You are on page 1of 5

*Corresponding author.

E-mail address: nbig@lle.rochester.edu (N.P. Bigelow)


Physica B 280 (2000) 27}31
Complex quantum gases: spinor Bose}Einstein condensates of
trapped atomic vapors
H. Pu, C.K. Law', N.P. Bigelow*
Department of Physics and Astronomy and Laboratory for Laser Energetics, The University of Rochester, Rochester, NY 14627, USA
'Department of Physics, The Chinese University of Hong Kong, Shatin, Hong Kong, China
Abstract
We discuss the energy eigenstates, ground and spin mixing dynamics of a spin-1 spinor Bose}Einstein condensate for
a dilute atomic vapor con"ned in an optical trap. Our results go beyond the mean "eld picture and are developed within
a fully quantized framework. 2000 Elsevier Science B.V. All rights reserved.
Keywords: Bose}Einstein condensation; Spinor condensates
1. Introduction
A Bose}Einstein condensate (BEC) is a fundamental
quantum state of simple Bose particles in which there is
a macroscopic occupation of a single quantum state of
the system [1]. One important description of the BEC is
in terms of a single many-particle wave function, or order
parameter. When the constituent bosons have internal
degrees of freedom, such as spin, the quantum state and
its properties becomes more complex and can display
a rich spectrum of interesting behavior. In the "eld of
low-temperature physics, a remarkable example of such
richness can be seen in the elegant system of super#uid
`He [2]. More recently, a new example has appeared
with the realization of an optically trapped BEC of so-
dium atoms in an F"1 electronic state [3,4]. In the
optical trap, the spin components m
'
"0, $1 are free to
evolve as the atoms interact and spin exchange e!ects
give rise to complex evolution and non-trivial ground
state properties. What has made the alkali spinor BEC
particularly interesting is that optical and magnetic "elds
can be used to probe and manipulate the system. Fur-
thermore, because the alkali vapor BECs are very dilute,
a theoretical treatment based on measurable funda-
mental atomic parameters is possible, involving no phe-
nomenological parameters.
In the last year, several groups [5}9] have turned their
theoretical e!orts to studying the alkali spinor BEC
problem with a frequent emphasis on the mean-"eld
limit. Our group has treated this problem from a some-
what di!erent point of view [8], centered on the tech-
niques familiar to quantum optics and using both
semi-classical and fully quantum frameworks. In this
paper we present new results on the energy eigenstates
and the spin mixing dynamics for a spin-1 spinor BEC.
We emphasize a fully quantum treatment and relate our
results to those found using a semi-classical, mean-"eld
approach.
2. The Hamiltonian and its energy eigenstates
Assuming that the interaction between atoms
preserves angular momentum and rotation symmetry in
hyper"ne spin space [5], the second quantized Hamil-
tonian for a dilute gas systemis given by H"H
`
#H
^
where
H
`
"
?

dr
?
(#<
'
)
?
#
z
'
2

?@

dr
?

@
, (1)
0921-4526/00/$- see front matter 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 1 - 4 5 2 6 ( 9 9 ) 0 1 4 2 9 - 5
H
^
"
z

2
dr(

#2

"

"
#2

"

"
!2

#2
"

"

#2

"

"
). (2)
Here
H
( j"0,$1) is the "eld operator for the jth spin
component, is the kinetic energy operator and <
'
is
the trapping potential which is assumed to be the same
for all three components. We have used z
'
and z

to
denote the coupling strengths. The separation of H
`
and
H
^
in the full Hamiltonian is useful because in most
systems z
'
<z

. Note that the spin-mixing interaction


which is the focus of this paper is described by H
^
.
We introduce angular momentum (magnetization)
density operators de"ned by: M
>
"(2[

"
#

"

], M

"(2[

"
#
"

] and M
X
"

. They satisfy the angular commuta-


tion relation: [M
X
(r), M

(r')]"$o(r!r')M

(r) and
[M
>
(r), M

(r')]"2o(r!r')M
X
(r). In this way we can
rewrite H
^
in a concise form
H
^
"
z

2
[M`!2N] dr, (3)
where N"

#
"

"
#

is the number
density operator which commutes with M's.
The angular momentum represented in Eq. (3) tells us
about the ground state structure of the condensate gener-
ated by the spin-mixing interaction H
^
. To see this we
assume that at zero temperature most atoms occupy
a single spatial mode with structure (r) and hence
that the "eld operators are e!ectively described within
a single mode approximation,

G
+a(
G
. (4)
Here a(
G
is an annihilation operator which obeys the
usual boson commutation relations. In Eq. (4), (r) is the
ground state solution of the Gross}Pitaevskii equation,
(#<
'
#z
'
N`)"j resulting from H
`
alone,
with N being the total particle number in the condensate
and j being the chemical potential. Such a single mode
approximation is valid when z
'
<z

(which is the case


for both sodium and rubidium atoms [5]) such that
H
`
is the dominant part of the full Hamiltonian [11].
Using Eq. (4), the Hamiltonian is simpli"ed as a problem
of three non-linear coupled oscillators [10]. H
`
and
H
^
now become
H
'
"jNK!z'
'
NK(NK!1), (5)
H

"z'

(K`!2NK). (6)
Here z'
H
,(z
H
/2)j dr(x)" ( j"a, s), NK is the total num-
ber operator, and is the angular momentum operator
de"ned by: K
>
,(2(a(

a(
"
#a(
"
a(

), K

,(2(a(
"
a(

#
a(

a(
"
) and K
X
,(a(

a(

!a(

a(

). We see that the


angular momentum structure of H
^
is still preserved in
the single-mode approximation, i.e., with M` replaced by
K`. Since the Hamiltonian conserves the total number of
particles, H
'
is a constant operator. Hence, we need only
consider H

to discuss the ground state of the condensate.


The energy eigenstates of the single-mode Hamiltonian
are angular momentum eigenstates of K` and K
X
, i.e.,
K`l, m,"l(l#1)l, m, with l"0, 2, 4,
2
, N if N is even,
and l"1, 3, 5,
2
, N if N is odd. The energy spectrum of
H

, E'J'

"z'

[l(l#1)!2N], indicates that the level spac-


ings are of the order z'

near l"0, and Nz'

near l"N. In
the case of z'

'0 (such as sodium), l"0, m"0, is the


ground state with no degeneracy. On the contrary, if
z'

(0 there are 2N#1 degenerate ground states


l"N, m"!N,!N#1,
2
, N,. In Ref. [8], we have
reported that all l"N, m, have sub-Poisson number
#uctuations in each component. This is in contrast to the
case of l"0, m"0, which has super-Poisson number
#uctuations. We note that the super-Poisson nature in
l"0, m"0, is quite familiar in the language of quan-
tum optics [8]. More recently, Ho and Yip [12] have
discussed this result from a viewpoint more familiar to
the condensed matter community and described this
property as an unexpected signature of a &super-frag-
mented' ground state of the spinor condensate.
Using the Fock states n

, n
"
, n

, de"ned by the
number operators n(
H
,a(
H
a(
H
for the three spin compo-
nents (i.e., n(
H
n

, n
"
, n

,"n
H
n

, n
"
, n

,), all the zero


m angular momentum states l, m"0, have the form
l, m"0,"
',`'

I"
AJ
I
k, N!2k, k,. (7)
In Fig. 1, we show the distribution AJ
I
for various
values of l. The states with low l share a similar feature in
that AJ
I
contains a wide range of k values. One interesting
feature is that AJ
I
exhibits rapid oscillations with k as
l increases. However, AJ
I
can become smooth and local-
ized as l approaches its maximum value N. Note how as
l increases the number #uctuations change from super-
Poissonian to sub-Poissonian.
We point out that most of the BEC angular mo-
mentum states are entangled states meaning that they
cannot be factorized into products of wave functions
associated each individual spin component. This is in
contrast to mean-"eld theory where each individual com-
ponent is assumed to be in a coherent state. In the "eld of
quantum information theory, entanglement of a pure
state can be measured by the entanglement entropy,
S,!Tr(j
H
ln j
H
) where j
H
is the reduced density matrix
of the sub-system j. Such an entropy should be distin-
guished from the familiar thermodynamic entropy. As an
example, the entropy of l, m"0, is given by
S"!
',`'

I"
AJ
I
` lnAJ
I
`. (8)
28 H. Pu et al. / Physica B 280 (2000) 27}31
Fig. 1. Fock state amplitudes associated with zero m angular
momentum states for N"1000.
The larger the S, the stronger the entanglement. Given
a "xed N, we "nd that l"0, m"0, has the largest
entropy among all of the available angular momentum
states. For example, l"0, m"0, has S"10.5 for a sys-
tem with N"10` atoms. This is close to the maximum
entropy S
`
"10.8 of a maximally entangled state
which is the state with equal amplitudes for all
k, N!2k, k, (assuming "xed N). In fact, we "nd that
S for l"0, m"0, approaches S
`
as N goes to in"nity.
Therefore, the ground state of a spinor condensate can
well be considered as a physically realizable example of
high quantum entanglement of three coupled systems.
Whether this is useful for quantum information applica-
tions is an open topic for further studies, but we point out
that S can well be an alternative indicator of fragmenta-
tion of a spinor condensate.
3. The spin-mixing dynamics
Having discussed the Hamiltonian and the energy
eigenstates of the spinor condensate, we now turn to the
spin-mixing dynamics of the system. In an earlier work,
we investigated the dynamics of the spinor condensate
assuming that it was initially in a Fock state. In the
present paper we emphasize instead the phase relation-
ship among di!erent spin states. For this purpose, we
choose an initial state given by
(0)"
1
(N!

Ca
"
#
S
(2
(a

#a

,
vac,, (9)
where C`#S`"1. Without loss of generality, we
choose S"S and C"Ce'F`. Here, 0 may be regarded
as the relative phase among individual spin states of the
spinor condensate. This state (9), can be experimentally
prepared, for example, by "rst preparing a BEC with all
atoms in the spin-0 state, then applying appropriate
Raman pulses to transfer equal number of particles into
spin-1 and spin-(!1) states. The phase 0 can be adjusted
by engineering the phases of Raman pulses.
Fig. 2 shows the population evolution of the system for
N"100. The initial populations in spin-0 and spin-($1)
are C` and S`/2, respectively. Under the spin-exchange
interaction, we see that the population in each individual
spin states experience collapse and revivals. The revivals
occur at z'

t"n/4 (n"1, 2, 3,
2
), hence the revival time
depends on the total particle number N through z'

. The
collapse is more complex. In the initial collapse regime,
the population oscillates with a frequency given by
f"4N(P
"
(1!P
"
) cos(0/2) (10)
with P
"
"C` being the initial population in spin-0. We
note that the same frequency appears in the semi-classi-
cal treatment where we regard each spin state being an
independent coherent state. However, in that case, the
population oscillation is not damped, whereas here it is
modulated by a damping envelope found to be Gaussian
in form: eR
`
R
`

. The collapse time t

depends on the value


of N and 0. z'

scales as 1/(N. A plot of z'

versus 0 is
presented in Fig. 3. Therefore, we point out that by
studying the dynamics of the population evolution, one
may determine not only the particle number, but also
have access to the relative phase of the spinor conden-
sate. Such a measurement would provide valuable in-
formation on quantum phase of the condensate, a subject
which is both interesting and controversial.
We also "nd that both the population oscillation am-
plitudes and t

are sensitive to the initial population P


"
[P

"(1!P
"
)/2]. For example, when 0"0, as P
"
ap-
proaches
`
, the oscillation amplitudes approach 0, while
t

PR. When P
"
"
`
, we found that the system does
not exhibit spin-mixing behavior. A closer examination
shows that under such conditions, the initial state (9) can
H. Pu et al. / Physica B 280 (2000) 27}31 29
Fig. 2. Populations as functions of time. The parameters for the
initial state are: (a) C`"
`
and 0"0; (b) C`"
`
and 0"/2;
(c) C`"
`
and 0"0. For all the "gures shown, N"100.
Fig. 3. Collapse time t

as a function of 0 for C`"


`
and
N"100.
The application of a magnetic "eld B introduces an extra
term into the Hamiltonian: H
"
"B) K. Such an interaction
does not change the total angular momentum l, however, it can
change the value of m if transverse magnetic "eld is present. For
a semi-classical treatment of the magnetic "eld e!ects on spinor
condensate (see Ref. [13]).
be expressed as
(0)"
,

J
X
,
C
J
X
l"N, m"l
X
,, (11)
i.e., it is a superposition of the angular momentum states
with total angular momentum l"N. Since such a state is
an eigenstate of K` (hence also H

), is a non-mixing
stationary state.
Similar non-mixing states were also found by us in
a semi-classical treatment, where the non-mixing states
were identi"ed as the eigenstates of the semi-classical
Hamiltonian for the spinor condensate. These non-mix-
ing states are particularly interesting because they repre-
sent situations in which the nonlinear atom}atom
interaction does not a!ect the condensate spin dynamics
and only appears as a constant energy shift. In addition,
if the condensate is prepared in one of these non-mixing
eigenstates, the spinor condensate responds linearly to
external magnetic "elds. This suggests that despite the
intrinsic nonlinearity of the condensate interactions, one
may realize macroscopic quantum control of the spinor
BEC, an interesting subject in the context of quantum
computing and quantum information. For example, one
can "rst prepare all the population in the spin-1 state,
which, in terms of the angular momentum state, corre-
sponds to l"N, m"!N,. Then, by applying appro-
priately engineered magnetic "elds, one may create an
arbitrary superposition of states with l"N, yet which
have di!erent m [13]. Recently, generation of arbitrary
superposition states involving ground state Zeeman sub-
levels of non-degenerate atomic Cs vapors was success-
fully demonstrated in Heinzen's group [14]. In their
experiment, the total states available is determined by the
number of Zeeman sublevels (in their case, the number of
Zeeman sublevels is 9 for the F"4 Cs ground state). If
we apply the similar approach to the angular momentum
states l, m, of the spinor condensate, the number of
available states can be on the order of total particle
number of the condensate, which in current experiments,
can be as large as tens of millions. Therefore, spinor
condensates can represent a very large Hilbert space and
hence have the potential to store enormously large
amount of quantum information. We thus believe that
spinor condensates may be a new and very interesting
platform suitable for quantum computing. More inves-
tigations along this line are underway.
30 H. Pu et al. / Physica B 280 (2000) 27}31
4. Summary
In summary we have described a fully quantum mech-
anical treatment of the spin-1 spinor Bose condensate.
Using an angular momentum algebra the energy eigen-
states were derived and discussed. These eigenstates are
identi"ed as collective spin states which are most conve-
niently represented by the angular momentum states
l, m, which are the simultaneous eigenvectors for K` and
K
X
. We have also described the spin-mixing dynamics of
the system initially prepared in a speci"c state selected
because it has a well-de"ned relative phase among indi-
vidual spin components and because this state can be
created and controlled experimentally. Finally, we have
pointed out the possibility of using spinor condensate as
a platform for quantum information storage and quan-
tum computing.
Acknowledgements
We thank S. Raghavan and Professors G. Shlyapnikov
and T.-L. Ho for many useful discussions. This research
was supported by NSF Grant No. PHY-9457897, and by
the David and Lucile Packard Foundation. C.K.L. ac-
knowledges the support from the Chinese University of
Hong Kong Direct Grant No. 2060148.
References
[1] K. Huang, Statistical Mechanics, Wiley, New York, 1987.
[2] D.M. Lee, R.C. Richardson, D.D. Oshero!, Rev. Mod.
Phys. 69 (1997) 645.
[3] D.M. Stamper-Kurn, M.R. Andrews, A.P. Chikkatur,
S. Inouye, H.-J. Miesner, J. Stenger, W. Ketterle, Phys.
Rev. Lett. 80 (1998) 2027.
[4] J. Stenger, S. Inouye, D.M. Stamper-Kurn, H.-J. Miesner,
A.P. Chikkatur, W. Ketterle, Nature (London) 396 (1998)
345.
[5] T.-L. Ho, Phys. Rev. Lett. 81 (1998) 742.
[6] T. Ohmi, K. Machida, J. Phys. Soc. Japan 67 (1998)
1822.
[7] Elena V. Goldstein, P. Meystre, Phys. Rev. A 59 (1999)
1509.
[8] C.K. Law, H. Pu, N.P. Bigelow, Phys. Rev. Lett. 81 (1998)
5257.
[9] W.-J. Huang, S.-C. Gou, Phys. Rev. A 59 (1999) 4608.
[10] L. Wang, R.R. Puri, J.H. Eberly, Phys. Rev. A 46 (1992)
7192.
[11] H. Pu, C.K. Law, S. Raghavan, J.H. Eberly, N.P. Bigelow,
Phys. Rev. A, in press.
[12] T.-L. Ho, S.K. Yip, preprint.
[13] H. Pu, S. Raghavan, N.P. Bigelow, submitted for
publication.
[14] D.J. Heinzen, G. Xu, Quantum Electronics and Laser
Science '99, OSA 1999 Technical Digest Series Conference
Edition, 1999, p. 177.
H. Pu et al. / Physica B 280 (2000) 27}31 31

You might also like