount of
supersymmetry. In all of these cases, the bulk -functions of the SM gauge couplings coincide.
This is because either GSM is contained in a simple gauge group or there is N = 4 supersymmetry. It is remarkable that regardless of which radii are large, unification in the bulk occurs. This
observation may indeed be relevant to the discrepancy between the GUT and the string scales.
However, to check whether this is really the case, logarithmic corrections from localized fields,
contributions from vector-like heavy fields and string thresholds have to be taken into account.
The 6D orbifold GUT limits of the model are displayed in Fig. 10. We note that the Standard
Model gauge group is obtained as an intersection of the local gauge groups at the orbifold fixed
points. For completeness, in Table D.13 we survey all possible orbifold GUT limits. For D 6,
we find that they are all consistent with gauge coupling unification in the bulk. The different
geometries differ, however, in the values of the gauge couplings at the unification scale as well
as the Yukawa couplings.
6. Supersymmetric vacuum configurations
In this section we discuss supersymmetric vacuum configurations of our model. In globally
supersymmetric theories, these require vanishing of the D- and F -terms. We start with the discussion of the D-terms.
6.1. D-flatness
In a supersymmetric gauge theory with an anomalous U(1), vanishing of the D-terms requires

i Ta i = 0,
Da =
i
Danom =

i
(i)
qanom
|i |2 +
gMP2
tr tanom = 0,
192 2
(6.1)
where Ta are generators of the gauge group and tanom is the generator of an anomalous U(1). In
particular, to have a vanishing FI D-term, there must exist at least one field whose anomalous
charge is opposite in sign to that of tr tanom .
In theories without an anomalous U(1), these conditions are satisfied if there exists a gauge
invariant monomial I (i ) [57]. The D-flat field configurations are found from

I
(6.2)
= c i ,
i
where c is a constant and x denotes the VEV of x. On solutions of this equation, gauge invariance of I simply means Da = 0 [57]. If an anomalous U(1) is present, D-flatness requires the
existence of I (i ) which is gauge invariant with respect to all symmetries apart from U(1)anom
and which carries a net anomalous charge whose sign is opposite to that of tr tanom (see, e.g.,
[12,61,62]).
Therefore, searching for a D-flat configuration amounts to finding gauge invariant monomials
with the above properties. Clearly, such monomials can be multiplied together while preserving
178
the required properties. We are particularly interested in the U(1)N gauge theory which is relevant
when the non-Abelian singlets si get VEVs. In this case, the gauge invariant monomial
(s1 )n1 (s2 )n2 (sk )nk ,
ni N,
represents the D-flat direction
|s1 |/ n1 = |s2 |/ n2 = = |sk |/ nk
(6.3)
(6.4)
in the field space. The overall scale of these VEVs is fixed by the FI D-term.
Starting with the anomalous U(1) D-term, an example of a gauge invariant monomial with a
negative anomalous charge is given by
I1 = (s12 )2 s39 s55 s56 .
(6.5)
Clearly, it is not unique (cf. Table D.6). In particular, it can be multiplied by a monomial with
zero anomalous charge. We find that every si that enters the mass matrices for the exotic states
also enters a gauge invariant monomial (see Table D.5). This shows that si can be given large
VEVs while having vanishing D-terms.
There is also another algorithm to check the D-flatness for the required singlet configuration.
Mass terms for vector-like exotic matter xi are generated by

W=
(6.6)
xi xj Mij (s)
ij
with
Mij (s) =

cij (k1 , . . . , kn )sk1 skn ,
(6.7)
n k1 ,...,kn
where ska are the singlets and cij (k1 , . . . , kn ) are some coefficients. Any monomial of fields in
the superpotential is gauge invariant and represents a D-flat direction. Thus multiplying all of the
monomials together, we again get a flat direction. However, we do not want to give VEVs to the
exotic matter fields xi since this would break the Standard Model gauge group. So, one needs to
replace those with some SM-singlets which have the same total U(1)-charges:

xi xj = U(1)-charges(sl1 sln ).
U(1)-charges
(6.8)
Mij =0
This is just one equation. In our case, there are many singlet monomials satisfying this equation
and one of them is
s7 (s19 )4 s26 s36 s39 (s40 )5 (s48 )18 s55 (s56 )3 (s57 )7 (s64 )2 (s68 )42 (s69 )27 .
(6.9)
To cancel the FI term, one has to multiply it with the monomial I1 which has a negative anomalous charge. This shows that one can give VEVs to all singlets involved in the decoupling of
extra matter consistently with the D-flatness.
6.2. Some of the F -flat directions
The requirement Fi = 0 for a singlet si is most easily satisfied if si is an F -flat direction, i.e.,
W
=0
si
(6.10)
179
for arbitrary values of si . (When si is not a flat direction, Fi = 0 is satisfied only at special values
of si .) The existence of such flat directions usually requires that the VEVs of some other singlets
appearing in the superpotential be zero.
Many exactly F -flat directions can be obtained from the selection rule (4.8c) for the superpotential couplings,

R 3 = 1 mod 2.
(6.11)
As seen from Tables D.2 and D.3, all of the non-Abelian singlets in U , T2 , T4 sectors have R 3 =
0. Thus they cannot couple among themselves consistently with the rule (6.11). Furthermore,
they cannot couple to a single state in T1,3 since the latter have R 3 = 1/2 and at least two of
such states are needed to have an allowed coupling. That means that the F -terms are proportional
to a VEV of some state in T1,3 :
Fi
W
singlet from T1,3 = 0,
si
(6.12)
as long as all singlets in T1,3 have zero VEVs. Thus one immediately gets 39 exactly F -flat
directions associated with si from the
U, T2 , T4
(6.13)
sectors. By this we mean that the 39 fields are allowed to attain non-zero VEVs simultaneously,
without referring to the number of real variables parametrizing such VEVs.
One can also show that these directions are D-flat.10 In particular, each non-Abelian singlet
from U , T2 , T4 enters a gauge invariant monomial which involves only U , T2 , T4 singlet states.
Furthermore, it is possible to construct a monomial with a negative net anomalous charge. An
example is (see also Table D.7)
I = s34 s35 s40 s39 s67 .
(6.14)
That means one can give non-zero VEVs to the U , T2 , T4 singlets while preserving supersymmetry. Some of such states presumably correspond to the blowing-up modes of the orbifold
which allow one to interpolate between a smooth CalabiYau manifold and an orbifold.
These flat directions allow us to decouple many exotic states but not all. One can perhaps
increase the dimensionality of the F - and D-flat space by including non-Abelian flat directions
or by other considerations. We also note that, for practical purposes, flatness is only required up
to a certain order in superpotential couplings and one may exploit approximately flat directions.
In any case, flat directions are not necessary for the decoupling. As we discuss below, supersymmetric field configurations are in general more complicated and allow for the decoupling of
the exotic states.
6.3. General supersymmetric field configurations
Given a set of 69 states si , supersymmetric field configurations are given by the sets of VEVs
si which satisfy Fi = Da = 0. Naively, it appears that the number of constraints, that is 69 plus
the number of the gauge group generators, is larger than the number of variables, 69. The system
seems to be overconstrained. However, this is not the case. As well known, complexified gauge
10 The requirement of the D-terms cancellation fixes some of the field VEVs.
180
transformations allow us to eliminate the D-term constraints ([63], [64, Chapter VIII]), such that
the number of variables equals the number of equations. In what follows, we demonstrate this
for Abelian and non-Abelian cases.
6.3.1. Abelian case
Consider a supersymmetric U(1)N gauge theory with n charged fields zi . The superpotential
can be written as

W=
(6.15)
I (a) (z1 , . . . , zn ).
(a)
I (a)
are gauge invariant monomials (some of which may be reducible, i.e. a product of
Here
lower order monomials),
I (z1 , . . . , zn ) = cz1k1 znkn
(6.16)
with c being a constant and

k1 Q1 + + kn Qn = 0,
(6.17)
where Qi = (qi1 , . . . , qiN ) is an N -vector of U(1) charges of the fields zi .

Supersymmetry is preserved in the vacuum if
Fi = 0,
Da = 0,
i = 1, . . . , n, a = 1, . . . , N.
Start with the F -terms. Fi = 0 can be written

Fi (z)
(6.18)
as11
W
=0
zi
(6.19)
for all i. Since there are n such equations and n variables, there are solutions. In general, there
are solutions with zi = 0 (for example, when W is a non-trivial polynomial).12
Consider a solution with zi = 0. Note that Fi (z) is not gauge invariant, but transforms as zi1 .
As a consequence, if {zk0 } is a solution to Fi (z) = 0, then the transformation
zk0 zk = zk0 (1 )qk (2 )qk (N )qk ,
1
leaves the F -terms vanishing,

N
1
2
Fi z0 Fi (z ) = Fi z0 (1 )qi (2 )qi (N )qi = 0,
(6.20)
(6.21)
where i are arbitrary complex numbers and qia is the ath U(1) charge of zi . Therefore, given a
solution zi0 to the F -term equations, it can be rescaled as above to give a family of solutions. In
fact, it can be rescaled in such a way that all the D-terms vanish:

2
N
1
2
Da (z ) =
(6.22)
qia |zi |2 =
qia zi0 |1 |2qi |2 |2qi |N |2qi = 0
i
11 Here we define the F component such that it has the quantum numbers of z .
12 In particular, this is generally the case in string orbifold models. The reason is that if a superpotential W is allowed by
0
string selection rules, W0N is also allowed for some integers N . For example, in the Z6 case, one has W W0 + W07 + .
Such superpotentials allow for non-trivial solutions to the F -term equations (consider, e.g., W0 = z1 z2 z3 ).
181
for a = 1, . . . , N . The N rescaling parameters |i | are found from the above N equations. In
terms of the rescaled variables zi , these solutions are encoded in the gauge invariant monomials
z1k1 znkn such that

|z1 |/ k1 = |z2 |/ k2 = = |z3 |/ kn
(6.23)
is a D-flat direction. This latter equation allows to find i most easily and also shows that sensible
solutions to Eq. (6.22) exist, i.e., |i |2 > 0.13
Let us now turn to the D-term of an anomalous U(1). The complexified gauge transformation
which leaves Fi = 0 intact is
(i)
zi0 zi = qanom zi0 ,

(i)
gMP2
(i) 0 2
Danom (z ) =
zi ||2qanom +
qanom
tr tanom .
192 2
(6.24)
As long as there is a field whose anomalous charge is opposite in sign to that of tr tanom , the
(i)
D-term can be cancelled. Suppose tr tanom < 0, then Danom < 0 for qanom 0 and Danom > 0
(i)
(i)
for qanom . Therefore, there is a solution to Danom = 0 for finite qanom .
It is now clear that the D-term constraints can be satisfied by an appropriate choice of complexified gauge transformations. This means that the number of SUSY conditions Fi = 0 equals
the number of variables zi , such that (non-trivial) solutions generally exist. Such solutions can
be points (up to gauge transformations) or low-dimensional manifolds in the field space.
6.3.2. Non-Abelian case
Let us now consider the case of a non-Abelian gauge theory following Ref. [64, Chapter VIII].
This situation arises in our construction when one assigns VEVs to the doublets of the hidden
sector SU(2). As in the Abelian case, if {zk0 } is a solution to the F -term equations Fi (z) = 0, then

a Ta z 0
z = exp i
(6.25)
a
is also a solution, where Ta are the group generators

and a are complex parameters. This is
because Fi transforms as zi1 , i.e., F (z ) = exp(i a Ta )F (z0 ).

The D-terms, Da (z) = i zi Ta zi , transform in the adjoint representation under this transformation. There is always a group element which transforms vector Da into (x, 0, . . . , 0)

corresponding to the direction of one of the Cartan generators Ta , i.e., Da Da = i zi Ta zi .
Writing (Ta )ij = i ij with real i , the only non-vanishing D-term is

i |zi |2 .
Da (z ) =
(6.26)
i
The complexified gauge transformation along this direction,

zi zi = exp(i )zi ,
(6.27)
13 Note that if 2 fields z

1,2 with identical charges are present, one has to be cautious. As far as the D-term equations
k k
k +k
go, these two fields can be treated as one, i.e. z11 z22 z21 2 and q1a |z1 |2 + q2a |z2 |2 q2a |z2 |2 .
182
with real , leaves Fi (z ) = 0 and transforms the D-term into

Da z =
i e2i |zi |2 .
(6.28)
In the non-degenerate case, Da (z ) for .14 Therefore, there is a solution to
Da = 0 for finite and hence finite zi .

6.3.3. Summary and applications
Employing complexified gauge symmetry, we have shown that the system of equations Fi =
Da = 0 in globally supersymmetric models is not overconstraining. In particular, solutions to
Fi = 0 exist since the number of equations equals the number of complex variables and in general
some of these solutions are non-trivial. Once a non-trivial solution to Fi = 0 is found, it can be
transformed using complexified gauge symmetry to satisfy Da = 0. This conclusion is based
on the observation that the F -term equations constrain gauge invariant monomials, while such
monomials are also associated with D-flat directions.
Consequently, supersymmetric field configurations in orbifold models form low-dimensional
manifolds or points (up to gauge transformations). In such configurations, SM singlets generally
attain non-zero VEVs, typically not far below the string scale. As a result, when such VEVs play
a role of the mass terms for vector-like exotic states, the decoupling of the latter can be made
consistently with supersymmetry.
The above considerations apply to globally supersymmetric models at the perturbative level.
In practice, we expect supergravity as well as non-perturbative effects to play a role in selecting
vacua. However, it would be very difficult to quantify such effects at this stage. We note that, in
existing literature, it is rather common to amend the global SUSY conditions Fi = Da = 0 by
W = 0 (see e.g. [50]), which implies a vanishing cosmological constant in supergravity. Such
a condition should however be imposed on the total superpotential which includes, in particular, non-perturbative potentials for moduli. Thus, requiring W = 0 does not set any immediate
constraint on the charged matter VEVs. At this stage, we include only the most important supergravity effect, that is gaugino condensation in the hidden sector, which we discuss in the next
section.
7. Spontaneous supersymmetry breaking
7.1. Hidden sector gaugino condensation
As supersymmetry is broken in nature, realistic models should admit spontaneous supersymmetry breakdown. An attractive scheme for that is hidden sector gaugino condensation [6568].
In this case, a hierarchically small supersymmetry breaking scale, which is favoured by phenomenology, is explained by dimensional transmutation.
The basic idea is that one or more gauge groups in the hidden sector become strongly coupled at an intermediate scale. This leads to confinement and gaugino condensation. Under certain
circumstances, that is if the dilaton is stabilized, gaugino condensation translates into supersym14 An example of the degenerate case is an SU(2) theory with 2 fundamental multiplets h
n
1,2 and W = (h1 h2 ) . The
solution to the F -term equations is h2 = ch1 such that all gauge invariant monomials vanish. The D-terms vanish only
for h1,2 = 0 corresponding to || .
183
metry breaking. In particular,

1/3 1013 GeV,
(7.1)
leads to the gravitino mass in the TeV range, m3/2 /MP2 . The condensation scale
1/3 is given by the Landau pole of the condensing gauge group,

1
1
= MGUT exp
(7.2)
.
2 g 2 (MGUT )
For certain gauge groups and matter content, can be in the right range.
Gaugino condensation leads to supersymmetry breaking only if the dilaton is stabilized at
a realistic value. Models with a single gaugino condensate and a classical Khler potential
suffer from the notorious dilaton run-away problem. That is, gaugino condensation creates a
non-perturbative superpotential for the dilaton W exp(aS) (where a = 3/2) which
leads to S at the minimum of the scalar potential. There are two common options to avoid
this problem: employ multiple gaugino condensates or use non-perturbative corrections to the
Khler potential. The first option is not available in our model as the hidden sector SU(2) either
does not condense or its condensation scale is too low, and we are left with a single SU(4). Thus,
we use the second option. In this case, the classical Khler potential for the dilaton is amended
by non-perturbative corrections,
+ Knp .
K = ln(S + S)
(7.3)
The functional form of Knp has been studied in the literature [6974]. For a favourable choice
of the parameters, this correction allows one to stabilize the dilaton at a realistic value, Re S 2,
while breaking supersymmetry [71,7376]. Supersymmetry is broken spontaneously by the dilaton F -term,

,
FT 0,
(7.4)
MP
where T is the heterotic T -modulus. In what follows, we will estimate the gaugino condensation
scale in our model without going into details of the dilaton stabilization mechanism.
The condensing gauge group in our case is SU(4). The condensation scale depends on the
The
matter content. If all the singlets have zero VEVs, there are 5 6-plets and 4 pairs of 4 + 4.
corresponding beta function is
FS

1
(7.5)
12 #(6) #(4 + 4 ) ,
2
16
where #(R) counts the number of representations R. With the above matter content, SU(4) is
asymptotically free but the condensation scale is too low. In a general field configuration, the
6-plets and the pairs 4 + 4 receive large masses (see Eqs. (D.3) and (D.4)) and are all decoupled.
In this case, the beta function becomes
SU(4) =
SU(4) =
3
.
4 2
(7.6)
The condensation scale is then 1010 1011 GeV. There are many factors that can affect it. In
particular, there are string threshold corrections [7781] which lead to different gauge couplings
in the visible and hidden E8 . The corresponding gauge kinetic functions are given by [77,80,81]
fvis/hid = S T ,
(7.7)
184
Fig. 11. Scale dependence of the hidden sector SU(4) gauge coupling for different threshold corrections .
where is a small parameter and, for simplicity, we have taken a large T limit. The gauge
couplings are found from
Re f = g 2 .
In the visible sector, the apparent gauge coupling unification requires
hidden sector gauge coupling is
2
(MGUT ) = Re fhid 2(1 ),
ghid
(7.8)
2
gGUT
2, whereas the
(7.9)
where parametrizes string threshold corrections. The corresponding condensation scale is

1
MGUT exp (1 ) .
(7.10)
For between 0 and 0.3, the condensation scale ranges between 5 1010 and 1013 GeV (cf.
Fig. 11). Thus a TeV scale gravitino mass can in principle be obtained.
There are of course other factors that can affect the above estimate. For example, the SUSY
breaking scale depends on coefficients entering a particular dilaton stabilization mechanism. Also
the identification of the Landau pole with 1/3 is not precise. The main point, however, is that
the model contains the necessary ingredients for gaugino condensation and SUSY breaking in
the phenomenologically interesting range.
7.2. Soft SUSY breaking terms
The Khler stabilization mechanism leads to a specific pattern of the soft terms, the so-called
dilaton dominated scenario [82]. The resulting soft terms are universal and given by

1
1
AY + h.c. ,
Lsoft = Ma a + h.c. m2
(7.11)
2
6
where a are the gauginos, are the scalars and Y are the Yukawa couplings. Dilaton dominated SUSY breaking implies the following relations among the soft breaking parameters (see
e.g. [83]):
A = M.
M = 3m3/2 , m = m3/2 ,
(7.12)
185
This is a restricted version of mSUGRA with the only independent parameter being the gravitino
mass m3/2 . Here we do not discuss the and B terms which depend on further details of the
model.
The dilaton dominated scenario has a number of phenomenologically attractive features. In
particular, due to flavour universality in the soft breaking sector, it avoids the SUSY FCNC
problem. Also, most of the physical CP phases, e.g., arg(A M), vanish which ameliorates the
SUSY CP problem. Other phenomenological aspects have been discussed in Ref. [84].
The above considerations are based on the assumption that the dilaton is stabilized via nonperturbative corrections to the Khler potential. Dilaton and other moduli stabilization is a
difficult issue and there may exist other possibilities which could lead to other patterns of the
soft terms.
8. BL symmetry and phenomenology
Realistic string vacua must satisfy a number of phenomenological constraints, in addition to
those imposed by the spectrum and the gauge group of the MSSM. In particular, the proton
should be sufficiently stable as well as flavour structures should be realistic. This constrains
vacuum configurations for the SM singlets. In generic vacua, there are baryon number violating
operators already at the renormalizable level, so in order to avoid rapid proton decay one must
be able to tune the VEVs and suppress such operators. This appears rather artificial and one may
ask whether there is a deeper reason behind it.
In this section, we explore vacua preserving the BL symmetry at the high energy (GUT)
scale, which appear phenomenology attractive. In this case, the renormalizable R-parity violating
couplings
WR/ = i i u + ij k i j ek + ij k i qj dk + ij k u i dj dk
(8.1)
are prohibited, leading to suppression of proton decay. The BL symmetry fits naturally into the
concept of local GUTs: it is related to the SO(10) BL generator, although there are differences.
Finally, BL can be broken at an intermediate scale which might induce small R-parity violating
couplings and could be related to the smallness of the neutrino masses.
Having suppressed BL violation, we further study whether the required singlet VEV configurations allow for the decoupling of the exotic matter, realistic flavour structures and a small
-term.
In this section, we study a particular singlet VEV configuration for which we are able to prove
the D-flatness, but not the F -flatness. Since the phenomenological analysis is intractable for the
general case, we hope that the considerations below will provide some guidance to the further
search for realistic vacua.
8.1. Vacuum configurations with unbroken BL
The first step is to obtain singlet VEV configurations which preserve

GSM U(1)B L SU(4) .
Here we keep the hidden sector SU(4) unbroken which is needed for gaugino condensation.
Let us now identify a BL generator. An obvious option would be to use the BL of SO(10).
SO(10)
This however leads to anomalous BL symmetry, tB L tanom = 0. It is possible to modify this
186
Table 8.1
BL charges of the relevant matter fields
Field
BL charges
qi
{ 13 , 13 , 13 }
u i
di
{ 13 , 13 , 13 }
di
i
i
ei
{ 23 , 23 , 23 , 13 }
{0, 1, 1, 0, 0, 0, 1, 1}
{0, 0, 0, 1, 0}
{1, 1, 1}
{ 13 , 13 , 23 , 13 , 13 , 23 , 23 }
generator such that the resulting BL is non-anomalous and the BL charges for the members
of the 16-plets are the standard ones. Requiring further that the hidden sector SU(2) doublets hi
be neutral under U(1)B L , fixes15

2 2 2
1 1 1 1
tB L = 0, 1, 1, 0, 0, , ,
(8.2)
, , , , 0, 0, 0, 0 .
3 3 3
2 2 2 2
The BL charges of matter fields are shown in Tables 8.1, D.10D.12. The qi and u i states
have the standard charges, while only four out of seven di have the right charge (1/3) to be
identified with the down type anti-quarks. The di states with exotic BL charges as well as
one linear combination of the di s with charge 1/3 pair up with four di s and decouple from
the low energy spectrum. Similar considerations apply to the lepton sector. The lepton doublets
carry charge 1, while the Higgs doublets are neutral. One pair of i and i with qB L = 0 must
remain in the massless spectrum and is identified with the physical Higgs bosons.
Among the 69 SM singlets si , 30 are neutral under BL, 21 have charge +1 and 18 have
charge 1 (cf. Table D.10). This excess of positively charged si leads to a net number of three
right-handed neutrinos.
Let us now consider configurations in which only states neutral under GSM U(1)B L
[SU(4)] are allowed to develop VEVs. Such states include si with zero qB L and the SU(2)
doublets hi . For our purposes, it suffices to restrict ourselves to a certain subset of these fields.
In particular, we assume that
s2 , s5 , s7 , s9 , s20 , s23 , s34 , s41 , s48 , s58 , s59 , s62 , s65 , s66 ,
h1 , h3 , h6 , h8 , h9 , h10 , h11 , h12 , h13
= 0,
(8.3)
while
{si } = {s1 , s3 , s12 , s14 , s16 , s18 , s19 , s22 , s24 , s39 , s40 , s53 , s54 , s57 , s60 , s61 ,
h2 , h4 , h5 , h7 , h14 }
(8.4)
develop non-zero VEVs. We find that such a configuration is D-flat since every field from {si }
enters a gauge invariant monomial consisting exclusively of {si } states (Table D.8). Also, it is
possible to construct a monomial out of these states which has a negative net anomalous charge
(Table D.9). The set {si } does not represent an F -flat direction. To preserve supersymmetry, we
15 This is the only phenomenologically viable U(1)
B L generator, up to an irrelevant component in the t9 direction.
187
assume that there exist non-trivial field configurations in {si } with vanishing F -terms. Then, as
described in Section 6, complexified gauge transformations allow us to satisfy Da = 0 at the
same time. The set {si } breaks all extra U(1)s but U(1)B L .
8.2. Decoupling the exotic states
The first question is whether it is possible to decouple all of the exotic states by giving VEVs
to the set {si } only. To answer this question, we recalculate the mass matrices for the exotic
matter. The relevant superpotential couplings are of the form
ij
ij
W = xi xj Mx (s )
with Mx (s ) =
si1 sin ,
(8.5)
and xi , xj being the vector-like pairs. Including the couplings up to order 10, the resulting mass
matrices are
0 0 s 6 0 0 s 6 s 6
0 0 s 6 0 0 s 7 s 7
ij
Md (s ) =
(8.6)
,
0 0 s 6 0 0 s 7 s 7
s 8
3
s
s
ij
M (s ) = s
0 0 s 6 s 6 0 0
0 0 0 0 s 8 0
0 0 0 0 s 6 0
0 0 0 0 s 6 0
s 8
s 8
0
s 6
0
s 6
0
0
0
0
0
s 6
6
s
(8.7)
Here the columns in Md correspond to dj and the rows to di ; in M , the columns correspond
ij
ij
ij
ij
to j and the rows to i . Mm (s ), Ms (s ), Mf (s ), and Mw (s ) are given by Eqs. (D.5)(D.8)
in Appendix D.4. As before, an entry s N implies that there is an allowed coupling involving N
states si . For instance, the d1 , d3 mass term includes
ij
Wd1 d3 = d1 d3 (s16 s40 h4 h14 h5 h14 + ).
ij
(8.8)
Although the form of the mass matrices is quite restricted, all of them have the maximal rank,
apart from the i , j matrix whose rank is 4. This means that all of the exotic states are decoupled
is massless, as required.
and one Higgs pair ,
Clearly, some of the zeros of the mass matrices are dictated by the BL symmetry (see Tables 8.1, D.11, D.12). The massless down type anti-quarks are 3 linear combinations of the 4 di
states with qB L = 1/3, namely d1 , d2 , d4 and d5 . The remaining linear combination couples
to d4 and becomes heavy. Likewise, the physical lepton doublets are the 3 linear combinations
of 2 , 3 , 7 and 8 which do not couple to 4 . Interestingly, this type of structure has recently
been explored in the context of orbifold GUTs [24]. It was shown that a mixing between chiral
and vector-like states can lead to realistic flavour patterns.
Not all texture zeros can be understood from the BL symmetry. For instance, BL does not
forbid the 1 , j >1 couplings such that additional input is needed. As we shall see, these zeros
are crucial for identification of the Higgs doublets.
188
Fig. 12. An idealized picture of gauge-top Yukawa unification. Here i = gi2 /(4 ), t = Yt2 /(4 ) and we have assumed
that Yt = gi at the GUT scale.
8.3. Higgs doublets and flavour structure

In our model, the only renormalizable BL conserving Yukawa coupling which involves SM
matter is
W = gq1 u 1 1 .
(8.9)
This is a superpotential of the type

(8.10)
U 1 U2 U3
with the untwisted superfields Ui formed out of the compactified components of the E8 gauge
multiplets in 10D. For example, for the scalar components we have U 1 A4 + iA5 , etc., where
Ai are the gauge field components in the compact directions. This can be understood by recalling
that the gauge supermultiplet in 10D decomposes into 1 vector and 3 chiral N = 1 multiplets in
4D. The above superpotential results from the kinetic term of the gauge supermultiplet in 10D
with the corresponding Yukawa coupling being the gauge coupling at the string scale.
As long as 1 has a significant component in the physical up-type Higgs doublet, the superpotential (8.9) naturally leads to a heavy top quark. The top Yukawa coupling is then of the order
of the gauge coupling at the string scale,
Yt g.
(8.11)
This remarkable top Yukawa-gauge unification (Fig. 12) stems from the fact that the top quark
is a gaugino in 10D. The other quark Yukawa couplings vanish at the renormalizable level. We
note that a large top Yukawa coupling has also been obtained in earlier analyses of the fermion
masses [22,8587].
This attractive mechanism is at work when 1 u . Inspection of the i j mass matrix (8.7)
shows that the first 3 rows are linearly dependent and the 1 1 coupling appears only at order
5 while the i 1 couplings with i = 2, 3 occur already at order 3. Thus, if the relevant si VEVs
are below the string scale, one expects at least mild suppression of the (1, 1) entry. Then, the
massless up-type Higgs is dominated by 1 ,

u 1 +
(8.12)
i i , |i | 1,
i=2,3
189
whereas the down-type Higgs is a linear combination of 4 and 5 ,

d = 4 a + 5 b,
(8.13)
with |a|, |b| of order one. Our choice of the vacuum configuration si (Eq. (8.4)) was in part
motivated by these considerations.
At this stage, the Higgs doublets u,d are massless. In contrast to the conventional 4D GUTs,
no finetuning is required to keep the doublets light while decoupling the color triplets. If the BL
symmetry gets broken at an intermediate scale, a small -term will be generated.
Having identified the Higgs doublets and the top quark, we turn to the discussion of the remaining Yukawa couplings. The relevant superpotential is
ij
WYukawa = Yu (s )u qi u j + Y ia (s )d qi da + Yeib (s )d ei b ,
(8.14)
d
where a {1, 2, 4, 5} and b {2, 3, 7, 8}. The Yukawa matrices at order 10 are
g s 6 s 4
ij
Yu (s ) = s 3 0 s 7 ,
s 7
0
Ydia (s ) = s 5
0
0
Yeib (s ) = s 5
0
s 7
s 6
0
s 5
s 6
0
s 5
s 2 s 2
s 5 s 5 ,
s s
0 0
0 0.
(8.15)
(8.16)
(8.17)
0 0
The low-energy 3 3 Yd and Ye Yukawa matrices16 are obtained by integrating out one linear
combination of da (a = 1, 2, 4, 5) which pairs up with d4 and one linear combination of b
(b = 2, 3, 7, 8) which pairs up with 4 . Their precise form depends on various coefficients, so let
us only discuss their main features.
The quark Yukawa matrices have the full rank such that, in general, there are no massless
eigenstates. The lepton Yukawa matrix has rank 2 implying that the electron is massless to order
8 in the superpotential. Also, there appears one massless pair of f -plets in the hidden sector,
which somewhat lowers the gaugino condensation scale.
Assuming that the relevant si have VEVs below the string scale, the Yukawa couplings are
hierarchical and resemble the FroggattNielsen structure [40]. However, the hierarchy appears
due to the string selection rules rather than the U(1) charge assignment only as in the original
FroggattNielsen mechanism.
It is remarkable that the up-type quarks tend to be heavier than the down-type quarks, which
in turn are heavier than the leptons. We also note that the Yukawa matrices generally contain
non-trivial CP phases due to complex si VEVs.
8.4. Proton stability and BL breakdown
The BL symmetry enforces absence of renormalizable operators leading to proton decay.
However, non-renormalizable BL conserving operators such as
W = ij(1)kl qi qj qk l + ij(2)kl u i u j dk el
16 Here we neglect corrections to the Khler potential.
(8.18)
190

(1)
(2)
also induce proton decay. The resulting constraint on ij kl and ij kl involving the first two generations is very tight [88,89],
108
(8.19)
.
MP
The operators (8.18) are induced both directly and by integrating out the vector-like matter.
For example, integrating out a heavy pair d2 d6 from the superpotential W = q2 q2 d2 + q3 3 d6
yields W q2 q2 q3 3 . These operators can be suppressed either by tuning the si VEVs or by an
additional, perhaps discrete, symmetry [90,91]. This issue will be discussed elsewhere.
Breaking BL is a difficult issue. It has to occur at an energy scale MB L well below Mstring .
In the following, we assume that the scale MB L is generated dynamically, without breaking
SUSY.
BL breaking VEVs fill in the zeros of the mass matrices, in particular, (8.7). This generates
the -term, W = u d . Assuming that apart from si , only states charged under BL get nonzero VEVs, its magnitude is (n 1)
ij(1,2)
kl
MBn+1
L .
n
Mstring
(8.20)
For an intermediate scale MB L , this can give a phenomenologically viable -term.

BL breakdown generates masses for the right-handed neutrinos. Our construction has the
necessary ingredients for the seesaw, i.e., neutrino Yukawa couplings and large Majorana neutrino masses. A detailed analysis of this issue will be presented elsewhere.
Finally, small R-parity violating couplings are generated. Their magnitude is given by
(MB L /Mstring )m with m depending on the type of the coupling. For m 2, an intermediate
scale MB L suppresses proton decay sufficiently [92].
To conclude, in this section we have studied a vacuum configuration with conserved BL at
the string scale. This suppresses renormalizable R-parity violating couplings as well as the term. Furthermore, flavour structures la FroggattNielsen arise as a consequence of the string
selection rules.
9. Outlook
Guided by the idea of local grand unification we have constructed an orbifold compactification of the heterotic string which leads to the supersymmetric Standard Model gauge group and
particle content. The model has large vacuum degeneracy. For certain vacua with unbroken BL
symmetry, the resulting phenomenology is particularly attractive. In this case, one pair of Higgs
doublets is massless automatically, with the subsequently generated -term being due to BL
breaking. The top quark Yukawa coupling is of the order of the gauge coupling and the arising
pattern of Yukawa couplings is reminiscent of the FroggattNielsen textures.
These results can be the first steps towards a fully realistic theory. They immediately lead,
however, to further questions which concern detailed properties of SUSY vacua, BL breakdown, incorporation of the seesaw mechanism, identification of R-parity and proton decay.
Furthermore, effects of string threshold corrections and other contributions on gauge coupling
unification have to be studied. Eventually, one would like to determine quantitatively the Yukawa
couplings for specific supersymmetric vacua.
On the conceptual side, a deeper understanding of the decoupling of exotic states is particularly desirable. Orbifolds often represent special points in the moduli space of more general
191
CalabiYau compactifications. Non-zero vacuum expectation values of specific Standard Model

singlets correspond to other points in the moduli space where the orbifold singularities have been
blown up. Since these vacuum expectation values also generate mass terms, at least some of the
unwanted exotic states should be absent in compactifications on smooth manifolds. The orbifold
limit of CalabiYau compactifications is well understood for the standard embedding [93] but
remains to be studied in detail for non-standard embeddings which are relevant to the models
presented in this paper.
Finally, it is important to search for other models in the framework of the E8 E8 and the
SO(32) heterotic string with localized 16-plets of SO(10) [94,95]. It would also be very interesting to understand the connection between orbifold compactifications and compactifications on
CalabiYau manifolds endowed with vector bundles [36,37,9698], which have many phenomenologically appealing features.
Acknowledgements
We would like to thank R. Blumenhagen, M. Cvetic, A. Hebecker, T. Kobayashi, W. Lerche,
M. Lindner, J. Louis, F. Plger, S. Raby, S. Ramos-Snchez, S. Stieberger, S. Theisen, P.K.S. Vaudrevange and in particular H.P. Nilles for valuable discussions. We are indebted to P.K.S. Vaudrevange for pointing out an error in the string selection rules presented in an earlier version of
this paper.
This work was partially supported by the EU 6th Framework Program MRTN-CT-2004503369 Quest for Unification and MRTN-CT-2004-005104 ForcesUniverse as well as the
virtual institute VIPAC of the Helmholtz society.
Appendix A. Sample calculations
In this appendix, we present details of the spectrum calculation for our model. These calculations are straightforward but tedious. For practical purposes, it is convenient to automatize them
by means of a computer algebra system.
A.1. Gauge group
The 4D gauge group is obtained by subjecting the E8 roots p (p 2 = 2) to the projection
conditions
p V6 Z,
(A.1a)
p W2 Z,
(A.1b)
p W3 Z.
(A.1c)
Consider condition (A.1a). The roots of the first E8 surviving the twist are

p 1, 1, 06 , 1, 1, 06 , 03 , 1, 1, 0, 0, 0 , 03 , 1, 1, 0, 0, 0 ,
(A.2)
where the underline denotes permutations and the superscripts indicate repeated entries. The
simple roots are by definition the smallest linearly independent positive roots (cf. [99]). For a
suitable choice of positivity, they read

{psr } = 1, 1, 06 , 1, 1, 06 ,
3

(A.3)
0 , 1, 1, 03 , 04 , 1, 1, 02 , 05 , 1, 1, 0 , 06 , 1, 1 , 06 , 1, 1 .
192
Table A.1
(k) , (k) and c(k) in Z6-II orbifolds with v6 = 16 (1, 2, 3; 0)
k
(k)
(k)
c(k)
1 (5, 4, 3)
6
1 (2, 1, 3)
3
1 (1, 2, 1)
2
1 (1, 2, 3)
3
1 (1, 2, 3)
6
1 (1, 2, 3)
3
1 (1, 2, 1)
2
1 (2, 1, 3)
3
11/36
2
3
4
2/9
1/4
2/9
i p , one finds that the simple roots in the first line correCalculating the Cartan matrix Aij = psr
sr
spond to the raising operators of two SU(2) factors whereas those in the second line correspond
to SO(10). Thus, the gauge group after twisting is SO(10) SU(2)2 , which also corresponds to
the local gauge symmetry at the origin (cf. Table D.1).
The Wilson line projections (A.1b) and (A.1c) lead to the simple roots
5

0 , 1, 1, 0 , 06 , 1, 1 , 03 , 1, 1, 03 ,
(A.4)
which correspond to the gauge groups SU(3) and SU(2) in 4D. All E8 Cartan generators survive
the projection. They give rise to the Cartan generators of SU(3) and SU(2) and to five U(1)
generators. The latter can be represented by vectors perpendicular to the simple roots (Eq. (5.5)).
The surviving subgroup of the second E8 is obtained analogously.
A.2. Untwisted sector
The untwisted sector states are obtained from the projection
p V6 q v6 Z,
p V6
/ Z,
(A.5a)
p W2 Z,
(A.5b)
p W3 Z,
(A.5c)
with p 2 = 2. There are 118 weights transforming in the first E8 which survive the first projection
(A.5a). They include

1 1 1
1 5
1 5
1 1 1
(A.6)
, , , odd
, , , , odd
,
2 2 2
2
2 2 2
2
where odd ( 12 )5 denotes 5 entries 12 with an odd number of signs. The Dynkin labels
of these representations are obtained by multiplying the above weights by the simple roots (A.3).
One finds that (A.6) is (16, 1, 2) of SO(10) SU(2)2 (cf. [100]).
1) of
The Wilson line projections eliminate some of the states such that the 4D result is (3,
SU(3) SU(2) plus non-Abelian singlets. The second E8 states are determined analogously.
A.3. T1
The first step is to solve the mass equations (2.41). For convenience, the quantities (k) ,
and c(k) appearing in (2.41) are listed in Table A.1. Consider now the (, 0) sector, i.e.
(k)
193
V(,0) = V6 . For N = 0, the shifted E8 E8 momenta psh p + V6 with p E8 E8 are17

1
1 5
1 7
{psh } =
0, 0, , odd
(A.7)
,0
.
6
2
3
Using the Dynkin labels, it is straightforward to show that these weights transform as 16 of the
local SO(10). The corresponding SO(8) lattice shifted momenta are given by

1 1
1
1 1 1
qsh =
(A.8)
, , 0; , , , ; 0 .
3 6
2
6 3 2
They describe the fermion and the boson of an N = 1 left-chiral superfield. As stated in Section 2, solutions to the mass equation in the T1 sector are twist invariant and all appear in the 4D
spectrum. The above 16-plet thus produces one complete generation of the SM matter.
Apart from the 16-plet, the massless spectrum contains one (2, 1) and two (1, 2) representations under the local SU(2)2 . Other T1 states are obtained by solving the mass equations for
Vf = V6 + n2 W2 + n3 W3 with 0 n2 1 and 0 n3 2.
A.4. T2
Consider the ( 2 , 0) sector. The local gauge group is given by the E8 roots satisfying V( 2 ,0)
p = 0 mod 1, where V( 2 ,0) = 2V6 . This yields SO(14) [SO(14)]. The corresponding massless
matter at the origin is
(14; 1) (1; 14) (1; 1) (1; 1),
(A.9)
where the non-Abelian singlets have non-zero oscillator numbers. On the right-moving side, one
has 4 solutions to the mass equations with v3 = 2v6 which combine into an N = 2 multiplet. As
explained in Sections 2 and 3, the next step is to form linear combinations of the massless states
which produce Z6 eigenstates. These are then subject to the conditions (2.62) with q {0, 12 , 1}
and psh W2 Z. The resulting spectrum is chiral.
A.5. T3
The local gauge shifts are Vf = 3(V6 + n2 W2 ) with 0 n2 1. The corresponding local
gauge groups and matter are shown in Table D.1. Again, one must impose projection conditions
(2.64), now with q {0, 13 , 1} and psh W3 Z.
A.6. T4
The T4 states are obtained analogously to the T2 states, with the only difference being the
local shift Vf = 4(V6 + n3 W3 ) and v = 4v6 .
A.7. T5
The fermionic component of the massless right mover has q4 = +1/2. The massless states are
CP-conjugates of the T1 sector and no left-chiral superfields arise in T5 .
17 An efficient way to solve automatically the mass equations is presented in [101].
194
Appendix B. Additional material for the selection rules

This appendix contains additional information on the string selection rules of Section 4 and
outlines of proofs of some statements.
B.1. Sublattices
The space group rule states that i of the space group elements have to add up to zero up to
shifts in the corresponding sublattices. For concreteness, these sublattices are listed in Table B.1.
B.2. On Eq. (4.10)
The coupling among n states must satisfy Eq. (4.9). Using the multiplication law for the space
group, one has

k1 , 1 k2 , 2 kn , n = k1 k2 kn , 1 + k1 2 + + k1 kn1 n .

The rule nr=1 kr = 0 mod 6 is then obvious. Further, by shifting the i ,

i i + 1 ki i ,
(B.1)
one can always achieve

1 + k1 2 + + k1 kn1 n 1 + 2 + + n .

Thus, i = 0 up to the sublattice ki .
(B.2)
B.3. On the selection rules in the G2 plane

Consider the
selection rule for a coupling of string states. If T1 states are involved,
G2 plane
the sublattice ki = (1 ki ) is the entire lattice and all fixed points can couple. Similarly,
there is no restriction when T3 and T2 (or T4 ) sectors are present simultaneously.
Suppose now that the coupling involves only the T2 states. The corresponding fixed points are
x, y, z (Section 3). x is at the origin and is Z6 invariant, while y and z are interchanged under
-twisting. The couplings consistent with the space group selection rule for the G2 plane are xn ,
xyz, y3 , z3 and higher couplings built out of these blocks. In terms of -eigenstates, this means
that the coupling of any number of q = 0 states to a single q = 0 state is prohibited, while
the others are allowed, i.e. {q (1) , . . . , q (n) }
/ permutations{x, 0, . . . , 0} with x = 0. Similar
considerations apply to couplings of the type T4 . . . T4 , T2 T2 . . . T4 T4 and T3 . . . T3 .
Table B.1
Sublattices (1 k ). The integers n, m are varied independently in each plane. Note that (1 6k ) = (1 k )
Sublattice
G2 plane
SU(3) plane
SO(4) plane
(1 1 )
ne1 + me2
3ne1 + me2
2ne1 + 2me2
ne3 + (n 3m)e4
ne3 + (n 3m)e4
2ne5 + 2me6
2ne5 + 2me6
(1 2 )
(1 3 )
195
Appendix C. Models with 3 local 16-plets

In this appendix, we discuss the obstacles to obtaining 3 equivalent families of 16-plets in
ZN orbifold models with N 6. Triplication of families could in principle be a result of the
presence of 3 equivalent fixed points, which support SM matter in the first twisted sector T1 .18
We however find that this simple possibility cannot be realized, at least in ZN6 .
First of all, in the Z3 orbifold one does not have a local 16-plet because one cannot break E8
to SO(10) by a Z3 twist. Then, in Z4 orbifolds there is no triplication due to geometry, i.e. the
number of equivalent fixed points is under no circumstances divisible by 3. The next simplest
possibility is the Z6 which we examine in detail below.
For Z6 orbifolds, all possible local shifts V6 are listed in [55], together with the corresponding
local groups and local T1 states. Among them, there are only 5 local shifts V6 which have a local
SO(10) and a 16-plet. There are 3 of them in Z6-I models [v6 = 16 (1, 1, 2)],
V6 =

1

1

1
2, 2, 2, 05 2, 1, 1, 05 , 3, 3, 2, 05 2, 2, 06 , 4, 1, 1, 05 08 ,
6
6
6
(C.1)
and 2 in Z6-II models [v6 = 16 (1, 2, 3)],

1

1
2, 2, 2, 05 1, 1, 06 , 3, 3, 2, 05 2, 07 .
(C.2)
6
6
These local shifts can be accompanied by Wilson lines W2 , W2 , and W3 , depending on the geometry of the orbifold [102].
We demand that the SO(10) be broken to SU(3) SU(2) U(1)2 by the orbifold action.
This requires at least two different Wilson lines. The Z6-I models allow for only one Wilson
line W3 , which destroys the triplication, and hence we discard them. The Z6-II models allow for
combinations of (W2 , W2 ), (W2 , W3 ), and (W2 , W2 , W3 ). Among them, only the first one can
produce three equivalent fixed points with local SO(10) symmetry and a 16-plet (cf. Fig. 9). We
therefore concentrate on these models, namely,
V6 =

1
W2 = any,
W2 = any,
2, 2, 2, 05 1, 1, 06 ,
(C.3)
6

1
V6 = 3, 3, 2, 05 2, 07 ,
(C.4)
W2 = any,
W2 = any.
6
Naively, one may think that the number of models to be studied is enormous. However, employing symmetry transformations of the local shifts and Wilson lines which produce equivalent
models, one can show that most of the models are redundant. These symmetries, which include
lattice translations and Weyl reflections, have been used in Ref. [103] for a systematic classification of inequivalent models in Z3 orbifolds. We have performed a similar classification of the
Z6 models and found that there are at most 69 inequivalent models of type (C.3) and at most 129
inequivalent models of type (C.4). At this stage, we have only required modular invariance and
SO(10) breakdown to SU(3) SU(2) U(1)2 .
As the next step, we have studied the massless spectrum of these models and identified quantum numbers of exotic states. Remarkably, we found that all of these models contain exotic states
V6 =
18 One could also entertain the possibility of obtaining 3 equivalent families from higher twisted sectors. However, such
states are subject to additional projection conditions which usually destroy either the equivalence of families or their
GUT structure.
196
which are chiral with respect to SU(3)c SU(2)L U(1)Y .19 Such states cannot be decoupled
and, therefore, the low energy spectrum contains exotic particles beyond the MSSM. We thus
conclude that geometric triplication of 16-plets is not possible in ZN6 orbifolds.
Appendix D. Tables
D.1. States of the model of Section 5
D.1.1. Survey of local GUTs
Table D.1
Local GUT groups and representations. Non-Abelian singlets and U(1) factors are omitted. The brackets [ ] indicate
subgroups of the second E8 . For different k, n3 and n2 the groups are in general embedded differently into E8 . The local
GUTs can be inferred from the tables of Ref. [55]
k
n3
n2 = 0
n2 = 1
SO(10) SO(4) [SO(14)]

(16, 1, 1; 1) 2 (1, 2, 1; 1) (1, 1, 2; 1)
SO(12) [SO(8) SU(4)]
1) (1; 1, 4)
(1; 8,
SU(7) [SO(8) SU(4)]
(1; 1, 4)
SO(8) SU(4) [SU(7)]

(1, 4; 1)
SO(8) SU(4) [SU(7)]
(1, 4; 1)
SO(8) SU(4) [SO(10) SO(4)]
1, 1, 2)
(1, 4;
0...2
SO(14) [SO(14)]
(14; 1) (1; 14)
SO(14) [SO(14)]
(14; 1) (1; 14)
SO(14) [SO(14)]
(14; 1) (1; 14)
E7 SU(2) [SO(16)]
(1, 2; 16)
SO(16) [E7 SU(2)]

(16; 1, 2)
SO(14) [SO(14)]
(14; 1) (1; 14)
SO(14) [SO(14)]
(14; 1) (1; 14)
SO(14) [SO(14)]
(14; 1) (1; 14)
D.1.2. Spectrum of the model of Section 5

Table D.2
All SM non-singlet representations in terms of left-chiral states. The U(1) charges refer to the basis of generators (5.5).
The H -momenta Ri are listed for the bosonic components
e1
n3
n2
n2
R1
R2
R3
irrep
u 1
(1, 1; 1, 1)
1; 1, 1)
(3,
q1
(3, 2; 1, 1)
qY
1
23
1
6
q2
q3
q4
q5
q6
q7
q8
q9
1
2
1
2
1
2
12
1
2
1
2
12
12
12
12
12
1
2
19 In SO(10), there are two distinct choices of U(1) which exchange the definitions of up-type and down-type rightY
handed quarks. We have checked both possibilities.
197
Table D.2 (continued)

k
n3
n2
n2
R1
R2
R3
irrep
qY
q2
12
1
2
16
12
3
2
1
2
3
2
1
2
12
3
2
1
2
3
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
6
1
6
1
6
1
6
1
6
1
6
16
5
6
16
5
6
16
5
6
16
5
6
16
5
6
16
5
6
14
14
14
14
14
14
7
12
7
12
7
12
7
12
7
12
7
12
5
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
5
12
5
12
5
12
5
12
5
12
5
12
1
12
1
12
1
12
1
12
1
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
5
12
1
12
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
23
1
3
1
6
23
1
3
2
3
23
16
16
1
3
1
3
2
3
23
16
16
1
3
1
3
13
1
6
1
2
13
1
6
q3
q4
1
1
(1, 2; 1, 1)
(1, 2; 1, 1)
e2
16
13
12
(1, 1; 1, 1)
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
(1, 2; 1, 1)
1; 1, 1)
(3,
(3, 2; 1, 1)
12
23
1
3
1
6
(1, 1; 1, 1)
(1, 2; 1, 1)
1; 1, 1)
(3,
12
23
(1, 1; 1, 1)
1
3
1
6
1
2
1
2
(1, 2; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
12
1
2
(1, 2; 1, 1)
(1, 1; 1, 1)
12
(1, 1; 1, 1)
1
2
(1, 2; 1, 1)
(1, 1; 1, 1)
12
(1, 1; 1, 1)
1
2
1
3
1
6
16
1
3
16
16
1
6
1
6
13
13
1
6
1
6
1
6
1
6
1
3
13
1
6
1
6
1
2
1
2
12
1
2
1
2
12
1
2
13
23
13
13
23
23
2
u 2
d1
q2
e3
3
u 3
d2
q3
s1+
m1
s1
s2
s2+
m2
s3
s3+
m3
s4+
m4
s4
s5
s6
s5+
s6+
m5
m6
s7
s8
s7+
s8+
m7
m8
d3
d1
1
2
1
2
4
1; 1, 1)
(3,
1; 1, 1)
(3,
(3, 2; 1, 1)
(1, 1; 1, 1)
(1, 2; 1, 1)
(1, 1; 1, 1)
12
(1, 1; 1, 1)
12
(1, 1; 1, 1)
(1, 1; 1, 1)
1
2
1
2
(1, 2; 1, 1)
(1, 2; 1, 1)
(1, 1; 1, 1)
12
(1, 1; 1, 1)
12
(1, 1; 1, 1)
(1, 1; 1, 1)
1
2
1
2
(1, 2; 1, 1)
(1, 2; 1, 1)
1; 1, 1)
(3,
1
3
1
3
12
(3, 1; 1, 1)
(1, 2; 1, 1)
0
0
1
2
0
0
0
0
1
2
1
2
0
0
0
0
1
2
1
2
q5
q6
1
1
0
1
1
0
2
3
1
6
5
6
2
3
1
6
5
6
16
16
23
23
5
6
5
6
1
6
16
23
23
5
6
5
6
16
q7
2
3
2
3
1
3
q8
q9
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
23
13
0
(continued on next page)
198
2
n3
n2
n2
0
0
(1, 2; 1, 1)
1; 1, 1)
(3,
1; 1, 1)
(3,
(1, 1; 1, 1)
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
d4
d5
1
2
13
13
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
1
3
1
3
1
3
13
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
(1, 2; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
1
3
1
3
23
13
1; 1, 1)
(3,
23
23
23
23
23
23
23
23
13
13
13
13
13
13
13
13
0
0
(3, 1; 1, 1)
1; 1, 1)
(3,
(3, 1; 1, 1)
d2
d7
d3
6
5
1
2
1
2
7
8
1
2
d4
(1, 2; 1, 1)
d6
1
6
1
3
1
3
1
3
16
13
23
23
23
23
23
1
2
1
2
1
2
1
3
1
3
1
2
23
13
13
13
13
13
(1, 2; 1, 1)
13
s9
s9+
s10
+
s10
s11
+
s11
s12
+
s12
s13
+
s13
s14
+
s14
s15
+
s15
s16
+
s16
q2
4
qY
R2
5
3
R3
irrep
R1
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 2; 1, 1)
(1, 2; 1, 1)
(1, 2; 1, 1)
(1, 2; 1, 1)
(3, 1; 1, 1)
12
1
2
12
1
2
1
2
1
2
12
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
q3
q4
q5
q6
q7
q8
13
23
13
23
13
1
2
16
23
2
3
2
3
1
6
13
13
13
13
13
23
23
2
3
2
3
2
3
13
13
1
3
1
3
1
3
1
2
1
6
1
3
13
13
1
6
1
2
1
2
1
2
3
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
12
1
2
1
2
1
2
12
14
1
4
14
1
4
1
4
1
4
14
1
4
1
4
1
4
1
4
1
4
1
4
1
4
1
4
1
4
1
2
1
2
0
0
0
0
0
1
2
1
2
1
2
3
2
0
0
1
2
1
2
1
2
1
2
12
32
1
4
14
1
4
14
1
4
14
1
4
14
1
4
14
1
4
14
1
4
14
1
4
14
23
23
23
23
2
3
1
6
16
16
23
1
3
1
3
1
3
1
3
1
3
2
3
2
3
23
23
23
1
3
1
3
13
13
13
1
3
1
3
1
3
1
3
1
3
16
16
16
1
3
1
2
1
2
1
2
3
2
1
2
12
1
2
12
1
3
1
3
1
3
13
12
1
2
1
2
12
13
q9
12
32
1
2
1
2
1
2
1
2
1
2
3
2
1
3
1
6
1
6
1
6
13
0
12
12
12
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
Table D.3
Same as Table D.2 for SM singlets
k
n3
n2
n2
R1
R2
R3
s1
s2
irrep
qY
q2
q3
q4
q5
q6
q7
q8
q9
(1, 1; 1, 1)
12
(1, 1; 1, 1)
1
2
1
2
12
199

k
n3
n2
n2
R1
R2
R3
irrep
qY
q2
q3
q4
q5
q6
q7
q8
q9
f1
1)
(1, 1; 4,
1
2
1
2
s3
f2
12
12
12
52
(1, 1; 1, 1)
1)
(1, 1; 4,
13
2
3
2
3
13
13
13
13
13
2
3
2
3
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
(1, 1; 1, 1)
52
(1, 1; 1, 1)
12
12
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
1
2
1
2
12
12
1
2
(1, 1; 1, 1)
1
2
1
2
1
2
1
2
1
2
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
6
16
16
23
1
3
1
3
16
16
16
23
1
3
1
3
23
1
3
1
3
23
1
3
1
3
13
16
13
16
13
2
3
2
3
13
13
13
13
2
3
2
3
13
23
23
1
3
2
3
23
1
3
16
1
3
2
3
1
3
1
6
16
16
1
3
2
3
1
3
1
3
2
3
13
13
2
3
1
3
s22
s23
w2
s24
s25
h2
s26
s27
h3
s28
s29
h4
16
16
16
5
6
5
6
11
6
11
6
16
16
16
5
6
5
6
11
6
11
6
16
16
16
5
6
5
6
5
6
1
6
16
16
5
6
5
6
5
6
1
6
16
16
16
16
16
s30
13
23
13
13
13
23
23
23
s4
s5
s6
s7
s8
s9
s10
s11
s12
s13
s14
s15
s16
s17
s18
s19
w1
s20
s21
h1
s31
s32
s33
(1, 1; 1, 1)
(1, 1; 1, 1)
12
12
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
1
2
1
2
12
12
1
2
(1, 1; 1, 1)
(1, 1; 1, 1)
16
1
3
1
3
2
3
23
1
3
1
3
16
1
3
1
3
2
3
23
1
3
1
3
1
6
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
13
13
13
13
(1, 1; 6, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 6, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
1
2
1
2
1
2
1
2
1
2
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
2
3
2
3
2
3
2
3
2
3
2
3
(1, 1; 1, 1)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
0
0
0
0
0
0
52
0
0
0
0
0
0
56
56
56
56
56
56
56
56
56
56
56
56
5
6
5
6
5
6
5
6
5
6
5
6
0
0
2
3
2
3
2
3
2
3
0
0
0
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
200

k
n3
n2
n2
s34
h5
f3
1
2
1
2
1
2
s35
R3
R2
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
(1, 1; 1, 1)
(1, 1; 1, 2)
1)
(1, 1; 4,
(1, 1; 1, 1)
13
1
6
1
6
13
1
6
1
6
56
1
6
2
3
1
6
16
16
5
6
16
16
5
6
16
16
1
2
1
2
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 2)
1
2
1
2
1
2
12
1
2
1
2
1
2
12
(1, 1; 1, 1)
12
1
2
12
1
2
12
1
2
12
1
2
(1, 1; 1, 1)
h6
1
2
1
2
1
2
1
2
s42
s43
s44
s45
s46
s47
s48
w3
1
2
1
2
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
0
s36
f1
s37
s38
f2
s39
s40
s41
s51
s52
h9
h10
13
13
13
13
13
13
13
13
s53
23
13
23
23
23
23
23
23
23
23
13
13
13
13
13
13
13
13
s49
s50
h7
h8
h11
f3
s54
h12
f4
s55
s56
1
2
1
2
s57
irrep
R1
0
0
0
0
0
0
qY
(1, 1; 1, 1)
(1, 1; 4, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 4, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 6, 1)
q2
q3
q4
0
1
2
1
2
1
2
1
2
0
12
12
12
1
2
1
2
12
1
2
1
2
12
0
0
0
53
5
6
5
6
53
5
6
5
6
5
6
5
6
53
5
6
5
6
56
56
56
56
56
56
56
q7
13
13
13
13
13
1
6
1
3
13
1
6
1
3
13
13
2
3
1
3
2
3
13
13
2
3
1
3
13
1
6
q8
q9
23
13
23
1
3
23
23
1
3
23
4
3
23
1
3
43
2
3
2
3
43
2
3
2
3
2
3
13
13
1
6
1
3
13
1
6
1
3
13
13
13
1
3
2
3
1
3
1
3
23
1
3
1
3
16
0
0
1
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
23
1
3
1
3
1
3
1
3
1
3
1
3
2
3
23
1
3
23
23
23
23
23
1
3
1
3
1
3
2
3
1
3
(1, 1; 1, 2)
(1, 1; 4, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 4, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
23
(1, 1; 1, 1)
q6
q5
2
3
2
3
2
3
2
3
5
6
1
6
2
3
56
1
6
1
6
1
6
13
1
6
1
6
1
6
1
6
1
6
1
6
1
6
56
1
6
0
0
0
0
0
0
5
3
201
s58
n3
n2
n2
R1
R2
R3
23
13
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
23
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
s59
h13
s60
s61
s62
h14
s63
s64
f4
1
2
1
2
1
2
s65
w4
s66
w5
s67
s68
s69
1
2
1
2
1
2
irrep
(1, 1; 1, 1)
qY
q2
q3
q4
q5
q6
q7
q8
16
1
2
1
2
1
2
16
56
16
16
1
3
16
16
16
5
6
23
16
5
6
1
6
5
6
1
6
16
16
16
56
56
5
3
56
56
56
56
5
3
5
6
5
6
5
6
5
6
5
6
5
6
5
6
5
6
1
3
1
3
23
1
3
1
3
1
3
23
1
3
1
3
16
1
3
1
6
1
3
1
6
1
3
2
3
1
3
43
5
6
1
6
23
16
5
6
1
6
16
1
3
1
6
1
6
1
6
1
6
1
6
5
6
1
6
1
6
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
1
3
16
13
1
6
13
1
6
13
2
3
13
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 2)
(1, 1; 1, 1)
(1, 1; 1, 1)
1)
(1, 1; 4,
(1, 1; 1, 1)
(1, 1; 6, 1)
(1, 1; 1, 1)
(1, 1; 6, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
(1, 1; 1, 1)
0
1
2
1
2
1
2
1
2
0
1
2
1
2
1
2
12
1
2
1
2
1
2
1
2
2
3
1
3
2
3
43
2
3
1
3
2
3
2
3
1
3
23
1
3
23
1
3
2
3
23
4
3
q9
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
Table D.4
Anomalous charges of the SM singlets si and hi
Field
Charge
Field
Charge
Field
Charge
s1
s2
s3
s4
s5
43
s6
0
1
s7
2
3
s8
s9
43
s10
s11
s12
43
s13
s14
2
3
s15
s16
43
s17
s18
s19
s20
s21
s31
2
3
2
3
4
3
5
3
5
3
5
3
s33
2
3
4
3
2
3
1
3
1
3
2
3
5
3
s34
h5
2
3
4
3
2
3
4
3
5
3
5
3
2
3
5
3
s35
s36
s37
s38
s39
73
s40
13
s41
4
3
h6
s42
43
s43
s44
1
3
5
3
s45
43
h1
s24
s26
s28
s30
s22
s25
s27
s29
s23
h2
h3
h4
s32
202

Field
Charge
Field
Charge
5
3
Field
Charge
4
3
s46
s47
s49
s50
h7
h8
s51
s52
h9
h10
s53
h11
s54
h12
s56
23
s57
s59
h13
s61
7
3
1
3
5
3
4
3
13
7
3
s63
s64
s65
5
3
5
3
1
3
4
3
13
43
s66
s67
53
s68
s69
43
4
3
s55
s58
s60
h14
s48
s62
D.2. Monomials
Table D.5
Examples of gauge invariant monomials
s1 s19(23) s20(24) s57(60)
s4(11) s26(28) s57(60)
s5(9,12,16) s21(25) s68
s31(33) s55
s36(38) s63
s41 s57(60)
s44(47) s67
s2 s18(22) s20(24) s57(60)

s4(11) s27(29) s39 s40
s6(10,13,17) s7(14) s55
s34 s53(54)
s39 s59(62)
s42(45) s69
s48 s65(66)
s3 s6(10,13,17) s26(28) s57(60)

s5(9,12,16) s8(15) s56
s30(32) s56
s35(37) s64
s40 s58(61)
s43(46) s68
s49(51) s50(52)
Table D.6
Examples of gauge invariant monomials carrying negative net anomalous charge
s5(9,12,16) s5(9,12,16) s39 s55 s56
s34 s39 s39 s43(46) s55
s39 s39 s40 s42(45) s63
s5(9,12,16) s34 s39 s49(51) s55

s35(37) s39 s40 s42(45) s57(60)
s39 s55 s56 s63 s67
s34 s35(37) s39 s40 s67

s35(37) s55 s56 s57(60) s67
Table D.7
Examples of gauge invariant monomials carrying negative net anomalous charge for si from U, T2 , T4
s34 s35(37) s39 s40 s67
s35(37) s55 s56 s57(60) s67
s34 s39 s39 s43(46) s55

s39 s39 s40 s42(45) s63
s35(37) s39 s40 s42(45) s57(60)

s39 s55 s56 s63 s67
203
Table D.8
Examples of gauge invariant monomials involving only the singlets of Eq. (8.4)
s12 s40 s40 s61 h2 h4 h7 h14
s16 s40 s40 s61 h2 h4 h7 h14
s1 s16 s19 s40 s57 h2 h4 h14 h14
s1 s16 s19 s40 s60 h2 h4 h14 h14
s14 s16 s40 s40 s61 h2 h4 h5 h14
s16 s39 s40 s40 s61 h2 h2 h4 h14
s40 s40 s53 s60 s61 h2 h2 h5 h14
s40 s40 s54 s60 s61 h2 h2 h5 h14
s16 s24 s39 s40 s40 s61 h2 h2 h4 h7
s3 s24 s40 s60 h2 h4 h4 h5 h7 h7
s16 s18 s19 s24 s24 s39 s54 s60 s60 h2 h2 h4 h5
s16 s19 s22 s24 s24 s39 s54 s60 s60 h2 h2 h4 h5
Table D.9
Examples of gauge invariant monomials carrying negative net anomalous charge and involving only the singlets of
Eq. (8.4)
s12 s12 s39 s40 h5 h14
s12 s16 s39 s40 h5 h14
s12 s12 s24 s39 s40 h5 h7
s12 s12 s24 s39 s39 s40 h2 h5
s12 s12 s39 s40 s40 s61 h5 h14
s3 s40 s40 s57 h5 h5 h7 h7
s3 s40 s40 s60 h5 h5 h7 h7
s12 s12 s14 s40 s40 h5 h5 h7 h14
s12 s12 s40 s53 s57 h5 h5 h14 h14
s12 s12 s40 s54 s57 h5 h5 h14 h14
s3 s12 s39 s40 s57 h4 h5 h5 h14
Table D.11
BL charges of the si
i
10
11
12
13
14
15
16
qB L (si+ )
qB L (si )
1
2
1
2
1
2
1
2
12
12
12
12
1
2
3
2
3
2
1
2
1
2
32
3
2
12
32
3
2
1
2
1
2
1
2
1
2
1
2
1
2
32
3
2
1
2
1
2
1
2
1
2
1
2
1
2
D.3. BL charges
Table D.10
BL charges of the si
i
qB L
1
0
2
0
3
0
4
1
5
0
6
1
7 8
0 1
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
0 1 1 0 1 0 1 0 1 0 0 0 1 0 0
i
24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
qB L 0 1 1 1 1 1 1 1 1 1 0 1 1 1 1 0 0 0 1 1 1 1 1
i
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69
qB L 1 0 1 1 1 1 0 0 1 1 0 0 0 0 0 0 1 1 0 0 1 1 1
204
Table D.12
BL charges of the mi
i
qB L (mi )
12
12
1
2
1
2
12
1
2
12
1
2
D.4. Mass matrices

D.4.1. Mass matrices for generic singlet vevs
0
s5
s5
s5
s5
0
0
s5
s5
0
0
s6
s6
s1
s4
s1
s5
s5
s5
s6
s6
s4
s1
s6
s5
s6
s1
s1
s5
s4
s5
s4
s5
s1
s5
0
s5
s5
s5
s5
0
s5
s6
s5
s5
0
s5
s5
s5
s5
s6
s6
s6
s5
s5
s1
s6
s5
s5
s5
s5
s5
s5
s5
s5
s6
s6
s6
s6
s6
s6
s6
s1
s5
s6
s5
s5
s6
s5
s6
s5
s6
s5
s6
s5
s1
s6
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s5
s6
s5
s5
s3
s5
s5
s5
s5
s3
s3
s6
5
s
s5
ij
Mm (s) = 6
s
s1
6
s
s5
5
s
s5
1
s
s5
6
s
s5
6
s
s5
ij
Ms (s) = 5
s
s1
1
s
s5
5
s
s5
5
s
s6
Mf (s) =
0
0
s6
(D.1)
s6
s6
s6
s6
s6
s6
s6
s6
s5
s5
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s1
s5
s5
s6
s6
s6
s6
0
s5
s5
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s1
s5
s6
s5
s6
s6
s5
s6
s1
s5
s6
s1
s6
s6
s5
s5
s5
s5
s1
s5
s5
s5
s1
s5
s5
s5
s6
0
s6
s5
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s1
s5
s1
s1
s6
s5
s5
s5
s5
0
s6
s5
s5
s5
s6
s5
s1
s5
s5
0
s6
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s1
s5
s5
s5
s5
s5
s5
s6
s5
0
s1
s6
s5
0
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s5
s1
s5
s3
,
s6
s6
s5
s1
s5
s5
,
s5
s6
s5
s6
s6
s6
s6
s6
s6
,
s5
s5
s5
s5
5
s
s5
s5
s5
(D.2)
(D.3)
s1
s5
Mw (s) = s 5
5
s
s5
s5
s1
s5
s5
s5
s5
s5
s5
s3
s3
s5
s5
s3
s3
s3
205
s5
s5
s3 .
s3
(D.4)
s3
D.4.2. Mass matrices for the BL preserving vacuum of Section 8
0
0
s 8
ij
Mm (s ) =
0
0
s 5
0
0
8
s
s 8
ij
Ms (s ) =
0
0
0
0
0
ij
Mf (s ) =
0
0
s 5
s
,
7
s
s 7
0
0
s 8
0
0
s 8
0
0
s 8
0
0
0
0
0
s
s 5
s
s 5
0
0
0
0
s 5
s
s 5
s 5
0
s
0
s 5
0
0
s
0
0
s 7
0
s 7
s 7
0
s 7
0
0
s 7
0
s 7
0
0
s 8
s 8
0
0
0
0
0
0
0
0
0
0
0
0
s 6
0
0
0
0
0
0
0
0
s 6
0
0
0
0
0
0
s 8
0
0
0
s 8
0
0
0
0
0
0
0
0
0
0
0
0
s 8
0
0
0
s 8
0
0
s 6
0
0
0
0
0
0
0
0
0
0
s 6
0
0
0
s 8
0
s 8
0
s 8
0
s 8
0
0
0
s 8
0
s 8
0
s 8
0
s 8
0
0
0
0
0
0
s 6
s 5
0
s 8
0
0
s 5
0
s 8
0
0
0
0
s 8
0
0
0
0
0
0
0
s 8
0
s 5
0
0
s 8
0
s 5
0
0
s 8
0
0
0
s 8
s 8
s 8
0
0
s 5
s 5
0
0
0
0
0
0
s 8
0
0
0
0
0
s 5
s 5
s 8
0
0
0
0
0
s
s 8
s 6
0
s 8
s
s 6
0
s 6
s 6
0
0
0
0
0
s 6
s 5
s 5
0
0
s 5
s 5
0
0
s 8
0
0
s 8
0
s 8
s 8
0
0
0
0
s 5
s 5
s 8
0
0
0
s 5
s 5
s 8
0
0
0
s 5
s 5
0
s 5
s 5
0
0
0
0
0
0
0
s
s 8
s 6
s 8
s
s 6
s 5
s 5
s 8
s 8
s 6
s 6
s 6
s 6
s
s 6
,
0
0
(D.5)
0
0
0
0
0
8
s
,
0
s 6
s 6
0
(D.6)
(D.7)
206
s
s 5
Mij
s ) = s 8
w (
0
0
s 5
s
s 8
s 7
s 7
s 8
s 8
0
s 6
s 6
0
s 7
s 6
s 6
s 6
0
s 7
s 6 .
s 6
(D.8)
s 6
D.5. Survey of orbifold GUT limits

Table D.13
Survey of the orbifold GUTs in different dimensions. The bullet indicates small compact dimensions. U(1) factors and
subgroups of the second E8 are omitted
dim.
Plane
Conditions
SUSY, bulk groups
10
N = 4, E8
p W2 Z
N = 4, SO(16)
N = 4, E8
N = 4, E8
p W3 Z
N = 4, E6 SU(3)
p W2 Z
N = 4, SO(16)
p W2 Z
N = 4, SO(16)
N = 4, E8
p W2 , p W3 Z
N = 4, SU(6) SO(4)
p W3 Z
N = 4, E6 SU(3)
p 2V6 , p W3 Z
N = 2, SU(6)
p 3V6 , p W2 Z
N = 2, SU(8)
p W2 , p W3 Z
N = 4, SU(6) SO(4)
p 2V6 , p W2 , p W3 Z
N = 2, SU(3) SU(3)
p 2V6 , p W3 Z
N = 2, SU(6)
p V6 , p W2 , p W3 Z
N = 1, SU(3) SU(2) GSM
G2
SU(3)
SO(4)
Appendix E. List of frequently used symbols

ea
f
I
n2
lattice vectors, see Eq. (2.1)

fixed point, see Eq. (2.7)
left-moving fermions, see Eq. (2.10)
localization quantum number in the SO(4) plane
n2
n3
p
i
i
q
q
Ri
Vf
VN
vN
i
XL,R
Zi
zi
207
localization quantum number in the SO(4) plane

localization quantum number in the SU(3) plane
p E8 E8 : E8 E8 root lattice vector (momentum), see Eq. (2.37)
right-moving fermions, see Eq. (2.9)
complex NSR fermions, see Eq. (2.11)
q SO(8) : SO(8) weight (momentum), see Eq. (2.36)
additional quantum number in Tk>1 twisted sectors of non-prime orbifolds, see
Eq. (2.58)
invariant H -momenta, see Eq. (4.7)
twist, see Eq. (2.3)
local gauge shift, see Eq. (2.32)
gauge shift vector, see Eq. (2.18)
twist vector, see Eq. (2.4)
string coordinates, see Eq. (2.9)
complex string coordinates, see Eq. (2.11)
complex coordinates of the torus
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
H. Georgi, S.L. Glashow, Phys. Rev. Lett. 32 (1974) 438441.

J.C. Pati, A. Salam, Phys. Rev. D 10 (1974) 275289.
H. Georgi, in: C.E. Carlson (Ed.), Particles and Fields 1974, AIP, New York, 1975, p. 575.
H. Fritzsch, P. Minkowski, Ann. Phys. 93 (1975) 193266.
D.J. Gross, J.A. Harvey, E.J. Martinec, R. Rohm, Phys. Rev. Lett. 54 (1985) 502505.
D.J. Gross, J.A. Harvey, E.J. Martinec, R. Rohm, Nucl. Phys. B 256 (1985) 253.
L.J. Dixon, J.A. Harvey, C. Vafa, E. Witten, Nucl. Phys. B 261 (1985) 678686.
L.J. Dixon, J.A. Harvey, C. Vafa, E. Witten, Nucl. Phys. B 274 (1986) 285314.
L.E. Ibez, H.P. Nilles, F. Quevedo, Phys. Lett. B 187 (1987) 2532.
L.E. Ibez, H.P. Nilles, F. Quevedo, Phys. Lett. B 192 (1987) 332.
L.E. Ibez, J.E. Kim, H.P. Nilles, F. Quevedo, Phys. Lett. B 191 (1987) 282286.
J.A. Casas, E.K. Katehou, C. Muoz, Nucl. Phys. B 317 (1989) 171.
J.A. Casas, C. Muoz, Phys. Lett. B 214 (1988) 63.
Y. Kawamura, Prog. Theor. Phys. 105 (2001) 9991006, hep-ph/0012125.
G. Altarelli, F. Feruglio, Phys. Lett. B 511 (2001) 257264, hep-ph/0102301.
L.J. Hall, Y. Nomura, Phys. Rev. D 64 (2001) 055003, hep-ph/0103125.
A. Hebecker, J. March-Russell, Nucl. Phys. B 613 (2001) 316, hep-ph/0106166.
T. Asaka, W. Buchmller, L. Covi, Phys. Lett. B 523 (2001) 199204, hep-ph/0108021.
L.J. Hall, Y. Nomura, T. Okui, D.R. Smith, Phys. Rev. D 65 (2002) 035008, hep-ph/0108071.
T. Kobayashi, S. Raby, R.-J. Zhang, Phys. Lett. B 593 (2004) 262270, hep-ph/0403065.
S. Frste, H.P. Nilles, P.K.S. Vaudrevange, A. Wingerter, Phys. Rev. D 70 (2004) 106008, hep-th/0406208.
T. Kobayashi, S. Raby, R.-J. Zhang, Nucl. Phys. B 704 (2005) 355, hep-ph/0409098.
W. Buchmller, K. Hamaguchi, O. Lebedev, M. Ratz, Nucl. Phys. B 712 (2005) 139156, hep-ph/0412318.
T. Asaka, W. Buchmller, L. Covi, Phys. Lett. B 563 (2003) 209216, hep-ph/0304142.
W. Buchmller, K. Hamaguchi, O. Lebedev, M. Ratz, hep-ph/0512326.
W. Buchmller, K. Hamaguchi, O. Lebedev, M. Ratz, Phys. Rev. Lett. 96 (2006) 121602, hep-ph/0511035.
W. Lerche, D. Lst, A.N. Schellekens, Nucl. Phys. B 287 (1987) 477.
R. Bousso, J. Polchinski, JHEP 0006 (2000) 006, hep-th/0004134.
L.E. Ibez, The search for a Standard Model SU(3) SU(2) U(1) superstring: an introduction to orbifold
constructions, Based on lectures given at the XVII GIFT Seminar on Strings and Superstrings, 1987.
[30] D. Bailin, A. Love, Phys. Rep. 315 (1999) 285408.
[31] G.B. Cleaver, A.E. Faraggi, D.V. Nanopoulos, Phys. Lett. B 455 (1999) 135146, hep-ph/9811427.
[32] A.E. Faraggi, C. Kounnas, S.E.M. Nooij, J. Rizos, Nucl. Phys. B 695 (2004) 4172, hep-th/0403058.
208
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
A.E. Faraggi, S. Frste, C. Timirgaziu, hep-th/0605117.

R. Blumenhagen, M. Cvetic, P. Langacker, G. Shiu, Annu. Rev. Nucl. Part. Sci. 55 (2005) 71139, hep-th/0502005.
P. Anastasopoulos, T.P.T. Dijkstra, E. Kiritsis, A.N. Schellekens, hep-th/0605226.
V. Braun, Y.-H. He, B.A. Ovrut, T. Pantev, hep-th/0512177.
V. Bouchard, R. Donagi, Phys. Lett. B 633 (2006) 783791, hep-th/0512149.
G. Honecker, T. Ott, Phys. Rev. D 70 (2004) 126010, hep-th/0404055;
G. Honecker, T. Ott, Phys. Rev. D 71 (2005) 069902, Erratum.
M. Dine, N. Seiberg, E. Witten, Nucl. Phys. B 289 (1987) 589.
C.D. Froggatt, H.B. Nielsen, Nucl. Phys. B 147 (1979) 277.
Y. Katsuki, et al., Nucl. Phys. B 341 (1990) 611640.
M.B. Green, J.H. Schwarz, E. Witten, Cambridge Monographs On Mathematical Physics, Cambridge Univ. Press,
Cambridge, 1987, 596p.
L.J. Dixon, D. Friedan, E.J. Martinec, S.H. Shenker, Nucl. Phys. B 282 (1987) 1373.
T. Kobayashi, N. Ohtsubo, Int. J. Mod. Phys. A 9 (1994) 87126.
F. Quevedo, Lectures on superstring phenomenology, hep-th/9603074.
A. Hebecker, M. Ratz, Nucl. Phys. B 670 (2003) 326, hep-ph/0306049.
S. Hamidi, C. Vafa, Nucl. Phys. B 279 (1987) 465.
O. Lebedev, H.P. Nilles, S. Raby, S. Ramos-Sanchez, M. Ratz, P.K.S. Vaudrevange, A. Wingerter, in preparation.
A. Font, L.E. Ibez, H.P. Nilles, F. Quevedo, Nucl. Phys. B 307 (1988) 109;
A. Font, L.E. Ibez, H.P. Nilles, F. Quevedo, Nucl. Phys. B 310 (1988) 764, Erratum.
A. Font, L.E. Ibez, H.P. Nilles, F. Quevedo, Phys. Lett. B 210 (1988) 101;
A. Font, L.E. Ibez, H.P. Nilles, F. Quevedo, Phys. Lett. B 213 (1988) 564, Erratum.
A. Font, L.E. Ibez, F. Quevedo, A. Sierra, Nucl. Phys. B 331 (1990) 421474.
D. Lst, S. Theisen, Lect. Notes Phys. 346 (1989) 1346.
A. Font, L.E. Ibez, H.P. Nilles, F. Quevedo, Phys. Lett. B 213 (1988) 274.
Y. Katsuki, et al., DPKU-8904.
Y. Katsuki, et al., DPKU-8810-REV.
M. Cvetic, L.L. Everett, J. Wang, Phys. Rev. D 59 (1999) 107901, hep-ph/9808321.
F. Buccella, J.P. Derendinger, S. Ferrara, C.A. Savoy, Phys. Lett. B 115 (1982) 375.
R. Gatto, G. Sartori, Commun. Math. Phys. 109 (1987) 327.
E. Witten, Nucl. Phys. B 471 (1996) 135158, hep-th/9602070, footnote 3.
A. Hebecker, M. Trapletti, Nucl. Phys. B 713 (2005) 173203, hep-th/0411131.
G. Cleaver, M. Cvetic, J.R. Espinosa, L.L. Everett, P. Langacker, Nucl. Phys. B 525 (1998) 326, hep-th/9711178.
G. Cleaver, M. Cvetic, J.R. Espinosa, L.L. Everett, P. Langacker, Nucl. Phys. B 545 (1999) 4797, hep-th/9805133.
B.A. Ovrut, J. Wess, Phys. Rev. D 25 (1982) 409.
J. Wess, J. Bagger, Supersymmetry and Supergravity, Princeton Univ. Press, Princeton, USA, 1992, 259p.
H.P. Nilles, Phys. Lett. B 115 (1982) 193.
S. Ferrara, L. Girardello, H.P. Nilles, Phys. Lett. B 125 (1983) 457.
J.P. Derendinger, L.E. Ibez, H.P. Nilles, Phys. Lett. B 155 (1985) 65.
M. Dine, R. Rohm, N. Seiberg, E. Witten, Phys. Lett. B 156 (1985) 55.
S.H. Shenker, The strength of nonperturbative effects in string theory, presented at the Cargese Workshop on
Random Surfaces, Quantum Gravity and Strings, Cargese, France, 28 May1 June, 1990.
T. Banks, M. Dine, Phys. Rev. D 50 (1994) 74547466, hep-th/9406132.
J.A. Casas, Phys. Lett. B 384 (1996) 103110, hep-th/9605180.
P. Bintruy, M.K. Gaillard, Y.Y. Wu, Nucl. Phys. B 481 (1996) 109, hep-th/9605170.
P. Bintruy, M.K. Gaillard, Y.-Y. Wu, Nucl. Phys. B 493 (1997) 2755, hep-th/9611149.
P. Bintruy, M.K. Gaillard, Y.-Y. Wu, Phys. Lett. B 412 (1997) 288295, hep-th/9702105.
T. Barreiro, B. de Carlos, E.J. Copeland, Phys. Rev. D 57 (1998) 73547360, hep-ph/9712443.
W. Buchmller, K. Hamaguchi, O. Lebedev, M. Ratz, Nucl. Phys. B 699 (2004) 292308, hep-th/0404168.
L.E. Ibez, H.P. Nilles, Phys. Lett. B 169 (1986) 354.
L.J. Dixon, V. Kaplunovsky, J. Louis, Nucl. Phys. B 355 (1991) 649688.
P. Mayr, S. Stieberger, Nucl. Phys. B 407 (1993) 725748, hep-th/9303017.
H.P. Nilles, S. Stieberger, Nucl. Phys. B 499 (1997) 328, hep-th/9702110.
S. Stieberger, Nucl. Phys. B 541 (1999) 109144, hep-th/9807124.
V.S. Kaplunovsky, J. Louis, Phys. Lett. B 306 (1993) 269275, hep-th/9303040.
A. Brignole, L.E. Ibez, C. Muoz, in: G.L. Kane (Ed.), Perspectives on Supersymmetry, World Scientific, 1998,
pp. 125148, hep-ph/9707209.
209
[84] S.A. Abel, B.C. Allanach, F. Quevedo, L. Ibez, M. Klein, JHEP 0012 (2000) 026, hep-ph/0005260.
[85] L.E. Ibanez, Phys. Lett. B 181 (1986) 269.
[86] J.A. Casas, C. Munoz, Nucl. Phys. B 332 (1990) 189;
J.A. Casas, C. Munoz, Nucl. Phys. B 340 (1990) 280, Erratum.
[87] G. Cleaver, M. Cvetic, J.R. Espinosa, L.L. Everett, P. Langacker, J. Wang, Phys. Rev. D 59 (1999) 055005, hepph/9807479.
[88] I. Hinchliffe, T. Kaeding, Phys. Rev. D 47 (1993) 279284.
[89] H. Murayama, D.B. Kaplan, Phys. Lett. B 336 (1994) 221228, hep-ph/9406423.
[90] H.K. Dreiner, C. Luhn, M. Thormeier, Phys. Rev. D 73 (2006) 075007, hep-ph/0512163.
[91] R. Tatar, T. Watari, hep-th/0602238.
[92] H.K. Dreiner, in: G.L. Kane (Ed.), Perspectives on Supersymmetry, World Scientific, 1998, pp. 462479, hepph/9707435.
[93] J. Polchinski, String Theory, Superstring Theory and Beyond, vol. 2, Cambridge Univ. Press, Cambridge, 1998,
531p.
[94] K.-S. Choi, S. Groot Nibbelink, M. Trapletti, JHEP 0412 (2004) 063, hep-th/0410232.
[95] H.P. Nilles, S. Ramos-Snchez, P.K.S. Vaudrevange, A. Wingerter, JHEP 0604 (2006) 050, hep-th/0603086.
[96] V. Braun, Y.-H. He, B.A. Ovrut, JHEP 0604 (2006) 019, hep-th/0601204.
[97] V. Bouchard, M. Cvetic, R. Donagi, Nucl. Phys. B 745 (2006) 6283, hep-th/0602096.
[98] R. Blumenhagen, S. Moster, T. Weigand, hep-th/0603015.
[99] R.N. Cahn, Semisimple Lie Algebras and their Representations, Benjamin/Cummings, 1984, 158 p.
[100] R. Slansky, Phys. Rep. 79 (1981) 1128.
[101] P.K.S. Vaudrevange, Geometrical aspects of heterotic orbifolds, Diploma thesis, University of Bonn, 2004,
http://www.th.physik.uni-bonn.de/nilles/db/thesis/vaudrevange.ps.
[102] T. Kobayashi, N. Ohtsubo, Phys. Lett. B 257 (1991) 5662.
[103] J. Giedt, Ann. Phys. 289 (2001) 251, hep-th/0009104.
Nuclear Physics B 785 [FS] (2007) 211233
Spectrum of local boundary operators from boundary

form factor bootstrap
M. Szots a , G. Takcs b,
a Etvs University, Budapest, Hungary
b Theoretical Physics Research Group of the Hungarian Academy of Sciences, Hungary
Received 29 March 2007; accepted 23 April 2007

Available online 29 April 2007
Abstract
Using the recently introduced boundary form factor bootstrap equations, we map the complete space of
their solutions for the boundary version of the scaling LeeYang model and sinh-Gordon theory. We show
that the complete space of solutions, graded by the ultraviolet behaviour of the form factors can be brought
into correspondence with the spectrum of local boundary operators expected from boundary conformal field
theory, which is a major evidence for the correctness of the boundary form factor bootstrap framework.
2007 Published by Elsevier B.V.
1. Introduction
The bootstrap program aims to classify and explicitly solve (1 + 1)-dimensional integrable
quantum field theories by constructing all of their Wightman functions. For bulk theories, the
first stage is the S-matrix bootstrap: the scattering matrix, connecting asymptotic in and out
states, is determined from its properties such as factorizability, unitarity, crossing symmetry and
the YangBaxter equation supplemented by the maximal analyticity assumption which results in
the complete on-shell solution of the theory, i.e., the spectrum of excitations and their scattering
amplitudes (for reviews see [1,2]). The second step is the form factor bootstrap, which allows one
to determine matrix elements of local operators between asymptotic states (form factors) using
their analytic properties originating from the already known S-matrix. The form factors are then
E-mail address: takacs@elte.hu (G. Takcs).

0550-3213/$ see front matter 2007 Published by Elsevier B.V.
doi:10.1016/j.nuclphysb.2007.04.026
212
M. Szots, G. Takcs / Nuclear Physics B 785 [FS] (2007) 211233
used to build the correlation (Wightman) functions via their spectral representations, yielding a
complete off-shell description of the theory (see [3] for a review).
The first step of an analogous bootstrap program for (1 + 1)-dimensional integrable boundary
quantum field theories, the boundary R-matrix bootstrap, was developed in the pioneering work
by Ghoshal and Zamolodchikov [4]; it makes possible the determination of the reflection matrices
and provides complete description of the theory on the mass shell.
For the second step matrix elements of local operators between asymptotic states have to be
computed. In a boundary quantum field theory there are two types of operators, the bulk and the
boundary operators, where their names indicate their localization point. The boundary bootstrap
program, namely the boundary form factor program for calculating the matrix elements of local
boundary operators between asymptotic states was initiated in [5].1
In this work we make further progress in understanding the boundary form factor bootstrap
by classifying and counting the solutions. In the bulk case it was proposed by Cardy and Mussardo that the space of solutions of form factor axioms is to be identified with the space of local
operators [8]. The program of counting the solutions was performed for several interesting models [912] and it was shown that the results agree with the spectrum of local operators expected
from the Lagrangian description in terms of the ultraviolet limiting conformal field theory. Here
we take the models treated in [5] and perform a similar counting of solutions, which can then
be brought into correspondence with the spectrum of boundary operators predicted by boundary
conformal field theory.
The outline of the paper is as follows. In Section 2 we give a short overview of the boundary
form factor axioms and also introduce some basic notions concerning form factor solutions. In
Section 3 we investigate the scaling LeeYang model with boundary, which has two physically
inequivalent boundary conditions. We start with the simpler boundary condition (denoted by 1)
and develop all the concepts necessary to perform the counting of the form factor solutions. Then
we apply these to the other boundary condition (denoted by ) too, while in Section 4 we treat
the boundary sinh-Gordon model using the same tools. We give our conclusions and discuss
major open problems in Section 5.
2. The boundary form factor bootstrap and the classification of solutions
2.1. The boundary form factor axioms
The axioms satisfied by the form factors of a local boundary operator were derived in [5].
Here we only list them without much further explanation. Let us suppose that we treat an integrable boundary quantum field theory in the domain x < 0, with a single scalar particle of
mass m, which has a two-particle S-matrix S( ) (using the standard rapidity parametrization)
and a one-particle reflection factor R() off the boundary, satisfying the boundary reflection factor bootstrap conditions of Ghoshal and Zamolodchikov [4]. For a local operator O(t) localized
at the boundary (located at x = 0, and parameterized by the time coordinate t ) the form factors
are defined as
1 There exists no analogous framework for bulk operators in the presence of the boundary yet. Large distance expansion
for their correlators in terms of bulk form factors can be given using the boundary state formalism, but so far this approach
has only been used to compute one-point functions [6,7].

out 1 , 2 , . . . , m |O(t)|1 , 2 , . . . , n in
O
= Fmn
(1 , 2 , . . . , m ; 1 , 2 , . . . , n )eimt (
1
2
cosh i
cosh j )
213
m
< 0, using the asymptotic in/out state forfor 1 > 2 > > n > 0 and < < <
malism introduced in [13]. They can be extended analytically to other values of rapidities. With
the help of the crossing relations derived in [5] all form factors can be expressed in terms of the
elementary form factors
out 0|O(0)|1 , 2 , . . . , n in
= FnO (1 , 2 , . . . , n ),
which can be shown to satisfy the following axioms:

I. Permutation:
FnO (1 , . . . , i , i+1 , . . . , n ) = S(i i+1 )FnO (1 , . . . , i+1 , i , . . . , n ).
II. Reflection:
FnO (1 , . . . , n1 , n ) = R(n )FnO (1 , . . . , n1 , n ).
III. Crossing reflection:
FnO (1 , 2 , . . . , n ) = R(i 1 )FnO (2i 1 , 2 , . . . , n ).
IV. Kinematical singularity
O
(
i Res Fn+2
=

+ i, , 1 , . . . , n ) = 1
n

S( i )S( + i ) FnO (1 , . . . , n )
i=1
or equivalently described as
O
i Res Fn+2
( + i, , 1 , . . . , n )
=

n

S( i )R( )S( + i ) FnO (1 , . . . , n ).
= R()
i=1
V. Boundary kinematical singularity

n

i
i
g
O
i Res Fn+1 +
S
, 1 , . . . , n =
1
i FnO (1 , . . . , n ),
=0
2
2
2
i=1
where g is the one-particle coupling to the boundary

ig 2
, i .
2 i
2
VI. Bulk dynamical singularity
R()
O
O
i Res Fn+2
( + iu, iu, 1 , . . . , n ) = Fn+1
(, 1 , . . . , n ),
=
corresponding to a bound state pole of the S matrix

i 2
, 2iu
2iu
(in a theory with a single particle, the only possible value is u = /3).
S( )
(2.1)
214
VII. Boundary dynamical singularity

O
(1 , . . . , n , ) = g F O (1 , . . . , n ),
i Res Fn+1
=iv
which corresponds to a pole in the reflection factor describing a boundary excited state:
R()
i g 2 /2
,
iv
iv.
We further assume maximum analyticity, i.e., that the form factors have only the minimal singularity structure consistent with the above axioms.
2.2. Solution of the axioms
2.2.1. The general ansatz
The general form factor solution can be written in the following form [5]:
Fn (1 , 2 , . . . , n ) = Gn (1 , 2 , . . . , n )
n
r(i )
i=1
f (i j )f (i + j ),
(2.2)
i<j
where f is the minimal bulk two-particle form factor (2PFF) satisfying the conditions
f ( ) = S( )f ( ),
f (i + ) = f (i )
and having the minimum possible number of singularities in the physical strip 0 < together
with the slowest possible growth at infinity [14], and r is the minimal boundary one-particle form
factor (1PFF) satisfying
r() = R()r( ),
r(i + ) = R( )r(i ),
(2.3)
plus analytic conditions similar to those of f , but in this case in the strip 0 < /2.
The functions Gn are totally symmetric and meromorphic in the rapidities i . They are also
even and periodic in them with the period 2i, so they can only be functions of the variables
yi = ei + ei .
Let us now turn to the analysis of the singularity structure. In a theory with only one particle,
the only possible singularity of the S-matrix in the physical strip is located at = 2i/3 corresponding to the self-fusion of the particle (plus the crossed channel pole for the same process at
i/3) and the relevant fusion coupling is defined as
2 = i Res S( ).
= 2
3 i
We suppose that the 2PFF function f is chosen such that it has a pole at = 2i/3 so that it encodes this singularity (for an example see (3.2)). We further assume that the boundary dynamical
singularities (but not the kinematical ones!) are similarly contained in the 1PFF function r. Then
the functions Gn only have singularities at the positions of the kinematical singularities, and so
they can be written in the form
Qn (y1 , y2 , . . . , yn )
,
Gn (1 , 2 , . . . , n ) =
i yi
i<j (yi + yj )
where the Qn are entire functions symmetric in their arguments.
215
2.2.2. Two-point functions and scaling weights

The two-point functions can be written using a spectral representation:

1
0|O(t)O(0)|0 =
d1 d2 dn Fn Fn+
(2)n
n=0
1 >2 >>n >0

n

exp imt
cosh i ,
(2.4)
i=1
where time translation invariance was used and the form factors Fn , Fn+ are defined as
Fn = 0|O(0)|1 , 2 , . . . , n in = FnO (1 , 2 , . . . , n )
and
Fn+ = in 1 , 2 , . . . , n |O(0)|0 = FnO (i + n , i + n1 , . . . , i + 1 ),
which, for unitary theories, is the complex conjugate of the previous one: Fn+ = Fn . In the
Euclidean time = it the form factor expansion of the correlator converges rapidly for large
separations since multi-particle terms are exponentially suppressed.
We are interested in operators which can be classified according to their scaling dimensions,
which means that the two-point function must have a power-like short-distance singularity
0|O( )O(0)|0 =
,
(2.5)
2
where is an exponent determined by the ultraviolet scaling weights of the local fields. More
precisely let us consider the conformal operator product expansion
O( )O(0)
Ci
OO
Oi (0),
2hhi
hi
where h is the ultraviolet weight of the field O and the hi are the weight of the Oi . It is obvious
that
2 = 2h hmin ,
where hmin is the minimum of the weights of the operators appearing in the expansion. In many
cases it is hmin = 0 (corresponding to the weight of the identity) and then h and are identical.
It is well known that the presence of a power-like singularity means that the form factors
themselves can only grow exponentially in the rapidity variables [15] and so the functions Qn
are symmetric polynomials in the variables yi . The singularity axioms give recursion relations
for the polynomials Qn , for which we introduce the abbreviated notation
Qn = K[Qn+2 ],
Qn = D[Qn+1 ],
Qn = B[Qn+1 ],
(2.6)
where K, D and B denote the recursion relation resulting from the bulk kinematical, bulk dynamical and boundary kinematical singularity axioms, respectively. We give explicit forms of
these relations for specific models later.
If the form factors grow asymptotically as

Fn (1 + , 2 + , . . . , n + ) ex|| as | |
(2.7)
216
(where we assume that the exponent x is independent of the level n),2 then the leading shortdistance behaviour of the individual terms can be easily shown to be 1/ 2x and so naively = x.
In fact, leading logarithmic corrections in can (and in many cases do) sum up to an anomalous
contribution to the ultraviolet exponent and therefore we call x the naive (or engineering)
dimension of the form factor solution.
2.2.3. Towers and local operators
Following the ideas in the work of Koubek and Mussardo on classification of the bulk form
factor solutions [9], the set of elementary form factors corresponding to a local operator forms a
tower of form factors, graded by the number of particles:

O FO = FnO (1 , . . . , n ) nN .
Due to the recursion relations (2.6) the solutions (2.2)

Qn (y1 , y2 . . . , yn )
Fn (1 , 2 , . . . , n ) =
r(i )
f (i j )f (i + j )
i yi
i<j (yi + yj )
n
i=1
i<j
of the form factor equations can also be classified into towers which consist of a single form factor
at each level n such that the solutions at different levels are linked together via the relations (2.6).
As we show later in the explicit examples, the recursion relations are such that the naive scaling
dimension x is independent of the level n and therefore it can be assigned to the tower itself.
Therefore the space of form factor solutions can be rearranged into a space of towers T , which
can in turn be graded by the naive scaling dimension. Due to the linearity of the form factor
axioms, the space of towers is a linear space, i.e., whenever F = {Fn }nN and F = {Fn }nN are
two towers, so is their general linear combination
F + F = {Fn + Fn }nN .
While the space of towers (T ) is always infinite-dimensional, the subspaces corresponding to a
given value of x can be finite-dimensional; in that case it makes sense to count the towers graded
by x. We can introduce the character of this space by

dx q x ,
X(q) =
(2.8)
x
where dx is the linear dimension of those towers which have naive dimension x.
The space L of scaling local boundary operators is also a graded linear space, where the grading is given by the ultraviolet scaling weight . The character of that space can be obtained from
boundary conformal field theory. Our aim here is to bring the two graded spaces into correspondence with each other via comparing their characters, which is actually nothing else than the
counting of linearly independent local operators classified by their ultraviolet scaling behaviour.
3. Scaling LeeYang model with boundary
The scaling LeeYang model with boundary is a combined bulk and boundary perturbation
of the boundary version of the M2,5 Virasoro minimal model which was investigated in detail
2 For all models considered here we show later that the space of solutions is spanned by so-called simple towers for
which the asymptotic exponent x is indeed independent of the level n.
217
in [16]. The conformal field theory has central charge c = 22/5 and the Virasoro algebra has
two irreducible representations Vh with highest weight h = h1,1 = 0 and h = h1,2 = 1/5 [17].
We define the truncated characters of these representations by
r,s (q) = TrVh q
L0 hr,s
d(n)q n ,
n=0
where d(n) gives the degeneracy of the level n descendants. They can be represented as the
following fermionic sums [18]
1,1 (q) =
1,2 (q) =

n=1

n=1
1
(1 q 2+5n )(1 q 3+5n )
1
.
(1 q 1+5n )(1 q 4+5n )
(3.1)
Boundary conformal field theory was developed in [1921] and the interested reader is referred to them for details. Applying the formalism to the conformal LeeYang model it can be
seen that there are two conformally invariant boundary conditions. On one of them, denoted by 1
in [16], the spectrum of boundary fields is given by the vacuum representation V0 of the Virasoro
algebra, and therefore it does not have any relevant boundary fieldsthus can have no boundary
perturbation either. In the other case, denoted in [16], the spectrum of boundary fields is given
by the direct sum V0 V1/5 and therefore it has a nontrivial relevant boundary field with
scaling dimension 1/5 and the general perturbed boundary conformal field theory action can
be written as3
A,(B ) = A +
0
dy
dx (x, y) + B
dy (y),
where A denotes the action for M(2/5) with the boundary condition imposed at x = 0,
and and B denote the bulk and boundary couplings respectively. There is a unique nontrivial
relevant bulk perturbation given by the spinless field with scaling dimensions h = h = 1/5.
The action of A,1 is similar, but the last term on the right-hand side is missing. For > 0 the
bulk behaviour is described by an integrable massive theory having only a single particle with
mass m with the following S-matrix [22]:

sinh( 2 + ix
1
2
1
2 )
S( ) =
, [x] = (x)(1 x).
=
;
(x) =
ix
3
3
3
sinh( 2 2 )
3
The pole at = 2i
3 corresponds to the property, i.e., the particle appears as a bound state
of itself. The minimal bulk two-particle form factor only has a zero at = 0 and a pole at = 2i
3
in the strip 0 m( ) < and is of the form [23]:
f ( ) =
y 2
v(i )v(i + ),
y +1
y = e + e ,
(3.2)
3 The field L = , albeit relevant, does not induce any nontrivial perturbation because it is a total derivative.
y
1
218
where

dt i t sinh 2t sinh 3t sinh 6t
v( ) = exp 2
e
t
sinh2 t
0
and it satisfies
sinh
,
sinh i sin 3

cosh + 1/2
f +i
f i
=
f ( )
3
3
cosh + 1
f ( )f ( + i) =
and consequently also f ( ) O(1) for large .

In the boundary theory with the perturbed boundary, the reflection amplitude of the particle
depends on the strength of the coupling constant of the boundary field as [16]

1
2 b+1 b1
1
,
R() = R0 ( )R(b, ) =
2
6
3
6
6
with the one-particle coupling
g = i2(3)
1/4
1/2 3 + 2 sin b
6
2 3
,
3 2 sin b
6
where the relation between the dimensionless bootstrap parameter b and the dimensionful Lagrangian parameters and B is known explicitely [24], while in the case of the 1 boundary the
reflection amplitude is the parameter independent expression

1
2
1
R()1 =
2
6
3
and the one-particle coupling is
1/2

.
g1 = i2(3)1/4 2 3
3.1. Counting towers for the boundary condition 1
3.1.1. The recursive equations for the form factors
The 1PFF function corresponding to the parameter free reflection factor R()1 is
r1 ( ) = i sinh u( ),
where
u( ) = exp
0

dt
t
1
i
t sinh 5t6 + sinh 2t sinh 3t
2 cosh cos
t sinh 2t
2
2
sinh2 t
and asymptotically behaves as r1 e2 when . Following the general ideas in Section 2.2.1 we introduce the ansatz
r1 (i ) f (i j )f (i + j )
,
Fn (1 , . . . , n ) = 4n Hn Qn (y1 , . . . , yn )
(3.3)
yi
y i + yj
i
i<j
219
where

Hn =
i3 4
n
(3.4)
2 2 v(0)
Then one finds the following recursion equations for the Qn s [5]:
D:
Q2 (y+ , y ) = Q1 (y),
Qn+2 (y+ , y , y1 , . . . , yn ) =
n

(y + yi )Qn+1 (y, y1 , . . . , yn ),
n > 0,
i=1
K:
Q2 (y, y) = 0,
Qn+2 (y, y, y1 , . . . , yn ) = Pn (y|y1 , . . . , yn )Qn (y1 , . . . , yn ),
B:
n > 0,
Q1 (y) = 0,
Qn+1 (0, y1 , . . . , yn ) = Bn (y1 , . . . , yn )Qn (y1 , . . . , yn ),
n > 0,
(3.5)
where

n
n

1
Pn (y|y1 , . . . , yn ) =
(yi y )(yi + y+ ) (yi + y )(yi y+ ) , (3.6)
2(y+ y )
i=1
i=1
n

n

1
yi + 3
yi 3
Bn (y1 , . . . , yn ) =
(3.7)
2 3 i=1
i=1
and
y+ = z + 1 z1 ,
y = 1 z + z1 ,
=e 3 ,
(3.8)
with the auxiliary variable z defined as a solution of y = z + z1 (i.e., writing y = 2 cosh we

obtain z = e ). The symmetry of the expressions (3.5), (3.6) in y ensures that the resulting
relations do not depend on which of the two possible solutions of the relation y = z + z1 is
chosen (switching from z = e to z = e only interchanges y+ with y ).
Note also that Qn+2 (y+ , y , y1 , . . . , yn ) only depends on y rather than separately on y because due to its symmetry in all variables it can only depend on the combinations y+ + y = y
and y+ y = y 2 3. Furthermore it is easy to see that Pn is a polynomial in all of its variables
(since the expression inside the parentheses vanishes if y+ = y and is therefore divisible by
y+ y ), and using the previous argument again it only depends on y. However, elimination of
y makes the formulae rather cumbersome and therefore we prefer to keep them in our equations.
It is also interesting to note that the boundary kinematical recursion is actually almost entirely contained in the bulk recursion. Using the bulk kinematical recursion with y = 0 results
in
Qn+2 (0, 0, y1 , . . . , yn ) = Pn (0|y1 , . . . , yn )Qn (y1 , . . . , yn ),
while using the boundary kinematical recursion twice leads to
Qn+2 (0, 0, y1 , . . . , yn ) = Bn (0, y1 , . . . , yn )Bn (y1 , . . . , yn )Qn (y1 , . . . , yn )
and therefore it must be true that
Pn (0|y1 , . . . , yn ) = Bn (0, y1 , . . . , yn )Bn (y1 , . . . , yn ),
(3.9)
220
which is indeed satisfied by the expression (3.6), (3.7). This means that the square of the
boundary kinematical recursion is contained in the bulk one, as already discussed in [5]. The
only independent information that the boundary kinematical singularity axiom carries is the sign
of the one-particle coupling g: all the other axioms depend only on the one-particle reflection
factor R, which in turn contains only g 2 using Eq. (2.1).
Indeed it was shown by Dorey et al. in [25] that although the fundamental reflection factors of
the boundary condition 1 and (b = 0) are identical, these boundary conditions are physically
different. Their one-particle couplings differ by a sign
1/2

g1 = i2(3)1/4 2 3
= g(b=0)
and, as a result the whole bootstrap structure changes: the 1 boundary condition has no boundary
excited state, while the (b = 0) does. This is also reflected in the different expression for their
1PFF function r as discussed after Eq. (3.10). We note that while the sign of g is not manifest in
the reflection factor itself, it affects many physical quantities besides the spectrum such as finite
size corrections to the energy levels [16,26] and one-point functions of the bulk fields [7,27].
3.1.2. Asymptotic behaviour of form factors and simple towers
It can be easily deduced that the asymptotic behaviour of the ansatz (3.3) is given by

Fn (1 + , . . . , n + )
exn ||
where xn = deg Qn
n(n 3)
,
2
where deg P denotes the total degree of the multivariable polynomial P defined by
P (y1 , . . . , yn ) deg P
as .
Note that
deg D(Qn+1 ) = deg B(Qn+1 ) = deg Qn+1 n + 1,
deg K(Qn+2 ) = deg Qn+2 2n + 1.
As a result, the solutions of the recursion relation can be expanded in the basis of simple towers
which are defined to be towers with the property
Qn = 0,
n < n0 ,
deg Qn+1 = deg Qn + n 1,
n n0 ,
for some n0 N.
For a simple tower the exponent xn describing the asymptotic growth of the form factor tower
is independent of the level n. Therefore we can use the naive dimension to introduce a grading
on the space of towers: the subspace of towers of grade x is spanned by the simple towers with
naive scaling dimension x. For a simple tower given by the polynomials {Qn }nN the scaling
dimension can be expressed as
x = deg Qn
n(n 3)
.
2
221
3.1.3. The recursion kernel and the number of independent towers

Here we carry over an approach developed by Koubek [10,11] to count solutions of the bulk
form factor equations to the boundary case. The idea is to classify the so-called kernels of the
recursion relations (2.6) first and use these to get the dimensions of the spaces of simple towers.
The kernel can be defined as polynomials (at a given level n) that are taken to zero by the
appropriate recursion relation in (2.6). The kernel of each recursion at level n can be generated
by multiplying an elementary kernel polynomial with an arbitrary symmetric polynomial of n
variables. The generating polynomials are

K: KnK (y1 , . . . , yn ) =
(yi + yj ),
1i<j n
D:
B:
KnB (y1 , . . . , yn ) =
KnB (y1 , . . . , yn ) =

yi2 + yi yj + yj2 3 ,
1i<j n
n
yi .
i=1
The common kernel of the three recursions at level n is generated by the least common multiple
of these polynomials which is just equal to their product:
Kn (y1 , . . . , yn ) = KnK (y1 , . . . , yn )KnB (y1 , . . . , yn )KnD (y1 , . . . , yn ),
with degree
n(3n 1)
2
and a basis of the kernel polynomials at level n is given by
deg Kn =
k(n)
k(n)
Kn ,
l
1
0 < k1 k2 kl n,
(n)
where k are the elementary symmetric polynomials of n variables and degree k defined by the
generating relation
n
n

(n)
(z + yi ) =
znk k (y1 , . . . , yn ).
i=1
k=0
The number of linearly independent towers can be counted using the observation that every
independent tower starts from a kernel polynomial, because each new tower appears at some
level (number of particles) n as an ambiguity of the solution of the recursion relations. This is
true even for the towers starting at the lowest level n = 1 if we note that the kernel at level 1 is
generated by
K1 (y1 ) = y1 .
(n)
(n)
The naive scaling dimension of the simple tower starting from the polynomial k1 kl Kn is
n(n 3)
= n(n + 1) + k1 + + kl .
2
Now we can write down the generating function X(q) defined in (2.8) easily:
(n)
(n)
(n)
xk1 kl = deg k1 kl Kn
X(q) =

n=1 m=0
(m|n)q n(n+1)+m ,
222
where (m|n) denotes the number of partitions of the number m such that none of the summands
is greater than n. Using the identity
(m|n)q m =
m=0
n

i=1
1
,
1 qi
we obtain
X(q) =

n=1
q n(n+1)
n
i
i=1 1 q
and with the help of a RogersRamanujan identity (following [11])

X(q) =
n=1
(1 q 2+5n )(1 q 3+5n )
1 = 1,1 (q) 1
using (3.1). Therefore the result of the counting of the form factor towers, graded by their naive
scaling dimension, exactly matches the conformal spectrum of the operators localized on the 1
boundary, which is given by the vacuum representation of the Virasoro algebra (with highest
weight h1,1 = 0). The subtraction 1 corresponds to the identity operator which only has trivial
(vanishing) form factors.
Note that in claiming the agreement we tacitly supposed that the naive scaling dimension x
of the towers can be identified directly with the conformal weights of the scaling operators, i.e.,
that there are no anomalous dimensions. This was checked for the form factor tower with the
lowest possible dimension x = 2 in [5], where performing the spectral sum we obtained that the
ultraviolet dimension of the corresponding operator was indeed = 2.
3.1.4. Some explicit solutions
Finally we give the explicit forms of the lowest levels of some form factor towers. The n = 1
generating kernel polynomial4 is K1 = 1 , and all kernel polynomials can simply be written
as 1n . The first three towers starting at level 1 are
x = 2:
QT1 = 1 ,
x = 3:
T
QT
n = 1 Qn ,
2
Qn T = 12 QTn .
x = 4:
QT2 = 1 ,
QT3 = 12 ,
QT4 = 12 (2 + 3),
...,
The upper index T shows that the tower corresponds to the (off-critical version of the) boundary stress-energy operator T = L2 I (already found in [5]), while the other towers are its first
and second derivatives, respectively. In fact, there is a single tower for each value of x less than
6 which corresponds to the fact that up to this level (due to the presence of null-vectors) the
conformal vacuum representation of the c = 22/5 Virasoro algebra contains only the vectors
Ln1 L2 I (n = 0, 1, 2, 3). At x = 6 the conformal representation contains another linearly independent vector, and indeed a new tower appears, which starts from the level 2 generating kernel
polynomial:
2

2
Q(6)
2 = 1 2 1 2 3 ,
4 We omit the upper index n of (n) specifying the number of variables since it is always the same as the number of
k
particles, i.e., the level of the form factor tower.
223

(6)
Q3 = 1 (1 2 3 ) 12 2 3 ,

2

Q(6)
4 = 1 (2 + 3) (1 2 3 ) 1 2 3 + 1 4 .
The counting argument in the last subsection ensures that the new towers always appear in such
a way that the space of simple towers graded by x matches the space of linearly independent
vectors in the vacuum representation.
3.2. Counting towers for the boundary condition
In this case the 1PFF is [5]
r ( ) =
i sinh
(sinh i sin
(b+1)
)(sinh
6
i sin (b1)
)
6
(3.10)
u( ).
The main difference from the I case is the presence of the denominators, which correspond to
boundary excited states. Note that at b = 0 the reflection factor R() is identical to R()1 , but
the corresponding 1PFFs are different. This is related to the different interpretation of the pole
structure of the R-matrices discussed in detail in [25]: for the boundary condition 1 the boundary
excited state is absent and the corresponding pole at = i 6 is explained by a ColemanThun
diagram. In this case we take the following ansatz
r (i ) f (i j )f (i + j )
Fn (1 , . . . , n ) = Hn Qn (y1 , . . . , yn )
(3.11)
.
yi
y i + yj
i
i<j
The recursion relations in this case are [5]

2
Q1 (y),
D: Q2 (y+ , y ) = y 2 3
n

2
Qn+2 (y+ , y , y1 , . . . , yn ) = y 2 3
(y + yi )Qn+1 (y, y1 , . . . , yn ),
n > 0,
i=1
K:
Q2 (y, y) = 0,
2

2
y 12 Pn (y|y1 , . . . , yn )Qn (y1 , . . . , yn ),
Qn+2 (y, y, y1 , . . . , yn ) = y 2 1
n > 0,
B:
Q1 (0) = 0,
Qn+1 (0, y1 , . . . , yn ) = 1 1 Bn (y1 , . . . , yn )Qn (y1 , . . . , yn ),
n > 0,
where Pn and Bn are given by Eqs. (3.6)(3.8) and
k (b) = 2 cos (b + k).

6
The bulk and boundary kinematical recursions again satisfy the appropriate generalization of the
compatibility relation (3.9).
For the degrees of the recursions we have
deg D(Qn+1 ) = deg B(Qn+1 ) = deg Qn+1 n 1,
deg K(Qn+2 ) = deg Qn+2 2n 3
and so simple towers can now be defined by the property
Qn = 0,
n < n0 ,
224
deg Qn+1 = deg Qn + n + 1,
n n0 .
Due to the additional factors in the denominator the asymptotic behaviour of r ( ) for
is O(1), and so the naive scaling dimension of a simple tower corresponding to the polynomials
{Qn }nN is
n(n + 1)
.
2
The kernel is generated by the same polynomials as before, but now the naive scaling dimension
(n)
(n)
corresponding to k1 kl Kn is
x = deg Qn
(n)
(n)
(n)
xk1 ...kl = deg k1 kl Kn
n(n + 1)
= n(n 1) + k1 + + kl
2
and as a result we obtain

X(q) =
(m|n)q
n(n1)+m
n=1 m=0
=
=

n=1

n=1
n=1
q n(n+1)
n
i=1 1 q

n=1
q n(n1)
n
i
i=1 1 q
qn
i
i=1 1 q
n

1
1
+
1
2+5n
3+5n
1+5n
(1 q
)(1 q
)
(1 q
)(1 q 4+5n )
n=1
= 1,1 (q) + 1,2 (q) 1,

using again a RogersRamanujan identity following [11]. This is consistent with the fact that
in the conformal limit the boundary operator content of the boundary condition is given by
the direct sum of the vacuum module with highest weight h1,1 = 0 and the only other nontrivial
representation with highest weight h1,2 = 1/5.
Note that in this case to achieve the agreement we must suppose that the naive scaling dimensions of the towers corresponding to operators of the identity representation do not get any
anomalous corrections, while those corresponding to operators in the model with highest weight
h1,2 = 1/5 get exactly the right contribution for their conformal weight to match with that predicted by conformal field theory. The true scaling weight of the operator can only be computed by
examining appropriate two-point functions. This makes the operator identification very difficult
for higher levels as shown below using the example of the boundary stress-energy tensor.
For x = 0 the operator identification is easy, since there is a unique tower with the required
asymptotic properties:

2
,
Q2 = 1 2 + 3 3
Q1 = 1 ,

2
2
Q3 = 1 1 2 + 1 2 + 1 (2 + 3)(1 2 3 ) , . . . .
(3.12)
In [5] the two-point function of this tower was compared to the perturbed conformal field theory
prediction and we found that its identification with the operator is justified (this also settles the
case for all the derivatives of ). We remark that the anomalous dimension of the correlator in
this case turns out to be
2
1
2h = ,
2 = =
5
5
225
which is due to the fact that itself appears in the operator product, and so the above result
is in accordance with the discussion under Section 2.2.2 since hmin = 1/5 (cf. also [5] and also
the analogous bulk situation discussed by Zamolodchikov in [23]). For x = 1 there is again a
single operator
Q
n = 1 Qn ,
whose scaling dimension can be known exactly in terms of the scaling dimensions of the lowest
tower, since it is a derivative operator. However, we find two towers at x = 2. One is the second
derivative
2
Qn
= 12 Qn ,
(3.13)
while the other one starts from the n = 2 generating kernel polynomial:

QT2 = 1 2 12 2 3 ,

QT3 = 1 (1 2 3 ) 1 1 12 2 3 + 1 3 , . . . ,
which is consistent with the fact that both the Virasoro module h = 1/5 and the module h = 0
have a single linearly independent vector at level two: L21 and T = L2 I, respectively. The
tower (3.13) can be directly identified with L21 since the action of L1 is exactly given by the
derivative, but the identification of T is more involved. In general we expect that
T T + + + 2
(3.14)
(where the proportionality sign means that the normalization of the operator must also be fixed).
Note that T cannot mix with higher towers because that would spoil the short-distance behaviour.
The identification of T is an open question, to which we return in the conclusions.
4. Sinh-Gordon model with Dirichlet boundary conditions
The sinh-Gordon theory in the bulk is defined by the Lagrangian
1
m2
L = ( )2 2 (cosh b 1).
2
b
It can be considered as the analytic continuation of the sine-Gordon model for imaginary coupling = ib. The S-matrix of the model is

B
B
B
2b2
S( ) = 1 +
.
=
, B=
2
2
2
8 + b2
In this case there is no self-fusion pole, and so bulk dynamical singularities are absent. The
minimal bulk two-particle form factor belonging to this S-matrix is [28]

B x
x
sinh(1
)
sinh
dx 2 x(i ) sinh xB
4
2 2
2
sin
,
f ( ) = N exp 8
x
2
sinh2 x
0
satisfying
f ( )f ( + i) =
sinh
sinh + i sin B
2
and therefore f ( ) O(1) as .
(4.1)
226
Sinh-Gordon theory can be restricted to the negative half-line, but integrability is only maintained by imposing either the Dirichlet
(0, t) = 0D ,
or the two parameter family of perturbed Neumann

b
VB (0, t) = M0 cosh
(0, t) 0 M0
2
boundary conditions [4]. The latter interpolates between the Neumann and the Dirichlet boundary
conditions, since for M0 = 0 we recover the Neumann, while for M0 the Dirichlet boundary condition with 0D = 0 . The reflection factor which depends on two continuous parameters
can be written as [29]

1
1 B
B
E1 F 1
+
1
R() = R0 ( )R(E, F, ) =
2
2
4
4
2
2
and can be obtained as the analytic continuation of the first breather reflection factor in boundary sine-Gordon model which was calculated by Ghoshal in [30]. The relation of the bootstrap
parameters E and F to the parameters of the Lagrangian is known both from a semiclassical
calculation [29,31] and also in an exact form in the perturbed boundary conformal field theory
framework [32]. Imposing Dirichlet boundary condition instead of the general one corresponds
to removing the F dependent factor from R().
4.1. Recursion relations for the form factors with Dirichlet boundary conditions
Boundary form factors for sinh-Gordon theory with Dirichlet boundary condition have already
been investigated to some extent in [5], and many solutions were constructed by Castro-Alvaredo
[33] at the self-dual point B = 1. Here we aim to classify all the solutions for general coupling.
The 1PFF corresponding to Dirichlet boundary condition is [5]
rE ( ) =
where
sinh
u(, B),
sinh i sin

u(, B) = exp 2
0
x
2

dx
i
x
cos
1
x
2

B x
x
xB
+ sinh 1
+ sinh
sinh
4
2 2
2
sinh2 x
cosh
and
=
(E 1).
2
The asymptotics of the 1PFF is rE e as . The one-particle coupling to the boundary is

given by
gE =
+ sin B
2(1 + cos B
4 ) cos
4
.
1 sin
sin B
2
227
Writing the n-particle form factors in the general form

rE (i ) f (i j )f (i + j )
,
Fn (1 , . . . , n ) = Hn Qn (y1 , . . . , yn )
yi
y i + yj
i
i<j
and choosing the ratio of the Hn s appropriately the recursion equations take the form:
K: Q2 (y, y) = 0,

Qn+2 (y, y, y1 , . . . , yn ) = y 2 4 cos2 Pn (y|y1 , . . . , yn )Qn (y1 , . . . , yn ),
n > 0,
B:
Q1 (0) = 0,
Qn+1 (0, y1 , . . . , yn ) = 2 cos Bn (y1 , . . . , yn )Qn (y1 , . . . , yn ),
n > 0,
(4.2)
B
2
where Pn is given by Eqs. (3.6), (3.8) with = ei ,

n

n

1
B
B
Bn (y1 , . . . , yn ) =
yi 2 sin
yi + 2 sin
2
2
4 sin B
2
i=1
i=1
and the bulk and boundary kinematical recursions again satisfy the appropriate generalization of
the compatibility relation (3.9). Note that for E = 0 the right-hand side of the boundary kinematical relation is zero:
B:
Qn+1 (0, y1 , . . . , yn ) = 0
and so one could dispense with the boundary kinematical recursion choosing an ansatz of the
form

f (i j )f (i + j )
Fn (1 , . . . , n ) = Hn Qn (y1 , . . . , yn )
rE=0 (i )
,
y i + yj
i
i<j
as in [5]. In this paper, however, we choose to treat the two cases together.
4.2. Counting towers and the spectrum of local operators
The recursion relations (4.2) have the following degrees:
deg K(Qn+2 ) = deg Qn+2 2n 1,
deg B(Qn+1 ) = deg Qn+1 n
and so simple towers are defined as

Qn = 0,
n < n0 ,
deg Qn+1 = deg Qn + n,
n n0 ,
while the naive scaling dimension of a simple tower corresponding to the polynomials {Qn }nN
is given by
n(n 1)
.
2
Due to the absence of bulk dynamical singularities, the generating kernel polynomial at level n
is the product of the ones for the bulk and boundary kinematical recursions
x = deg Qn
Kn (y1 , . . . , yn ) = KnK (y1 , . . . , yn )KnB (y1 , . . . , yn ),
228
with degree
deg Kn =
n(n + 1)
2
and the naive scaling dimension of the simple tower starting from the polynomial k1 kl Kn
is
n(n 1)
= deg k(n)
k(n)
Kn
= n + k1 + + kl .
xk(n)
l
1 kl
1
2
The generating function (2.8) takes the form
(n)
X(q) =
(m|n)q
n+m
n=1 m=0
(n)q =
n
n=1

j =1
(n)
1
,
1 qj
where (n) is simply the number of all partitions of n.

On the other hand, the independent boundary operators for the Dirichlet boundary condition
(t, x = 0) = 0 are simply given by the differential monomials

xk1 (t, x) xkl (t, x)x=0 ,
which have scaling weight k1 + + kl . Therefore the number of operators with a given scaling
weight n N is indeed equal to the number of integer partitions of n.
4.3. General boundary conditions
When both E and F are nonzero, the 1PFF reads
rEF ( ) =
sinh
u(, B),
(sinh i sin )(sinh i sin )
(E 1), = (F 1),
2
2
with asymptotics rEF O(1) when , and therefore the scaling dimension of a simple
tower is
n(n + 1)
x = deg Qn
2
and so each tower starting from an elementary kernel polynomial Kn has naive scaling dimension
x = 0. This results in infinitely many towers corresponding to any integer value of the naive
scaling dimension!
This is in fact not so surprising if we consider the expected spectrum of scaling operators.
These are of the form

tk1 (t, x) tkl (t, x)e(t,x) x=0
(the x-derivatives of the field can be expressed with exponential operators using the boundary
condition). These can be thought to be organized into families of descendants of the exponential
operators
e(t,x=0) ,
the descendant level given by k1 + + kl , so the number of descendants at some level n is
the number of partitions of the integer n. However, the exponential operator has naive (classical)
scaling dimension zero: its scaling weight is a fully quantum effect. Although at first it seems that
229
there exists a continuum of such operators, recall that they can be expressed in terms of powers
of the field:
e
(t,0)

k
k=1
k!
k (t, 0),
which shows that they depend only on countably many independent operators, the powers k
(which are not scaling fields in themselves, their correlation function involves logarithms) so one
expects only a countable infinity of form factor towers with naive scaling dimension 0, which is
indeed what we found.
In this paper we cannot go further with the identification between the form factor towers and
the local operators because some further tools are necessary in order to classify the towers of zero
naive scaling dimension arising from the kernel solutions; see the conclusions in the next section
for details. We only wish to note that our results in the free field limit are identical to those in [5],
and that the whole situation above is very similar to the case of the bulk sinh-Gordon model [9].
5. Conclusions and open problems
In this paper we performed the classification and counting of the solutions of the boundary
form factor bootstrap based on their ultraviolet behaviour. We showed that in the boundary version of the scaling LeeYang model and in the boundary sinh-Gordon model the results are in
perfect agreement with the expectations based on boundary conformal field theory in the first
case and on the Lagrangian formulation in the second. This gives an additional and crucial piece
of evidence for the consistency of the bootstrap axioms developed in [5].
Our discussion, however, has made several open problems manifest, that are worthwhile to be
pursued in the future.
5.1. The issue of anomalous dimensions
The first is the question of anomalous dimensions. It was shown (using the example of the
boundary operator and its descendants in the LeeYang case) that the naive dimension obtained
from the asymptotic growth of the form factors of some tower is not necessarily identical to the
exact ultraviolet weight of the corresponding operator. In fact the counting of towers depends
very much on the additional assumption that anomalous contributions to the scaling dimensions
do not mess up the spectrum,5 and so the space of simple towers with a given value of the naive
scaling dimension x can be brought into correspondence with the space of local operators of a
given value for the exact ultraviolet weight h.
This issue is also well known in the bulk and there the computation the ultraviolet weight
can be tackled by two different approaches. One of them uses the cumulant expansion of the
logarithm of the two-point function [34] (for a very nice discussion see also [35]). The spectral
5 It could happen, at least in principle, that the grading of the operators is changed by the anomalous dimensions,
so that the classification according to conformal descendants does not match the grading provided by the naive scaling
dimension.
230
expansion (2.4) can be written in the following form for the Euclidean two-point function

1
(m ) =
d1
d2
dn em i cosh i fn (1 , . . . , n ),
n!
n=0
1
fn =
Fn Fn+ ,
(4)n
where we used that the functions fn are symmetric and even in all the rapidity variables due to
the symmetry properties of the form factors (note that in a unitary theory they are also positive
semidefinite, but that is not necessary for the following argument). Supposing that the leading
n = 0 term in the short-distance limit is a constant and normalizing it to 1 we can write a similar
expansion for the logarithm of the correlation function

1
log (m ) =
d1
d2
dn em i cosh i hn (1 , . . . , n ),
n!
n=1
where the hn are the cumulants of the functions fn defined recursively by

f1 (1 ) = h1 (1 ),
f2 (1 , 2 ) = h2 (1 , 2 ) + h1 (1 )h1 (2 )
f3 (1 , 2 , 3 ) = h3 (1 , 2 , 3 ) + h1 (1 )h2 (2 , 3 ) + h1 (2 )h2 (1 , 3 ) + h1 (3 )h2 (1 , 2 )

+ h1 (1 )h1 (2 )h1 (3 ) . . . .
Defining
h n (1 , . . . , n ) = lim hn (1 + , . . . , n + ),
(5.1)
it is easy to see that the functions h n depend only on the differences of the rapidities (for spinless
operators in the bulk this is already true for fn and therefore also hn and in that case h n hn ).
Following the same argument as in the bulk case we arrive at the representation

1
=
d2
dn h n (0, 2 , . . . , n ),
n!
n=1
(5.2)
for the short-distance exponent defined in (2.5).

The main problem with this approach is that the limit (5.1) exists only for boundary operators
for which the naive scaling dimension is zero, and therefore it is of very limited use (in the bulk
it obviously exists for any spinless operator, but a similar problem appears when trying to apply
this approach to operators with nonvanishing Lorentz spin). An example to which the cumulant
expansion (5.2) can be applied is given by the operator in the scaling LeeYang model with
boundary condition, but in that case we already have a much more detailed comparison with the
conformal prediction directly via the two-point function, which was carried out in [5]. Therefore
it is an interesting problem to develop some method to extract the ultraviolet dimension directly
using only the spectral densities fn , generalizing the above argument.
The other approach in the bulk is to use a sum rule such as the so-called -theorem developed
in [36]. Unfortunately, the arguments used in the bulk case are not directly applicable to boundary
theories, because the boundary component of the energymomentum tensor does not obey any
231
conservation law in itself, but it remains to be seen whether there exists some other way of
deriving a sum rule.
A further possibility along this line of thought would be to examine the operator product
of T (the operator that corresponds to the conformal boundary Virasoro current off criticality)
with the boundary operator O in question using the spectral representation for their two-point
function, since it is known from boundary conformal field theory that they have an operator
product expansion of the form
hO O(0)
+ less singular terms,
2
but to perform this we would need to identify the form factors corresponding to T ; the discussion
below just highlights the difficulties related to operator identification.
T ( )O(0) =
5.2. Operator identification

It is also an open problem how to identify individual operators with their corresponding towers. In some cases this is straightforward, like for the operator with the lowest conformal weight
(an excellent example is again the operator in the scaling LeeYang model with boundary
condition, or the boundary stress energy tensor T in the case of the 1 boundary condition [5]). It
is also rather easy to identify form factors of derivative operators since
n

cosh i F O (1 , . . . , n )
F O (1 , . . . , n )
i=1
(once the form factors of O are known). For other operators, however, the identification is not at
all straightforward. Take the example of the boundary stress energy tensor T for the boundary
condition in the scaling LeeYang model (Section 3.2) where the new tower at x = 2, corresponding to the appearance of T , mixes nontrivially with other towers according to (3.14). It
is not easy to disentangle the mixing and so produce a direct identification. There are certainly
some approaches to try which we leave to future investigations.
One possibility is to calculate the mutual two-point correlation functions of the towers T ,
, , 2 and compare it numerically to the conformal perturbation theory prediction for the
operators T , , L1 , L21 (in fact only the correlators involving T are necessary). Using that,
one can then fit the mixing (and normalization) coefficients numerically.
Another approach to operator identification in general is to use a boundary extension of the approach developed by Delfino and Niccoli for the bulk scaling LeeYang model in [37] who used
the fact that part of the conformal descendants can be generated using derivation (L1 ) and the
charges which remain conserved off-criticality. This is very efficient for low levels: it makes possible the precise identification of towers with the appropriate operators up to descendant level 7.
It remains to be seen whether this method can be extended to the boundary case.
In general, however, it is also difficult to identify towers corresponding to the primary fields.
As we have shown in Section 4.3, this is a very complicated issue in the boundary sinh-Gordon
model with general boundary conditions, where an infinity of x = 0 towers must be matched
against the spectrum of exponential fields. In the bulk case form factors of primary fields satisfy
a factorization property [3,34,38], for which a general argument was given in [36]. Factorization
makes possible the identification of towers corresponding to primary fields, and can also be
extended to descendants. It remains to be seen whether such clustering property can be extended
to the boundary form factor bootstrap: at first sight, it seems more promising to take an approach
232
which relates clustering to the operator product expansion as in the paper [39] by Balog and
Weisz instead of one appealing to conformal holomorphic factorization (as done by Cardy et
al. in [36]), which is broken by the presence of the boundary condition. In this regard it is also
interesting to examine the relation between the bulk and boundary form factor bootstrap, to which
we now turn.
5.3. Connection to the bulk form factor bootstrap
Isolating the leading asymptotic coefficient of the n-particle form factor FnO in the large rapidity limit
FnO (1 + , . . . , n + )
FnO (1 , . . . , n )ex + subleading terms,
it is obvious that FnO (1 , . . . , n ) is a function of rapidity differences only, and it is also easy to
verify that it satisfies the bulk form factor axioms as a consequence of FnO satisfying the boundary form factor axioms. Therefore FnO (1 , . . . , n ) can identified with the n-particle form factor
As an example, using the asymptotic behaviours of the

fnO (1 , . . . , n ) of some bulk operator O.
2PFF function f in (3.2) and of the 1PFF function r in (3.10), the form factor tower F (3.12)
in the LeeYang model with boundary condition in this limit gives rise (up to normalization)
to the bulk form factor tower of the trace of stress energy tensor found by Zamolodchikov [23].
As an application of this correspondence it is easy to see that the bulk cumulant expansion for the
anomalous dimension of in [34] is just twice the expression (5.2) evaluated for the boundary
field . This is consistent with the fact that in the bulk = 2/5 while for the boundary case
= 1/5.
This correspondence may be related in some way to the bulk-boundary operator product expansion in boundary conformal field theory [20], and it is possible that (extending the example
above) the exact scaling dimension of operators can be identified this way using known results
for the bulk form factors. It could also explain why the detailed structure of the counting of form
factor towers (kernels, naive scaling dimensions, character identities to use, etc.) is so similar to
the bulk case as it was treated in [911].
Acknowledgements
G.T. wishes to thank Z. Bajnok, L. Palla and also F. Smirnov for useful discussions. This
research was partially supported by the Hungarian research funds OTKA T043582, K60040 and
TS044839. G.T. was also supported by a Bolyai Jnos research scholarship.
References
[1] G. Mussardo, Phys. Rep. 218 (1992) 215379.
[2] P.E. Dorey, Exact S-matrices, in: Z. Horvth, L. Palla (Eds.), Etvs Summer School in Physics: Conformal Field
Theories and Integrable Models, Budapest, Hungary, 1318 August, 1996, in: Lecture Notes in Physics, Springer,
1996.
[3] F.A. Smirnov, Form-factors in completely integrable models of quantum field theory, Adv. Ser. Math. Phys. 14
(1992) 1208.
[4] S. Ghoshal, A.B. Zamolodchikov, Int. J. Mod. Phys. A 9 (1994) 38413886, hep-th/9306002;
S. Ghoshal, A.B. Zamolodchikov, Int. J. Mod. Phys. A 9 (1994) 4353, Erratum.
[5] Z. Bajnok, L. Palla, G. Takcs, Nucl. Phys. B 750 (2006) 179212, hep-th/0603171.
233
[6] G. Mussardo, Spectral representation of correlation functions in two-dimensional quantum field theories, hepth/9405128.
[7] P. Dorey, M. Pillin, R. Tateo, G. Watts, Nucl. Phys. B 594 (2001) 625659, hep-th/0007077.
[8] J.L. Cardy, G. Mussardo, Nucl. Phys. B 340 (1990) 387402.
[9] A. Koubek, G. Mussardo, Phys. Lett. B 311 (1993) 193201, hep-th/9306044.
[10] A. Koubek, Nucl. Phys. B 428 (1994) 655680, hep-th/9405014.
[11] A. Koubek, Nucl. Phys. B 435 (1995) 703734, hep-th/9501029.
[12] F.A. Smirnov, Nucl. Phys. B 453 (1995) 807824, hep-th/9501059.
[13] Z. Bajnok, G. Bhm, G. Takcs, J. Phys. A 35 (2002) 93339342, hep-th/0207079;
Z. Bajnok, G. Bhm, G. Takcs, Nucl. Phys. B 682 (2004) 585617, hep-th/0309119.
[14] M. Karowski, P. Weisz, Nucl. Phys. B 139 (1978) 455476.
[15] G. Delfino, G. Mussardo, Nucl. Phys. B 455 (1995) 724758, hep-th/9507010.
[16] P. Dorey, A. Pocklington, R. Tateo, G. Watts, Nucl. Phys. B 525 (1998) 641663, hep-th/9712197.
[17] J.L. Cardy, Phys. Rev. Lett. 54 (1985) 13541356.
[18] R. Kedem, T.R. Klassen, B.M. McCoy, E. Melzer, Phys. Lett. B 307 (1993) 6876, hep-th/9301046;
A. Berkovich, Nucl. Phys. B 431 (1994) 315348, hep-th/9403073.
[19] J.L. Cardy, Nucl. Phys. B 324 (1989) 581596.
[20] J.L. Cardy, D.C. Lewellen, Phys. Lett. B 259 (1991) 274278.
[21] D.C. Lewellen, Nucl. Phys. B 372 (1992) 654682.
[22] J.L. Cardy, G. Mussardo, Phys. Lett. B 225 (1989) 275278.
[23] Al.B. Zamolodchikov, Nucl. Phys. B 348 (1991) 619641.
[24] P. Dorey, I. Runkel, R. Tateo, G. Watts, Nucl. Phys. B 578 (2000) 85122, hep-th/9909216.
[25] P. Dorey, R. Tateo, G. Watts, Phys. Lett. B 448 (1999) 249256, hep-th/9810098.
[28] A. Fring, G. Mussardo, P. Simonetti, Nucl. Phys. B 393 (1993) 413441, hep-th/9211053.
[29] E. Corrigan, A. Taormina, J. Phys. A 33 (2000) 87398754, hep-th/0008237.
[30] S. Ghoshal, Int. J. Mod. Phys. A 9 (1994) 48014810, hep-th/9310188.
[31] E. Corrigan, Boundary bound states in integrable quantum field theory, in: Proceedings of 4th Annual European
TMR Conference on Integrability Non-Perturbative Effects and Symmetry in Quantum Field Theory, Paris, France,
713 September 2000, hep-th/0010094.
[32] Al.B. Zamolodchikov, Talk at the 4th Bologna workshop on conformal and integrable models, unpublished;
Z. Bajnok, L. Palla, G. Takcs, Nucl. Phys. B 622 (2002) 565592, hep-th/0108157.
[33] O.A. Castro-Alvaredo, J. Phys. A 39 (2006) 1190111914, hep-th/0606269.
[34] F.A. Smirnov, Nucl. Phys. B 337 (1990) 156180.
[35] H. Babujian, M. Karowski, Int. J. Mod. Phys. A 19S2 (2004) 3449, hep-th/0301088.
[36] G. Delfino, P. Simonetti, J.L. Cardy, Phys. Lett. B 387 (1996) 327333, hep-th/9607046.
[37] G. Delfino, G. Niccoli, J. Stat. Mech. 0504 (2005) P004, hep-th/0501173.
[38] G. Mussardo, P. Simonetti, Int. J. Mod. Phys. A 9 (1994) 33073338, hep-th/9308057.
[39] J. Balog, P. Weisz, Construction and clustering properties of the 2d non-linear sigma-model form factors: O(3),
O(4), large n examples, hep-th/0701202.
Fermions in self-dual vortex background

on a string-like defect
Yu-Xiao Liu , Li Zhao, Xin-Hui Zhang, Yi-Shi Duan
Institute of Theoretical Physics, Lanzhou University, Lanzhou 730000, PR China
Received 22 February 2007; accepted 11 May 2007
Abstract
By using the self-dual vortex background on extra two-dimensional Riemann surfaces in 5 + 1 dimensions, the localization mechanism of bulk fermions on a string-like defect with the exponentially decreasing
warp-factor is obtained. We give the conditions under which localized spin 1/2 and 3/2 fermions can be
obtained.
PACS: 11.10.Kk; 04.50.+h
Keywords: Fermionic zero modes; String-like defect
1. Introduction
It is now widely believed that extra dimensions play an important role in constructing a unified
theory of all interactions and provides us with a new solution to hierarchy problem [1,2]. The
possible existence of such dimensions got strong motivation from theories that try to incorporate
gravity and gauge interactions in a unique scheme, in a reliable manner. The idea dates back
to the 1920s, to the works of Kaluza and Klein [3] who tried to unify electromagnetism with
Einstein gravity by assuming that the photon originates from the fifth component of the metric.
Recently, co-dimension two models in six dimensions have been a topic of increasing interest
[47]. Apart from model construction, the question of solving the cosmological constant problem
E-mail addresses: liuyx@lzu.edu.cn (Y.-X. Liu), zhl03@lzu.cn (L. Zhao), zhangxingh03@lzu.cn (X.-H. Zhang),
ysduan@lzu.edu.cn (Y.-S. Duan).
doi:10.1016/j.nuclphysb.2007.05.018
Y.-X. Liu et al. / Nuclear Physics B 785 [FS] (2007) 234245
235
has been the primary issue addressed in several articles [8]. Other aspects such as cosmology,
brane gravity [9], fermion families and chirality [10], etc., have been discussed by numerous
authors. A list of some recent articles on codimension two models is provided in Ref. [11].
In the brane world scenario, our universe is regarded as a 3-brane embedded in a higherdimensional spacetime with non-factorizable warped geometry. It is a priori assumed that all
the matter fields are constrained to live on the three-brane, whereas gravity is free to propagate
in the extra dimension. Then a key ingredient for realizing the brane world idea is localization of
various bulk fields on a brane by a natural mechanism. In other words, in this scenario various
fields we observe in our universe are nothing but the zero modes of the corresponding bulk fields
which are trapped on our brane by some ingenious mechanism.
This localization mechanism has been recently investigated within the framework of a local
field theory. Ever since Goldberger and Wise [12] added a bulk scalar field to fix the location
of the branes in five dimensions, investigations with bulk fields have become an active area of
research. It has been shown that the graviton [2] and the massless scalar field [13] have normalizable zero modes on branes of different types, that the Abelian vector fields are not localized
in the RandallSundrum (RS) model in five dimensions but can be localized in some higherdimensional generalizations of it [6]. In contrast, in [13,14] it was shown that fermions do not
have normalizable zero modes in five dimensions, while in [6] the same result was derived for a
compactification on a string [7] in six dimensions. Subsequently, Randjbar-Daemi et al. studied
localization of bulk fermions on a brane with inclusion of scalar backgrounds [15] and minimal
gauged supergravity [16] in higher dimensions and gave the conditions under which localized
chiral fermions can be obtained.
Since spin half fields can not be localized on the brane [2,6] in five or six dimensions by
gravitational interaction only, it becomes necessary to introduce additional non-gravitational interactions to get spinor fields confined to the brane or string-like defect. Fermionic zero modes
in the absence of gravity and in four dimensions in vortex background were studied in [17] and
extended to the case of six-dimensional spacetimes in gravity, gauge and vortex backgrounds
in [18]. The aim of the present article is to study localization of bulk fermions on a string-like
defect with codimension 2 in self-dual vortex background. In this article, we first review the solutions to Einsteins equations with a warp factor in a 6-dimensional spacetime, which has been
studied by many groups [4,7,19,20]. Then, we shall prove that spin 1/2 and 3/2 fields can be
localized on a defect with the exponentially decreasing warp factor if the self-dual vortex and
gravitational backgrounds are considered.
2. Self-dual vortex on a two-dimensional curved space
This paper is focused on braneworld models with codimension greater than one. In particular,
we shall be exclusively concerned with bulk spacetimes in six dimensions generically represented
by the line element
ds 2 = gMN dx M dx N = g (x, y) dx d x + ij (y) dy i dy j ,
(1)
where M, N denote 6-dimensional spacetime indices, , = 0, 1, 2, 3 and i, j = 1, 2 for our

4-dimensional spacetime and the two-dimensional extra space K 2 , respectively, ij is the metric
on K 2 .
To generate the vortex solution, we introduce the generalized Abelian Higgs Lagrangian

M

1
MN
2
2 2
LAH = g FMN F
(2)
+ D (DM ) v
,
4
2
236
where g = det(gMN ), FMN = M AN N AM , = (y k ) is a complex scalar field on extra

1
dimensions, = ( ) 2 , AM is a U(1) gauge field, and DM = M ieAM is gauge-covariant.
In Eq. (2), v is the vacuum expectation value of the Higgs field determining the masses of the
Higgs and of the gauge boson
mH = 2v,
(3)
mV = ev.
The AbrikosovNielsenOlesen vortex solution on K 2 could be generated from the Higgs field .
In the generalized Abelian Higgs model, if the system admits a Bogomolnyi limit [21], one can
arrive at the first-order Bogomolnyi self-duality equations in a curved space [22]:

B = e 2 v 2 ,
(4)
Di i ij j k Dk = 0.
(5)
The complex Higgs field can be regarded as the complex representation of a two-dimensional
vector field = ( 1 , 2 ) over the base space, it is actually a section of a complex line bundle on
the base manifold. Substituting = 1 + i 2 and Di = i ieAi into Eq. (5) and splitting the
real part form the imaginary part, we obtain the following equations

i j = eAi j k k il lm j k m k eAm j .

(6)
From Eq. (6), by calculating i i , we can obtain the expression of the gauge potential

1
i i ij j k k ln .
eAi =
(7)
2
2i
If we define the unit vector
a
(a, b = 1, 2),
na =
(8)

and note the identity

1
i i ,
ab na i nb =
(9)
2
2i
Eq. (7) further simplifies to:
eAi = ab na i nb ij j k k ln .

(10)
In curved space, the magnetic field is defined by B = 1 ij i Aj , according to Eq. (10), we
have

e B = ij ab i na j nb ij j k i kl l ln .
(11)
So the first self-duality equation (4) can be generalized to

e2 2 v 2 = ij ab i na j nb ij j k i kl l ln .
(12)
According to Duans -mapping topological current theory [23], it is easy to see that the first
term on the RHS of Eq. (12) bears a topological origin, and the topological term just describes
the non-trivial distribution of n [24]. Noticing i na = i a / + a i (1/) and the Green
(a = a ), it can be proved that [25]
function relation in -space: a a ln() = 2 2 ()

N

Wk (
y yk ),
= 2
ab i n j n = 2 ()J
y
ij
k=1
(13)
237
where J (/y) is the Jacobian and Wk = k k is the winding number around the kth vortex, the
positive integer k is the Hopf index and k = 1 is the Brouwer degree, yk are the coordinates
of the kth vortex. So the first Bogomolnyi self-duality equation (4) should be

Wk (
y yk ) ij j k i kl l ln .
e2 2 v 2 = 2
N
(14)
k=1
Obviously the first term on the RHS of Eq. (14) describes the topological self-dual vortex.
Now let us discuss the case of flat space for the self-duality equation (14). In this special case,
ij = ij and Eq. (14) reads as
N

Wk (
y yk ) i i ln .
e2 2 v 2 = 2
(15)
k=1
While the corresponding conventional self-duality equation is [26]

e2 2 v 2 = i i ln .
(16)

Comparing our equation (15) with Eq. (16), one can see that the topological term 2 N
k=1 Wk
(
y yk ), which describes the topological self-dual vortex, is missed in the conventional equation. Obviously, only when the field = 0, the topological term vanishes and the conventional
equation is correct. So, the exact self-duality equation should be Eq. (15) for flat space and
Eq. (14) for curved one. As for conventional self-dual nonlinear equation (16), a great deal of
work has been done by many physicists on it, and a vortex-like solution was given by Jaffe [27].
But no exact solutions are known.
In the following sections, we first give a brief review of a string-like defect solution to Einsteins equations with sources, then we study fermionic zero modes coupled with the vortex
background and the localization of fermions on the string-like defect with an exponentially decreasing warp factor.
3. Review of a string-like defect
Let us consider Einsteins equations with a bulk cosmological constant and an energy
momentum tensor TMN in general six dimensions:
1
RMN gMN R = gMN + 62 TMN ,
2
(17)
where 6 denotes the 6-dimensional gravitational constant with a relation 62 = 8GN = 8/M4 ,
GN and M being the 6-dimensional Newton constant and the 6-dimensional Planck mass scale,
respectively, the energymomentum tensor is defined as

TMN =
(18)
d 6 x gLm .
g g MN
We shall consider the most general metric ansatz for a warped brane embedded in six dimensions obeying four-dimensional Poincare invariance
ds 2 = eA(r) g (x) dx dx + dr 2 + R02 eB(r) d 2 ,
(19)
238
where the radial coordinate r is infinitely extended (0 < r < ) and the compact coordinate
ranges from 0 2 , R0 is an additional parameter characterizing the extra compact direction. Moreover, we shall adopt the ansatz for the energymomentum tensor respecting the
spherical symmetry:
T = t0 (r),
Trr = tr (r),
T = t (r),
(20)
where ti (i = 0, r, ) are functions of only the radial coordinate r.

Under these anstze, Einsteins equations (17) and the conservation law for energy
momentum tensor M TMN = 0 reduce to
eA R 3(A
)2 2A
B
2 + 262 tr = 0,
(21)
1
eA R + 4A
5(A
)2 (B
)2 2 + 262 t = 0,
(22)
2
eA R + 2B
+ 6A
10(A
)2 (B
)2 4 + 462 t0 = 0,
(23)
1
tr
= 2A
(tr t0 ) + B
(tr t ),
(24)
2
where R are the scalar curvatures associated with the metric g , and the prime denotes the
derivative with respect to r. Here we define the cosmological constant on the 3-brane by the
equation
1
R g R = g .
(25)
2
It is now known that there are many interesting solutions to these equations (see, for instance,
[19]). Here, we shall consider the brane solutions with a warp factor
A(r) = cr,
(26)
where c is a constant. A specific solution occurs when we have the spontaneous symmetry breakdown tr = t [19]:
ds 2 = ecr g dx dx + dr 2 + R02 ec1 r d 2 ,
(27)
where

2 2
6 t > 0,
5
2
c1 = c 62 t ,
c
R = 4 = 0.
c2 =
(28)
(29)
(30)
This special solution would be utilized to analyze localization of fermionic fields on a string-like
defect.
4. Localization of fermions
In this section, we have the physical setup in mind such that local cosmic string sits at
the origin r = 0 and then ask the question of whether various bulk fermions with spin 1/2 and
3/2 can be localized on the brane with the exponentially decreasing warp factor by means of
the gravitational interaction and vortex backgrounds. Of course, we have implicitly assumed
that various bulk fields considered below make little contribution to the bulk energy so that the
solution (27) remains valid even in the presence of bulk fields.
239
4.1. Spin 1/2 fermionic field

In this subsection we study localization of a spin 1/2 fermionic field in gravity (27) and
vortex backgrounds. It will be shown that provided that if the vortex background satisfies certain
condition, there is a localized zero mode on the string-like defect.
Let us consider the Dirac action of a massless spin 1/2 fermion coupled to gravity and vortex
backgrounds:

Sm = d D x gi M DM ,
(31)
from which the equation of motion is given by
M (M + M ieAM ) = 0,
(32)
M N is the spin connection with M,

N , . . . denoting the local Lorentz
where M = 14 M
M N
M
M
indices, and are the curved gamma matrices and the flat ones, respectively. From the
M M with eM being the vielbein, we have the relations:
formula M = eM
M
1
= e 2 cr e ,
r = rr r ,
= R01 e 2 c1 r .
1
(33)
M N in the covariant derivative D is defined as

The spin connection M
M

1

M N
N
N
= eN M M eN
N eM
M
2

1
M
M
N eM
eN N M eN
2
1
R
eP M eQN (P eQR Q eP R )eM
.
2
So the non-vanishing components of M are
(34)
1
1
= c1 r .
= cr ,
(35)
4
4
In what follows, to illustrate how the vortex background affects the fermionic zero modes, we
first discuss the simple case that the Higgs field is only relative to r, and then solve the general
Dirac equation for the vacuum Higgs field solution = v.
Case 1: = (r) = 1 (r) + i 2 (r).
In this case, Eq. (10) reduces to:
eAr = ab na r nb ,
1
eA = R01 e 2 c1 r r
(36)
ln .
The Dirac equation then becomes

1
1
e 2 cr e D + r r c c1 + iab na r nb
4

1
iR0 r R01 e 2 c1 r r ln + = 0,
(37)
(38)
240
where e D = e ( ieA ) is the Dirac operator on the 4-dimensional braneworld in

the background of the gauge field A . We are now ready to study the above Dirac equation for
6-dimensional fluctuations, and write it in terms of 4-dimensional effective fields. Since is a
6-dimensional Weyl spinor we can represent it by [16]
(4)
=
(39)
,
0
where (4) is a 4-dimensional Dirac spinor. Our choice for the 6-dimensional constant gamma
matrices M , M = 0, 1, 2, 3, r , , are

0
0 5
0 i
r
=
=
,
,
=
(40)
i 0

0
5 0
where are the 4-dimensional constant gamma matrices and 5 the 4-dimensional chirality
matrix. Imposing the chirality condition 5 (4) = + (4) , the Dirac equation (38) can be written
as

1
1
2 c1 r
cr
a
b
2
e e D + r c c1 + iab n r n R0 e r ln
4
1
+ iR01 e 2 c1 r (4) = 0.
(41)
Now, from the equation of motion (41), we will search for the solutions of the form

eil ,
(4) (x, r, ) = (x)(r)
(42)
where (x) satisfies the massless 4-dimensional Dirac equation e D = 0. For s-wave
solution, Eq. (41) is reduced to

1
2 c1 r
a
b
r c c1 + iab n r n R0 e r ln (r) = 0.
(43)
4
The solution of this equation is given by

r
r
1
2
a
b
dr ec1 r r ln .
(r) exp cr + c1 r i dr ab n r n R0
4
So the fermionic zero mode reads

r
r
1
2
a
b
dr ec1 r r ln .
exp cr + c1 r i dr ab n r n R0
0
4
(44)
(45)
Now we wish to show that this zero mode is localized on the defect sitting around the origin
r = 0 under certain conditions. The condition for having localized 4-dimensional fermionic field
is that (r) is normalizable. It is of importance to notice that normalizability of the ground state
wave function is equivalent to the condition that the coupling constant is non-vanishing.
Substituting the zero mode (45) into the Dirac action (31), the effective Lagrangian for then
becomes

(0)
e D ,
Leff = dr d g i M DM = I1/2 g i
(46)
241
where

I1/2

r
1
2
dr ec1 r r ln .
dr exp cr 2R0
2
(47)
In order to localize spin 1/2 fermion in this framework, the integral (47) should be finite. From
Eq. (47), one can see that whatever the form of Ar (r) is, the effective Lagrangian for (x)
has the same form. If the vortex background vanishes, this integral is obviously divergent for
c > 0 while it is finite for c < 0. If the vortex background does not vanish, the requirement that
the integral (47) should be finite for c > 0 is easily satisfied. For example, a simple choice is
= er and c1 > 0. These fermionic zero modes are generically normalizable on the brane in
the self-dual vortex background if the integral I1/2 does not diverge.
Case 2: the vacuum solution 2 = v 2 .
For the vacuum solution, 2 = v 2 , i.e. = vei , we have nr = cos , n = sin , and
eAr = 0, eA = 1. The Dirac equation then becomes

1
1
cr
r
2
e e D + r c c1 + ( + i) = 0.
(48)
4
Repeating the deduction as the above case and imposing the chirality condition 5 (4) = (4) ,
one can get the fermionic zero modes

r
1
1
1
dr e 2 c1 r .
exp cr + c1 r R0
4
The effective Lagrangian for (x) then becomes

(0)
D ,
Leff = dr d g 0 i M DM 0 = I1/2 g i
(49)
(50)
where

I1/2

r
1
1
1
c1 r
2
dr e
dr exp cr 2R0
.
2
(51)
In order to localize spin 1/2 fermion on a string-like defect with the exponentially decreasing
warp-factor (i.e. c > 0) in this framework, the integral (51) should be finite. It is easy to see that
the condition is c1 > 0, i.e.
< 462 t ,
< 62 t ,
for t > 0,
for t < 0.
(52)
This situation is a little different from the above case.

4.2. Spin 3/2 fermionic field
Next we turn to spin 3/2 field, in other words, the gravitino. Let us start by considering the
action of the RaritaSchwinger gravitino field:

Sm = d D x g M i [M N R] DN R ,
(53)
242
where the square bracket denotes the anti-symmetrization, and the covariant derivative is defined
R = eR ( eM + M N e
with the affine connection MN
N N ) by
M
M M N
R
R + M N + AM N .
DM N = M N MN
(54)
From the action (53), the equations of motion for the RaritaSchwinger gravitino field are given
by
[M N R] DN R = 0.
(55)
For simplicity, from now on we limit ourselves to the flat brane geometry g = . After
taking the gauge condition = 0 and A = 0, the non-vanishing components of the covariant
derivative are calculated as follows:
1
1
D = cecr r + cr ieA ,
2
4
1
1
D r = r + c + cr r ieA r ,
2
4
1
Dr = r + c ieAr ,
2
Dr r = r r ieAr r ,
1
D = + c1 r ieA ,
4
1
D r = r + c1 r r ieA r ,
4
1
D = c1 R0 ec1 r r .
2
Again we represent M as the following form
(4)
M
,
M =
0
(56)
(57)
(4)
where M is the 4D RaritaSchwinger gravitino field.

(4)
(4)
Imposing the chirality condition 5 = + , and substituting Eqs. (56) and (57) into the
equations of motion (55), we will look for the solutions of the form

(4) (x, r, ) = (x)u(r)
(58)
eil ,

eil ,
r(4) (x, r, ) = r (x)u(r)
(59)
where (x) satisfies the following 4-dimensional equations = = [ ] (
ieA ) = 0. Then the equations of motion (55) reduce to

1
1
1 12 c1 r
A (r) u(r) = 0,
r c c1 ieAr (r) + eR0 e
(60)
2
4
from which u(r) is easily solved to be

r
r
1
1
1
u(r) exp cr + c1 r + ie dr Ar (r) eR01 dr e 2 c1 r A (r) .
2
4
(61)
243
In the above we have considered the s-wave solution and r = 0.

Let us substitute the zero mode (61) into the RaritaSchwinger action (53). It turns out that
the effective Lagrangian becomes

Leff = dr d g M i [M N R] DN R
= I3/2 i [ ] ( ieA ) ,
(62)
where the integral I3/2 is defined as

I3/2

r
1
1
dr exp cr 2eR01 dr e 2 c1 r A (r) .
2
(63)
As in the above subsection, to illustrate how the vortex background affects the fermionic zero
modes, we first discuss the simple case that the Higgs field is only relative to r, and then solve
the general Dirac equation for the vacuum Higgs field solution = v.
Case 1: = (r) = 1 (r) + i 2 (r).
In this case, Eq. (10) reduces to Eqs. (36) and (37), and the integral I3/2 can be expressed as

I3/2

r
1
2
c1 r
dr e r ln .
dr exp cr 2R0
2
(64)
It is easy to see that this expression is equivalent to I1/2 in (47) up to an overall constant factor so
we encounter the same result as in the corresponding case for spin 1/2 field. So the zero modes
for spin 3/2 field are generically normalizable on the brane in the self-dual vortex background if
the integral I3/2 (64) does not diverge.
Case 2: the vacuum solution 2 = v 2 .
In this case, eAr = 0, eA = 1. Again, changing the chirality condition to 5 (4) = (4) ,
the integral I3/2 takes the form

I3/2

r
1
1
1
c1 r
2
dr e
dr exp cr 2R0
,
2
(65)
which is equivalent to I1/2 in (51) up to an overall constant factor so it is also finite for c > 0 and
c1 > 0.
5. Discussions
Using the generalized Abelian Higgs model and -mapping theory, we investigate the selfdual vortex on an extra two-dimensional curved Riemann surface, and obtain the inner topological structure of the self-dual vortex. Under the gravity and vortex backgrounds, we have
investigated the possibility of localizing the spin 1/2 and 3/2 fermionic fields on a brane with
the exponentially decreasing warp factor. We first give a brief review of a string-like defect solution to Einsteins equations with sources, then check localization of fermionic fields on such
a string-like defect with the background of self-dual vortex from the viewpoint of field theory.
It has been found that the vortex background affects the fermionic zero modes, and that spin
1/2 and 3/2 fields can be localized on a defect with the exponentially decreasing warp factor if
self-dual vortex and gravitational backgrounds are considered.
244
Acknowledgements
This work was supported by the National Natural Science Foundation of the Peoples Republic of China and the Fundamental Research Fund for Physics and Mathematic of Lanzhou
University.
References
[1] V.A. Rubakov, M.E. Shaposhnikov, Phys. Lett. B 125 (1983) 136;
K. Akama, Lect. Notes Phys. 176 (1982) 267, hep-th/0001113;
M. Visser, Phys. Lett. B 159 (1985) 22, hep-th/9910093;
G.R. Dvali, M.A. Shifman, Phys. Lett. B 396 (1997) 64;
N. Arkani-Hamed, S. Dimopoulos, G.R. Dvali, Phys. Lett. B 429 (1998) 263, hep-ph/9803315.
[2] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hep-ph/9905221;
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690, hep-th/9906064.
[3] T. Kaluza, Sitzungsber. Preuss. Akad. Wiss. Berlin, Math. Phys. K 1 (1921) 966;
O. Klein, Z. Phys. 37 (1926) 895.
[4] A. Chodos, E. Poppitz, Phys. Lett. B 471 (1999) 119, hep-th/9909199;
A.G. Cohen, D.B. Kaplan, Phys. Lett. B 470 (1999) 52, hep-th/9910132;
R. Gregory, Phys. Rev. Lett. 84 (2000) 2564, hep-th/9911015.
[5] F. Leblond, R.C. Myers, D.J. Winters, JHEP 0107 (2001) 031;
I.I. Kogan, S. Mouslopoulos, A. Papazoglou, G.G. Ross, Phys. Rev. D 64 (2001) 124014;
Z. Chacko, A.E. Nelson, Phys. Rev D 62 (2000) 085006;
E. Ponton, E. Poppitz, JHEP 0102 (2001) 042;
P. Kanti, R. Madden, K.A. Olive, Phys. Rev. D 64 (2001) 044021;
C.P. Burgess, J.M. Cline, N.R. Constable, H. Firouzjahi, JHEP 0201 (2002) 014;
R. Koley, S. Kar, Class. Quantum Grav. 24 (2007) 79, hep-th/0611074.
[6] I. Oda, Phys. Lett. B 496 (2000) 113, hep-th/0006203.
[7] T. Gherghetta, M. Shaposhnikov, Phys. Rev. Lett. 85 (2000) 240, hep-th/0004014.
[8] S.M. Carroll, M.M. Guica, hep-th/0302067;
I. Navarro, Class. Quantum Grav. 20 (2003) 3603;
J. Garriga, M. Porrati, Phys. Lett. B 580 (2004) 87;
J. Vinet, J.M. Cline, Phys. Rev. D 71 (2005) 064011;
J.M. Schwindt, C. Wetterich, Phys. Lett. B 628 (2005) 189.
[9] U. Guenther, P. Moniz, A. Zhuk, Phys. Rev. D 68 (2003) 044010;
P. Bostock, R. Gregory, I. Navarro, J. Santiago, Phys. Rev. Lett. 92 (2004) 221601;
E. Papantonopoulos, A. Papazoglou, JCAP 0507 (2005) 004;
E. Papantonopoulos, A. Papazoglou, JHEP 0509 (2005) 012;
G. Kofinas, Phys. Lett. B 633 (2006) 141;
J.M. Schwindt, C. Wetterich, Nucl. Phys. B 726 (2005) 75.
[10] R. Erdem, Eur. Phys. J. C 25 (2002) 623, hep-ph/0011188.
[11] M. Peloso, L. Sorbo, G. Tasinato, Phys. Rev. D 73 (2006) 104025;
S. Aguilar, D. Singleton, Phys. Rev. D 73 (2006) 085007;
B.M.N. Carter, A.B. Nielsen, D.L. Wiltshire, JHEP 0607 (2006) 034;
N. Kaloper, D. Kiley, JHEP 0603 (2006) 077;
E. Papantonopoulos, gr-qc/0601011.
[12] W.D. Goldberger, M.B. Wise, Phys. Rev. D 60 (1999) 107505;
W.D. Goldberger, M.B. Wise, Phys. Rev. Lett. 83 (1999) 4922.
[13] B. Bajc, G. Gabadadze, Phys. Lett. B 474 (2000) 282, hep-th/9912232.
[14] S.L. Dubovsky, V.A. Rubakov, P.G. Tinyakov, Phys. Rev. D 62 (2000) 105011;
Y. Grossman, N. Neubert, Phys. Lett. B 474 (2000) 361;
C. Ringeval, P. Peter, J.P. Uzan, Phys. Rev. D 65 (2002) 044416;
S. Ichinose, Phys. Rev. D 66 (2002) 104015;
R. Koley, S. Kar, Class. Quantum Grav. 22 (2005) 753;
R. Koley, S. Kar, Phys. Lett. B 623 (2005) 244.
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
S. Randjbar-Daemi, M. Shaposhnikov, Phys. Lett. B 492 (2000) 361, hep-th/0008079.

S.L. Parameswaran, S. Randjbar-Daemi, A. Salvio, Nucl. Phys. B 767 (2007) 54, hep-th/0608074.
R. Jackiw, C. Rebbi, Phys. Rev. D 13 (1976) 3398.
Y.Q. Wang, T.Y. Si, Y.X. Liu, Y.S. Duan, Mod. Phys. Lett. A 20 (2005) 3045;
Y.S. Duan, Y.X. Liu, Y.Q. Wang, Mod. Phys. Lett. A 21 (2006) 2019;
Y.X. Liu, Y.Q. Wang, Y.S. Duan, Commun. Theor. Phys. (2007), in press;
Y.X. Liu, L. Zhao, Y.S. Duan, JHEP 0704 (2007) 097, hep-th/0701010.
I. Olasagasti, A. Vilenkin, Phys. Rev. D 62 (2000) 044014, hep-th/0003300.
M. Cvetic, J. Wang, Phys. Rev. D 61 (2000) 124020, hep-th/9912187;
C. Csaki, J. Erlich, T.J. Hollowood, Y. Shirman, Nucl. Phys. B 581 (2000) 309, hep-th/0001033;
P. Berglund, T. Hubsch, D. Minic, JHEP 0009 (2000) 015, hep-th/0005162;
M. Chaichian, M. Gogberashvili, A.B. Kobakhidze, hep-th/0005167;
A. Chodos, E. Poppitz, D. Tsimpis, Class. Quantum Grav. 17 (2000) 3865, hep-th/0006093.
E.B. Bogomolnyi, Sov. J. Nucl. Phys. 24 (1976) 449.
S. Kim, Y. Kim, J. Math. Phys. 43 (2002) 2355, math-ph/0012045;
Y. Verbin, S. Madsen, A.L. Larsen, Phys. Rev. D 67 (2003) 085019, gr-qc/0302010.
Y.S. Duan, M.L. Ge, Kexue Tongbao 20 (1976) 282;
Y.S. Duan, M.L. Ge, Scientia Sinica 11 (1979) 1072.
G. t Hooft, Nucl. Phys. B 79 (1974) 276;
A.M. Polyakov, JETP Lett. 20 (1974) 194.
Y.S. Duan, S. Li, G.H. Yang, Nucl. Phys. B 514 (1998) 705.
G.V. Dunne, Aspects of ChernSimons theory, hep-th/9902115, and references therein.
A. Jaffe, C. Taubes, Vortices and Monopoles, Birkhuser, 1980.
245
The Bethe ansatz approach for factorizable centrally

extended su(2|2) S-matrices
M.J. Martins , C.S. Melo
Universidade Federal de So Carlos, Departamento de Fsica, C.P. 676, 13565-905 So Carlos (SP), Brazil
Received 9 March 2007; received in revised form 10 May 2007; accepted 23 May 2007
Abstract
We consider the Bethe ansatz solution of integrable models interacting through factorized S-matrices
based on the central extension of the su(2|2) symmetry. The respective su(2|2) R-matrix is explicitly related
to that of the covering Hubbard model through a spectral parameter dependent transformation. This mapping
allows us to diagonalize inhomogeneous transfer matrices whose statistical weights are given in terms of
su(2|2) S-matrices by the algebraic Bethe ansatz. As a consequence of that we derive the quantization
condition on the circle for the asymptotic momenta of particles scattering by the su(2|2) su(2|2) S-matrix.
The result for the quantization rule may be of relevance in the study of the energy spectrum of the AdS5 S 5
string sigma model in the thermodynamic limit.
PACS: 05.50.+q; 02.30.Ik
Keywords: Bethe ansatz; S-matrix
1. Introduction
Nowadays there exists a considerable amount of evidences that integrable structures appear in
both planar N = 4 supersymmetric YangMills theory [1,2] and free IIB superstring theory on
the AdS5 S 5 curved space [3,4]. Integrability in the planar N = 4 gauge theory1 is related to the
fact that the spectrum of the anomalous dimensions of the conformal operators can be obtained by
E-mail address: martins@df.ufscar.br (M.J. Martins).

1 For a review on integrable properties of general four-dimensional gauge theories see [5].
doi:10.1016/j.nuclphysb.2007.05.021
M.J. Martins, C.S. Melo / Nuclear Physics B 785 [FS] (2007) 246262
247
diagonalizing long-range integrable spin chains [6,7]. For the AdS5 S 5 string theory, signatures
of integrability have been found on the context of semi-classical string states [810]. There is,
however, indications that integrability may survive when quantum corrections are taken into
account [11]. There exists also arguments in favor of absence of particle production at tree level
for the AdS5 S 5 gauge-fixed sigma model [12] and that the respective quantum world-sheet
energies may arise from quantization conditions of the form of Bethe ansatz equations [1316].
Assuming that such gauge and string theories are indeed quantum integrable it is natural to
expect that much of their properties could be inferred within the factorized S-matrix framework [17]. In the N = 4 gauge theory the S-matrix is supposed to encode the interactions of
a long-range spin chain and it is expected to satisfy the dynamical form of the YangBaxter
equation [18,19]. From the AdS5 S 5 string theory perspective the S-matrix should describe the
scattering amplitudes between the world-sheet excitations and therefore has to obey the standard
YangBaxter equation [20,21]. The basic form of these two matrix operators have recently been
argued to be constrained by the requirement that they are invariant under the centrally extended
su(2|2) su(2|2) superalgebra [1820]. The only remaining freedom consists in an Abelian
phase factor that has been argued to be constrained with the help of crossing symmetry [21].
The next step would then be to determine the quantization conditions for the particle momenta
excitations on a ring of size L. This requisite is necessary to perform non-perturbative investigations of the energy and momenta eigenstates associated to the AdS5 S 5 gaugedfixed sigma
model in the infinite volume limit. The solution of this problem is directly related to the diagonalization of an inhomogeneous two-dimensional vertex model of statistical mechanic whose
Boltzmann weights are the matrix elements of the underlying factorized S-matrix. In fact, this
task has recently been discussed by Beisert [19] in the context of the N = 4 gauge theory. The
respective transfer matrix eigenvalues were proposed by means of the analytical Bethe ansatz
method and apparently also with the help of a mapping to the Hubbard model. We recall that the
construction of eigenvectors is unfortunately beyond the scope of the analytical Bethe ansatz.
It seems therefore of interest to tackle the problem of diagonalization of transfer matrices
based on centrally extended su(2|2) S-matrices by an independent, first principle approach, such
as the algebraic Bethe ansatz method [23,24]. This framework can provide us in an unambiguous
way both the transfer matrix eigenvalues and eigenvectors as well as the respective Bethe ansatz
equation. To this end, the string theory point of view appears to be the ideal one, since the
respective S-matrix satisfies canonical properties expected from (1 + 1)-dimensional integrable
theories [20]. The exact knowledge of the eigenvectors, in a near future, could be of utility in the
understanding of the Hilbert space structure of the AdS5 S 5 string theory in the thermodynamic
limit.
In this paper we are going to consider the latter approach in order to present the Bethe ansatz
solution of the monodromy problem associated to the su(2|2) su(2|2) string S-matrix. We
have organized this paper as follows. In next section we describe the centrally extended su(2|2)
R-matrix in a convenient basis. We show that this operator is related to the R-matrix of the covering Hubbard model by means of a unitary transformation depending on the spectral parameters.
In Section 3, with the help of this mapping, we formulate the diagonalization of an inhomogeneous transfer matrix based on the su(2|2) R-matrix weights by the algebraic Bethe ansatz
approach. In Section 4 we use this result to derive the quantization rule on a ring of size L for
the particle momenta within the asymptotic Bethe ansatz framework [22]. Our conclusions are
summarized in Section 5 and in Appendix A we have collected Shastrys R-matrix.
248
2. The su(2|2) R-matrix

In this section we are going to discuss a family of solutions of the YangBaxter equation,
R23 (1 , 2 )R12 (1 , 3 )R23 (2 , 3 ) = R12 (2 , 3 )R23 (1 , 3 )R12 (1 , 2 ),
(1)
where the R-matrix Rab (, ) acts on the tensor product of two Z2 graded spaces Va(0) Va(1) and
(0)
(1)
Vb Vb . As usual, the th element of the even V (0) and odd V (1) subspaces are distinguished
by its Z2 parity ,

(0)
= 0 for V (1) (even),
(2)
1 for V (odd).
Here we shall consider the case where the graded vector space is that of a central extension
of the su(2|2) superalgebra [25] on its fundamental four-dimensional representation [19,26]. In
what follows, it will be convenient to represent the R-matrix R12 (, ) in terms of its matrix
elements and we shall use the following representation,
R12 (, ) =
4
c,d
Rb,a
eac ebd ,
(3)
a,b,c,d=1
where eab denotes standard Weyl matrices.

One important feature of Eq. (1) is that it is insensitive to the parities of the elements of the
su(2|2) superalgebra. In principle, from a given R-matrix satisfying Eq. (1) it is possible to define
two types of S-matrices [27].2 One of them is the standard S-matrix,
S12 (, ) = P12 R12 (, ),
(4)
while the other possibility is the so-called graded S-matrix,

(g)
S12 (, ) = P12 R12 (, ),
(5)
where the standard and the graded permutators are given by

P12 =
4

a,b=1
eab eba ,
(g)
P12 =
4
(1)a b eab eba .
(6)
a,b=1
The S-matrices defined by Eqs. (4), (5) satisfy the canonical form of the YangBaxter equation,
S12 (1 , 2 )S13 (1 , 3 )S23 (2 , 3 ) = S23 (2 , 3 )S13 (1 , 3 )S12 (1 , 2 ),
(7)
where in the case of the graded S-matrix the tensor products take into account the graduation of
the subspaces.
The framework of R-matrices, however, will be the most convenient one for the purposes of
this paper. In fact, it is the R-matrix that plays the decisive role in an algebraic formulation of
the Bethe ansatz. It will also help us to make a clear relationship between the standard su(2|2)
S-matrix discussed by Arutynov et al. [20] and the R-matrix of the Hubbard model [28]. This
2 Strictly speaking it is required that R cd (, ) = 0 for any + + + = 1, 3. This condition is satisfied by the
a
c
b
d
ab
su(2|2) R-matrix to be discussed in this section.
249
connection is expected, since the existence of such mapping has been predicted by Beisert [19]
in the context of the N = 4 gauge theory. It should be stressed, however, that in this paper we
intend to search a connection between R-matrices satisfying the same canonical YangBaxter
relations (1).
In order to accomplish this task one has to perform certain unitary transformations in the
su(2|2) S-matrix S12 (, ) given by Arutynov et al. [20]. More specifically, we shall use the
property that the YangBaxter equation (7) is invariant by spectral dependent transformations of
the following type,
1
S12 (, ) = G1 ()G2 ()S12 (, )G1
1 ()G2 (),
(8)
where G() is an arbitrary invertible matrix.

It turns out that the suitable transformation we need to make the above mentioned connection
can be decomposed as
1
0
0
0
1 0 0 0
0
0 0 0 0 1
0 t ()
G() =
(9)
,
0
0
t () 0
0 0 1 0
0
0
0
1
0 1 0 0
where t () plays the role of an external field.
By transforming the centrally extended su(2|2) standard S-matrix given by Arutynov et al.
[20] according to Eqs. (8), (9) and by taking into account Eq. (4) we find that the corresponding
R-matrix can be written as
R 12 (, ) = a1 (, )[e11 e11 + e14 e41 + e41 e14 + e44 e44 ]
+ a2 (, )[e11 e44 + e14 e41 + e41 e14 e44 e11 ]
+ a3 (, )[e22 e22 + e23 e32 + e32 e23 + e33 e33 ]
+ a4 (, )[e22 e33 + e23 e32 + e32 e23 e33 e22 ]
+ a5 (, )[e21 e12 + e31 e13 + e24 e42 + e34 e43 ]
+ a6 (, )[e12 e21 + e13 e31 + e42 e24 + e43 e34 ]
1
[e12 e43 + e13 e42 + e42 e13 e43 e12 ]
+ a7 (, )
t ()t ()
+ a8 (, )t ()t ()[e21 e34 + e31 e24 + e24 e31 e34 e21 ]
t ()
+ a9 (, )
[e22 e11 + e22 e44 + e33 e11 + e33 e44 ]
t ()
t ()
+ a10 (, )
[e11 e22 + e11 e33 + e44 e22 + e44 e33 ].
t ()
The ten distinct weights ai (, ) are obtained directly from [20] and they are given by
[x () x + ()] ()
,
[x + () x ()] ()
[x () x + ()][x () + x + ()][x () x + ()] ()
a2 (, ) =
,
[x () x + ()][x ()x () x + ()x + ()] ()
a3 (, ) = 1,
[x () x + ()][x () x + ()][x () + x + ()]
,
a4 (, ) =
[x () x + ()][x ()x () x + ()x + ()]
a1 (, ) =
(10)
250
a5 (, ) =
a6 (, ) =
a7 (, ) =
a8 (, ) =
a9 (, ) =
a10 (, ) =
[x () x ()]
(),
[x + () x ()]
[x + () x + ()] 1
,
[x () x + ()] ()
[x () x + ()][x () x + ()][x + () x + ()]

,
[x () x + ()][x ()x () 1]()
[x () x + ()][x () x + ()][x + () x + ()]

,
[x () x + ()][x ()x () 1]2 ()()
[x () x + ()][x () x + ()]
,
[x + () x ()]
[x () x + ()][x () x + ()] ()
,
[x + () x ()]
()
(11)
+
. The functions x () depend on both the string sigma model coupling
where () = xx ()
()
constant g and the world-sheet rapidity . They are constrained to satisfy the following relations [18,19],
x + ()
= ei ,
x ()
x + () +
1
1
i
x ()
= .
x + ()
x () g
(12)
Before proceeding we note that the constant part of the transformation (9) only relabels the
Weyl basis. This means that for t () = 1 the operator S12 (, ) = P12 R 12 (, ) is just the original su(2|2) S-matrix discussed by Arutynov et al. [20] written in a different basis. We emphasize
that in this new basis the Grassmann parities associated to the R-matrix (10)(12) is

1 for = 2, 3,
=
(13)
0 for = 1, 4.
The first step to compare the R-matrix (10)(12) with the graded R-matrix of the Hubbard
model [28,29] is to rewrite the restriction (12) in terms of Shastrys original coupling constraint.
(s)
Let us denote this latter R-matrix by R12 (, ) which for sake of completeness has been summarized in Appendix A. It turns out that a convenient parameterization for the variables x ()
is3
x + () = i
a() 2h()
,
e
b()
x () = i
b() 2h()
,
e
a()
(14)
where a(), b() and h() are the free-fermion weights and the coupling entering Shastrys
R-matrix defined by Eq. (A.3).
By substituting Eq. (14) into the R-matrix weights (11) and by comparing them with Shastrys
weights given by Eqs. (A.2) we see that a perfect matching occurs for a special choice of function t (), namely
+ 1/4
x ()
.
t () = () =
(15)
x ()
3 We recall that this type of parameterization has first appeared in [28,30] to simplify the transfer matrix eigenvalues
of the covering Hubbard model.
251
We see that for this particular value of t () the Boltzmann weights of the R-matrix (10)(12)
are brought to the most possible symmetrical form. For the specific value (15) we found the
following relationship,
(s)
R12
(, )
R 12 (, )
=
.
1,1
1,1
R(,
)1,1 R (s) (, )1,1
(16)
The above discussion reveals us that both the centrally extended su(2|2) R-matrix and Shastrys graded R-matrix are indeed special cases of a family of R-matrices R 12 (, ) having a
spectral dependent free-parameter. This result brings extra support to the claim by Beisert [19]
that the symmetry underlying the integrability of the Hubbard model should be that of the su(2|2)
superalgebra. Indeed, our result makes it precise the way one can go from the su(2|2) R-matrix
to that of the Hubbard model within a common integrable structure satisfying the canonical
YangBaxter relation (1). This comparison will be of great help to formulated the corresponding
algebraic Bethe ansatz solution in next section.
3. The algebraic Bethe ansatz
In this section we shall study the problem of diagonalizing an inhomogeneous row-to-row
transfer matrix by an algebraic Bethe ansatz,

T , {pi } | = , {pi } |,
(17)
where p1 , . . . , pN are the inhomogeneities.
We will consider the situation in which the Boltzmann weights S12 (, ) of the transfer matrix
T (, {pi }) take into account the graduation of the degrees of freedom,
(g)
S12 (, ) = P12 R 12 (, ),
(18)
where the parities of the graded permutator are defined in Eq. (13).
As usual the transfer matrix T (, {pi }) can be written as the supertrace of an operator denominated monodromy matrix [27],
4

(1) T , {pi } ,
T , {pi } = StrA TA , {pi } =
(19)
=1
where T (, {pi }) denotes the matrix elements on the auxiliary space A C 4 of the following
ordered product of S-matrices,

TA , {pi } = SAN (, pN )SAN1 (, pN1 ) SA1 (, p1 ).
(20)
An essential ingredient to establish an algebraic Bethe solution is the existence of a reference
state | such that the action of the monodromy operator (20) in this state gives as a result a
triangular matrix. In our case this state is easily built by the following product of local vectors,

1
n

0
|j , |j = ,
| =
(21)
0
j =1
0 j
which is an exact eigenstate of T (, {pi }).
252
In order to construct further eigenvectors other than | we need the help of the quadratic
algebra satisfied by the monodromy operator, namely

s

s

R 12 , T , {pi } T , {pi } = T , {pi } T , {pi } R 12 (, ),
(22)
s
where the symbol stands for the supertensor product [27] whose parities are defined in
Eq. (13).
From this algebra one can in principle derive suitablecommutation rules between the monodromy matrix elements acting on the quantum space nj=1 C 4 . The diagonal monodromy
operators define the transfer matrix eigenvalue problem while the off-diagonal ones play the role
of creation and annihilation fields over the pseudovaccum |. A convenient representation of
T (, {pi }) in terms of these fields turns out to be

B() B()
F ()

T , {pi } = C() A() B ()
(23)
,
C() C () D() 44

are two-component row (column) vectors, A()
is a 2 2
where B()
(B ()) and C () (C())

matrix. The fields B(),
B () and C(),
C () are creation and annihilation operators over the
state |, respectively.
The construction of the eigenvectors in terms of the creation fields will depend much on the
form of the Boltzmann weights of the R-matrix R 12 (, ). Considering that the main structure
of R 12 (, ) resembles that of the Hubbard model one expects that the algebraic Bethe ansatz
solution of the eigenvalue problem (17)(20) will be similar to that developed for the Hubbard
chain [30]. For arbitrary t (), however, not all the weights of R 12 (, ) are exactly the same as
that of Shastrys graded R-matrix. It turns out that some of them are crucial in the construction
of the eigenvectors and this means that we need to implement few adaptations on the results of
[30] before using them in our situation. In what follows, we will describe such modifications in
terms of the general matrix elements R(,

)c,d
b,a in order to make the construction of [30] more
widely applicable.
The structure of the eigenvectors is that of multiparticle states parameterized by variables
1 , . . . , m1 that are fixed by Bethe ansatz equations. Formally, they can be written by the following scalar product [30],

| = m1 (1 , . . . , m1 ) . F|,
(24)
m 1
where the components of the vector F j =1 Cj2 shall be fixed by a second Bethe ansatz. The
vector m1 (1 , . . . , m1 ) carries the dependence on the creation fields and obeys a second order
recursion relation given by
1 ) m1 1 (2 , . . . , m1 )
m1 (1 , . . . , m1 ) = B(
2,3
1,1
m1
m1

R(1 , j )4,1
R(k , j )1,1
+
4,1
2,1
j =2 R(1 , j )4,1 k=2 R(k , j )2,1

k =j
F (1 )m1 2 (2 , . . . , j 1 , j +1 , . . . , m1 )B(j )
j
1
k , j )2,2
R(
2,2
k=2
k , j )1,1
R(
1,1
rk k+1 (k , j ).
(25)
253
The auxiliary four-dimensional vector and the 4 4 R-matrix r12 (, ) are given by
1
0
0
0
) 0
) b(,
0 a(,
= ( 0 1 1 0 ) ,
r12 (, ) =
(26)
,
) a(,
0 b(,
) 0
0
0
0
1
) are
where the expressions for the elements a(,
) and b(,
a(,
) =
) =
b(,
2,3
4,1
2,3
4,1
R(,
)3,2
R(,
)4,1
R(, )4,1
R(, )3,2
2,2
4,1
R(,
)2,2
R(, )4,1
3,2
4,1
3,2
4,1
R(,
)4,1
R(,
)3,2
R(, )4,1
R(, )3,2
2,2
4,1
R(,
)2,2
R(, )4,1
(27)
From the structure of the R-matrix (10) we see that the eigenvectors dependence on the free2,3
)
. The functions a(,
) and b(,
parameter t () is encoded only through the weight R(,
)4,1
does not carry any dependence on this extra parameter since it is canceled out in the product
2,3
4,1
. From now on the analysis becomes fairly parallel to that already carried
R(, )3,2
R(,
)4,1
out for the Hubbard model [30]. In particular, one expects that the auxiliary R-matrix r12 (, )
should be that of an isotropic six-vertex model. This can be seen with the help of the following
change of variables,
1
i
i
= x () +
+ .
2g
x () 2g
as
After some simplifications it is possible to rewrite r12 ( , )

2
2

1
i
)
=
e e +
e e ,
r12 (,
( )
g
+ gi
= x + () +
x + ()
=1
(28)
(29)
=1
which is exactly the R-matrix of the rational six-vertex model.

As explained in [30] the state | defined by the expressions (24)(27) becomes an eigenvector of the transfer matrix T (, {pi }) under the requirement that the vector F is an eigenstate
of yet another inhomogeneous transfer matrix whose weights are that of the isotropic six-vertex
model (29). The form of the eigenvalues (, {pi }) depends also on extra variables 1 , . . . , m2
that are needed in the diagonalization of such inhomogeneous six-vertex transfer matrix. Its final
expression in terms of the matrix elements is

, {pi }; {j , l } =
N

1,1
R(,
pi )1,1
j =1
i=1
N

1,1
m1

R(j , )1,1
j , )2,1
R(
2,1
j , )2,1
R(
2,1
m1

2,2
R(,
j )2,2
2,1
R(,
j )2,1
i=1
m
2

j =1
m1

N

4,2
m1

R(, j )4,2
m2

1
1
j )
+
b(,
l , )
l )
b(
b(,
l=1
j =1
l=1
i=1
4,1
R(,
pi )4,1
j =1
4,1
R(,
j )4,1
(30)
254
provided that the rapidities {j } and {j } satisfy the nested Bethe ansatz equations,
1,1
N

R(j , pi )1,1
2,1
i=1 R(j , pi )1,2

m1

j =1
l , j ) =
b(
m1

k=1
k =j
m2
j , k )2,2 R(
k , j )2,1
R(
2,2
1,2
j , k )2,1 R(
k , j )1,1
R(
1,2
1,1
m2

l , k )
b(
,
k , l )
b(
1
,
b(l , j )
l=1
j = 1, . . . , m1 ,
l = 1, . . . , m2 .
(31)
k=1
k =l
From the above expressions we see that both the eigenvalues (30) and the Bethe ansatz equa
tions (31) depend only on Boltzmann weights R(,
)c,d
b,a that are independent of free-parameter
t (). This is not surprising since we have shown that this family of models are related by the
transformation (8) which clearly preserves transfer matrix eigenvalues. The eigenvectors, however, carry a non-trivial dependence on the parameter t () explicitly exhibited in expression (25).
For practical purposes it is relevant to present the expressions for the eigenvalues and Bethe
ansatz equations in terms of the kinematical string variables x (). Since they do not depend on
t () the simplifications that we need to perform on Eqs. (30), (31) follow closely that carried out
for the Hubbard model [30]. Ommiting here such technicalities we find the eigenvalues are given
by

m1
N

x () x + (j )
x (pi ) x + () (pi )
()
, {pi }; {j , l } =
x + (pi ) x () ()
x + () x + (j )
i=1
j =1
m

N
1

x () x + (j )
x + () x + (pi ) 1
()
x () x + (pi ) ()
x + () x + (j )
j =1
i=1
m2

x + () +
1
x + ()
1
x + ()
l +
i
2g
i
x + () +
l 2g
l=1

m1
m2

x + (j ) x +1()
+
()
x + (j ) x 1() l=1
j =1

x () +
x () +
1
x ()
1
x ()
l
l +
i
2g
i
2g

1
x + (pi ) x + () (pi )
x ()x + (pi )
x + (pi ) x () ()
1 x ()x1 (p )
i=1
i
+

m1

x (j ) x +1()
()
x + (j ) x 1()
j =1
+
N 1

(32)
and the corresponding Bethe ansatz equations become,

N +

x (j ) x (pi )
i=1
x + (j ) x + (pi )
m1
l x + (j )

j =1
l x + (j )
(pi ) =
1
x + (j )
1
x + (j )
m2 x + ( ) +

j
l=1
i
2g
i
2g
x + (j ) +
1
x + (j )
1
x + (j )
m2

l k +
k=1
k =l
l k
i
g
i
g
l +
l
i
2g
i
2g
j = 1, . . . , m1 ,
l = 1, . . . , m2 .
(33)
255
We conclude this section with the following comment. Our result for the eigenvalues (32) is
consistent with that proposed by Beisert [19] provided that one takes into account the following
change of variables x () x () and g g. Though the Bethe ansatz equations (33) are
+
invariant by such transformation the ratio xx ()
defining the pseudo-momenta is now inverted.
()
While this ambiguity can appear in analytical Bethe ansatz analysis it certainly does not occur in
our algebraic Bethe ansatz framework.
4. The asymptotic Bethe ansatz
The purpose of this section is to derive the quantization rule for the momenta of the particles
interacting through a su(2|2) su(2|2) factorizable S-matrix by using the asymptotic Bethe
ansatz [22]. This form of scattering has been argued to be the one relevant for the world-sheet
excitations of the AdS5 S 5 sigma model [1820]. Formally, the respective scattering matrix
can be written as,
S12
(p1 , p2 ) = S12 (p1 , p2 ) S12 (p1 , p2 ),
(34)
where p1 , . . . , pN denote the momenta of an interacting N -particle system.

The basic structure of the building block S-matrix S12 (p1 , p2 ) is constrained by the invariance
of the solution of the YangBaxter equation under the centrally extended su(2|2) symmetry. In
principle, one can choose either the standard or the graded su(2|2) S-matrix discussed in Sec (p , p ) plays the role of an operator
tion 2. Our context here, however, is that the S-matrix S12
1 2
describing the monodromy of the N -particle wave function. When periodic boundary conditions
are considered one has to take into account the different particle statistics under cyclic permutations. This compatibility condition leads us to single out the graded su(2|2) S-matrix,
(g)

S12 (p1 , p2 ) = S0 (p1 , p2 ) P12 R 12 (p1 , p2 ) ,
(35)
where S0 (p1 , p2 ) is a scalar factor that cannot be determined on basis of the su(2|2) invariance.
The possible functional form of S0 (p1 , p2 ) has been argued to be restricted by unitarity and
an extension of the crossing property to S-matrices depending on both momenta p1 and p2 [21].
This factor was first proposed on the context of the AdS5 S 5 string spectrum [14] and since
then has been further investigated by several authors [3133]. The general structure of this factor
is believed to be given by the expression,

2 x + (p2 ) x (p1 ) 1
S0 (p1 , p2 ) =
x (p2 ) x + (p1 ) 1
1
x + (p1 )x (p2 )
1
x (p1 )x + (p2 )
2
(p1 , p2 ) ,
(36)
where (p1 , p2 ) is the so-called string dressing term.

This dressing factor can in general be expressed in terms of a standard phase-shift (p1 , p2 ) =
exp[i(p1 , p2 )] where the phase (p1 , p2 ) is an anti-symmetric function of the two momenta p1
and p2 . For recent closed formula representations of the gauge independent part of this function
see for instance [34]. Here we recall that the property (p1 , p2 ) = (p2 , p1 ) together with the
(p , p ) obeys the
expression for the R-matrix given in Section 2 implies that the S-matrix S12
1 2
unitarity condition, namely
S12
(p1 , p2 )S21
(p2 , p1 ) = Id,
where Id is the identity matrix.
(37)
256
We now proceed under the assumption of the existence of an Abelian phase (p1 , p2 ) such
that the S-matrix (34) indeed encodes the interactions of the many-body problem associated to
the AdS5 S 5 string sigma model. Under this hypothesis it is possible to find the respective
particles momenta quantization by means of the asymptotic Bethe ansatz framework [22]. In this
method one assumes that there exists particle coordinate regions |xi xj | Rc where the particles do not interact and that possible off-mass-shell effects can be neglected. In such asymptotic
regions the wave function is simply the linear combination plane waves with asymptotic momenta p1 , . . . , pN . More precisely, the state vector, in an asymptotic region where the particles
coordinates are ordered as 0 xQ1 < xQ2 < < xQN L, has the form of a generalized Bethe
ansatz [35,36],

| =
dx1 dxN
N
A1 ,...,N (P|Q)e
j =1 pPj xQj
N

k (xk )|0,
(38)
k=1
where
creates a particle with internal quantum number k on the vacuum |0. The sum
runs over all N! permutations P of numbers {1, . . . , N } that index the particles.
The interaction between the particles permit them to cross the various regions and the corresponding amplitudes A1 ,...,N (P|Q) are related to one another by the S-matrix (34). For
Q)
differing by the permutation
instance, the amplitudes of two distinct regions (P|Q) and (P|
of neighboring ith and j th particles are connected by
k (xk )
Q)
= S (pi , pj )ii ,jj A1 ,...,i ,j ,...,N (P|Q),
A1 ,...,i ,j ,...,N (P|
(39)
where S (pi , pj )ii ,jj are the matrix elements of the S-matrix (34).
The total energy E and momenta P of the Bethe wave function (38) are given by the free
particle expressions, namely
P=
N
pk ,
k=1
E=
N
(pk ),
(40)
k=1
where (pi ) is the one-particle dispersion relation.

From Eqs. (38), (39) one clearly sees that the role of the scattering theory is to provide the conditions to match the wave function in adjacent free distinct regions. Therefore, the information
that the internal wave function degrees of freedom satisfy both commuting and anticommuting
rules of permutation should then be encoded in the S-matrix (34). This feature is guaranteed
when one takes as S12 (p1 , p2 ) the graded su(2|2) S-matrix equation (35). The next step is to
quantize the asymptotic momenta pk by imposing periodic boundary conditions to the wave
function (38), (39) on a ring of size L. This is accomplished by the successive use of Eq. (39)
in order to relate different regions in the configuration space. The graded approach assures us
strictly periodic boundary conditions in all sectors for both bosonic and fermionic variables. It
turns out that the one-particle momenta pk are required to satisfy the following condition,
eipk L =
( = pk , {pi })
1,1
S 1,1
(pk , pk )
k = 1, . . . , N,
(41)
where (, {pi }) are the eigenvalues of the transfer matrix operator T (, {pi }) given by

()

()
()
T () , {pi } = StrA SAN (, pN )SAN1 (, pN1 ) . . . SA1 (, p1 ) .
(42)
257
There is no need of extra effort to compute the eigenvalues (, {pi }) because of the tensor
()
product character of the S-matrix S12 (p1 , p2 ). They are simply given in terms of the product of
the eigenvalues of the transfer matrix diagonalized in Section 2. More specifically we have,
N

2
(1) (1)
(2) (2)
, {pi } =
S0 (, pi ) , {pi }; j , l
, {pi }; j , l
,
(43)
i=1
()
()
where {j } and {j }, = 1, 2, denote the Bethe ansatz roots used to diagonalize the transfer
matrix based on the two su(2|2) S-matrices of the tensor product (34).
Taking into account the explicit expression given in Eq. (32) as well as the structure of the
Abelian factor (36) one finds that the nested Bethe ansatz equations for one-particle momenta pk
are

1
(1)
(2)
N

m
m

2
x (pi ) x + (pk ) 1 x + (pk )x (pi )
ipk L+N 12 12
iP
e
=e
(pk , pi )
+
1
x (pi ) x (pk ) 1
x (p )x + (p )
i=1
k
i =k
2 m
1 x + (() ) x (p )

k
j
()
=1 j =1
N x + (() ) x (p )

i
j
i P2
i=1
+
x + (()
j ) x (pi )
()
x + (j ) x + (pk )
() +
x (j ) +
()
m2
l=1
x + (()
j )+
k = 1, . . . , N,
1
()
x + (j )
l () +
i
2g
l ()
i
2g
()
x + (j )
()
j = 1, . . . , m1 , = 1, 2,
()
()
l
m1
j =1
()
x + (j )
()
()
x + (j )
(44)
(45)
1
()
x + (j )
1
()
x + (j )
i
2g
i
2g
()
m2
()
()
k +
l
()
k=1
k =l
()
i
g
i
g
l = 1, . . . , m()
2 , = 1, 2.
(46)
Interesting enough, part of the Bethe ansatz equations depend on the total momenta of a given
()
sector as well as of the number of the Bethe rapidities N and m1 . The latter feature suggests that
m
(1)
(2)
L = L N + 21 + 21 play the role of an effective scale from which densities should probably
be measured. It turns out that this scale hides the angular momenta on S 5 of the AdS5 S 5 theory
in the light-cone gauge [32]. This fact can be viewed by bringing the compact Bethe equations
(44)(46) close to the form of those originally proposed by Beisert and Staudacher [13]. We
start the analysis by Eq. (44) that carries the explicit dependence on the size of the system. This
equation is translated to that written by Beisert and Staudacher [13] by using the following roots
identification,
x (pk ) =
x4,k
(1)
g
,
x + j =
x1,j
k = 1, . . . , K4 ,
j = 1, . . . , K1 ,
(47)
(1) x3,j
x + K1 +j =
,
g
j = 1, . . . , K3 ,
(48)
258
(2) x5,j
x + j =
,
g
(2)
g
x + K5 +j =
,
x7,j
j = 1, . . . , K5 ,
j = 1, . . . , K7 ,
(49)
and xi,j
where N K4 , m1 K1 + K3 and m1 K5 + K7 . We recall that the rapidities x4,k
for i = 1, 3, 5, 7 denote the Bethe roots used by Beisert and Staudacher [13].
By taking into account the above relabeling of Bethe roots it follows that Eq. (44) is equivalent
to
(1)
(2)
K K
K K
ipk (LK4 + 3 2 1 + 5 2 7 )
=e
iP

K4 +
1

x4,k x4,i
g2
+
x4,k
x4,i
g2
+
x4,k
x4,i
K3
K1 1

x4,k x3,j
g2
x4,k
x1,j
x3,j
g2
+
x4,k
x1,j
K5
K7 1

x4,k x5,j
g2
x4,k
x7,j
x5,j
g2
+
x4,k
x7,j
x4,k
i=1
i =k
x+
j =1 4,k
x+
j =1 4,k
+
x4,i
j =1
j =1
2
(pk , pi )
k = 1, . . . , K4 .
(50)
Before proceeding we note that the right-hand side of Eq. (50) reveals us the presence of the
7)
1)
+ (K5 K
[13]. Recall that in the notation of
angular momenta charge J = L K4 + (K3 K
2
2
[13] the length L refers to effective scale of the system conjugated to the momenta variables.
Considering the momenta ambiguity mentioned in Section 3 one has to set here J = L. This
fact is in agreement with the expected meaning of J as the world sheet thermodynamic scale of
the gauge-fixed AdS5 S 5 model [32]. In order to complete the mapping of the nested Bethe
equations (45), (46) one has to make the following extra identifications,
(1)
u j =
u2,j
,
g
(1)
(2)
j = 1, . . . , K2 m2 ,
u j =
u6,j
,
g
(2)
j = 1, . . . , K6 m2 , (51)
as well as the definition ui,j = xi,j + xgi,j for i = 1, 3, 5, 7.

By using all these identifications, Eq. (45) can be transformed to the following four type of
Bethe equations,
eiP /2
K4 1

g2
x4,i
x1,j
g2
+
x4,i
x1,j
i=1
eiP /2
iP /2
K4

x4,i x3,j
x+
i=1 4,i
x3,j
K4

x4,i x5,j
i=1
eiP /2
+
x4,i
x5,j
g2
x4,i
x7,j
g2
+
x4,i
x7,j
K2

u1,j u2,l +
l=1
K4 1

i=1
u1,j u2,l
K2

u3,j u2,l +
u u2,l
l=1 3,j
K6

u5,j u6,l +
l=1
u5,j u6,l
j = 1, . . . , K1 ,
i
2
i
2
j = 1, . . . , K3 ,
i
2
i
2
j = 1, . . . , K5 ,
K6

u7,j u6,l +
l=1
i
2
i
2
u7,j u6,l
i
2
i
2
j = 1, . . . , K7 ,
(52)
259
while Eqs. (46) become

K1

u2,l u1,j +
j =1
u2,l u1,j
K5

u6,l u5,j +
j =1
u6,l u5,j
K3
i
2
i
2 j =1
u2,l u3,j +
K7
i
2
i
2 j =1
u6,l u7,j +
u2,l u3,j
u6,l u7,j
i
2
i
2
K2

u2,l u2,k + i
=
,
u2,l u2,k i
l = 1, . . . , K2 ,
i
2
i
2
K6

u6,l u6,k + i
,
u6,l u6,k i
l = 1, . . . , K6 .
k=1
k =l
(53)
k=1
k =l
Direct comparison between Eqs. (50), (52), (53) for P = 0 with those presented by Beisert
and Staudacher in Table 5 of Ref. [13] shows us that they are indeed equivalent in the case of the
grading 1 = 2 = +1 choice.4 Here we emphasize that on the context of the AdS5 S 5 string,
however, is the charge J that plays the role of the thermodynamic scale. As a consequence of
that the thermodynamic limit J has to be taken without making reference to any particular
sector of the theory. This is of special importance in nested Bethe ansatz systems since all the
Bethe roots levels can in principle contribute to the physical properties in the infinite volume
limit. Recall that this feature has recently been pointed out to be relevant to unveil the possible
origin of the dressing phase of the N = 4 YangMills theory [37].
5. Conclusion
In this work we have studied Bethe ansatz properties of integrable models associated to centrally extended su(2|2) superalgebras. The su(2|2) R-matrix has been shown to be in the same
family of Shastrys R-matrix [28,29] with the help of spectral parameter transformation. This
connection made possible the solution of the eigenvalue problem associated to the su(2|2) transfer matrix within the algebraic Bethe ansatz method.
The motivation to study this type of system arose from its conjectured relationship with scattering properties of the world-sheet excitations of the AdS5 S 5 string sigma model. In this
context we have been able to derive the Bethe ansatz equations for the respective particle momenta on the circle. They presented an unusual dependence on the total momenta and the number
of certain rapidities that parameterize the Hilbert space. We have argued that the latter feature
encodes the information that the effective AdS5 S 5 thermodynamic scale is governed by the
angular momenta J charge. This identification allows, in principle, to take the thermodynamic
limit considering the nested Bethe ansatz equations altogether. It remains to be investigate the
configurations of the Bethe roots that dominate the J limit which is crucial for the understanding of the elementary excitations of the spectrum such as the existence of possible bound
states [38,39].
Acknowledgements
We would like to thank N. Beisert and S. Frolov for useful discussions. M.J. Martins would
like to thank Instituut voor Theoretische Fysica, Amsterdam, where this work started, for the
hospitality. This work has been supported by the Brazilian Research Agencies FAPESP and
CNPq.
4 We also recall that here g =
/(4
) where is the t Hooft coupling.
260
Appendix A. Shastrys R-matrix

In this appendix we present the explicit expression of Shastrys graded R matrix [28,29].
Following the notation of [30] this R-matrix is given by
(s)
R 12 (, ) = 2 (, )[e11 e11 + e44 e44 ] + 4 (, )[e11 e44 + e44 e11 ]
+ 1 (, )[e22 e22 + e33 e33 ] + 3 (, )[e22 e33 + e33 e22 ]

+ 7 (, )[e14 e41 + e41 e14 ] 6 (, )[e23 e32 + e32 e23 ]
i8 (, )[e21 e12 + e24 e42 + e31 e13 + e34 e43 ]
i9 (, )[e12 e21 + e13 e31 + e42 e24 + e43 e34 ]
i10 (, )[e12 e43 e13 e42 e42 e13 + e43 e12 ]
+ i10 (, )[e21 e34 e31 e24 e24 e31 + e34 e21 ]
+ 5 (, )[e11 e22 + e11 e33 + e44 e22 + e44 e33 ]
+ 5 (, )[e22 e11 + e22 e44 + e33 e11 + e33 e44 ].
(A.1)
The ten non-null Boltzmann weights are

1 (, ) = e[h()h()] a()a() + e[h()h()] b()b() 5 (, ),

2 (, ) = e[h()h()] a()a() + e[h()h()] b()b() 5 (, ),
3 (, ) =
e[h()+h()] a()b() + e[h()+h()] b()a()

a()b() + a()b()

cosh[h() h()]
5 (, ),
cosh[h() + h()]
e[h()+h()] a()b() + e[h()+h()] b()a()

a()b() + a()b()

cosh(h() h())
5 (, ),
cosh(h() + h())
[h()+h()]
a()b() e[h()+h()] b()a()
e
6 (, ) =
a()b() + a()b()
2
cosh[h() h()]
b () b2 ()
5 (, ),
cosh[h() + h()]
[h()+h()]
a()b() + e[h()+h()] b()a()
e
7 (, ) =
a()b() + a()b()
2
cosh[h() h()]
b () b2 ()
5 (, ),
cosh[h() + h()]

8 (, ) = e[h()h()] a()b() e[h()h()] b()a() 5 (, ),

9 (, ) = e[h()h()] a()b() + e[h()h()] b()a() 5 (, ),

b2 () b2 ()
cosh[h() h()]
10 (, ) =
5 (, ),
a()b() + a()b() cosh[h() + h()]
4 (, ) =
(A.2)
261
where the weight 5 (, ) is an overall normalization. The functions a(), b() and the coupling
h() satisfy the constraints,
a 2 () + b2 () = 1,

a()b()
.
sinh 2h() =
2g
(A.3)
References
[1] J.A. Minahan, K. Zarembo, JHEP 0303 (2003) 013.
[2] N. Beisert, C. Kristjansen, M. Staudacher, Nucl. Phys. B 664 (2003) 131;
N. Beisert, M. Staudacher, Nucl. Phys. B 670 (2003) 439.
[3] I. Bena, J. Polchinski, R. Roiban, Phys. Rev. D 69 (2004) 046002.
[4] G. Mandal, N.Y. Suryanarayana, S.R. Wadia, Phys. Lett. B 543 (2002) 81.
[5] A.V. Belitsky, V.M. Braun, A.S. Gorsky, G.P. Korchemsky, Int. J. Mod. Phys. A 19 (2004) 4715.
[6] N. Beisert, V. Dippel, M. Staudacher, JHEP 0407 (2004) 075;
A. Rej, D. Serban, M. Staudacher, JHEP 0603 (2006) 018.
[7] N. Beisert, Phys. Rep. 405 (2004) 1.
[8] G. Arutynov, M. Staudacher, JHEP 0403 (2004) 004;
G. Arutynov, S. Frolov, J. Russo, A.A. Tseytlin, Nucl. Phys. B 671 (2003) 3;
G. Arutynov, J. Russo, A.A. Tseytlin, Phys. Rev. D 69 (2004) 086009.
[9] V.A. Kazakov, A. Marshakov, J.A. Minahan, K. Zarembo, JHEP 0405 (2004) 024;
V.A. Kazakov, K. Zarembo, JHEP 0410 (2004) 060.
[10] A.A. Tseylin, Semiclassical strings and AdS/CFT, hep-th/0409296;
J. Plefka, Living Rev. Rel. 8 (2005) 9.
[11] I. Swanson, Nucl. Phys. B 709 (2005) 443;
N. Berkovits, JHEP 0503 (2005) 041.
[12] T. Klose, T. Mc Loughlin, R. Roiban, K. Zarembo, World-sheet scattering in AdS5 S 5 , hep-th/0611169.
[13] N. Beisert, M. Staudacher, Nucl. Phys. B 727 (2005) 1;
M. Staudacher, JHEP 0505 (2005) 054.
[14] G. Arutynov, S. Frolov, M. Staudacher, JHEP 0410 (2004) 016.
[15] T. McLoughlin, I. Swanson, Nucl. Phys. B 702 (2004) 86;
S. Frolov, J. Plefka, M. Zamaklar, J. Phys. A: Math. Gen. 39 (2006) 1307.
[16] T. Klose, K. Zarembo, J. Stat. Mech. (2006) P05006.
[17] A.B. Zamolodchikov, A.B. Zamolodchikov, Ann. Phys. 120 (1979) 253.
[18] N. Beisert, The su(2|2) dynamic S-matrix, hep-th/0511082.
[19] N. Beisert, J. Stat. Mech. 0701 (2007) P017.
[20] G. Arutynov, S. Frolov, M. Zamaklar, The ZamolodchikovFaddeev algebra for AdS5 S 5 superstring, hep-th/
0612229.
[21] R.A. Janik, Phys. Rev. D 73 (2006) 086006.
[22] B. Sutherland, Phys. Rev. Lett. 75 (1995) 1248;
B. Sutherland, Rocky Mauntain J. Math. 8 (1978) 413.
[23] E.K. Sklyanin, L.A. Takhtadzhan, L.D. Faddeev, Teor. Mat. Fiz. 40 (1979) 194;
L.A. Takhtajan, L.D. Faddeev, Russ. Math. Surveys 34 (1979) 11.
[24] V.E. Korepin, G. Izergin, N.M. Bogoliubov, Quantum Inverse Scattering Method, Correlation Functions and Algebraic Bethe Ansatz, Cambridge Univ. Press, Cambridge, 1992.
[25] K. Iohara, Y. Koga, Comment. Math. Helv. 76 (2001) 110.
[26] Y.-Z. Zhang, M.D. Gould, J. Math. Phys. 46 (2005) 013505.
[27] P.P. Kulish, E.K. Sklyanin, J. Sov. Math. 19 (1982) 1596.
[28] B.S. Shastry, J. Stat. Phys. 30 (1988) 57.
[29] M. Wadati, E. Olmedilla, Y. Akutsu, J. Phys. Soc. Jpn. 36 (1997) 340;
E. Olmedilla, M. Wadati, Phys. Rev. Lett. 60 (1977) 1595.
[30] M.J. Martins, P.B. Ramos, Nucl. Phys. B 522 (1998) 413;
P.B. Ramos, M.J. Martins, J. Phys. A: Math. Gen. 30 (1997) L195.
262
[31] R. Hernandez, E. Lopez, JHEP 0607 (2006) 004;

L. Freyhult, C. Kristjansen, Phys. Lett. B 638 (2006) 258;
N. Gromov, P. Vieira, Constructing the AdS/CFT dressing factor, hep-th/0703266.
[32] G. Arutynov, S. Frolov, Phys. Lett. B 639 (2006) 378.
[33] N. Beisert, R. Hernandez, E. Lopez, JHEP 0611 (2006) 070.
[34] N. Beisert, B. Eden, M. Staudacher, J. Stat. Phys. Mech. 0701 (2007) P021;
N. Dorey, D.M. Hofman, J. Maldacena, On the singularities of the magnon S-matrix, hep-th/0703104.
[35] C.N. Yang, Phys. Rev. Lett. 19 (1967) 1312;
M. Gaudin, Phys. Lett. A 24 (1967) 55.
[36] M. Gaudin, La Fonction DOnde de Bethe, Masson, 1983;
N. Andrei, K. Furuya, J.H. Lowenstein, Rev. Mod. Phys. 55 (1983) 331.
[37] K. Sakai, Y. Satoh, Origin of dressing phase in N = 4 super-YangMills, hep-th/0703177;
A. Rej, M. Staudacher, S. Zieme, Nesting and dressing, hep-th/0702151.
[38] D.H. Hofman, J. Maldacena, J. Phys. A 39 (2006) 13119.
[39] N. Dorey, J. Phys. A 39 (2006) 13119;
H.-Y. Chen, N. Dorey, K. Okamura, JHEP 0611 (2006) 035.
YangBaxter R-operators and parameter permutations

S. Derkachov a , D. Karakhanyan b , R. Kirschner c,
a St. Petersburg Department of Steklov Mathematical Institute of Russian Academy of Sciences,
Fontanka 27, 191023 St. Petersburg, Russia

b Yerevan Physics Institute, Br. Alikhanian st. 2, 375036 Yerevan, Armenia
c Institut fr Theoretische Physik, Universitt Leipzig, PF 100 920, D-04009 Leipzig, Germany
Received 14 March 2007; received in revised form 16 May 2007; accepted 23 May 2007
Abstract
We present an uniform construction of the solution to the YangBaxter equation with the symmetry
algebra s(2) and its deformations: the q-deformation and the elliptic deformation or Sklyanin algebra. The
R-operator acting in the tensor product of two representations of the symmetry algebra with arbitrary spins
1 and 2 is built in terms of products of three basic operators S1 , S2 , S3 which are constructed explicitly.
They have the simple meaning of representing elementary permutations of the symmetric group S4 , the
permutation group of the four parameters entering the RLL-relation.
1. Introduction
The YangBaxter equation and its solutions play a key role in the theory of the completely
integrable quantum models [15]. The general solution of the YangBaxter equation (R-matrix)
is the operator R(u) acting in a tensor product V1 V2 of two linear spaces. There exists the
natural construction of the solution to the YangBaxter equation. The solution of the problem
consists of two steps:
on the first stage we construct the solution to so-called RLL-relation,
E-mail addresses: derkach@euclid.pdmi.ras.ru (S. Derkachov), karakhan@lx2.yerphi.am (D. Karakhanyan),

roland.kirschner@itp.uni-leipzig.de (R. Kirschner).
doi:10.1016/j.nuclphysb.2007.05.022
264
S. Derkachov et al. / Nuclear Physics B 785 [FS] (2007) 263285
on the second stage we prove that operator R12 (u v) constructed as the solution of the
RLL-relation obeys the general YangBaxter equation.
Choosing one of the three spaces involved to carry a fundamental representation of the symmetry
algebra the YangBaxter equation is reduced to the simpler defining equation for the R-matrix [6]
R12 (u v)L1 (u)L2 (v) = L2 (v)L1 (u)R12 (u v),
(1.1)
where L(u) is the Lax matrix. In the cases of the symmetry algebra s(2) and its q-deformation
the R-matrix can be obtained by the following method [5,6]: The R-matrix is a function of the
Casimir operator and the defining RLL-equation is reduced to a recurrence relation for the function of one variable.
We shall give an uniform construction of the solution of the RLL-relation extending to the
three cases of the symmetry algebra s(2) and its two deformations, the q-deformation and the
elliptic deformation or Sklyanin algebra [7]. It is well known that the resulting R-matrices are
building blocks in the construction of integrable spin chain models. Whereas in questions related
to condensed matter physics the sites of the chain carry usually finite-dimensional representations
the case of infinite-dimensional representations turned out to be relevant in studies of QCD and
super YangMills theories, in particular in calculating the anomalous dimensions of composite
operators or of the Regge singularities of multi-gluon exchange (cf. [8]). It is not known so
far whether the deformed symmetry cases are relevant in the YangMills context. The explicit
formulation of the results in analogous form for all cases may be useful to find new applications.
The main idea of our construction is quite simple: The R-matrix represents some particular
element of the group S4 of permutations of four parameters entering the RLL-relation and it
can be constructed from the simple building blocks, the operators S1 , S2 , S3 corresponding
to the elementary permutations. The Lax matrix L(u) depends on two parameters: the spin of
representation and the spectral parameter u. It is useful to introduce two related parameters u1
u
u
and u2 , u1 = 2
+ ; u2 = 2
1 , where is a free parameter in the case of the Sklyanin
algebra. In the cases of no deformation and q-deformation it can be fixed as = 12 without loss
12
of generality. After extracting the operator of permutation P12 from the R-matrix R12 = P12 R
the defining equation is rewritten in check formalism,
12 L1 (u1 , u2 )L2 (v1 , v2 ) = L1 (v1 , v2 )L2 (u1 , u2 ) R
12 .
R
(1.2)
12 interchanges the set of parameThis equation admits the natural interpretation: The operator R
ters (u1 , u2 ) of the first Lax-matrix with the set of parameters (v1 , v2 ) of the second Lax-matrix.
It corresponds to the special permutation s in the group of permutations S4 of four parameters,
s

( u1 , u2 , v1 , v2 ) ( v1 , v2 , u1 , u2 ).
Arbitrary permutations in the group S4 can be constructed from the elementary transpositions s1 ,
s2 and s3 which interchange only two nearest neighboring components in the set (u1 , u2 , v1 , v2 ).
We look for the operators Si representing these elementary transpositions
s1
s3

( u1 , u2 , v1 , v2 );
s2

(u1 , u2 , v1 , v2 );
s1

S1 L1 ( u1 , u2 ) = L1 (u2 , u1 )S1 ,
s2
s3

S3 L2 ( v1 , v2 ) = L2 (v2 , v1 )S3 ,

S2 L1 (u1 , u2 )L2 (v1 , v2 ) = L1 (u1 , v1 )L2 (u2 , v2 )S2 .
265
These equations appear to be much simpler than the initial defining equation for the R-operator
and the solution can be obtained in a closed form. Finally we construct the R-matrix as the
composite object out of the simplest building blocks, the operators S1 , S2 , S3 .
Note that there are different possible levels of resolution into more elementary permutations:
On the first level we have the group of permutation S2 of two pairs of parameters (u1 , u2 )
and (v1 , v2 ) and the R-matrix is the elementary operator corresponding to this permutation.
On the second level we allow the separate permutations u1 v1 or u2 v2 so that our
(1) and R
(2) corresponding
group of permutations is S2 S2 . On this level the operators R
to these permutations are the elementary building blocks but the R-matrix is a composite
object.
On the third level we allow any permutations of parameters so that our group of permutations
becomes S4 . On this level the operators Si , i = 1, 2, 3, are elementary building blocks.
(1) , R
(2) and R-matrix are composite objects. So we have the chain of
Now the operators R
inclusions with increasing symmetry
S 2 S 2 S 2 S4 .
Thus the R-matrix is factorized into operators representing more elementary permutations.
We shall construct the elementary operator factors Si , i = 1, 2, 3, and proof that they generate a
representation of the permutation group for generic values of the representation parameters u1 ,
u2 , v1 , v2 . Then the YangBaxter relation for the general R-matrix composed out of these factors
is just a consequence of the permutation group relations.
The solution of the YangBaxter relation by factorization into elementary parameter permutations has been given previously in [17] for the symmetry s(N ). Here we show that the method
extends to the deformations of s(2). The construction of Si and the permutation group proofs
will be done uniformly for the cases of no deformation and of the quantum and elliptic deformations emphasizing the analogies. Guided by the analogies we are able to present a relatively
simple solution also for the non-trivial elliptic case.
(i) [16,18,19] act within the tensor prodFor generic values of the spins 1 , 2 the operators R
uct of the corresponding representation spaces, whereas the operators Si map to tensor products
with changed spin values. Moreover, in the case of finite-dimensional representations only the
R-matrix itself leaves the tensor product space invariant. At some special values of the representation parameters the elementary operators Si become singular on parts of their spectra or
develop kernels. Then the permutation group properties become invalid. In this paper we restrict
ourselves to generic parameter values and postpone those more subtle questions related to representation theory to the next paper. There we plan also to consider the uniform construction of
Baxter Q operators in this approach.
The presentation is organized as follows. In Section 2 we give the uniform expression for the
fundamental R-matrices and Lax matrices for the three cases: no deformation, q-deformation and
elliptic deformation of the symmetry algebra s(2). In Section 3 we discuss the connection between RLL-relations and the group of permutations S4 . In Section 4 we construct the operators
S1 , S2 , S3 representing the elementary transpositions and in Section 5 we prove the Coxeter relations for them. In Section 6 we construct the R-matrix and prove that it obeys the YangBaxter
(1) and R
(2) and prove the corresponding
relation. In Section 7 we construct the operators R
YangBaxter relations for them. Finally, in Section 8 we summarize.
266
2. YangBaxter R matrices related to s2

The YangBaxter relation,
R12 (u v)R13 (u)R23 (v) = R23 (v)R13 (u)R12 (u v),
(2.1)
of operators acting on the tensor product V1 V2 V3 where Rij is acting non-trivially only
on Vi , Vj , has the three well-known solutions in the case Vi C2 related to s2 ,
1
wa (u + ) a a .
2
3
R12 (u) =
(2.2)
a=0
Here a denote the Pauli matrices, 0 = 1. There are the three cases of no deformation, quantum
deformation and elliptic deformation. The parameter can be fixed in the first two cases, 2 = 1,
but is variable in the last case. The weight functions are depending on case,
no deformation
w0 (u) = 2u;
w1 (u) = w2 (u) = w3 (u) = 1,
(2.3)
quantum deformation
w0 (u) =
q u q u
1
2
q q
12
w3 (u) =
q u + q u
1
q 2 + q 2
w1 (u) = w2 (u) = 1,
(2.4)
elliptic deformation
wa (u) =
a+1 (u| )
.
a+1 (| )
(2.5)
The relation (2.1) is the source of a rich algebraic structure and integrable quantum systems. In
the case V1 = C2 and V2 arbitrary one calls R12 (u + ) = L(u) Lax matrix,

3
1 w0 (u)S0 + w3 (u)S3 w1 (u)S1 + iw2 (u)S2
1
wa (u) a Sa =
, (2.6)
L(u) =
2
2 w1 (u)S1 iw2 (u)S2 w0 (u)S0 w3 (u)S3
a=0
where the operators Sa generate the universal enveloping algebra U(s2 ) in the first case, its
quantum deformation Uq (s2 ) (deformation parameter q) in the second case and its elliptic deformation(parameters , ), Sklyanin algebra, in the third case. In the first two cases S0 , S3 are
not independent and they are related to the generators in conventional notation
no deformation
S0 = 1,
S3 = 2S,
S1 iS2 = 2S ,
where the conventional generators for the representation with spin are
S = ,
S = z ,
S + = z2 2z.
(2.7)
After substitution in the Lax matrix (2.6) the functions (2.3) and generators in terms of z,
as in (2.7) we obtain that the Lax matrix decomposes into factors depending on one of
the two elementary operators z or only. The spectral parameter u and spin enter in the
combinations u1 = u + and u2 = u 1

1 0
1
u2
L(u1 , u2 ) =
z u1
0 1
z
267

0
;
1
quantum deformation
S0 =
q S + q S
1
2
q +q
;
1
S3 =
q S q S
1
2
q q
12
S1 iS2 = 2S .
(2.8)
The conventional generators for the representation with spin are

1
S = [z]q ,
z
S = z ,
S + = z[z 2]q ,
(2.9)
q
where [x]q is the usual notation for q-numbers [x]q = qqq
1 . Substituting the functions
a
(2.4) and q deformed generators S from (2.9) in the Lax matrix (2.6) we obtain a factorized
form with features analogous to the undeformed case [9,10,19]

z+1
u

1
1
q
q 2
0
z1
;
L(u1 , u2 ) =
zq u1 zq u1
q u2 z1
0
q z1
x
the generators Sa of the Sklyanin algebra can be expressed in terms of z, as follows

Sa =

(i)a,2 a+1 ()
a+1 2(z ) e a+1 2(z ) e .
1 (2z)
(2.10)
Note that in order to simplify the formulae in the elliptic case we change the variable z z
in comparison to the standard notations [7,10] and absorb the model parameter which indeed plays important role in XYZ model in contrast with simpler XXX and XXZ case, but
in present context we do not touch the subtle questions related to the elliptic curve parametrizing the model under consideration. The needed properties of the -functions are listed in
Appendix A. Also in this case an analogous factorization of the Lax matrix can be written
after substituting in (2.6) the generators in terms of z, [9,10]

1
(z u1 )3 (z + u1 )3
L(u1 , u2 ) =
(z u1 )4 (z + u1 )4
1 (2z)

e
(z + u2 )4 (z + u2 )3
0
,
(z u2 )4 (z u2 )3
0 e
where u1 =
u
+ ; u2 = 2
1 and for simplicity we use the notation

(x)3 = 3 x ;
(x)4 = 4 x .
2
2
u
2
We obtain the uniform expressions of Lax factorization.

Proposition 1. In terms of a Heisenberg pair z, the Lax matrices (2.6) can be written in factorized form as
Lc (u1 , u2 ) = [u1 ]c V1
c (u1 , z)Dc Vc (u2 , z),
(2.11)
where the subscript c = o, q, e labels the three cases of no deformation (o), quantum deformation (q) and elliptic deformation (e). The first factor is a usual number [u]o = u, q-number
268

u
q
[u]q = qqq
1 or elliptic number [u]e = 1 (2u) and the others are 2 2 matrices. The matrices
Dc do not depend on the parameters u1 , u2 , u1 = 2uc + ; u2 = 2uc 1 , with 2o = 2q = 1,
2e = 2 and are given in terms of operator of differentiation (o), dilatation by q 1 (q) or
shift by 1 (e)

z+1

1
q
e
0
0
;
Dq =
;
De =
.
Do =
0 1
0
q z1
0 e
u
The matrices Vc depend on the representation parameters and on z,

u

u 0
q
z1
;
Vq (u, z) =
;
Vo (u, z) =
z 1
q u z1

(z + u)4 (z + u)3
.
Ve (u, z) =
(z u)4 (z u)3
3. The RLL-relation and the permutation group
Let us consider the reduction of the general YangBaxter equation (2.1) in the special case
when V3 = C2 . The operator R13 (u) is reduced to the Lax matrix L1 (u), operator R23 (v) is
reduced to L2 (v) so that one obtains the defining RLL-relation for the operator R12 (u v) (1.1).
It is useful to extract the permutation operator from the R-matrix
12 ,
R12 = P12 R
12 the RLL-relation takes the form (1.2)
where P12 (z1 , z2 ) = (z2 , z1 ). For the operator R
where L1 is to be substituted as in (2.11) with z, replaced by z1 , 1 and L2 with z, replaced by
z2 , 2 . The parameters are u1 = 2uc + 1 , u2 = 2uc 1 1 and v1 = 2v c + 2 , v2 = 2v c 1 2 .
12 (u v) is some operator depending on the Heisenberg pairs z1 , 1 and z2 , 2 .
In Eq. (1.2) R
12 (u v) in (1.2) corresponds to the special
As it is mentioned in Introduction the operator R
permutation s in the group of permutations S4 of four parameters u = (u1 , u2 , v1 , v2 ).
12 (u v);
sR
s(u1 , u2 , v1 , v2 ) = (v1 , v2 , u1 , u2 ).
The arbitrary permutation from the group S4 can be constructed from the elementary transpositions s1 , s2 and s3 ,
s1 u = (u2 , u1 , v1 , v2 );
s2 u = (u1 , v1 , u1 , v2 );
s3 u = (u1 , u2 , v2 , v1 )
which interchange only two nearest neighboring components in the set u = (u1 , u2 , v1 , v2 ). For
example the permutation s has the following decomposition s = s2 s1 s3 s2 . It is natural to search
the operators Si (u1 , u2 , v1 , v2 ) = Si (u) representing these elementary transpositions
S1
S3

( u1 , u2 , v1 , v2 );
S2

(u1 , u2 , v1 , v2 ).
These operators obey the following defining equations

S1 (u)L1 (u1 , u2 ) = L1 (u2 , u1 )S1 (u;
S3 (u)L2 (v1 , v2 ) = L2 (v2 , v1 )S3 (u),
S2 (u)L1 (u1 , u2 )L2 (v1 , v2 ) = L1 (u1 , v1 )L2 (u2 , v2 )S2 (u)
(3.1)
(3.2)
and our first step will be the explicit construction of these operators in Section 4. Having these
operators we can construct the R-matrix.
269
12 ,
Proposition 2. The operator R
12 (u v) = S2 (s1 s3 s2 u)S1 (s3 s2 u)S3 (s2 u)S2 (u),
R
(3.3)
obeys the relation (1.2) provided the operators Si obey the relations (3.1), (3.2).
This expression for the R-matrix corresponds to the particular decomposition of the permutation s: s = s2 s1 s3 s2 . In the last section we shall present equivalent expressions corresponding
to another decompositions. We shall see that the operators Si have the special dependence on
parameters
S1 (u) = S1 (u1 u2 );
S2 (u) = S2 (u2 v1 );
S3 (u) = S3 (v1 v2 )
(3.4)
12 (u v) depends on the difference of spectral parameters as it should be.

so that the operator R
We have the correspondence
si Si (u);
si sj Si (sj u)Sj (u)
(3.5)
and to prove that we obtain the representation of the permutation group S4 it remains to prove
the corresponding defining (Coxeter) relations for the generators
si si = 1 Si (si u)Si (u) = 1;
s1 s3 = s3 s1 S1 (s3 u)S3 (u) = S3 (s1 u)S1 (u),
(3.6)
s1 s2 s1 = s2 s1 s2 S1 (s2 s1 u)S2 (s1 u)S1 (u) = S2 (s1 s2 u)S1 (s2 u)S2 (u),
(3.7)
s2 s3 s2 = s3 s2 s3 S2 (s3 s2 u)S3 (s2 u)S2 (u) = S3 (s2 s3 u)S2 (s3 u)S3 (u).
(3.8)
In Section 5 we prove that the obtained operators Si (u) obey these defining relations. After
this we shall prove that YangBaxter relation (2.1) for the operator (3.3) is consequence of the
Coxeter relations (3.6), (3.7) and (3.8).
4. Elementary permutation operators S1 , S2 and S3
4.1. The operator S2
Consider the defining condition for S2 (3.2) where L1 is to be substituted as in (2.11) with z,
replaced by z1 , 1 and L2 with z, replaced by z2 , 2 ,
S2 [u1 ]V1 (u1 , z1 )D1 V(u2 , z1 ) [v1 ]V1 (v1 , z2 )D2 V(v2 , z2 )
= [u1 ]V1 (u1 , z1 )D1 V(v1 , z1 ) [u2 ]V1 (u2 , z2 )D2 V(v2 , z2 ) S2 .
This equation suggests the ansatz S2 = S2 (z1 , z2 ) as a multiplication operator independent of
1 , 2 . Then the operator S2 commutes with the matrices V1 (u1 , z1 ) and V(v2 , z2 ) so that they
can be cancelled and we immediately obtain a much simpler defining equation for the function
S2 (z1 , z2 ),
1
[v1 ] D1
1 S2 (z1 , z2 )D1 V(u2 , z1 )V (v1 , z2 )
= [u2 ] V(v1 , z1 )V1 (u2 , z2 ) D2 S2 (z1 , z2 )D1

2 .
It remains to solve this equation in each case.
(4.1)
270
In the case of no deformation the defining condition (4.1) results in

u2
v1
1 2 ln S2
0
0
1 1 ln S2
.
=
0
1
0
1
z2 z1 v1
z2 z1 u2
The wanted function of z1 , z2 has to obey two relations,
1 S2 (z1 , z2 ) = 2 S2 (z1 , z2 );
(z2 z1 )2 ln S2 (z1 , z2 ) = u2 v1 .
This results in the permutation operator S2 up to normalization

S2 (z1 , z2 ) = (z2 z1 )u2 v1 .
(4.2)
In the case of quantum deformation the defining condition (4.1) in matrix form is
u2 z2 v1

q z1 q
q u2 zz21 q v1
S2 (q 1 z1 , z2 )
0
0
S2 (qz1 , z2 )
q u2 + zz21 q v1 q u2 + zz21 q v1
v1 z2 u2

v
q z1 q
q 1 zz21 q u2
S2 (z1 , qz2 )
0
=
.
0
S2 (z1 , q 1 z2 )
q v1 + zz2 q u2 q v1 + zz2 q u2
1
Equating the diagonal elements of the matrices resulting on both sides leads to the conclusion

z2
S2 (z1 , z2 ) = z1u2 v1
.
z1
The resulting equations from the two off-diagonal elements are compatible and lead to the
difference equation

(qx) 1 xq v1 u2 = q 1 x 1 xq u2 v1 .
Therefore, in the q deformed case the permutation operator S2 has the form
S2 = z1u2 v1
( zz21 q 1u2 +v1 ; q 2 )

( zz21 q 1+u2 v1 ; q 2 )
(4.3)

2k
We use the standard notation (x; q 2 ) =
k=0 (1 xq ). In [20] this expression appeared as
the appropriate generalization to the q deformed case of (z2 z1 )u2 v1 (4.2) in the role of
conformal propagators.
In the case of elliptic deformation the defining equation (4.1) results in the systems of four
difference equations
(z1 + z2 + u2 v1 ) (z1 z2 + u2 + v1 )S2 (z1 1, z2 )
= (z1 + z2 + v1 u2 ) (z1 z2 + v1 + u1 )S2 (z1 , z2 + 1),
(z1 + z2 u2 + v1 ) (z1 z2 u2 v1 )S2 (z1 + 1, z2 )
= (z1 + z2 v1 u2 ) (z1 z2 v2 u2 )S2 (z1 , z2 1),
(z1 + z2 + u2 + v1 ) (z1 z2 + u2 v1 )S2 (z1 1, z2 )
= (z1 + z2 + u2 + v1 ) (z1 z2 + v1 u2 )S2 (z1 , z2 1),
(z1 + z2 u2 v1 ) (z1 z2 u2 + v1 )S2 (z1 + 1, z2 )
= (z1 + z2 u2 v1 ) (z1 z2 v1 + u2 )S2 (z1 , z2 + 1),
where (x) 1 (x| ). This leads us to look for the solution in the form
S2 (z1 , z2 ) = + (z1 + z2 ) (z1 z2 ).
271
In each equation one of the factors drops out and we obtain coinciding difference equations for
these factors. The result for the permutation operator S2 in the elliptic case is
S2 = e2i(u2 v1 )z1
(z1 + z2 + u2 v1 + 1) (z1 z2 + u2 v1 + 1)
,
(z1 + z2 u2 + v1 + 1) (z1 z2 u2 + v1 + 1)
(4.4)
where the function (x) is closely related to the elliptic gamma-function and is defined in (A.8).
We emphasize the common features allowing to do analogous steps in the three cases for
deriving the uniform expressions of operator of parameter permutation S2 .
Proposition 3. The permutation operator is defined by the relation with two Lax matrices
S2 (u1 , u2 ; v1 , v2 )L1 (u1 , u2 )L2 (v1 , v2 ) = L1 (u1 , v1 )L2 (u2 , v2 )S2 (u1 , u2 ; v1 , v2 ),
where Li involves the generators of representation with spin i and is expressed in terms of the
Heisenberg pair zi , i and u1 = 2uc + 1 , u2 = 2uc 1 1 ; v1 = 2v c + 2 , v2 = 2v c 1 2 ,
with 20 = 2q = 1, 2e = 2. S2 is represented as an operator of multiplication, depending on
z1 , z2 and on u2 v1 as given in (4.2), (4.3), (4.4).
Notice that the result for the undeformed case is given by a particular symmetric two-point
function or conformal propagator. The results in the other cases are related to the corresponding
deformed propagators.
4.2. The operators S1 , S3 and the intertwining operator
Let us consider the defining equations (3.1) for the operators S1 and S3 . The permutation of
parameters u1 = 2uc + 1 and u2 = 2uc 1 1 is equivalent to the change of the spin 1
1 1 and similarly the permutation of parameters v1 = 2v c + 2 and v2 = 2uc 1 2 is
equivalent to the change of the spin 2 1 2 . In the Lax matrix (2.6) only the generators Sa depend on the spin so that the defining equations (3.1) can be rewritten in terms of the
generators only. The meaning of this equations is the following: The operator S1 intertwines the
representations of spin 1 and of spin 1 1 and the operator S3 intertwines the representations
of spin 2 and of spin 1 2 ,
S1 Sa1 = Sa11 S1 ;
S3 Sa2 = Sa12 S3 .
Let us consider the intertwining operator W for some , then S1 and S3 are special cases thereof.
W Sa = Sa1 W.
(4.5)
We study subsequently the three cases of no deformation, quantum deformation and elliptic
deformation.
In the case of no deformation Eqs. (4.5) can be rewritten in terms of conventional generators
W (z ) = (z + 1 + )W ;
W z(z 2) = z(z + 2 + 2)W ;
W = W.
The general solution to the first equation has the form W = z21 (z). The two remaining equations are compatible and lead to the difference equation:
(x + 1) =
x +1
(x);
x 2
(x) =
(x + 1)
,
(x 2)
x = z.
272
In the case of quantum deformation the equations (4.5) again can be rewritten in terms of
conventional generators
W (z ) = (z + 1 + )W ;
W z[z 2]q = z[z + 2 + 2]q W ;
1
1
W [z]q = [z]q W.
z
z
The general solution of the first equation is the same as in previous case W = z21 (z)
and two remaining equations result in the difference equation with the known solution
(q 2x2 ; q 2 ) (2+1)x
q x+1 q x1
q
(x);
(x)
=
.
q x2 q x+2
(q 2x+2 ; q 2 )
In the elliptic case the intertwining operator W has been constructed by A. Zabrodin [10].
For the generic the solution of Eqs. (4.5) is given in [10] in terms of the very well poised
elliptic hypergeometric series
(x + 1) =
W=
e2i(2+1)z (2z)
(2z + 2(2 + 1))

[z 2 1 + 2k][z 2 1]k [2 1]k (2+12k)
.
e
[z 2 1][z + 1]k [k]!
(4.6)
k=0
We emphasize the common features allowing to do analogous steps in the three cases for
deriving the uniform expressions for the intertwining operator W .
Proposition 4. The intertwining operator W (u1 , u2 ) is defined by the relation with Lax matrix
W (u1 , u2 )L(u1 , u2 ) = L(u2 , u1 )W (u1 , u2 ),
where L(u1 , u2 ) involves the generators of representation with spin and L(u2 , u1 ) involves the
generators of representation with spin 1 expressed in terms of the Heisenberg pair z, and
u1 = 2uc + , u2 = 2uc 1 ; u1 u2 = 2 + 1.
In all cases the operator W depends on the difference u1 u2 : W (u1 , u2 ) = W (u1 u2 )
a2
q 2 (q 2z+1a ; q 2 ) az
q
,
Wq (a) = a
z
(q 2z+2 ; q 2 )
1
(z + 1)
Wo (a) = a
;
z (z + 1 a)
e2iaz+ia (2z) [z a + 2k][z a]k [a]k (a2k)

We (a) =
.
e
(2z + 2a)
[z a][z + 1]k [k]!
k=0
Note that for future convenience we change slightly the normalization of the operators Wq
and We . The operator S1 is obtained from the operator W (u1 u2 ) by substitution z z1 and
the operator S3 is obtained from the operator W (v1 v2 ) by substitution z z2 .
5. Coxeter relations for the elementary permutation operators
In this section we shall prove that the obtained operators Si obey the Coxeter relations (3.6),
(3.7) and (3.8) for the permutation group S4 . The operators Si have simple dependence on parameters
S1 (u) = S1 (u1 u2 );
S2 (u) = S2 (u2 v1 );
S3 (u) = S3 (v1 v2 )
273
so that the defining equations can be represented in the more simple form
Si (a)Si (a) = 1;
S1 (a)S2 (a + b)S1 (b) = S2 (b)S1 (a + b)S2 (a),
S1 (a)S3 (b) = S3 (b)S1 (a);
(5.1)
S2 (a)S3 (a + b)S2 (b) = S3 (b)S2 (a + b)S3 (a).
(5.2)
There are two evident equalities. The operator S3 (a) differs from the operator S1 (a) only by
a change of variable z1 z2 and therefore the operators commute, S1 (a)S3 (b) = S3 (b)S1 (a).
The equality S2 (a)S2 (a) = 1 can be checked easily because the operator S2 reduces to multiplication on a given function.
5.1. The case of elliptic deformation
The transposition operators have the form
2
e2iaz1 +ia (2z1 )

S1 (a) =
Sk (a, z1 )e(a2k)1 ;
(2z1 + 2a)
k=0
[z1 a + 2k][z1 a]k [a]k

,
Sk (a, z1 ) =
[z1 a][z1 + 1]k [k]!
(z1 + z2 + 1 + a) (z1 z2 + 1 + a)
S2 (a) = e2iaz1
,
(z1 + z2 + 1 a) (z1 z2 + 1 a)
where we use the following notations for the elliptic numbers
[z] = 1 (2z);
[z]0 = [z],
[z]k = [z] [z + 1] [z + k 1],
k = 1, 2, . . . .
Let us calculate the product S1 (a)S1 (a). Multiplying two power series and using the formula (A.10) for the shifted -functions we obtain the power series of the general form with
the composite coefficients
S1 (a)S1 (a) =
SN e2N 1 ;
N=0
SN =
N

Sk (a, z1 )SNk (a, z1 + a 2k)
k=0
[z1 + 1]2k
.
[z1 a + 1]2k
We have to prove that S0 = 1 and SN = 0 for N = 1, 2, . . . . Using the formulae (A.12) for the
elliptic numbers we transform this expression to the canonical form
SN =
N
[z1 + 2N ][z1 ]N [1 a N ]N [z1 a + 2k][z1 a]k
[z1 ][z1 1 + a]N [N]N

[z1 a][k]!
k=0
[a]k
[z1 + N ]k
[N ]k
.
[z1 + 1]k [1 a N ]k [z1 a + 1 + N ]k
The key formula which allows to calculate the sum of this special type is the FrenkelTuraev
summation formula [13,14]
274

N

[A + 2k][A]k
[A][k]!
k=0
[C]k
[D]k
[B]k
[A + 1 B]k [A + 1 C]k [A + 1 D]k
[E]k
[N]k
[A + 1 E]k [A + 1 N]k
[A + 1]N
[A + 1 C B]N [A + 1 D B]N
=
[A + 1 B]N
[A + 1 C]N
[A + 1 D]N
[A + 1 D C]N
.
[A + 1 D C B]N
(5.3)
Here the coefficients are restricted by the condition B + C + D + E = 2A + N + 1. Our special

case corresponds to the substitutions
A = z1 a;
B = a;
C = z1 + N;
D=E=
A+1
1a
= z1 +
2
2
and we have
N

[z1 a + 2k][z1 a]k
k=0
[z1 a][k]!
[z1 + N ]k
[N ]k
[a]k
[z1 + 1]k [1 a N ]k [z1 a + 1 + N ]k
[ z21 +
[z1 a + 1]N
[1 N ]N
[z1 + 1]N
[1 a N ]N [ z21 +
1+a
2 ]N
1a
2 ]N
[ z21 +
[ z21 +
1a
2 A N ]N
1+a
2 A N ]N
We obtain the needed equality SN = 0 for N = 1, 2, . . . because the factor [1 N ]N = 0 for

N = 1, 2, . . . and the equality S0 = 1 can be easily checked.
Next we prove the second equality in (5.1) and the second equality in (5.2) can be proven in a
similar way. Using (A.10) the second product can be transformed to the power series of the form
S2 (b)S1 (a + b)S2 (a)

SN (a + b, z1 )
=P
N=0
2
[ z1 +z
2 +
2
[ z1 +z
2 +
1b
2 ]N
1b2a
]N
2
2
[ z1 z
2 +
2
[ z1 z
2 +
1b
2 ]N
e(a+b2N )1 ,
1b2a
]
N
2
where
P = ei(b
2 a 2 )
(z1 + z2 + 1 + 2a + b) (z1 z2 + 1 + 2a + b)
(2z1 )
.
(z1 + z2 + 1 b)
(z1 z2 + 1 b) (2z1 + 2a + 2b)
To calculate the first product we multiply two power series, use the formula (A.10) and obtain
the power series of the general form
S1 (a)S2 (a + b)S1 (b) = P
[z1 a b + 2N ] SN e(a+b2N )1
N=0
with the following coefficients

SN =
N

Sk (a, z1 )SNk (b, z1 + a 2k)
k=0
2
[ z1 +z
2 +
2
[ z1 +z
2 +
1+b
2 ]k
1b2a
]k
2
2
[ z1 z
2 +
2
[ z1 z
2 +
[z1 a b + 1]2k
[z1 a + 1]2k
1+b
2 ]k
.
1b2a
]k
2
275
Using the formulae (A.12) it is possible to transform this expression to the form
SN =
N
[z1 a b]N [1 + b N ]N [z1 a + 2k][z1 a]k
[a]k
[z1 a b]
[N ]N
[z1 a][k]!
[z1 + 1]k
2
[ z1 +z
2 +
1+b
2 ]k
1b2a
]k
2
k=0
1+b
2 ]k
1b2a
]k
2
2
[ z1 z
2 +
2
2
[ z1 z
[ z1 +z
2 +
2 +
[N ]k
.
[z1 a + 1 + N]k
[z1 a b + N ]k
[1 + b N ]k
Next we use the FrenkelTuraev formula (5.3) for

A = z1 a;
B = a;
C=
E = z1 a b + N
z1 + z2 1 + b
+
;
2
2
D=
z1 z2 1 + b
+
;
2
2
in the calculation of the sum and obtain desirable relation

SN = SN (a + b, z1 )
2
[ z1 +z
2 +
2
[ z1 +z
2 +
1b
2 ]N
1b2a
]N
2
2
[ z1 z
2 +
2
[ z1 z
2 +
1b
2 ]N
1b2a
]N
2
This proves that the series expansions for the operators in both sides of the equality (5.1) coincide.
5.2. The case of quantum deformation
The operators have the form
a2
q 2 (q 22a+2z1 1 ; q 2 ) az1 1
q
;
S1 (a) = a
z1
(q 2+2z1 1 ; q 2 )
S2 (a) = z1a
( zz21 q 1a ; q 2 )
( zz21 q 1+a ; q 2 )
(5.4)
Using the q-binomial formula (A.1) we represent the operator S1 (a) as the sum
a2
q2
Sk (a)q (a2k)z1 1 ;
S1 (a) = a
z1
k=0
Sk (a) =
q 2k (q 2a ; q 2 )k
.
(q 2 ; q 2 )k
(5.5)
We prove the equality S1 (a)S2 (a + b)S1 (b) = S2 (b)S1 (a + b)S2 (a) and the second defining
equation of this type can be proven in a similar way. Using (A.3) the second product can be
transformed to the power series of the form
S2 (b)S1 (a + b)S2 (a)
=
b2 a 2
2
( zz21 q 1b ; q 2 )
( zz12 q 1b ; q 2 )N
S
(a
+
b)
q (a+b2N )z1 1 .
N
( zz21 q 1+b+2a ; q 2 )
( zz12 q 1b2a ; q 2 )N
N=0
To calculate the first product we multiply two power series, use the formula (A.3) and obtain the
power series of the general form
S1 (a)S2 (a + b)S1 (b) =
b2 a 2
2
( zz21 q 1b ; q 2 )
SN q (a+b2N )z1 1
( zz21 q 1+b+2a ; q 2 )
N=0
276
with the following coefficients

SN = q 2N
N

Sk (a)SNk (b)
k=0
( zz12 q 1+b ; q 2 )k
( zz12 q 1b2a ; q 2 )k
q 2bk .
Using the formula (A.4) it is possible to transform such expression to the form
SN = q 2N
N
( zz12 q 1+b ; q 2 )k
(q 2b ; q 2 )N (q 2a ; q 2 )k
(q 2N ; q 2 )k
q 2k .
(q 2 ; q 2 )N
(q 2 ; q 2 )k ( zz12 q 1b2a ; q 2 )k (q 2(1+bN ) ; q 2 )k
k=0
The key formula which allows to calculate this sum is the Jackson summation formula [15]
N

; q)N
( C ; q)N ( C
(A; q)k (B; q)k
(q N ; q)k
AB
qk = A
B
.
C
(q; q)k (C; q)k ( C q 1N ; q)k
(C; q)N ( AB
; q)N
k=0
(5.6)
We use this formula for q q 2 and

A = q 2a ;
B=
z1 b+1
q ;
z2
C=
z1 1b2a
q
;
z2
AB 22N
= q 2+2b2N
q
C
in the calculation of the sum and obtain the needed relation

SN = q
2N
( zz12 q 1b ; q 2 )N
(q 2b ; q 2 )N
(q ab ; q 2 )N
z
(q 2 ; q 2 )N
( z12 q 1b2a ; q 2 )N (q 2b ; q 2 )N
= SN (a + b)
( zz12 q 1b ; q 2 )N
( zz12 q 1b2a ; q 2 )N
5.3. The case of no deformation
The operators have the form ((a)k = a(a + 1) (a + k 1))
1
(z1 1 + 1)
;
S1 (a) = a
z1 (z1 1 + 1 a)
S2 (a) = (z2 z1 )
= z2a
k

(a)k
z1
. (5.7)
k!
z2
k=0
We again prove the equality S1 (a)S2 (a + b)S1 (b) = S2 (b)S1 (a + b)S2 (a) only. Using the equality
(z1 1 + A) z1n = z1n (z1 1 + A + n),
the first product can be transformed to the power series of the form
Nab

(a b)N
(z1 1 + 1 b + N )
z1
S1 (a)S2 (a + b)S1 (b) =

N!
z2
(z1 1 + 1 a b + N )
N=0
(z1 1 + 1)
.
(z1 1 + 1 b)
(5.8)
277
To calculate the second product we multiply two power series, use the formula (5.8) and obtain
the power series of the general form with the composite coefficients
Nab

z1
SN ;
S2 (b)S1 (a + b)S2 (a) =
z2
N=0
SN =
N

k=0
(a)k (b)Nk
(z1 1 + 1 + k)
.
k!
(N k)! (z1 1 + 1 a b + k)
Using the formula

(b)Nk
(b)N
(N )k
=
(N k)!
N!
(1 + b N )k
(5.9)
it is possible to transform such expression to the form (d z1 1 )

SN =
N

(a)k
k=0
k!
(N )k
(b)N
(d + 1)k
(d + 1)
.
(d + 1 a b)k (1 + b N )k
N!
(d + 1 a b)
The key formula which allows to calculate the sum of this special type is the PfaffSaalschutz
summation formula [15]
N

(A)k (B)k
(N )k
(C A)N
(C B)N
k! (C)k (1 + A + B C N )k
(C)N
(C A B)N
(5.10)
k=0
We use this formula for

A = a;
B = d + 1;
C = d + 1 a b;
1+A+B C N =1bN
in the calculation of the sum and obtain the needed relation
(a b)N
(d + 1 b + N )
(d + 1)
SN =
.
N!
(d + 1 a b + N ) (d + 1 b)
6. The permutation group and the YangBaxter relation
Let us return to the permutation group. The transpositions s1 , s2 , s3 form the basis of the
permutation group of four parameters (u1 , u2 , v1 , v2 ). Any permutation can be decomposed into
the product of these operators sk . This representation is not unique because different decompositions can represent the same permutation. The equivalence of the permutation group S4
and the Coxeter group with the defining relations s12 = s22 = 1, s1 s3 = s3 s1 and s1 s2 s1 = s2 s2 s2 ,
s2 s3 s2 = s3 s2 s3 guarantees that two sequences representing the same permutation can be transformed each other by using the defining relations between generators only. In the general case
it is the existence theorem, but we shall follow the more constructive way and indicate the particular transformations we need each time. The operators S1 , S2 , S3 represent the corresponding
permutations of four parameters (u1 , u2 , v1 , v2 ) entering in the product of two Lax matrices
L1 (u1 , u2 )L2 (v1 , v2 ) and obey the same defining relations. This key property of the operators Si
shows that we deal with a representation of the permutation group.
278
Note that everything can be generalized to the product of an arbitrary number of Lax matrices.
We consider the product of the three Lax matrices L1 (u1 , u2 )L2 (v1 , v2 )L3 (w1 , w2 ) because this
example immediately leads to the YangBaxter equation. Let us join all six parameters in one set
u (u1 , u2 , v1 , v2 , w1 , w2 ) and consider the group of permutations S6 of six parameters. Now
we have five elementary transpositions si and corresponding operators Si (u)
S1
S3
S5

( u1 , u2 , v1 , v2 , w1 , w2 );
S2
S4

(u1 , u2 , v1 , v2 , w1 , w2 ),
which obey the following defining relations

S1 L1 (u1 , u2 ) = L1 (u2 , u1 )S1 ;
S3 L2 (v1 , v2 ) = L2 (v2 , v1 )S3 ;
S5 L3 (w1 , w2 ) = L2 (w2 , w1 )S5 ,

S2 (u)L1 (u1 , u2 )L2 (v1 , v2 ) = L1 (u1 , v1 )L2 (u2 , v2 )S2 (u),
S4 (u)L2 (v1 , v2 )L3 (w1 , w2 ) = L2 (v1 , w1 )L3 (v2 , w2 )S4 (u).
It is evident that all is effectively reduced to the case of the product of two Lax matrices.
The generators S1 , S2 and S3 are the same as in the case of the product of two Lax matrices L1 (u1 , u2 )L2 (v1 , v2 ). The generators S3 , S4 and S5 play the same role for the product
L2 (v1 , v2 )L3 (w1 , w2 ) and can be obtained from the S1 , S2 and S3 by the simple change of
variables and parameters
z1 , z2 z2 , z3 ;
(u1 , u2 , v1 , v2 ) (v1 , v2 , w1 , w2 ).
(6.1)
The defining relations

si si = 1 Si (si u)Si (u) = 1;
si sj = sj si Si (sj u)Sj (u) = Sj (si u)Si (u),
|i j | > 1,
si si+1 si = si+1 si si+1 Si (si+1 si u)Si+1 (si u)Si (u) = Si+1 (si si+1 u)Si (si+1 u)Si+1 (u)
are effectively reduced to the relations for the operators from the previous case also. The operator
12 (u v) is the solution of the equation
R
12 (u v)
12 (u v)L1 (u1 , u2 )L2 (v1 , v2 ) = L1 (v1 , v2 )L2 (u1 , u2 )R
R
and corresponds to the permutation s2 s1 s3 s2 :
12 (u v) = S2 (s1 s3 s2 u)S1 (s3 s2 u)S3 (s2 u)S2 (u).
R
(6.2)
23 (v w) is the solution of the equation

The operator R
23 (v w)
23 (v w)L2 (v1 , v2 )L3 (w1 , w2 ) = L2 (w1 , w2 )L3 (v1 , v2 )R
R
and corresponds to the permutation s4 s3 s5 s4 :
23 (v w) = S4 (s3 s5 s4 u)S3 (s5 s4 u)S5 (s4 u)S4 (u).
R
(6.3)
23 (v w) can be obtained by the same change of variables and parameters (6.1)

The operator R
12 (u v). There exist two equivalent decompositions of the special permufrom the operator R
tation
s(u1 , u2 , v1 , v2 , w1 , w2 ) = (w1 , w2 , v1 , v2 , u1 , u2 )
279
in terms of permutations s2 s1 s3 s2 and s4 s3 s5 s4 and corresponding operators are expressed in

terms of R-matrix
12 (v w)R
23 (u w)R
12 (u v),
s = s2 s1 s3 s2 s4 s3 s5 s4 s2 s1 s3 s2 R
23 (u v)R
12 (u w)R
23 (v w).
s = s4 s3 s5 s4 s2 s1 s3 s2 s4 s3 s5 s4 R
These operators correspond the same permutation and therefore there exists the transformation
from the one expression to the another.
12 (6.2) and R
23 (6.3) obey the YangBaxter relation
Proposition 5. The operators R
23 (u v)R
12 (u w)R
23 (v w) = R
12 (v w)R
23 (u w)R
12 (u v).
R
It is sufficient to give an example of the chain of transformations allowing to transform the
first decomposition of the permutation s to the second one
s2 s3 s1 s2 s4 s3 s5 s4 s2 s1 s3 s2
= s2 s3 s4 s1 s2 s3 s2 s1 s5 s4 s3 s2 = s2 s3 s4 s3 s1 s2 s1 s5 s3 s4 s3 s2
= s4 s3 s5 s2 s1 s4 s3 s4 s2 s5 s3 s4 = s4 s3 s5 s4 s2 s1 s3 s2 s4 s5 s3 s4 .
Repeating step by step this chain of transformations for the operators Si it is possible to transform
the operator in the right-hand side of YangBaxter relation to the operator in the left-hand side.
We omit the corresponding formulae for brevity.
(2)
(1) and R
7. The operators R
Recall the different levels of resolution of the YangBaxter operator into operators representing more elementary parameter permutations in acting on the product of Lax matrices as
discussed in Introduction. Besides of the operators Si representing the elementary transpositions
(k) , k = 1, 2, are important; they permute the parameters uk , vk in
the following operators R
acting on the product of Lax matrices.
Again everything can be generalized to the product of arbitrary number of Lax matrices.
We consider the product of the three Lax matrices L1 (u1 , u2 )L2 (v1 , v2 )L3 (w1 , w2 ) because this
(i) ,
example immediately leads to the analogon of the YangBaxter relations for the operators R
(i)
we effectively reduce the symmetry
i = 1, 2. Recall that working in terms of operators R
S6 S 3 S 3 .
Let us introduce the special permutations r1 = s2 s1 s2 , r3 = s4 s3 s4 and p3 = s2 s3 s2 , p5 =
s4 s5 s4 . The permutation r1 and r3 interchange only the parameters u1 v1 and v1 w1 correspondingly
r1 u = s2 s1 s2 u = (v1 , u2 , u1 , v2 , w1 , w2 );
r3 u = s4 s3 s4 u = (u1 , u2 , w1 , v2 , v1 , w2 )
and generate the group of permutations S3 of three parameters (u1 , v1 , w1 ). The corresponding
operators have the form
(1) (u) = S2 (s1 s2 u)S1 (s2 u)S2 (u);
R
12
(1) (u) = S4 (s3 s4 u)S3 (s4 u)S4 (u).

R
23
(7.1)
280
The permutation p3 and p5 interchange in a similar way u2 v2 and v2 w2 ,

p3 u = s2 s3 s2 u = (u1 , v2 , v1 , u2 , w1 , w2 );
p5 u = s4 s5 s4 u = (u1 , u2 , v1 , w2 , w1 , v2 )
and generate the group of permutations S3 of three parameters (u2 , v2 , w2 ). The corresponding
operators are
(2) (u) = S4 (s5 s4 u)S5 (s4 u)S4 (u).
R
23
(2) (u) = S2 (s3 s2 u)S3 (s2 u)S2 (u);

R
12
(7.2)
12 ,
Proposition 6. There are the following equivalent representations for the operator R
12 (u v) = R
(1) (pu)R
(2) (u) = R
(2) (ru)R
(1) (u)
R
12
12
12
12
= S2 (s1 s3 s2 u)S1 (s3 s2 u)S3 (s2 u)S2 (u).
The first and second expressions correspond to the decompositions of the permutation s on
the product of two commuting permutations r = s2 s1 s2 and p = s2 s3 s2 . The third expression
corresponds to the decomposition s = s2 s1 s3 s2 . The chain of the transformations is
rp = s2 s1 s2 s2 s3 s2 = s2 s1 s2 s2 s3 s2 = s2 s1 s3 s2 = s2 s3 s1 s2 = s2 s1 s2 s2 s3 s2 = pr
and there is the corresponding chain of transformations which proves the equivalence of all representations for the R-matrix
(1) (pu)R
(2) (u)
R
12
12
= S2 (s1 s2 s2 s3 s2 u)S1 (s2 s2 s3 s2 u)S2 (s2 s3 s2 u) S2 (s3 s2 u)S3 (s2 u)S2 (u)
= S2 (s1 s3 s2 u)S1 (s3 s2 u)S3 (s2 u)S2 (u) = S2 (s3 s1 s2 u)S3 (s1 s2 u)S1 (s2 u)S2 (u)
= S2 (s3 s2 s2 s1 s2 u)S3 (s2 s2 s1 s2 u)S2 (s2 s1 s2 u) S2 (s1 s2 u)S1 (s2 u)S2 (u)
(2) (ru)R
(1) (u).
=R
12
12
(2) (7.2) obey the relations
(1) (7.1) and R
Proposition 7. The operators R
(1) (u) = R
(1) (p3 u)R
(2) (u);
(2) (r1 u)R
R
12
12
12
12
(2) (r3 u)R

(1) (u) = R
(1) (p5 u)R
(2) (u),
R
23
23
23
23
(1) (u) = R
(1) (p5 u)R
(2) (u);
(2) (r1 u)R
R
23
12
12
23
(2) (r3 u)R

(1) (u) = R
(1) (p3 u)R
(2) (u),
R
12
23
23
12
(1) (r3 u)R

(1) (u) = R
(1) (r3 r1 u)R
(1) (r1 u)R
(1) (u),
(1) (r1 r3 u)R
R
23
12
23
12
23
12
(2) (p5 u)R
(2) (u) = R
(2) (p5 p3 u)R
(2) (p3 u)R
(2) (u).
(2) (p3 p5 u)R
R
23
12
23
12
23
12
These relations correspond to the relations for the group of permutations S3 S3 . The first
four are relations of commutativity for the transformations from two different groups S3 ,
p3 r1 = r1 p3 ;
p5 r3 = r3 p5 ;
p5 r1 = r1 p5 ;
p3 r3 = r3 p3 .
(i) mimics step by step the transforThe proof of the corresponding relations for the operators R
mations for the permutation group. We collect these chains of transformations in the same order
as they are listed above.
p3 r1 = s2 s3 s2 s2 s1 s2 = s2 s3 s1 s2 = s2 s1 s3 s2 = s2 s1 s2 s2 s3 s2 = r1 p3 ,
p5 r3 = s4 s5 s4 s4 s3 s4 = s4 s5 s3 s4 = s4 s3 s5 s4 = s4 s3 s4 s4 s5 s4 = r3 p5 ,
281
p5 r1 = s4 s5 s4 s2 s1 s2 = s2 s1 s2 s4 s5 s4 = r1 p5 ,
p3 r3 = s2 s3 s2 s4 s3 s4 = s3 s2 s3 s3 s4 s3 = s3 s2 s4 s3 = s3 s4 s2 s3 = s3 s4 s3 s3 s2 s3
= s4 s3 s4 s2 s3 s2 = r3 p3 .
The last two relations are the triple defining relations of each group S3 ,
r3 r1 r3 = r1 r3 r1 ;
p5 p 3 p 5 = p 3 p 5 p 3
and the chains of transformations have the form

r1 r3 r1 = s2 s1 s2 s4 s3 s4 s2 s1 s2 = s1 s2 s1 s4 s3 s4 s1 s2 s1 = s4 s1 s2 s3 s2 s1 s4
= s4 s1 s3 s2 s3 s1 s4 = s4 s3 s1 s2 s1 s3 s4 = s4 s3 s4 s2 s1 s2 s4 s3 s4 = r3 r1 r3 ,
p3 p5 p3 = s2 s3 s2 s4 s5 s4 s2 s3 s2 = s2 s3 s2 s5 s4 s5 s2 s3 s2 = s5 s2 s3 s4 s3 s2 s5
= s5 s2 s4 s3 s4 s2 s5 = s5 s4 s5 s2 s3 s2 s5 s4 s5 = s4 s5 s4 s2 s3 s2 s4 s5 s4 = r5 r3 r5 .
8. Discussion and summary
In the present analysis we did not specify to a series of group representations related to the
considered Lie algebra s2 . This would imply a restriction on the values of the representation
parameter , but this appears unnatural here because enters the relevant expressions in combination with the spectral parameter.
In the undeformed and q-deformed cases the representation can be realized as a module spanned by the monomials zm , m = 0, 1, . . . , not assuming a metric structure of a functional space. The constant function 1 represents the lowest weight vector. An invariant finitedimensional sub-module appears for positive integer values of 2. In the elliptic case the representation can be described in terms of entire even periodic functions and the finite-dimensional
representation appearing in the case of positive integer N = 2 is spanned by even products
of 2N -functions. These realizations are convenient for describing the generic representations
and their tensor products. The YangBaxter operator and its factors R(1) , R(2) are operating in
these classes of functions. However the operators of elementary permutations Si map to different
realizations involving e.g. functions with branch points.
Also the ambiguity of the solutions of difference equations by periodic factors is to be fixed
by specifying the function class.
We refer to the particular case of the principal series of SL(2, C) representations where all
operators in question can be defined as acting on functions on the complex plane [21] and the
mentioned difficulty does not appear. There one deals with actually two copies of s2 , one represented by operators acting holomorphically and the other anti-holomorphically on functions
defined on the complex plane. Correspondingly, the representations are labelled by the pair (, )
n1
n+1
taking values = 2 i, = 2 i, n integer and real. The structure of the operators

being products of a holomorphic and anti-holomorphic part ensures that they do not lead beyond
uniquely defined functions on the complex plane.
The simplest case of a symmetry algebra of the YangBaxter equation is s(2) and there are
two possible ways of generalization. First one can deform the algebraic structure but keep the
rank equal to one:
U (s2 ) Uq (s2 ) Sklyanin algebra.
282
In this paper we present the uniform expression for the solution of the YangBaxter equation
connected with all three cases. The R-operator acting in the tensor product of two representations
of the symmetry algebra with spins 1 and 2 can be constructed from the three basic operators
S1 , S2 , S3 . The operators S1 , S2 , S3 represent elementary permutations from the symmetric
group S4 permutation group of four parameters entering the RLL-relation.
The second way of generalization is to increase the rank going to the algebra s(n). The Weyl
group W of sn is the permutation group Sn generated by elements w1 , . . . , wn1 with defining
relations wi2 = 1, wi wi+1 wi = wi+1 wi wi+1 and wi wj = wj wi if |i j | > 1. The RLL-relation
has the same form (1.2)
12 (u v)L1 (u1 , u2 un )L2 (v1 , v2 vn )
R
12 (u v),
= L1 (v1 , v2 vn )L2 (u1 , u2 un )R
but now the Lax matrix depends on the n parameters u1 , u2 , . . . , un [17]. There are n 1 parameters which label the representation of the algebra s(n) (analog of spin ) and spectral
parameter u. There are 2n parameters in the RLL-relation and the permutation group is now S2n .
There are 2n 1 elementary transpositions in the group S2n
s1
sn1
sn+1
s2n1

u1 , u2 un1 , un , v1 , v2 vn1 , vn ;
sn

u1 , u2 un1 , un , v1 , v2 vn1 , vn .
The R-matrix again admits the decomposition in factors of operators representing the elementary
transpositions [17]. The operators, representing s1 , . . . , sn1 and sn+1 , . . . , s2n1 are the wellknown intertwining operators and are connected to the Weyl group of the algebra s(n). The
operator representing sn plays a special role and can be constructed explicitly [17].
It seems that the considered decomposition of the R-matrix is universal and gives some insight
into a possible structure of solutions of YangBaxter equation with general symmetry.
In terms of the elementary parameter permutation operators we construct the factor operators R(k) . The product of these factors over k = 1, . . . , n results in the YangBaxter R-operator
for s(n). Acting on the product of two Lax operators R(k) permutes their parameters uk , vk . For
a spin chain the products of these factor operators over the sites leads to a commuting family of
operators Qk which can be identified as the Baxter operators [16,18,19].
Acknowledgements
We thank A. Zabrodin for illuminating discussions and explanations. This work has been
supported by the RFFI grant 05-01-00922, partially by grant NSh-5403.2006.1 and by DFG
grant 436 Rus 17/9/06 (S.D.), by DFG grant 436 Arm 17/1/06, by Volkswagen Stiftung (D.K.)
and by NTZ of Leipzig University.
Appendix A
A.1. q-Special functions
The standard q-products are
(x; q) =
+

1 qk x ;
(x; q)n =
k=0
n1

k=0
283

(x; q)
1 qk x =
;
(xq n ; q)
q C, |q| < 1.
The q-binomial formula has the form
(Az; q) (A; q)k k

z .
=
(z; q)
(q; q)k
(A.1)
k=0
The function (x; q) obeys the recurrent relation (qx; q) = (1 x)1 (x; q) so that it is used to
solve the recurrent relation of the type
(qx) =
1 ax
(x);
1 bx
(x) =
(bx; q)
.
(ax; q)
(A.2)
There are useful formulae for the q-products which are used in the text
2
q
2
(yq 2n ; q 2 ) y n ( y ; q )n (y; q 2 )
=
,
(xq 2n ; q 2 ) x n ( q 2 ; q 2 )n (x; q 2 )
x
(A.3)
2b ; q 2 )
(q 2b ; q 2 )Nk
(q 2N ; q 2 )k
N
2k(b+1) (q
=
q
.
(q 2 ; q 2 )Nk
(q 2 ; q 2 )N
(q 2(1+bN ) ; q 2 )k
(A.4)
A.2. Elliptic special functions

The general -function with characteristics is [11]

2
a,b (z| ) =
ei(n+a) e2i(n+a)(z+b)
nZ
and we shall use the four standard functions

1 2
1
1
ei(n+ 2 ) e2i(n+ 2 )(z+ 2 ) ,
1 (z| ) = 1,1 (z| ) =
nZ
2 (z| ) = 1,0 (z| ) =
1 2
ei(n+ 2 )
e2i(n+ 2 )z ,
nZ
3 (z| ) = 0,0 (z| ) =
e2inz ,
e2i(n+ 2 )(z+ 2 ) .
ein
nZ
4 (z| ) = 0,1 (z| ) =
ein
nZ
The following identities are used to factorize the Lax matrix and for the derivation of the defining
equations for the operator S2 ,
21 (x + y)1 (x y) = 4 (x)3 (y) 4 (y)3 (x);
24 (x + y)4 (x y) = 4 (x)3 (y) + 4 (y)3 (x);
22 (x + y)2 (x y) = 3 (x)3 (y) 4 (y)4 (x);
23 (x + y)3 (x y) = 3 (x)3 (y) + 4 (y)4 (x),
284
where 3 (z) 3 (z| 2 ) and 4 (z) 4 (z| 2 ).

The elliptic gamma function [12] is defined by the double product

1 e2i( (n+1)+ (m+1)z)
.
1 e2i( n+ m+z)
(z|,
)
(A.5)
n,m=0
We need the following properties of this function

(z + |,
) = R(
) eiz 1 (z|
) (z|,
),
(z + |, ) = R( ) e
iz
(A.6)
1 (z| ) (z|, ),
(A.7)
i
where the constant R( ) does not depend on z: R( ) = ie 4 (1; e2i )1 .

We introduce the function (z) (z|, 2) which obeys the recurrence relation
(z + 2) = R( ) eiz (z) (z);
(z) 1 (z| ),
(A.8)
and can be used to solve the more general recurrence relation

(z + 2) =
(z + a)
(z);
(z + b)
(z) = ei(ba)z
(z + a)
,
(z + b)
(A.9)
needed to obtain the operator S2 . There is the useful formula for the shifted -functions
[ y2 + 1]m
(x 2m) (y)
,
= eim(xy)
(x) (y 2m)
[ x2 + 1]m
(A.10)
where we use the following notations for the elliptic numbers

[z] = 1 (2z);
[z]0 = [z],
[z]k = [z] [z + 1] [z + k 1],
k = 1, 2, . . . .
(A.11)
To transform the series with elliptic numbers one uses the transformation formulae
[A + 1]2k [A + 2k]Nk
[A]N [A + N]k
=
;
[A + 2k]
[A]
[b]Nk
[1 + b N ]N [N]k
.
=
[N k]!
[N ]N [1 + b N ]k
[B]2k [B + 2k]Nk = [B]N [B + N ]k ;

(A.12)
References
[1] R. Baxter, Partition function of the eight-vertex model, Ann. Phys. (N.Y.) 70 (1972) 193228;
R. Baxter, Eight-vertex model in lattice statistics and one-dimensional anisotropic Heisenberg spin chain I, Ann.
Phys. (N.Y.) 76 (1973) 124;
R. Baxter, Eight-vertex model in lattice statistics and one-dimensional anisotropic Heisenberg spin chain II, Ann.
Phys. (N.Y.) 76 (1973) 2547;
R. Baxter, Eight-vertex model in lattice statistics and one-dimensional anisotropic Heisenberg spin chain III, Ann.
Phys. (N.Y.) 76 (1973) 4871.
[2] L.D. Faddeev, L.A. Takhtadzhan, The quantum method of the inverse problem and the Heisenberg XYZ model,
Russian Math. Surveys 34 (1979) 1168 (English translation).
[3] P.P. Kulish, E.K. Sklyanin, On the solutions of the YangBaxter equation, Zap. Nauchn. Sem. LOMI 95 (1980) 129;
P.P. Kulish, E.K. Sklyanin, Quantum spectral transform method. Recent developments, Lect. Notes Phys. 151 (1982)
61.
285
[4] M. Jimbo, Introduction to the YangBaxter equation, Int. J. Mod. Phys. A 4 (1983) 3759;
M. Jimbo, YangBaxter equation in integrable systems, in: M. Jimbo (Ed.), Adv. Ser. Math. Phys., vol. 10, World
Scientific, Singapore, 1990.
[5] L.D. Faddeev, How Algebraic Bethe Ansatz Works for Integrable Model, Les-Houches Lectures, 1995, hep-th/
9605187.
[6] P.P. Kulish, N.Yu. Reshetikhin, E.K. Sklyanin, YangBaxter equation and representation theory, Lett. Math. Phys. 5
(1981) 393403.
[7] E.K. Sklyanin, On some algebraic structures related to YangBaxter equation, Funkts. Anal. Prilozh. 16 (1982)
2734;
E.K. Sklyanin, On some algebraic structures related to YangBaxter equation: Representations of the quantum
algebra, Funkts. Anal. Prilozh. 17 (1983) 3448.
[8] A.V. Belitsky, V.M. Braun, A.S. Gorsky, G.P. Korchemsky, Integrability in QCD and beyond, Int. J. Mod. Phys.
A 19 (2004) 4715, hep-th/0407232.
[9] I. Krichever, A. Zabrodin, Vacuum curves of elliptic L operators and Representations of Sklyanin algebra, solvint/9801022.
[10] A. Zabrodin, Commuting difference operators with elliptic coefficients from Baxters vacuum vectors, J. Phys. A:
Math. Gen. 33 (2000) 3825.
[11] D. Mumford, Tata Lectures on Theta I, Progress in Mathematics, Birkhuser, 1983.
[12] G. Felder, A. Varchenko, The elliptic gamma function and SL(3, Z) Z 3 , math.QA/9907061.
[13] I. Frenkel, V. Turaev, Elliptic Solutions of the YangBaxter Equation and Modular Hypergeometric Functions, The
Arnold-Gelfand Mathematical Seminars, Birkhauser Boston, Cambridge, MA, 1997, pp. 171204.
[14] H. Rosengren, Sklyanin invariant integration, math.QA/0405072.
[15] G. Gasper, Elementary derivation of summation and transformation formulas for q-series, math.CA/9605230.
[16] S. Derkachov, Factorization of the R-matrix. I., Zap. Nauchn. Sem. POMI 335 (2006) 134163, math.QA/0503396;
S. Derkachov, Factorization of R-matrix and Baxters Q-operator, math.QA/0507252.
[17] S. Derkachov, A. Manashov, R-matrix and Baxter Q-operators for the noncompact SL(N, C) invariant spin chain,
SIGMA 2 (2006) 084, nlin.SI/0612003.
[18] S. Derkachov, A. Manashov, Baxter operators for the quantum sl(3) invariant spin chain, J. Phys. A 39 (2006)
13171, nlin.SI/0604018.
[19] S. Derkachov, D. Karakhanyan, R. Kirschner, Baxter Q-operators of the XXZ chain and R-matrix factorization,
Nucl. Phys. B 738 (2006) 368390, hep-th/0511024.
[20] D. Karakhanyan, R. Kirschner, M. Mirumyan, Universal R operator with deformed conformal symmetry, Nucl.
Phys. B 636 (2002) 529, nlin.SI/0111032.
[21] S.E. Derkachov, G.P. Korchemsky, A.N. Manashov, Noncompact Heisenberg spin magnets from high-energy QCD.
I: Baxter Q-operator and separation of variables, Nucl. Phys. B 617 (2001) 375, hep-th/0107193.
Slave bosons in radial gauge: A bridge between

path integral and Hamiltonian language
Raymond Frsard a, , Henni Ouerdane a,1 , Thilo Kopp b
a Laboratoire CRISMAT, UMR CNRS-ENSICAEN 6508, 6 Boulevard Marchal Juin, 14050 Caen Cedex, France
b EP VI, Center for Electronic Correlations and Magnetism, Universitt Augsburg, D-86135 Augsburg, Germany
Received 26 January 2007; accepted 25 May 2007

Abstract
We establish a correspondence between the resummation of world lines and the diagonalization of the
Hamiltonian for a strongly correlated electronic system. For this purpose, we analyze the functional integrals for the partition function and the correlation functions invoking a slave boson representation in the
radial gauge. We show in the spinless case that the Greens function of the physical electron and the projected Greens function of the pseudofermion coincide. Correlation and Greens functions in the spinful case
involve a complex entanglement of the world lines which, however, can be obtained through a strikingly
simple extension of the spinless scheme. As a toy model we investigate the two-site cluster of the single
impurity Anderson model which yields analytical results. All expectation values and dynamical correlation
functions are obtained from the exact calculation of the relevant functional integrals. The hole density, the
hole auto-correlation function and the Greens function are computed, and a comparison between spinless
and spin 1/2 systems provides insight into the role of the radial slave boson field. In particular, the exact expectation value of the radial slave boson field is finite in both cases, and it is not related to a Bose
condensate.
PACS: 71.27.+a; 11.10.-z; 11.15.Tk
Keywords: Strongly correlated electrons; Field theories; Non-perturbative methods
E-mail address: raymond.fresard@ensicaen.fr (R. Frsard).

1 Present address: LASMEA, UMR CNRS-Universit Blaise Pascal 6602, 24 avenue des Landais, 63177 Aubire
Cedex, France.
doi:10.1016/j.nuclphysb.2007.05.025
R. Frsard et al. / Nuclear Physics B 785 [FS] (2007) 286306
287
1. Introduction
In contemporary solid state research, strongly correlated electrons comprise the most fascinating albeit intangible physical systems. They cover a wide range of phenomena, including
high temperature superconductivity, colossal magnetoresistance, aspects of the fractional quantum Hall effect, and even electronic reconstruction in oxide electronic devices which are built on
interfaces of strongly correlated films. Whereas their importance is generally perceived, a fundamental comprehension is still not achieved, especially for high-temperature superconductivity.
This unfortunate absence of a well established theoretical scheme or even solution is not
surprising: strong electronic correlations are based on (sufficiently strong) local interactions in
real space but the Fermi surface, the concept on which the physics of metals is firmly rooted, is
defined and understood in momentum space. Correspondingly, a theoretical investigation is either
built on momentum or real space approaches which allow to treat either the kinetic (band) term or
the interaction accurately. However, a momentum space weak coupling approach is insufficient
to generate the desired new energy scales and fix points whereas standard perturbation theory
from the highly degenerate local (atomic) limit suffers from severe drawbacks [1].
Nevertheless, many effective strong coupling theories expand, in a generalized sense, with
respect to local models. A model with a single local interaction term is the Anderson impurity
model. It is the prominent strong-coupling many-body model which can still be solved exactly
(with certain restrictions) and which has been understood in basically all aspects (see, for example [26]). It may justly be seen as the paradigm of a strongly correlated many-body system.
A successful scheme to investigate lattice models with on-site interactions originates from a
self-consistent extension of the Anderson model. The self-consistency is generated through a
dynamical mean-field theory (DMFT) which singles out a site with strong local interaction; this
site couples to an electronic bath, the effective medium, the local density of states of which
is calculated self-consistently [7,8]. Actually, the DMFT is exact for infinite space dimensions,
a limit which was introduced in Ref. [9] for correlated electron systems. However, it is missing
the spatial correlations. In recent years it has been devised to treat clusters which can couple to
various bath systems in order to investigate correlations with a spatial extension of the cluster
size [1014].
In our theoretical study we will focus on a different approach, the slave boson technique [15
17]. The formalism entails a local decomposition of electronic excitations into charge and spin
components. Electron creation and annihilation operators are thereby represented by composite
operators which separate into canonical operators with bosonic and fermionic character. However
these operators are enslaved in the sense that their respective number operators have to fulfill a
local constraint. The original idea was to decouple spin and charge degrees of freedom; other,
modified schemes attribute to each type of excitation a bosonic mode which allows to study
the correlated system in a saddle point approximation [1719]. This mean field approach has
been successful when set against numerical simulations: ground state energies [20] and charge
structure factors show excellent agreement [21], as the procedure is exact in the large degeneracy
limit [18,22].
While the saddle point approximation allows to calculate translationally invariant expectation
values in momentum space, the corresponding mean field solution is not a priori legitimate.
The objection is concerned with the local decomposition of the electron field into fermion and
slave boson components. This implies that the model acquires a local gauge invariance with the
consequence that Elitzurs theorem [23] prevents the (slave) bosonic fields to acquire a non-zero
288
expectation value. In fact, it is the phase fluctuations of the boson field which suppress the finite
value or condensation of these fields.
One alternative to avoid such a condensation has been devised by Kroha, Wlfle, and coworkers [24,25]. In their approach the local gauge invariance is guaranteed through Ward identities in
a conserving approximation. The projection onto the physical sector of the Fock space is achieved
with an Abrikosov procedure by sending the Lagrange multiplier of the constraint to infinity.2
The other alternative is to use a radial decomposition of the bosonic field [27], the details of
which were presented in a previous paper by two of the authors [28]. In the limit of large on-site
interaction, the bosonic fields in radial representation reduce to their respective (real) amplitude
as the time derivatives of the conjugate phase can be absorbed in a time-dependent Lagrange
multiplier field.
In this article we provide a scheme for the solution of cluster models in radial slave-boson
representation. We present in sufficient detail the calculation of correlation and Greens functions
for a two-site cluster of the single impurity Anderson model, in order to exemplify our scheme.
Although the model can be diagonalized without slave boson technique we esteem the explicit
solution in the radial decomposition of considerable significance. First, it relates the world line
expansion of slave boson path integrals to the quantum states in the Fock space, in particular
for entangled states. This is achieved through a decomposition of the fermionic determinant into
resolvents at each time step. Second, it allows to compare these exact results (e.g., for the slave
boson amplitude) to saddle point evaluations and to assess their validity [29].
The article is organized as follows: in Section 2, we introduce the functional integral formulation of the two-site cluster model. We give expressions for the action, partition function and hole
density as well as for the hole auto-correlation function in terms of radial slave bosons. The spinless system is studied first in Section 3 where, through the derivation of the partition function, we
show how to proceed from a world line representation to the representation with quantum states
in the Fock space. In Section 4, we show how our formalism allows to derive results for the
spinful case from a straightforward extension of the spinless case. The Greens function necessitates a distinct derivation of the fundamental connection between the slave-boson path integral
representation and the Hamiltonian scheme which is the object of Section 5.
2. Functional integral formulation of the two-site cluster model
2.1. Hamiltonian and radial slave boson representation
The single impurity Anderson model (SIAM) has been investigated with a variety of techniques and for many different purposes. One of them consists of testing a new approach, in
particular against exact results. Here we adopt a similar spirit in order to link the evaluation of
the path integral representation of thermodynamic and dynamical quantities to their computation
through a straightforward diagonalization of the Hamiltonian. For the SIAM it reads:

ck, d + h.c. + U
ck, (tk + c )ck, +
d d d + V
d d ,
H=
(1)
k,
k,
=,
where U is the on-site repulsion, which is hereafter taken as infinite. The operators ck,
(ck, )
and d (d ) describe the creation (annihilation) of the band electrons and impurity electrons,
2 Note that a similar procedure can be set up without introducing slave bosons [26].
289
respectively, with spin ; the kinetic energy in the band is denoted tk , and c and d are the band
and impurity energy levels, respectively. The hybridization is given by V .
The link between the two schemes is actually a complex procedure when the impurity is
coupled to an infinite bath. In order to lower this complexity we reduce the bath to a single
site, in which case the diagonalization of the Hamiltonian is easy and all relevant results can
be obtained analytically. Nevertheless, the problem is non-trivial when handled in the functional
integral formalism. The level of difficulty depends on the functional representation which is used.
For our purpose, a promising one is that of the slave boson (SB) representation in the radial gauge
[28]. It is based on the original representation by Barnes [15] and is augmented in that respect
that the underlying U (1) gauge symmetry, originally discussed by Read and Newns [27], is fully
implemented, as the phase of the bosonic field is integrated out from the outset. Accordingly, the
original field d is represented in terms of a real and a Grassmann field for each spin:
dn, = xn+1 fn, ,
dn,
(2)
= xn fn,
,
(3)
where xn and xn+1 are the slave boson field amplitudes at time steps n and n + 1, and fn, is the
auxiliary fermion field. The shift of one time step for x in the relation for dn, is necessary to
obtain a non-zero value of the Grassmann integration for the Greens functions d ( )d (0)
as clearly shown for its calculation in the atomic limit in Ref. [28]. More precisely, the path
integral is zero if xn+1 is replaced by xn in Eq. (2). Moreover, Eqs. (2) and (3) are required in
order to properly represent the hybridization term in the action as given below. Further detail on
this matter can be found in Ref. [28].
2.2. Action and partition function
Following Ref. [28], the path integral representation of the partition function of the two-site
cluster is given by:
N

d
n
Z = lim
(4)
D cn, , cn,
D fn, , fn,
dxn eS
N
2
n=1
where the action S may be written as the sum of a fermionic part, Sf , and a bosonic part, Sb ,
with
Sf =
Sb =
Sf, =
N

cn,
(cn, Lc cn1, ) + fn,
(fn, Ln fn1, )
n=1

+ V xn cn,
fn1, + fn,
cn1, ,
N

in (xn 1) + W xn (xn 1) ,
(5)
n=1
where Lc = e(c ) , Ln = e(d +in ) Ld ein , n is the time-dependent real constraint

field, n denotes the time steps, and /N , with = 1/kB T and N the number of time steps.
Here, Sf (Sf, ) is bilinear in the fermionic fields, and the corresponding matrix of the coefficients
will be denoted as [S] ([S0 ]). The positive real number W is sent to infinity at the end of the
calculation which guarantees the projection onto the physical subspace. The above treatment of
290
the bosonic field is specific to radial slave bosons: no phase variable appears, and the above form
cannot be obtained by transformations of the conventional functional integral in the Cartesian
gauge without further assumptions (see Ref. [28]).
Inserting Eq. (5) into Eq. (4) yields a particularly suggestive expression of the partition function Z in the functional integral formulation, of the kind:
Z = lim P1 PN det[S],
(6)
N
W
where det[S] is the determinant of the fermionic matrix defined by Eq. (5); Pn is defined as:
+
+
dn
Pn =
dxn e[in (xn 1)+W xn (xn 1)] ,

2
(7)
and acts as a projector from the enlarged Fock space spanned by the auxiliary fermionic fields
down to the physical one. Explicitly, the action of these projectors on the various contributions
resulting from det[S] are found to be:
Pn 1 = 1,
(8)
Pn xn = 1,
(9)
Pn Ln = Ld ,
(10)
Pn Ln xn = 0,
(11)
Pn L2n = 0,
Pn xn2 = 1.
(12)
(13)
As will be seen below no further property of Pn will be needed for our purpose. We note
that there is some freedom in writing the projectors Pn , and alternative expressions exist [28].
However the properties Eqs. (8)(13) are independent of the particular form of Pn .
3. Application to the spinless fermion case
We first consider a spinless fermion system for simplicity. Even though this is a noninteracting problem, the level of complexity of its path integral representation following from
Eqs. (4)(6), is equivalent to the one of a fully interacting problem. The matrix representation of
the action Sf, of such system is a 2N 2N square matrix whose explicit expression in the basis
{cn, , fn, } reads:
1
[L ]
2
[L2 ]
[S0 ] =
12
..
.
..
(14)
12
[LN ]
where 12 is the 2 2 identity matrix and [Ln ] are 2 2 blocks given by:

Lc V xn
,
[Ln ] =
(15)
V xn Ln
at time step n, 1 n N . Note that the matrix [S0 ] as defined in Eq. (14) has the same structure
as the action matrix S () in Chapter 2 of Ref. [30].
291
3.1. Partition function

The partition function Z0 of the spinless fermion system has a form similar to that of Eq. (6)
except that det[S] is replaced by det[S0 ]. Its calculation is straightforward since we only have to
evaluate:
Z0 = lim P1 PN
N
W
sgn(P )
{P }
2N

SP (n),n
(16)
n=1
where sgn(P ) is the signum function of permutations P in the permutation group S2N and Si,j
are the matrix elements of [S0 ]. Since the time step n is only involved in Si,2n3 and Si,2n2 we
may recast Eq. (16) into:

Z0 = lim
sgn(P )P2 (SP (1),1 SP (2),2 )
N
W {P }
PN (SP (2N 3),2N3 SP (2N2),2N2 )P1 (SP (2N1),2N1 SP (2N ),2N ).
(17)
At this point it is straightforward to verify that performing the projections only implies to
make use of Eqs. (8)(10). We are left with:
Z0 = lim

{P }
sgn(P )
2N

SP
(n),n = lim det S0
,
N
n=1
are the elements of the 2N 2N matrix [S

] defined as:
where Si,j
0
[L]
12
[L] 1
2

.
S0 =
..
..
.
.
[L]
(18)
(19)
12
In Eq. (19) the 2 2 matrix blocks [L] are similar to the blocks [Ln ] except that Ln becomes
Ld , and xn is replaced by 1:

Lc V
.
L =
(20)
V Ld
In the form of Eq. (18) it is now obvious that Z0 can be readily obtained. We notice that
Eq. (19) is the expected action matrix for this free fermionic problem. Besides, it is straightforward to extend the above calculation to the case of an arbitrary bath. Unfortunately the calculation
becomes considerably more involved in the spin 1/2 case, which leads us to develop another
strategy to that purpose. We first present it in the spinless case, before extending it to the spinful
case.
3.1.1. Generation and resummation of the world lines
Part of the difficulty in computing Z0 is that the time steps are mixed in the fermionic determinant, in contrast to the bosonic part of the action represented by the projectors Pn , Eq. (6).
Therefore, transforming this determinant into a form where the time steps are decoupled, is
desirable. Achieving this amounts to handle all the world lines following from the action in
292
Eq. (5) which represents a problem equivalent to a particle in a time-dependent field with a timedependent hopping amplitude.
In order to generate the dynamics of the world lines we first expand det[S0 ] along the first two
columns. We obtain:
2
2
2
2
2
+ Lc M3,2
+ V x2 M4,2
V x2 M1,3
L2 M1,4
det[S0 ] = 1 M1,2

2
+ Lc L2 (V x2 )2 M3,4
.
(21)
Here the notation is as follows: we construct a matrix similar to [S0 ], but we only include time
n is a minor of this matrix, where both the ith and j th rows, together with
steps 1 and m > n. Mi,j
the first and second columns, are removed.
At this stage we may proceed with the generation of the world lines. To that aim we need
to express the minors M 2 as linear combinations of the minors M 3 which in turn can also be
expressed as similar linear combinations of the minors M 4 , and so forth up to the time step N .
The recurrence relation that we have established takes the following form:
n+1
M1,2
M n
1,2
n+1
M n (n+1),1
3,2
]
[K
3,2
M n+1
M n
[K (n+1),2 ]
4,2
4,2
,
(22)
n =
(n+1),3
M
M n+1
]
[K
1,3
1,3
n
[K (n+1),4 ]
M1,4
M n+1
1,4
n
M3,4
n+1
M3,4
where the four matrix blocks
(n),1
= (1),
K
(23)

(n),2 (n),3
Lc
V xn
= K
=
,
K
(24)
V xn
Ln
(n),4
= (Lc Ln ),
K
(25)
at time step n, 1 n N , define the 66 block diagonal matrix [K (n) ]. 3 The matrices [K (n) ]
describe the evolution of the two-site system along the world lines at each time step n. Iterating
this procedure up to time step N yields the determinant det[S0 ] in the following scalar product
form:
1
Lc
N

(n)
V x1
K
det[S0 ] = ( 1, Lc , V x2 , V x2 , L2 , Lc L2 )
(26)
,
V x1
n=3
L1
Lc L1
where the row vector is identified from Eq. (21) and the column vector has been obtained from
the last time step of the iteration process. Since the minus signs in Eq. (26) cancel, they can be
discarded for further considerations.
3 In Eq. (25) a term of order 2 , that vanishes in the limit N , is neglected.
293
Fig. 1. Typical world lines of a two-site system. Thick (thin) lines denote the elementary processes with the electron
sitting on the impurity (band) site. Full (empty) circles denote the elementary processes with a hole sitting on the
band (impurity) site. Dashed lines represent hopping processes. The horizontal axis runs with the time steps, while in
the vertical direction different sites are displayed.
The above expression corresponds to the full resummation of the world lines, some of them
being represented in Fig. 1. The first contribution, labeled by 1, corresponds to the subspace
with zero electron while the last one, labeled by Lc Ld , corresponds to the subspace with two
electrons. In both cases the world lines are straight: namely no hopping process takes place,
and the system is right away in an eigenstate. In Eq. (26) they correspond to the terms involving
[K (n),1 ] and [K (n),4 ], respectively. The structure of the world lines in the one-electron subspace
is more intricate.
In order to gain an intuitive picture of these world lines let us first consider the trivial functional integral representation of an interactionless electron where the f -field directly represents
the physical electron. We begin with the processes where the electron is on the band site at
(0)
time step one. If it stays there during all time steps, the resulting contributions Zc to Z0
N
(0)
will be given by Zc = n=1 Lc , namely there is one factor Lc per time step and the world
line is straight. If, on the contrary, the electron hops onto the impurity at time step m, and
(2)
back to the band at time step m
, the corresponding contribution Zmm
,c to Z0 results in

m
1
N
(2)
Zmm
,c = ( m1
n=1 Lc )V ( n=m+1 Ld )V ( n=m
+1 Lc ). Higher order processes in V follow accordingly. Complementary processes are those where the electron resides on the impurity at time
step one. The contributions of all these processes to Z0 will be weighted by both the number of
hopping processes and the difference in energy between the two levels. Assuming d < c results
in Lc < Ld , and for world lines involving the same number of hopping process, the world line
containing the largest number of factors Ld yields the largest contribution.
If we now return to our representation the contributions of the world lines to the partition function follow in a similar fashion, except for that (i) the factor corresponding to hopping process
at time step m is given by V xm , and (ii) the particle line at time step n corresponding to the
electron sitting on the impurity site results in a factor Ln . Accordingly the contribution of the
294
world line labeled by Lc in Fig. 1 to det[S0 ] is:

m
1
N
m1

Lc V xm
Ln V xm
Lc ,
n=1
while the world line labeled by Ld in Fig. 1 yields

m
1
N
m1

Ln V xm
Lc V xm
Ln .
n=1
(27)
n=m
+1
n=m+1
(28)
n=m
+1
n=m+1
3.1.2. Connecting with the Hamiltonian language

We observe that the structure of Eq. (26) is not manifestly translationally invariant in time. For
this reason, we proceed to bring Eq. (26) into a form where no particular time step is singled out.
We note that the (row) column vector in Eq. (26) can be identified with the (rows) columns of
the matrix blocks [K (n) ]disregarding the cancelling minus signs. Accordingly, we can rewrite
Eq. (26) as:

(2),1
(1),1
),1
K12 K(3),1
K(N
KN 1
det[S0 ] =
2 3
N1 N
{2 ,...,N }
{2 ,...,N }
{2 ,...,N }
{2 ,...,N }
(2),2 (3),2
),2
K1
K2 3 K(N
K(1),2
N1 N
N1
2
(2),3
(1),3
(2),4
(1),4
),3
K22 K(3),3
K(N
KN 2
2 3
N1 N
),4
K12 K(3),4
K(N
KN 1 ,
2 3
N1 N
(29)
since the matrices [K (n) ] are block diagonal and symmetric. Observe that the first and last lower
index in line three of Eq. (29) is not 1 but 2.
In Eq. (29), the first sum is equal to 1; the second and third sums are the diagonal elements
Lc V xn
N
N
of the matrix product N
n=1 V xn Ln , respectively. The last sum is equal to Lc
n=1 Ln .
Therefore Eq. (29) reduces to the trace of a 4 4 matrix that is the product of the N block
diagonal matrices [Kn ] whose elements are [K (n),1 ], [K (n),2 ] and [K (n),4 ], respectively:
det[S0 ] = Tr
N

[Kn ],
(30)
n=1
where the matrix [Kn ] is given by:
1
Lc
V xn
[Kn ] =
V xn
Ln
(31)
Lc Ln
Here, it is of interest to note that we have established a direct link between the world line
picture embodied in the 6 6 matrices [K (n) ] and the simpler description in terms of quantum
states in the Fock space through the 4 4 matrices [Kn ]. Indeed, when performing the world
line expansion using Eq. (26), the propagation matrices [K (n) ] acquire an involved structure
295
following from the initial conditions attached to the world lines. This is best seen in the oneelectron sub-space: at time-step one, the electron may either be on site c or on site d (see Fig. 1),
and the corresponding dynamics is governed by the matrices [K (n),2 ] and [K (n),3 ]. Summing up
all these world lines yields the one-electron contribution to det[S0 ]. Thus, one needs to handle
two 22 matrices, one in each case. In contrast, a single 2 2 matrix needs to be treated on the
level of Eq. (30).
In the above form of the fermionic determinant, Eq. (30), the time steps are decoupled, which
greatly simplifies the projection onto the physical Hilbert space. Indeed, in terms of the matrix [q]
given by:
1

Lc V
[q] Pn [Kn ] =
(32)
,
V Ld
Lc Ld
we obtain the partition function Z0 as:
Z0 = lim Tr
N
N

n=1
[q] = lim Tr 14
N
c
V
V
d
c + d 2
(33)
Namely, we recover here the Hamiltonian matrix in the Fock space. Therefore the expansion
of det[S0 ] in minors together with the recurrence relations, Eq. (22), allows for a correspondence
between the ensemble of the world lines and the Hamiltonian matrix. It is apparent that the
complexity of this interrelation depends on whether or not the system is in an eigenstate at time
step one. If this is the case, there is one single straight world line, and the connection is obvious.
Otherwise there is a proliferation of world lines, here controlled by the matrices [K (n),2 ] and
[K (n),3 ], which recombine to yield the contribution resulting from the 2 2 block of the matrix
[Kn ]. A result equivalent to Eq. (33) was already obtained by Barnes [31], though in a totally
different fashion.
3.2. Hole density
In our path integral formalism, the expectation value of the amplitude of the slave boson field
at time step m is xm . It simply represents the hole density 1 nd which can be written as xm =
bm1 where b represents the original Barnes slave boson, and xm is finite. In contrast the
bm
= bm = 0, for each time step because

expectation values of the boson operators are zero: bm
of the fluctuations of their respective phase factor, in agreement with Elitzurs theorem. Note that
a = x = 0 for any
expectation values of higher order moments of xm are also non-zero: xm
m
real positive parameter a.
In the case of spinless fermions we calculate xm as:

Z0 xm = lim P1 PN det[S0 ]xm
N
W
= lim P1 PN xm Tr
N
W
N

n=1
[Kn ] .
(34)
296
If we introduce the 44 matrix [QX ] Pn (xn [Kn ]) for all n, and [Q] the 44 diagonal
matrix that satisfies [q] = [UQ ][Q][UQ ] , [UQ ] being the eigenvector matrix, Eq. (34) becomes:

Z0 xm = lim Tr [Q]N1 [UQ ] [QX ][UQ ] .
(35)
N
Using Eqs. (8)(10), and the 0 limit, we obtain that the matrix [QX ] reduces to the
representation of the hole density operator on the impurity in the Fock space:
[QX ]i,j = i,1 j,1 + i,2 j,2 .
(36)
Correspondingly, Eq. (35) expresses xm as the averaged value of the hole density operator on
the impurity, represented by its matrix elements in the basis of the eigenstates of the Hamiltonian
([UQ ] [QX ][UQ ]) and weighted by the Boltzmann factors Z10 [Q]N1 . Therefore, in contrast to
a Bose condensate, xm is generically finite and may only vanish for zero hole concentration.
Its numerical evaluation will be presented in a forthcoming paper [29].
3.3. Densitydensity correlation function
To obtain further insight into the approach, it is of interest to compute dynamical correlation
functions. Probably, the simplest one is provided by the hole density auto-correlation function on
the impurity site. When expressed in terms of eigenstates of the Hamiltonian, it takes the form:
2

1 (mN )E mE

|1 n |
,
e
e
1 nd (m) 1 nd () =
(37)
d
Z0
where the eigenvalues E can be obtained from the diagonalization of the Hamiltonian. The
latter can be read from, e.g., Eq. (33).
In our path integral representation, the auto-correlation function x1 xm may be written as:

Z0 x1 xm = lim P1 PN det[S0 ]x1 xm
(38)
N
W
= lim P1 PN x1 xm Tr
N
W
N

[Kn ] ,
n=1
which reduces to

Z0 x1 xm = lim Tr [QX ][q]m2 [QX ][q]Nm ,
N
(39)
after application of the projectors, Eqs. (8)(13). If we now introduce the eigenvalues and eigenvectors of the Hamiltonian, Eq. (39) can be recast into:

Z0 x1 xm = lim Tr [Q]Nm [UQ ] [QX ][UQ ][Q]m2 [UQ ] [QX ][UQ ] .
(40)
N
In this form we recognize the standard expression in Eq. (37). Indeed, the combination
[UQ ] [QX ][UQ ] represents the matrix elements of the hole density operator in the basis of the
eigenstates of the Hamiltonian, and the factors [Q] the exponential factors in Eq. (37).
4. Spin 1/2 system
We turn now to the spinful case. In this section, we show how results obtained for the spinless
system are relevant and useful to derive in a straightforward fashion the corresponding quantities
of the spin 1/2 system.
297
4.1. Partition function

Since up- and down-spins are decoupled, and since the fermionic contribution to the action
2
is identical for both of them, we can write the fermionic determinant as: det[S] = det[S
0 ] . This
allows to express the determinant as the product of the traces of the matrix products N
n=1 [Kn ].
By making use of the mixed product property of the Kronecker product: ([A][C]) ([B][D]) =
([A] [B])([C] [D]), we obtain:
det[S] = Tr
N

[Kn ] [Kn ].
(41)
n=1
The partition function is obtained by combining Eqs. (41) and (6). Since the time steps are
decoupled, the evaluation of the integrals over the x and fields is straightforward. The tensorial
products [Kn ] [Kn ] yield 16 16 matrices, which, after application of the projectors P1 PN ,
become block diagonal with a 4 4 zero block. They all reduce to the same 12 12 real symmetric matrix [k]. The matrix [k] represents the projection of the tensor product [Kn ] [Kn ] for
all n as

[k]i,j = Pn [Kn ]i1 ,j1 [Kn ]i2,j 2
(42)
with the convention
4(i1 1) + i2
i=
i = i2 + 10
if (1 i1 2 and 1 i2 4),
and (i1 = 3 and 1 i2 2),
if (i1 = 4 and 1 i2 2),
(43)
and similarly for j . The remaining matrix elements of the 16 16 matrix form a vanishing 4 4
separate block and have been discarded in Eq. (42). Finally we obtain a simple expression for the
partition function Z of the two-site single impurity Anderson model:
Z = lim Tr[k]N .
(44)
with
[k] =
1
Lc
V
V
Ld
Lc Ld
Lc
0
0
0
V
0
0
0
0
L2c
Lc V
0
0
Lc V
0
0
0
Lc V
Lc Ld
0
0
0
0
0
0
0
0
L2c Ld
0
0
0
0
V
0
0
0
Ld
0
0
0
0
Lc V
0
0
0
Lc Ld
0
0
0
0
0
0
0
0
Lc Ld
0
0
0
0
0
0
0
0
2
Lc Ld
(45)
When expanded to lowest order in , the blocks of [k] represent the Hamiltonian matrix in the
Fock space, in the same fashion as in Eq. (33). Diagonalizing these blocks yields the expected
expression of the partition function Z:
298

Z = 1 + 3 exp (c + d 2) + 2 exp (2c + d 3)

c + d
c d 2
2
exp
+V
+j
+2
2
2
j =1

3c + d
c d 2
2
+
exp
+ 2V
2 + j
.
2
2
(46)
j =1
Through the exact calculation of the functional integrals we have recovered this result that can
also be derived from the diagonalization of the Hamiltonian. Note that V is multiplied by different coefficients in the one-particle and two-particle states. In contrast to the spinless case Eq. (33)
the eigenvalues of the Hamiltonian matrix entering Eq. (45) result from entangled states, the entanglement being achieved by the projection onto the physical Fock space in Eq. (42). Note that
we here obtained the Hamiltonian matrix without having explicitly used any basis of the Fock
space. It naturally arose as the projected tensor product of the basis appropriate to the spinless
case.
4.2. Hole density and auto-correlation function
In the spinful case, the hole density xm is given by:

Zxm = lim P1 PN det[S]xm
N
W
= lim P1 PN xm Tr
N
W
N

[Kn ] [Kn ] .
(47)
n=1
For the spin 1/2 case, the counterpart of the matrix [QX ] is the matrix [KX ] Pn (xn [Kn ]
[Kn ]). It is a 12 12 matrix, the elements of which may be expressed as:
[KX ]i,j = i,1 j,1 + i,2 j,2 + i,5 j,5 + i,6 j,6 ,
(48)
using Eqs. (8)(13), in the 0 limit. In this form it represents the hole density operator on the
impurity site in the Fock space. Thus Eq. (47) becomes

Zxm = lim Tr [KX ][k]N1 ,
N
(49)
which has exactly the same form as for the spinless case and a similar interpretation applies.
As for auto-correlation functions such as x1 xm , we can adopt the same procedure to obtain

Zx1 xm = lim Tr [KX ][k]m2 [KX ][k]Nm ,
N
(50)
which, when again introducing the eigenstates of the Hamiltonian, can also be identified to the
ordinary expression in Eq. (37). On the formal level, the use of the slave boson representation
in the radial gauge greatly simplifies the evaluation of the dynamical correlation functions of the
operators that can be represented by the radial slave bosons.
299
5. Greens function
We turn to the impurity Greens function G (q p). When expressed in terms of the eigenvalues of the Hamiltonian, E which can be read from, e.g., Eq. (46) for spin 1/2it reads:

ZG (q p) =
(51)
e(qNp)E |d |
e(pq)E

|d | .
,
In the radial gauge the creation and annihilation operators are expressed in terms of auxiliary
fields as given in Eqs. (2) and (3). We obtain G (q p) as follows:
ZG (q p) = lim
N

N
W n=1

dn

=,

D fn,
, fn,
D cn,
, cn,
dxn eS fq, fp,

xq+1 xp ,
(52)
in the language of functional integrals. Note that the three other Greens functions involving the
, cq, dp,
and cq, cp,
, can be calculated in the same
three expectation values: dq, cp,
fashion, except that they are simpler to evaluate since they contain at most one amplitude of the
slave bosonic field, x, unlike the one we chose to study.
5.1. Derivation of the pseudofermion Greens function
Following standard procedures (see, e.g., Negele and Orland [30]), we cast Eq. (52) into the
form:

ZG (q p) = lim P1 PN det[S0 ]Gp,q xp xq+1 ,
(53)
N
W
where Gp,q is the minor of one of the matrix elements of the 2 2 block that shares the same row
labels as [Lp ] and column labels as [Lq+1 ] in the matrix [S0 ] as defined in Eq. (14) (if q = N then
the block to be considered is [L1 ]). The minor Gp,q is the unprojected pseudofermion Greens
function.
For the subsequent calculations we set p = N m + 1 and q = N . The minor GNm+1,N can
be calculated as:

GNm+1,N =
(54)
det S(a; m) ,
a
where [S(a; m)] is given by:

S(a; m) i,j = [S0 ]i,j + ai,2(Nm+1) j,2N .
(55)
To calculate det[S(a; m)], we find it convenient to move the last two columns to the left in the
matrix [S(a; m)] in Eq. (55). In the same fashion as for det[S0 ], we expand det[S(a; m)] along
the first two columns. Once the derivative of det[S(a; m)] with respect to a is calculated we
obtain the three contributions:

(56)
det S(a; m) = 1 M11,a + Lc M1Lc ,a + V x1 M1,a ,
a
300
where the three minors M11,a , M1Lc ,a and M1,a are defined in Appendix A. They can be ex
pressed as linear combinations of M21,a , ML2 c ,a and M2,a in the same fashion as for the minors
defined in the previous sections. They also follow a recurrence relation that reads:
Mn+1
Mn1,a
1,a
1
0
0
n
n+1
(57)
V xn+1
Lc
MLc ,a = 0
MLc ,a ,
n
0
V
x
L
n+1
n+1
M,a
Mn+1
,a
up to time step n = N m. Here we recognize the blocks defined in Eqs. (23) and (24) entering
the matrix [K (n+1) ] which we encountered during the evaluation of the partition function. Then,
m
Nm
Nm
are linear combinations of the minors MNm+1
and MNm+1
:
MN
1,a , MLc ,a and M,a
1,4
3,4
Nm

M1,a
1
0
MNm+1
1,4
Nm
M
0
(58)
.
Lc
Lc ,a =
Nm+1
M
3,4
0 V xNm+1
MNm
,a
and Mn are tightly related to M n and M n : they are also built on a matrix
The minors Mn1,4
3,4
1,4
3,4
similar to [S0 ], but with the difference that only the time steps m with m > n are included. Mni,j
is a minor of this matrix, where the first two columns and both the ith and j th rows are removed.
and MNm+1
are also given in Eq. (22) and
Accordingly, the recurrence relations for MNm+1
1,4
3,4
hence we also need to introduce the minor MNm+1

. Thus, their evaluation involves the blocks
1,3
(n+1)
of Eqs. (24) and (25), which enter the matrix [K
] jointly with the last three components of
the column vector in Eq. (26), taken at time step N . Therefore, to calculate GNm+1,N as defined
above we have to consider the following set of six minors: M1,a , MLc ,a , M,a for time steps
1 n N m, and M1,3 , M1,4 and M3,4 for time steps N m + 2 n N .
Combining the above steps, we can write the unprojected pseudofermion Greens function as:
Nm+2
!

(N )
(n)
GNm+1,N = Tr Kf
K>
n=N1

Kf(Nm+1)
"
1

(n)
K<
(59)
n=Nm
(N )
(Nm+1)
]) reprewhere the time steps enter in decreasing order. The 6 6 matrices [Kf ] ([Kf
senting the annihilation (creation) of a fermion in the world line language are given by
(N )
Kf i,j = V xN i,1 j,4 + LN i,2 j,5 + Lc LN i,2 j,6 ,
(Nm+1)
= 1i,5 j,1 + Lc i,6 j,2 + V xNm+1 i,6 j,3 .
Kf
(60)
i,j
(n)
In Eq. (59), [K< ] is a 6 6 block diagonal matrix the non-zero elements of which are the two
(n)
] is also 6 6 block
blocks [K (n),1 ] and [K (n),2 ] defined in Eqs. (23) and (24). The matrix [K>
diagonal matrix determined by
(n) (n) (n)
K> = K
(61)
K< .
301
Eq. (59) can also be interpreted on the basis of Fig. 1. In fact, computing dN, dNm+1,
can
be visualized as the resummation of particular subsets of world lines. They are naturally split into
(n)
sets involving Ld for N m + 2 < n < N 1, in which case [K> ] is controlling the dynamics
(n)
of the world lines; and sets excluding Ld for 1 < n < N m, in which case [K< ] controls the
dynamics. The transition from the first (second) subset to the second (first) is taken care of by
] ([Kf(N ) ]).
the matrix [Kf(Nm+1)
This expression for the Greens function in the world line language can also be related to its
counterpart in the Hamiltonian language. Indeed, the trace of the matrix product in Eq. (59) can
be written in terms of 4 4 matrices:

Nm+2
1

>
<
Kn
Kn
[Nm+1 ]
,
GNm+1,N = Tr [FN ]
(62)
n=Nm
n=N1
where
0
0
0
0
0
L c LN
,
0
V xNm+1
0
(63)
<
characterize the creation and annihilation of an electron, respectively. The matrices [Kn ] and
[Kn> ] result from [Kn ] as:

<
Kn i,j = Kn i,j Lc Ln i,4 j,4 ,

>
Kn i,j = Kn i,j i,1 j,1 .
(64)
0 0
0 0
[Nm+1 ] =
1 0
0 Lc
0
0
0
0 V xN
0
0
and [FN ] =
0
0
0
0
LN
0
0
0
5.2. Spinless case

Now that the unprojected pseudofermion Greens function has been converted to a compact
form, we can evaluate the physical Greens function which, in the spinless case, reads:
!
Nm+2

>
Z0 G0 (q p) = lim P1 PN Tr x1 xNm+1 [FN ]
Kn
N
W
[Nm+1 ]
1

"
Kn<
n=N1
(65)
n=Nm
With the application of the projectors P1 PN , we obtain:

m2 < Nm1
Z0 G0 (q p) = lim Tr [QX ][F] q >
,
[] q
N
(66)
where [QX ] is given by (36). The matrices [] and [F] are defined as [] PNm+1 (xNm+1
[N m+1 ]) and [F] PN ([FN ]), respectively. They can be read off from Eq. (63) if x is
replaced by 1, and Ln by Ld . Note that [] = [F ] only in the limit 0 when Lc 1
and Ld 1. In the limit 0, they coincide with the matrix representations of the operators f and f , respectively. The matrices [q < ] and [q > ] are given by [q < ] Pn ([Kn< ]) and
[q > ] Pn ([Kn> ]), respectively, and are easily related to [q]:
>
<
q i,j = [q]i,j i,1 j,1 .
q i,j = [q]i,j Lc Ld i,4 j,4 ,
(67)
302
we can now reshape Eq. (66) into the form:

Nm1

Z0 G0 (q p) = lim Tr Q<
[UQ ] [QX ][F][UQ ]
N

m2
[UQ ] [][UQ ] ,
Q>
(68)
where [Q< ] and [Q> ] are obtained from the diagonalization of [q < ] and [q > ], respectively. Note
that the same matrix [UQ ] diagonalizes the matrices [q], [q < ], and [q > ].
In Eq. (68), the product [UQ ] [QX ][F][UQ ] is the representation of the annihilation operator d in the basis of the eigenstates of the Hamiltonian, while [UQ ] [][UQ ] represents d , as
can be easily verified explicitly in the limit 0. The factors [Q> ] and [Q< ] determine the
time evolution. Therefore Eq. (68) can be easily identified with Eq. (51). Note that all eigenvalues seem to contribute to the Greens function in Eq. (51), and only the matrix elements of
the creation (annihilation) operators restrict the set of eigenvalues that effectively contribute to
the Greens function. In contrast, in Eq. (68) some of these restrictions are contained in the factors [Q> ] and [Q< ] which replace the full set of eigenvalues, that would be contained in the
matrix [Q].
It is tempting to compare the physical electron Greens function (including the factors x) to
the projected pseudofermion Greens function (without the factors x). Straightforward algebra
yields:

lim PNm+1 xNm+1 [Nm+1 ] = lim PNm+1 [Nm+1 ] ,
(69)
0
and

lim P1 PN x1 K1< [FN ] = lim P1 PN K1< [FN ] .
(70)
Therefore, as a particularity of the spinless case, the factors x in Eq. (65) play no role, and
both Greens functions coincide. Consequently the same result for the physical electron Greens
function would have been obtained by substituting x by 1 in Eqs. (2) and (3), and accordingly
in the fermionic contribution to the action Sf . Incidentally, such a procedure is in complete
agreement with the original suggestion by Kotliar and Ruckenstein [17] to modify the expression
of the physical electron operator by introducing square root factors, when extended to the spinless
case.
5.3. Spin 1/2 system
Again, as shown below, results obtained for the spinless system can be immediately applied
to derive the Greens function in the spinful case. Inserting Eq. (59) into Eq. (53) yields:
ZG (q p) = lim P1 PN x1 xNm+1
N
W
Tr

[KN ] [FN ]
Nm+2

>
[Kn ] Kn
n=N1

[KNm+1 ] [Nm+1 ]
1

n=Nm
[Kn ]
"
Kn<
(71)
which after application of the projectors P1 PN becomes:

m2 < Nm1
,
[] k
ZG (q p) = lim Tr [KX ][ ] k >
303
(72)
with the matrices [] = PNm+1 (xNm+1 [KNm+1 ][Nm+1 ]) and [ ] = PN ([KN ][FN ]).
Leaving out entries which do not contribute in the limit N they read:
[]i,j = 1i,3 j,1 + Lc (i,4 j,2 + i,7 j,5 ) + L2c i,8 j,6 ,
[ ]i,j = Lc i,1 j,3 + Lc Ld (i,2 j,4 + i,5 j,7 ) + L2c Ld i,6 j,8 .
[k < ]
[k > ]
[k < ] Pn ([Kn ] [Kn< ])
(73)
[k > ] P
>
and
are given by
and
The matrices
n ([Kn ] [Kn ]),
respectively. The asymmetry in the representation of the physical electron creation and annihilation operators (Eqs. (2) and (3)) is also apparent in Eq. (72). Indeed the operator d is represented
by the matrix [], and d by the product of the matrices [KX ][ ]. In contrast to the spinless case
the factors x in Eq. (71) play a role, and the projected pseudofermion Greens function and physical electron Greens function differ.
6. Conclusion
In summary we have established a new scheme which provides a fundamental connection
between the representation of expectation values and dynamical correlation functions in the
Hamiltonian language and their counterpart in the slave-boson path integral formulation. This
has been achieved for the U = spin 1/2 single impurity Anderson model through their exact
evaluation for a two site cluster. The new scheme allowed us to compute the partition function
and the hole density, expressed as the expectation value of the radial slave field x. Moreover the
Greens function and the hole auto-correlation function were evaluated within this scheme.
We verified that the exact expectation value of the slave boson amplitude field x is finite, as
postulated in mean-field calculations, even in this extreme quantum case. It is therefore not related to the condensation of a boson, which would necessarily vanish in such a calculation. The
suppression of the condensation originates in the use of the radial representation, where the phase
of the boson is integrated out in the first place. We note that higher slave boson correlation functions such as x n ( )x m (0) reduce to x( )x(0). Therefore the field x bears little resemblance
to ordinary complex bosonic fields. The corresponding calculations follow a similar scheme as
those for the partition function.
Through an independent calculation we obtained both the physical electron and pseudofermion Greens functions. In the spinless case, the projected pseudofermion Greens function is
finite, and it is identical to the one of the physical electron. Therefore a perturbation theory-like
factorization of the latter as a product of the boson and pseudofermion Greens functions does
not appear appropriate in general, but it may still be valid in particular frequency ranges, such as
the low frequency domain. In the latter case, a mean-field decoupling looks more appropriate. It
is likely to provide a better agreement with the exact result if the square root factors, originally
proposed in [17], are introduced.
It is also of great importance to understand that our formalism allows immediate and straightforward use of results, which were obtained for the spinless system, in order to derive those of
the spinful case: the proposed scheme first treats the coherent states of fermions in the two spin
sectors (up and down spin) separately. The world lines of particles with different spin projection
evolve independently. Only in a final step, when the full fermionic determinant is built from the
product of the determinants of the two spin species, the projection onto the physical space with
304
no double occupancies is straightforwardly achieved through the projection rules, Eqs. (8)(13),
applied to the entries of the determinant. Here, the projection is easily accomplished in the Fock
space which, when directly enforced for the world lines, produces a complicated entanglement
of coherent states. For larger systems the exact resummation of the world lines and their respective projectionas presented in, e.g., Eqs. (39), (50) and (66)is difficult on the analytical
level, but probably not on the numerical level. As an alternative to the exact calculation one may
consider a plain saddle point approximation scheme when tackling spatially extended systems.
Unfortunately the latter fails to reproduce the exact result even for the two-site problem, and it
will be necessary to determine appropriate quasiparticle weight factors from the two-site solution
within an effective slave boson approach, similar to the KotliarRuckenstein scheme [17]. This
challenge will be addressed in future work [34].
The extension of the above scheme to the spin rotation invariant formulation of the tJ model
where the phases of all the bosonic fields can be gauged away is desirable [18]. Work along this
line is in progress.
One may wish to extend such a calculation to other representations of this model, such as
the ones based on Hubbard X-operators [32]. This unfortunately poses another challenge, since
the angular part of the respective action is intrinsically off diagonal in time, which makes
the integral over the angular variables significantly more difficult. This also holds true for the
KotliarRuckenstein representation where one of the bosonic fields is complex [17,18,33]. Alternatively one may also consider weak-coupling approaches, such as the HubbardStratanovich
decoupling of the interaction term in the charge channel. Even if it were possible to evaluate the
partition function exactly, using the corresponding form of Eqs. (6), (7) and (41), it would still
require a major effort to obtain dynamical response functions. Nevertheless such a calculation
deserves further study.
Acknowledgements
We gratefully thank O. Juillet for stimulating discussions. R.F. is grateful for the warm hospitality at the EKM of Augsburg University where part of this work has been done. This work was
supported by the Deutsche Forschungsgemeinschaft (DFG) through SFB 484.
Appendix A. Expression of the minors M11,a , M1Lc ,a and M1,a
The minors M11,a , M1Lc ,a and M1,a in Eq. (56) are obtained from the expansion of
det[S(a; m)] given below for p = N m + 1. Their definition follows as:
M11,a by expanding det[S(a; m)] along the two first columns and eliminating the (2N 1)th
and 2pth lines.
M1,a by expanding det[S(a; m)] along the two first columns and eliminating the second and
2pth lines.
M1Lc ,a by expanding det[S(a; m)] along the two first columns and eliminating the first and
2pth lines.
In the expansion of det[S(a; m)] M11,a is multiplied by [S(a; m)]2N1,1 = 1 and [S(a;
m)]2p,2 = a,
M1,a
is multiplied by [S(a; m)]2,1 = V x1 and [S(a; m)]2p,2 = a, and M1Lc ,a
305
is multiplied by [S(a; m)]1,1 = Lc and [S(a; m)]2p,2 = a. The three minors above satisfy a recurrence relation given in Eqs. (57) and (58), while det[S(a; m)] reads:

det S(a; m)

L
V x1
1
0

c

V x
L1
0
1
1

Lc V x2

V x2 L2

.. ..

.
.

0
1
0
0
...
Lc V xp

0
a
...
V xp Lp
0
1
=
.

V xp+1
Lc

V xp+1 Lp+1

..
..

.
.

1
0

0
1

0
Lc V xN
1

0
V xN
LN
(A.1)
References
[1] S. Pairault, D. Snchal, A.-M.S. Tremblay, Eur. Phys. J. B 16 (2000) 85;
S. Pairault, D. Snchal, A.-M.S. Tremblay, Phys. Rev. Lett. 80 (1998) 5389.
[2] P.W. Anderson, J. Phys. C: Solid State Phys. 3 (1970) 2346.
[3] P. Nozires, J. Low. Temp. Phys. 17 (1974) 31.
[4] K.G. Wilson, Rev. Mod. Phys. 47 (1975) 773.
[5] N. Andrei, Phys. Rev. Lett. 45 (1980) 379.
[6] P.B. Wiegmann, Sov. Phys. JETP Lett. 31 (1980) 392.
[7] A. Georges, G. Kotliar, Phys. Rev. B 45 (1992) 6479.
[8] A. Georges, G. Kotliar, W. Krauth, M.J. Rozenberg, Rev. Mod. Phys. 68 (1996) 13.
[9] W. Metzner, D. Vollhardt, Phys. Rev. Lett. 62 (1989) 324.
[10] T. Maier, M. Jarrell, Th. Pruschke, M.H. Hettler, Rev. Mod. Phys. 77 (2005) 1027.
[11] G. Kotliar, S.Y. Savrasov, G. Palsson, G. Biroli, Phys. Rev. Lett. 87 (2001) 186401.
[12] G. Biroli, G. Kotliar, Phys. Rev. B 65 (2002) 155112.
[13] M. Potthoff, Adv. Solid State Phys. 45 (2005) 135.
[14] B. Kyung, S.S. Kancharla, D. Snchal, A.-M.S. Tremblay, M. Civelli, G. Kotliar, Phys. Rev. B 73 (2006) 165114.
[15] S.E. Barnes, J. Phys. F: Metal Phys. 6 (1976) 1375.
[16] P. Coleman, Phys. Rev. B 29 (1984) 3035.
[17] G. Kotliar, A.E. Ruckenstein, Phys. Rev. Lett. 57 (1986) 1362.
[18] R. Frsard, P. Wlfle, Int. J. Mod. Phys. B 6 (1992) 685;
R. Frsard, P. Wlfle, Int. J. Mod. Phys. B 6 (1992) 3087, Erratum.
[19] R. Frsard, G. Kotliar, Phys. Rev. B 56 (1997) 12909;
H. Hasegawa, Phys. Rev. B 56 (1997) 1196.
[20] R. Frsard, M. Dzierzawa, P. Wlfle, Europhys. Lett. 15 (1991) 325.
[21] W. Zimmermann, R. Frsard, P. Wlfle, Phys. Rev. B 56 (1997) 10097.
306
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
S. Florens, A. Georges, G. Kotliar, O. Parcollet, Phys. Rev. B 66 (2002) 205102.

S. Elitzur, Phys. Rev. D 12 (1975) 3978.
J. Kroha, P. Wlfle, T.A. Costi, Phys. Rev. Lett. 79 (1997) 261.
S. Kirchner, J. Kroha, P. Wlfle, Phys. Rev. B 70 (2004) 165102.
K. Baumgartner, H. Keiter, Phys. Status Solidi B 242 (2005) 377.
N. Read, D.M. Newns, J. Phys. C 16 (1983) 3273.
R. Frsard, T. Kopp, Nucl. Phys. B 594 (2001) 769.
H. Ouerdane, R. Frsard, T. Kopp, unpublished.
J.W. Negele, H. Orland, Quantum Many-Particle Systems, AddisonWesley, 1988.
S.E. Barnes, J. Phys. F: Metal Phys. 7 (1976) 2637.
E.O. Tngler, T. Kopp, Nucl. Phys. B 443 (1995) 516.
Th. Jolicur, J.C. Le Guillou, Phys. Rev. B 44 (1991) 2403.
R. Frsard, et al., unpublished.
Quantum phase transition in lattice model

of unconventional superconductors
Kenji Sawamura, Yuki Moribe, Ikuo Ichinose
Department of Applied Physics, Nagoya Institute of Technology, Nagoya 466-8555, Japan
Received 28 March 2007; received in revised form 20 April 2007; accepted 31 May 2007
Abstract
In this paper we shall introduce a lattice model of unconventional superconductors (SC) like d-wave
SC in order to study quantum phase transition at vanishing temperature (T ). Finite-T counterpart of the
present model was proposed previously with which SC phase transition at finite T was investigated. The
present model is a noncompact U(1) lattice-gauge-Higgs model in which the Higgs boson, the Cooper-pair
field, is put on lattice links in order to describe d-wave SC. We first derive the model from a microscopic
Hamiltonian in the path-integral formalism and then study its phase structure by means of the Monte Carlo
simulations. We calculate the specific heat, monopole densities and the magnetic penetration depth (the
gauge-boson mass). We verified that the model exhibits a second-order phase transition from normal to SC
phases. Behavior of the magnetic penetration depth is compared with that obtained in the previous analytical
calculation using XY model in four dimensions. Besides the normal to SC phase transition, we also found
that another second-order phase transition takes place within the SC phase in the present model. We discuss
physical meaning of that phase transition.
1. Introduction
GinzburgLandau (GL) theory plays a very important role in study of the superconducting
(SC) phase transition. In particular GL models defined on a lattice have been extensively studied
E-mail address: ikuo@nitech.ac.jp (I. Ichinose).

doi:10.1016/j.nuclphysb.2007.05.030
308
K. Sawamura et al. / Nuclear Physics B 785 [FS] (2007) 307321
in order to investigate effects of topological excitations like vortices nonperturbatively. These

models are nothing but gauge-Higgs models on a lattice that also play an important role in elementary particle physics.
At present many materials that have unconventional SC phase have been discovered [1].
Among them, some of the materials, including the high-temperature (T ) cuprates, the rare-earth
heavy-fermion materials, etc., exhibit d-wave SC. Order parameter (the Cooper-pair wave function) of the d-wave SC has nodes. Therefore in order to formulate a GL theory of the d-wave
SC, the Cooper-pair field, i.e., the Higgs field, must be put on links of the lattice instead of lattice
sites.
In the previous papers [2,3], we introduced a new type of lattice gauge-Higgs models that
describe d-wave SC phase transitions. We call these models U V models where the U -field
refers to the ordinary gauge field whereas the V -field is the Higgs field defined on links. In
Ref. [2], a compact-U(1)-gauge version was studied from the viewpoint of the tJ model that
is a canonical model of the high-Tc cuprates. In the slave-particle representation, the U -field is
an emergent compact U(1) gauge boson and the V -field is a spinon-pair field for the resonatingvalence-bond (RVB) order. By means of Monte Carlo (MC) simulations, we found that there
appears a phase transition line as the coupling between the U - and V -fields is increased. We
measured instanton densities and expectation value of the Wilson loops, and verified that the
observed phase transition is the one from a confinement to Higgs phases. There the confinement
phase corresponds to electron phase, whereas the Higgs phase to fractionalized spin-gap phase.
Numerical results show that the phase transition is of first order.
In Ref. [3], on the other hand, we studied finite-T phase transition of unconventional SC by
using the U V model. In this approach, the U -field is a noncompact U(1) gauge field corresponding to the usual electromagnetic field, whereas V is the electron Cooper-pair field. There
we considered the London limit of V , i.e., Vx,i U(1). Gauge-invariant gauge-boson mass (the
inverse correlation length of the magnetic field) is an order parameter for the normal to SC phase
transition. We observed the phase transition by means of the MC simulations and found that the
transition is of first order. However as the anisotropy of the layered structure is increased, signal
of the first-order phase transition is getting weaker, e.g., the discontinuity of the gauge-boson
mass at the transition point is getting smaller.
In the present paper, we shall continue to study critical properties of the unconventional SC by
using the U V model. In particular we focus on the quantum phase transition (QPT) at T = 0.
In Section 2, we shall derive the GL model for the QPT starting from a microscopic Hamiltonian of the Cooper pair. In Section 3, we shall show the numerical calculations of the U V
model in (2 + 1) dimensions ((2 + 1)D). We calculated the internal energy, specific heat,
monopole density of the U - and V -fields, and the magnetic penetration depth (the inverse
gauge-boson mass). We found that the model exhibits a second-order phase transition from the
normal to SC phases at certain critical value of the parameter gc , which controls quantum fluctuations. (Larger g suppresses quantum fluctuations more.) In Section 4, we report the result of
the MC simulations for the U V model in (3 + 1)D. The reason why we study both the (2 + 1)
and (3 + 1)D models is that most of the materials of the d-wave SC are quasi-2D and have a
layered structure. We found that the (3 + 1)D model also exhibits a second-order SC phase transition. However besides that transition, the model has another phase transition at larger gc > gc
within the SC phase. This phase transition is also of second-order. We shall discuss its physical
meaning. Section 5 is devoted for discussion.
309
2. GinzburgLandau model on a lattice

In this section, we shall derive the lattice GL model from the microscopic Hamiltonian by
using the path-integral formalism. To this end, let us first introduce a Cooper-pair field x,i on
links (x, i) of d-dimensional hypercubic lattice with lattice spacing a0 (= 1), where x denotes the
lattice site and i is the direction index i = 1, . . . , d. In terms of the electron annihilation operator
x, , where =, is the spin index,
x,i = x, x+i, x, x+i, .
(2.1)
In the d-wave SCs, amplitude of on-site Cooper pair is vanishingly small because of, e.g., the
strong on-site Coulomb repulsion. Furthermore, for example, in the dx 2 y 2 -wave SC state, x,1
x,2 < 0, i.e., the Cooper-pair field changes its sign under a /2 rotation in the 12 plane of the
real space. One should also notice that SC phase transition in the underdoped region of the highTc materials can be understood as a BoseEinstein condensation of a bosonic bound state of a
pair of charge carriers. From these observations, we propose the following quantum Hamiltonian
for the d-wave SCs,

H = t
x,i Ux+i,j
x,i x+i,j
x+j,i
Ux,j
+ H.c. C
x+j,i x,j
+ H.c.
x,i =j
+D
x,i =j
x,i x+i,j
Ux+j,i
Ux,j
x,i =j

+ H.c. +
Vx,y (Nx,i 0 )(Ny,j 0 ),
x,y,i,j
(2.2)
where t, C, D are parameters, the electromagnetic (EM) gauge field Ux,i is related to the EM
x+i
vector potential A
em as Ux,i = eiAx,i , Ax,i = x A
em d
and Nx,i is the number operator of the
. denotes an uniform background positive charge and V
Cooper-pair field, Nx,i = x,i
x,i
0
x,y
is a Coulomb potential including long-range interactions like,
2
e V ,
x = y,
Vx,y = 2 0
(2.3)
e /|x y|, |x y| 1,
for d = 3. As we are interested in the long-range interaction in Eq. (2.3) in the later discussion,
we have ignored the dependence on direction of the Cooper pairs in the Coulomb potential. The
t-term in Eq. (2.2) is nothing but the hopping of the Cooper pair and the C-term is a plaquette
term of the link field x,i that stabilizes the phase fluctuations of x,i . On the other hand, the Dterm determines relative phase of adjacent s. The Hamiltonian (2.2) might be derived from a
canonical model of the high-Tc cuprates like the tJ model by integrating out electrons just like
Gorikovs derivation of the GL Hamiltonian for the conventional s-wave SC [4,5]. It is obvious
that the Hamiltonian H in Eq. (2.2) is invariant under the following local gauge transformation,
Ux,i eix+i Ux,i eix ,
x,i eix+i x,i eix .
(2.4)
In the rest of this section, we shall derive the GL model from the Hamiltonian (2.2). If the reader
is not interested in the derivation, please go to the result Eq. (2.27) directly.
We employ the path-integral formalism by introducing imaginary time . Then the partition
function Z is given by

Z = [d dA]eS ,
(2.5)
310
with the following action S,

S = d
x,i
x,i + H (, U ) + Sem (A),
(2.6)
x,i
where
Sem (A)

Fij2 (x) =
(Ax,i Ax+j,i + Ax+i,j Ax,j )2 .
Let us decompose x,i into radial and phase variables as follows,

x,i = 0 + x,i eix,i .
(2.7)
By substituting Eq. (2.7) into Eq. (2.6), the first term of S becomes

x,i
x,i
S = d
x,i

=

2i x,i 0 x,i + O (x,i )2 ,
(2.8)
x,i
where we have used the periodic boundary condition on x,i and x,i (mod 2) in the
-direction, and x,i = x,i . Similarly the terms in H Eq. (2.2) are expressed as

x,i Ux+i,j
x+j,i
Ux,j
+ c.c.
Ht t
x,i =j
= 2t0
cos(x,i x+i,j Ax,j Ax+i,j ) + O(x,i ),
x,i =j
x,i x+i,j
x+j,i x,j
+ c.c.
HC C
x,i =j
= 2C02
cos(x,i x+i,j + x+j,i x,j ) + O(x,i ),
x,i =j
HD D
x,i x+i,j
Ux+j,i
Ux,j
+ c.c.
x,i =j
= 2D02
HN

x,y,i,j
cos(x,i x+i,j Ax+j,i Ax,j ) + O(x,i ),
x,i =j
Vx,y (Nx,i 0 )(Ny,j 0 )

4Vx,y 0 x,i y,j + O (x,i )3 .
(2.9)
x,y,i,j
From Eqs. (2.8) and (2.9), we can integrate over x,i in Eq. (2.5) and obtain

1
S
e
[d ]eS d HN = e d x,i,j Vx,y x,i y,j .
(2.10)
311
Furthermore, we can prove the following identity by straightforward calculation (see

Appendix A),

S
= [dA0 ]e d HA ,
e
(2.11)
with

HA =

1
1
dk d 2 e2 V (k) 1 A 0 (k)A 0 (k) + 2
( x,i eAx,0 )( x,j eAx,0 ),
e
x,i,j
(2.12)
where V (k) is the Fourier transformation of Vx,y and the scalar potential Ax,0 is the field that
mediates the Coulomb interaction (2.3). (A 0 (k) is the Fourier transformation of Ax,0 .)
Similarly we can show (see Appendix A),

d Ht
e
(2.13)
=
dV dV e d HV ,
where
HV =

1 1 iA
J e Vx,i Vx+j,i Vx,i eix,i + c.c. ,
2
(2.14)
x,i =j
with A Ax+i,j + Ax,j and J = 2t0 . As Vx+j,i eiA+ix,i , the field Vx,i can be regarded as
the Cooper-pair field.
Finally we shall integrate over x,i . To this end, we consider the following integral

1
d HV
= [d] exp
d 2
(x,i eAx,0 )( x,j eAx,0 )
e
e
x,i,j

Vx,i eix,i + c.c.
.
+ (d 1)
(2.15)
x,i
From Eq. (2.15), it is obvious that requirement of the invariance of HV under a local gauge
transformation,
x,i x,i + ex ( ),
Ax,0 Ax,0 + x ( ),
Vx,i eiex ( ) Vx,i ,
determines dependence on Ax,0 in HV . Then by simply putting Ax,0 = 0, we define

12 d x,i x,j
1 d x,i x,j
1
e
[d]e
O, Z = [d]e e2
.
O0 = Z
(2.16)
(2.17)
It is not difficult to show,1

e2
x,i (1 )y,i (2 ) 0 = xy lim e|1 2 | ,
0
(2.18)
1 It should be remarked here that the Green function like ei(x,i (1 )x,i (2 )) , which is invariant under a 0
independent local gauge transformation x,i ( ) x,i ( ) + x,i , is well-defined, whereas x,i (1 )y,i (2 )0 is
not.
312
where is the infrared cutoff. Then,

i( ( ) ( ))
e x,i 1 x,i 2 0 = exp e2 |1 2 | .
(2.19)
From Eqs. (2.15) and (2.19), the lowest-order terms of the cumulant expansion are obtained as

d2
Vx,i ( )Vx,i
(2 )ei(x,i ( )x,i (2 ))
HV A =0 = (d 1)2
0

= (d 1)2
= 1
d2 Vx,i ( )Vx,i
(2 )ee
2 |
2|

Vx,i ( )2 2
Vx,i ( )2 + ,
2
x,i
where
x,i
(2.20)
x,i
2
,
e2

4
2
2 = (d 1)2 d 2 ee | | = (d 1)2 6 .
e
1 = (d 1)2
d ee
2 | |
= (d 1)2
(2.21)
From Eq. (2.14), the spatial coupling of Vx,i in the Hamiltonian is given as (neglecting
V
the gauge field for simplicity) HV Vx,i
x+j,i + c.c., and therefore we replace as Vx,i
|x|
J Vx,i in the action where |x| = 1(1) for even (odd) site. After this replacement, the
()
first term of HV becomes

1
1 iA
e Vx,i Vx+j,i + c.c. =
|Dj Vx,i |2 (d 1)
|Vx,i |2 ,
(2.22)
2
2
x,i =j
x,i =j
x,i
where Dj Vx,i = Ux,j Ux+i,j Vx+j,i Vx,i .

Let us define the effective action SCP of the Cooper-pair field Vx,i and the gauge field as

SCP Sem
= [d dA]eS .
Z = [dV dA]e
(2.23)
Then up to the second-order terms,

2
1
J 2
|Dj Vx,i |2 +
D Vx,i ( )
SCP d
2
2
x,i =j
x,i

1
(J 1 + d 1)
|Vx,i |2 + dk d 2 e2 V (k) 1 A 0 (k)A 0 (k) ,
(2.24)
x,i
where D Vx,i = ( ieAx,0 )Vx,i .

At long distances, the last term of SCP in Eq. (2.24) becomes the usual relativistic kinetic
terms of Ax,0 ,

d dk k 2 A 0 (k, )A 0 (k, ),
(2.25)
whereas the effects of the short-range part of Vx,y generate terms like |Vx,i |4 . Similar quartic
terms also appear in the higher-order terms of the cumulant expansion. Furthermore, the second
313
and third terms of H in Eq. (2.2) are expressed in terms of Vx,i by the perturbative calculation in
powers of C, D in the cumulant expansion.
In the derivation of the GL model action SCP given in this section, only the lowest-order terms
of x,i are considered. It is obvious that higher-order terms like O((x,i )2 )-terms in Eq. (2.8)
give nonvanishing contributions to the GL model. However we think that the full GL model,
which contains all relevant terms for studying the SC phase transition, has a similar form to that
obtained in the above. The expansion in powers of the order-parameter field Vx,i , the invariance
under a /2-rotation in the spatial planes and the invariance under a local gauge transformation,
Ux,i eix+i Ux,i eix ,
Vx,i eix+i Vx,i eix ,
(2.26)
which originates from Eq. (2.4), restrict the form of the GL action. Effective field-theory model
for the granular s-wave SC has been derived in a similar way [6]. Then in order to study phase
structure, order of phase transitions, etc., we propose action of the GL model as follows, which
is a kind of the U V model proposed previously,

2
SGL = g c1
Ux, Vx+,i Ux+i, Vx,i
F
(x) c2
+ c.c.
x, =
d2
x, =i
Ux,i Ux+i,j Vx+j,i Vx,j
x,i =j

x,i
|Vx,i |
4

Vx,i Vx+i,j Vx+j,i Vx,j + c.c.

+ c.c. c3
x,i =j
|Vx,i | ,
2
(2.27)
x,i
where F (x) = Ax, + Ax+, Ax, Ax+, (, = 0, 1, . . . , d) and we have introduced

a lattice also in the -direction. It should be noticed that the gauge field Ux, (Ax, ) also has
the -component Ux,0 = eiAx,0 besides the spatial ones,2 whereas the Cooper-pair field Vx,i has
only spatial components. For numerical calculation, we fix values of the parameters c1 , . . . , in
Eq. (2.27) and study phase structure by varying g. For large g, the electromagnetic interactions
between Vx,i is getting weak, and the hopping of Vx,i is enhanced. Because of the uncertainty
relation between the number nx,i and phase x,i , nx,i x,i 1, an SC state, in which Vx,i s
are stabilized, is formed. For small g, on the other hand, charge density nx,i tends to have definite
distributions and as a result an insulating Wigner crystal or the normal metallic state appears. In
other words, the parameter g plays a role of 1/h .
3. Numerical results in (2 + 1)D
Let us turn to the numerical studies of the present lattice U V model for the SC phase transition in two and three spatial dimensions [7]. In order to investigate phase structure, we measured
the internal energy E and the specific heat C,

2
E SGL /V ,
(3.1)
C SGL SGL /V ,
where V = L3 is the system size. For investigation on 2D thin film SC materials, we consider
3D symmetric lattice as we are studying the QPT at T = 0. In the action SGL in Eq. (2.27),
2 It should be noticed that in the action (2.27) V
x,i couples with both Ux,0 and Ux+i,0 . This means that Vx,i has
electric charge at x and x + i as the original Cooper pair given by Eq. (2.1) does.
314
Fig. 1. System-size dependence of the specific heat C for = = 5 (left) and the London limit (right) of the (2 + 1)D
UV model. The peak develops as L is getting larger in each case. The result indicates that the second-order phase
transition occurs at around g = 0.902 (0.490) for the = = 5 (London limit) case.
we fix the values of c1 , . . . , and measured E and C as a function of g. As we explained in the

previous section, the parameter g controls quantum fluctuations, and therefore an observed phase
transition is nothing but a QPT.
We mostly studied the cases with the parameters c1 = c2 = c3 = d2 = 1, = = 3, 5, 10
and also the London limit of Vx,i . The negative value of d2 means that the configurations corresponding to the dx 2 y 2 -wave SC are enhanced for large g. In all these cases, we have observed
no anomalous behavior of E, whereas we found that C exhibits a sharp peak as g is varied. We
show the system-size dependence of C in Fig. 1 for the cases of = = 5 and the London limit.
It is obvious that the peak develops as L is getting larger in both cases. Similar behavior of C was
observed for the case with = = 3, 10. Then we conclude that a second-order phase transition
takes place as g is increased.
In order to verify that the observed phase transition is a SC transition, we measured gaugeinvariant gauge-boson mass MG (the inverse of the magnetic
penetration depth). MG is defined as

the inverse of the correlation length of the operator O(x) = i,j =1,2 ij sin(Fij (x)), where 12 =
21 = 1 [8]. In order to evaluate MG accurately, we introduce Fourier transformed operator of
O(x) in the 12 plane,

0) =
O(x
(3.2)
O(x)eip1 x1 +ip2 x2 .
x1 ,x2
Then we expect the following behavior,

0 + )O(x
0 ) e
O(x
2
p12 +p22 +MG
(3.3)
0+
In the numerical calculation, we put p1 = p2 = 1/L and verified that the correlator O(x
)O(x0 ) exhibits
the exponential decay as in Eq. (3.3). More precisely, we define MG as MG =
0 ).
sgn(2 p
2 ) 2 p
2 , where is the inverse correlation length of the correlator of O(x
In Fig. 2, we show the calculations of MG for = = 5 and the London limit cases. The
negative value of MG in Fig. 2 is the finite-size effect [8]. The results show that in both cases
MG develops continuously from the phase transition point obtained by the measurement of C.
At first-order phase transition points, the MG exhibits a sharp jump from vanishing to a finite
value [3]. Therefore we conclude that the phase transition observed by the measurement of C is
a second-order SC transition. We have observed similar behavior of MG for the other cases with
= = 3, 10.
315
Fig. 2. Gauge-boson mass MG for the cases of = = 5 (left) and London limit (right) of the (2 + 1)D UV model.
MG develops at the critical coupling obtained by the measurement of C. This result indicates that the SC phase transition
occurs and the order of the transition is of second order.
Fig. 3. Specific heat C for (3 + 1)D U V model with = = 5. System size is 84 . There exist two peaks at g 0.67
and 0.78.
4. Numerical results in (3 + 1)D

In this section, we shall show results of the numerical study of the (3 + 1)D cases. As most of
the materials of the unconventional SC have quasi-2D structure, studies on 2D and 3D systems
are useful to obtain the physical picture of the real materials. It is known that quasi-2D systems
except for thin film samples exhibit 3D properties in critical regions.
As in the (2 + 1)D case, we first show the calculation of C in Fig. 3 for the case = = 5.
There exist two peaks at g 0.67 and 0.78. Then we carefully studied the system-size dependence of each peak and the results are shown in Fig. 4. From the calculations in Fig. 4, we
conclude that the both peaks are signal of second-order phase transition as they have the systemsize dependence. Similar behavior of C was observed in the case with = = 10 (see Figs. 5
and 6) and the London limit.
In order to investigate physical meaning of the phase transitions, we measured the gaugeboson mass MG . Results are given in Figs. 7 and 8. MG starts to develop at the first critical
coupling gc 0.67. Furthermore at the second critical coupling gc 0.78, MG slightly changes
its behavior upwards. Therefore, the SC phase transition occurs at gc 0.67. On the other hand
316
Fig. 4. System-size dependence of two peaks in C for = = 5. Result indicates that both peaks are signal of second-order phase transition.
Fig. 5. Specific heat C for = = 10 with system size 84 . There are two peaks.
Fig. 6. System size dependence of two peaks in C for = = 10. Result indicates that both peaks are signal of second-order phase transition.
at gc 0.78, a certain mechanism suppressing fluctuations of the gauge field Ax, further starts
to work.
It is quite helpful to study the monopole density of U and V gauge fields for understanding
physical meaning of the second phase transition. In the ordinary s-wave SC state, vortices appear
as low-energy excitations. In the GL theory of the s-wave SC, the vector potential A
em and the
317
Fig. 7. Gauge-boson mass MG in (3 + 1)D system for the = = 5 case. MG starts to develop at the first critical
coupling gc 0.67. Furthermore at the second critical coupling gc 0.78, MG slightly changes its behavior upwards.
Fig. 8. Gauge-boson mass MG in (3 + 1)D system for the = = 10 case. MG starts to develop at the first critical
coupling gc 0.68. Furthermore at the second critical coupling gc 0.84, MG slightly changes its behavior upwards.
order parameter are given as follows for vortex configurations [9],

1
A
em =
grad ,
= | |ei ,
2e
where

1
1
A
em d l
= = 2n, n = 1, 2, . . . .
2e
2e
(4.1)
(4.2)
In the present U V model, similar vortex configurations exist. However as the V -field is defined
on links, the V -field has also monopole-like configurations just like the ordinary compact U(1)
gauge field, i.e., vortices can terminate at monopole and anti-monopole. The phase transition
from the SC to normal states occurs as a result of the condensation of (infinitely long) vortices,
but we can also expect that some phase transition, which is related with monopole dynamics,
occurs at a certain coupling constant.
In Fig. 9, we show the calculation of the U - and V -monopole densities. We employ the definition of the monopole density in the lattice gauge theory given in Ref. [10]. Please remark that the
V -field is a 3D vector field and its monopole configuration can be defined straightforwardly.
On the other hand, the U -field is a (3 + 1)D vector field, and we define the U -monopole just
focusing on its 3D spatial component.
318
Fig. 9. U - and V -monopole densities, U and V , in the (3 + 1)D U V model with = = 5.
Fig. 10. U - and V -monopole densities, U and V , in the (3 + 1)D U V model with = = 10.
From Fig. 9, it is obvious that the U -monopole density U is very low in the whole parameter
region of g as the U -field is the noncompact gauge field. It starts to decrease very rapidly at
around the first phase transition point gc 0.67. At the second phase transition point gc 0.78,
U is substantially vanishing. On the other hand, the V -monopole density V starts to decrease at
gc 0.67, but it has a finite value between the first and second phase transitions. After the second
transition at gc 0.78, V decreases very rapidly. Similar behavior is observed also in the case
of = = 10 and the London limit. (See Fig. 10.) This result indicates that a finite density
of short V-vortices connecting monopole and anti-monopole survive as low-energy excitations
in the parameter region between the first and second phase transition points. From the above
consideration, it is expected that the second phase transition is a very specific one to the U V
model and it does not exist in the ordinary s-wave SC.
5. Discussion
In this paper, we have derived the effective gauge model of the Cooper-pair field sitting on
links from the microscopic quantum Hamiltonian by the path-integral methods. This lattice GL
theory for unconventional SC is a lattice gauge model with dual gauge fields. There the electromagnetic field Ux, is a noncompact U(1) gauge field, whereas the link Higgs field Vx,i is a
compact U(1) gauge field. Then we studied phase structure of the model mostly focusing on the
QPT at T = 0 by means of the MC simulations. For the (2 + 1)D case, we found that there exists
a second-order phase transition for the normal to SC phases as the coupling constant g controlling quantum fluctuations is increased. The gauge-invariant gauge-boson mass start to develop at
319
the critical coupling obtained by the measurement of the specific heat C. On the other hand for
the (3 + 1)D case, we found two second-order phase transitions. From the results of the measurement of MG and the monopole densities U (V ) , we identify the first one as the normal-SC
phase transition and the second one as the transition of V -monopole suppression.
It is quite instructive to compare the above results to those obtained in our previous papers,
in which some related models are studied. In Ref. [2], we investigated the compact U(1) U V
model in 3 and (3 + 1)D in which Ux, is the compact U(1) gauge field. In particular in Section 6
of Ref. [2], we studied the QPT of the model and found that a first-order phase transition from
the confinement to Higgs (SC) phases occurs. On the other hand in the present paper, we found
two second-order phase transitions for the noncompact U V model in similar parameter regions
with the above. This result can be understood as follows; in the compact U V model, the U and
V -fields are (almost) symmetric though there is no time component of V -field. The monopole
proliferation transition of the U - and V -gauge fields takes place simultaneously as actually observed in Ref. [2]. On the other hand in the noncompact case, there is the asymmetry between
the U and V gauge fields. Therefore roughly speaking, the two second-order phase transitions
in the noncompact case coincide in the compact case and as a result a first-order phase transition
appears. Phenomena similar to the above are expected to occur rather generally. In fact in the
very recent paper [11], we studied multi-flavor Higgs U(1) lattice gauge models in various parameter regions and found that when two second-order phase transitions coincide at certain points
in the parameter region of the model, a single first-order transition appears. One may ask if there
exist parameters controlling the distance between the two second-order transitions in the present
model. We expect that the coefficient c3 in the action (2.27) controls the distance, e.g., the two
peaks get closer for larger value of c3 , as the topologically nontrivial excitations of the V -field
are suppressed for large c3 . This problem is under study and the result will be reported in a future
publication.
Preliminary result of the numerical study on this problem seems to support the above expectation. Furthermore, we found that the relative magnitude of two peaks in C changes as the value
of c3 is varied. For larger c3 , the second peak is getting larger. If the two second-order transitions
merge and as a result a single first-order phase transition appears as c3 is getting large, interesting
behavior of the amplitude of V -field v |Vx,j | is expected to be observed. From the behavior
of V in Figs. 9 and 10, it is expected that for g< gc the potential term in SGL , v 2 + v 4 ,
determines the expectation value of v as v 2

because the phase degrees of freedom of

Vx,j fluctuate strongly. As g is getting larger than gc , the other terms in SGL start to contribute
to v and as a result v is a sharp increasing function of g near g gc . At the first-order phase
transition point, on the other hand, v is expected to exhibit a sharp discontinuity and have a
hysteresis loop as V does. In any case, the above interpretation of the structure of the phase
transitions in the U V models is useful and also interesting.
In Ref. [3], we studied thermal phase transition of the noncompact U V model in 3D. There
not only the electromagnetic gauge field Ux,i but also the Cooper-pair field Vx,i are put on all
links of the cubic lattice, i.e., the direction index i = 1, 2 and 3. We found first-order phase
transitions from the normal to SC phases in the parameter regions c2 > 0, d2 = 0 and c2 = 0,
d2 < 0. As the results obtained in the present paper for c2 > 0, d2 < 0 show the existence of
the second-order QPT in 3D, it is possible that the thermal phase transition becomes of second
order in the parameter region like c2 > 0, d2 < 0, i.e., there exist tricritical points. This problem
is under study and results will be reported in a future publication.
320
Let us turn to the discussion on the QPT of the high-Tc cuprates given in the previous work.
In Ref. [12], the QPT of the cuprates was studied by assuming that the critical behavior can be
described by a 3D quantum XY model whose Hamiltonian is given by
HXY =
1
1
n x Vx,y n y
J0 cos(x+i x ),
2
2
x,i
(5.1)
x,i
where n x is the number operator conjugate to the phase of the Cooper pair x on the site x of
the cubic lattice, [n x , y ] = ix,y . Detailed investigation on the model (5.1) was given for the
short-range interaction case, Vx,y = V0 x,y , and an effective GL theory in terms of the order
parameter x eix was derived. Near the quantum critical point, the superfluid density s was
obtained as follows by means of a mean-field approximation neglecting fluctuations of x ,

8 V0 2
(J0 V0 /4d).
s =
(5.2)
7 J0
The critical point is identified as

V0
= 4d,
J0 c
(5.3)
and for VJ00 < 4d ( VJ00 > 4d), the system is in the SC (normal or insulating) phase. Though we
studied the case of the long-range interactions Eq. (2.3) in this paper, we can qualitatively expect
to the SC QPT is consisthe relation like V0 1/g, J0 g. Then the above result concerning
tent with that obtained

in
this
paper.
Furthermore
as
M
1/
in
the SC phase, Eq. (5.2)
G
s
predicts MG (V0 /J0 )c (V0 /J0 ) g gc , where gc is the critical coupling constant. The
numerical results obtained in this paper qualitatively coincide with this behavior of MG and are
consistent with the experiments [13], though in our model the Cooper pair is put on lattice links
instead of sites and behaves like a compact gauge field.
Appendix A. Derivation of the GL theory
In this appendix, we shall prove Eqs. (2.11) and (2.13). We first prove Eq. (2.11). To this end,
let us consider the following quantity,

1
1
HA = dk d 2 e2 V (k) 1 A 0 (k)A 0 (k) + 2
( x,i eAx,0 )( x,j eAx,0 )
e
x,i,j

1
1
dk
=
+ 1 A 0 (k)A 0 (k)
e2 V (k) 1
2
i,j

2
1
i (k)A0 (k) + 2 i (k)j (k)

e
e

1
1
1
dk
=
i (k) A0 (k)
j (k)
(k) A0 (k)
2
e (k)
e (k)
i,j

+ V 1 (k) i (k) j (k) ,
(A.1)
where (k) = [1 e2 V 1 (k)]1 . Therefore

eS = [dA0 ]e d HA .
321
(A.2)
Next we shall prove Eq. (2.13). To this end, let us consider the following integral,

1 1 iA
J e Vx,i Vx+j,i Vx,i eix,i + c.c.
HV =
2
x,i =j

1 1 iA
=
J e Vx,i J eiA eix+j,i Vx+j,i J eiA eix,i
2
x,i =j

J ei(x,i x+j,i A) + c.c. ,
where A Ax+i,j + Ax,j . Therefore

d Ht
e
=
dV dV e d HV .
(A.3)
(A.4)
References
[1] See for example, V.P. Minnev, K.V. Samokhin, Introduction to Unconventional Superconductivity, Gordon and
Breach, 1999.
[2] T. Ono, Y. Moribe, S. Takashima, I. Ichinose, T. Matsui, K. Sakakibara, Nucl. Phys. B 764 (2007) 168.
[3] T. Ono, I. Ichinose, Phys. Rev. B 74 (2006) 212503.
[4] L.P. Gorikov, JETP 9 (1959) 1364.
[5] For detailed derivation of the GL theory from the tJ model see, I. Ichinose, T. Matsui, Phys. Rev. B 45 (1992)
9976.
[6] S. Doniach, Phys. Rev. B 24 (1981) 5063;
M.P.A. Fisher, G. Grinstein, Phys. Rev. Lett. 60 (1988) 208.
[7] For practical methods of the numerical simulations, see for example, J.M. Thijssen, Computational Physics, Cambridge Univ. Press, 1999.
[8] S. Takashima, I. Ichinose, T. Matsui, Phys. Rev. B 72 (2005) 075112.
[9] See for example, D.R. Tilley, J. Tilley, Superfluidity and Supeconductivity, second ed., Adam Hilger, Bristol, 1986.
[10] T.A. DeGrand, D. Toussaint, Phys. Rev. D 22 (1980) 2478;
R.J. Wensley, J.D. Stack, Phys. Rev. Lett. 63 (1989) 1764.
[11] T. Ono, I. Ichinose, T. Matsui, arXiv: 0704.1323.
[12] M. Franz, A.P. Iyengar, Phys. Rev. Lett. 96 (2006) 047007.
[13] See for example, D.M. Broun, P.J. Turner, W.A. Huttema, S. zcan, B. Morgan, R. Liang, W.N. Hardy, D.A. Bonn,
cond-mat/0509223.

Nucl - Phys.B v.785

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nucl - Phys.B v.785

Uploaded by

Copyright:

Available Formats

Nuclear Physics B 785 (2007) 118

The strong coupling limit of the scaling function from

Received 25 May 2007; accepted 18 June 2007

E-mail address: kristjan@nbi.dk (C. Kristjansen).

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

Expanding for large z we get

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

coefficient from the large- expansion (3). As we shall

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

(xk , xj ) = ei(xk ,xj ) ,

where the charges qr (x) (with r 2) are defined by

r 1 (xk )r1 (xk )r1

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

and read [20,21]

and finally the string energy is then obtained as

We then rescale the variables x = x(u) and g 2 in the following way

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

and we assume that G(x) has a well-defined expansion in

and the string energy, E, is encoded in (y) in the following way

where G (x) = G (x),

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

The endpoints of C+ , a and b, are determined by

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

and from the second boundary equation (53) we find

and therefore in the limit (58) we have

x 2 y 2 (y 2 a 2 )1/2 (b2 y 2 )1/2

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

we can also write

where now the normalization condition reads

We also note the formula

Taking the limit (4), we get

5. The one-loop string energy

with V0 (x) given by Eq. (48) and with

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

5.1. The spin chain finite size correction

5.2. The HL phase

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

This integration is carried out in Appendix A. The result reads

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

is due to a highly non-trivial cancellation between terms

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

Fig. 1. Cuts on the complex plane in the integration of B(, ).

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

Finally, putting everything together we end up with

P.Y. Casteill, C. Kristjansen / Nuclear Physics B 785 (2007) 118

[26] I.K. Kostov, Mod. Phys. Lett. A 4 (1989) 217.

Nuclear Physics B 785 (2007) 1933

Screening masses in the SU(3) pure gauge theory

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

2. Screening masses from Polyakov loop correlators

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

R. Falcone et al. / Nuclear Physics B 785 (2007) 1933

be compared with the corresponding determinations on a 4 82 16 lattice of Ref. [35], which