Professional Documents
Culture Documents
HW ASSIGNMENT #1
(1) To measure the heat capacity of an object, all you usually have to do is put it in thermal
contact with another object whose heat capacity is known. Suppose that a 100 g chunk of
metal is immersed in boiling water (100 C). After a time, the metal is removed and quickly
transferred to a Styrofoam cup containing 250 g of water at 20 C. After a little while, the
temperature of the contents of the cup is found to be 24 C. You may assume that the heat
transferred from contents of the cup to its surroundings, and the heat capacity of the cup
itself are both negligible.
(a) How much heat is gained by the water in the Styrofoam cup?
(b) How much heat is lost by the metal?
(c) What is the heat capacity of the metal?
(d) What is the specific heat (in J/g K) of the metal?
(2) The heat capacity of Albertsons Rotini Tricolore is approximately 1.8 J/g C. Suppose
you toss 340 g of pasta at 15 C into 1.5 liters of boiling water. What effect does this have
on the temperature of the water, before the stove has time to provide any more heat?
Hint: You will need to look up some constants in order to solve this problem. Part of the
problem is in understanding what constants you need. Once you know, try Google.
p+
a 2
V2
V b = RT
where = N/NA is the number of moles. Find expressions for each of the following
quantities in terms of the specific volume v = V /, the temperature T , and constants.
Dont confuse the symbols and v!
(a)
p
V
(b)
T
p
T
(c)
V
1
V
V
T
(4) A peculiar molecule discovered in Santa Barbara, consisting of uranium, carbon, and
antimony (UCSb) is found in its gas phase to obey the modified van der Waals equation,
p (V b) = RT .
From calorimetry, the energy is determined to be
E(T, V, ) = 0 eT /T0 ,
which is volume-independent, with 0 = 1000 J/mol and T0 = 300 K.
(a) Find an expression for the molar specific heat cp .
(b) How much heat energy Q is required to raise = 5 moles of the UCSb gas from
TA = 300 K to TB = 600 K and at constant pressure?
(c) (Challenging!) Find an expression for the adiabatic speed of sound cS (v, T ) in terms
of the molar volume v = V /, the temperature T , and other constants (including the
molecular mass M ). For a discussion of sound in fluids (and gases), see 14.5.5 of my
lecture notes from Physics 110B (still available online).
(5) Which of the following quantities is a state function for driving? Explain your answers.
(a) travel time
(b) elevation
(c) aggravation
(6) For each of the following differentials, identify whether it is exact or inexact. If you
can, find the integrating factor L for which dU = eL dW
is exact.
(a) dW
= 2x2 y dx + x3 dy
(b) dW
=
y2 z
(x+y)2
dx +
x2 z
(x+y)2
(c) dW
= exy dx + 2x +
x
y
dy +
exy
xy
x+y
dz
dy
Qm = Qw = 4.2 ! 10 3 J
Cm =
Qm
= 55.3 J / K
T0 ! T1
Cm
= 0.55 J / g ! K
m
Problem 2.
Solution:
When you toss pasta into boiling water, there will be heat transferring from boiling water to
pasta. In this process, the total thermal energy should be conserved.
! TFinal = 92.5 o C
So if the stove provides no heat to the system before thermal equilibrium, the final
o
Problem 3.
Solution:
(a) % p +
$
,$ #p ' 2a+ 2 /
a" 2 &
a+ 2 '
(V
)
"
b)
=
"
RT
"
*
(V
*
+
b)
+
p
+
&
)
(
&
)=0
.
1
V2 '
V3 0
V2 (
-% #V (T
%
$ #p '
+RT
2a+ 2
"& ) =*
+
% #V (T
(V * +b) 2
V3
uRT
2a
' (p $
) !V %
! 2
" =
2
u
& (V #T (u ! b)
!
#
(b) % p +
$ #p '
a" 2 &
(V ) "b) = "RT " & ) (V * +b) = +R
2 (
% #T (V
V '
$ #p '
R
"& ) =
% #T (V (u * b)
!
#
(c) % p +
&
a" 2 &
2a$ 2 & %V )
a$ 2 )& %V )
(V
)
"
b)
=
"
RT
"
#
(V
#
$
b)
+
p
+
(
+
(
(
+( + = $R
V2 '
V 3 ' %T * p
V 2 *' %T * p
'
!
2a$ 2 & %V )
$RT & %V )
" # 3 ( + (V # $b) +
( + = $R
V ' %T * p
V # $b ' %T * p
"
!
where u =
1 $ #V '
R
& ) =
V % #T ( p 2ab * 2a + RTu
u2
u u*b
V
is the molar volume.
v
!
Problem 4.
! Solution:
# "E &
( +
$ "T ' p
(a) C p = %
! cp =
# "V & )*
p% ( = 0 eT /T0 + )R
$ "T ' p T0
" 0 T / T0
e
+R
T0
# "V &
( can be obtained by the similar method employed in Problem 3.
$ "T ' p
The term p%
TB
(b) Q =
# "c
TA
TB
p
dT =
% "$0
# '& T
TA
(
eT /T0 + "R*dT = E(TB ,V," ) + E(TA ,V," ) + "R(TB + TA )
)
= 3.58 !10 4 J
!
$ "p '
, ! is density defined by M /V = M 0 /u , M 0 is molar mass.
)
% "# ( dQ= 0
(c) c s = &
c s2 =
"p "u
u 2 % "p (
=$
' *
"u "#
M 0 & "u ) dQ= 0
(1)
" dT = #
pT0 #T /T0
e
du
$0
(2)
(u " b)dp + pdu = RdT , then substituating (2) into this expression, we get:
dp = "
!
(3)
c s2 =
2
RT # u & # RT0 "T /T0 &
e
%
( %1+
(
M0 $ u " b ' $
)0
'
Problem 5.
!
Solution:
(a) Travel time is not a state function because it depends on the path taken and not on the
endpoints alone.
(b) Elevation is a state function.
Problem 6.
Solution:
For a differential as: dW =
! A dx
i
!Ai !A j
=
!x j
!xi
(1)
dW = 2 x 2 ydx + x 3 dy
!Ay
!Ax
= 2 x 2 " 3x 2 =
!y
!x
! Inexact differential.
It is easy to find if we employ a factor
1
, the new differential becomes:
x
dU = 2 xydx + x 2 dy = d ( x 2 y )
Now it is an exact differental. L = ln x
(2)
dW =
y2z
x2 z
xy
dx
+
dy +
dz
2
2
x+ y
( x + y)
( x + y)
" xyz %
'
# x + y&
(3)
x
dW = e xy dx + (2 x + e xy )dy
y
!
Obviously, it is an inexact differential.
(1) One mole of a diatomic ( = 75 ) ideal gas is driven along a cycle depicted in fig. 1.
Figure 1: A three-stage cycle consisting of an isotherm (AB), an isobar (BC), and an isochore
(CA).
(2) Consider the Diesel cycle of fig. 2. Define the compression ratio r = VB /VD and the cutoff
ratio s = VA /VD . Derive a formula for the efficiency = W/QDA for an ideal gas Diesel
cycle in terms of r, s, and the ideal gas parameter = 1 + f2 . Compare your result with the
efficiency of an Otto cycle with the same compression ratio, and show that the Otto cycle
has a greater efficiency. However, an optimized Otto cycle with real fuel has a compression
ratio of about r = 8. Compare its efficiency with that of a Diesel cycle with r = 18 and
s = 2.
(3) Scientists discover a novel extraterrestrial molecule, illudium phosdex, in the Murchison meteorite. They are overjoyed. The thermodynamics of this new material follow from
the relation
a S5
E(S, V, N ) = 2 2 .
V N
S
N kB
(5) Challenging! Consider the modified ideal gas Carnot cycle in fig. 3. The upper isotherm
AB is broken into two separate stages. AA is an adiabatic free expansion, and all other
stages are quasistatic. Define the parameter
x=
VA VA
.
VB VA
When x = 0, the entire cycle is reversible, and when x = 1, the entire upper isotherm is
taken up by the adiabatic free expansion. Also define the expansion ratio
r=
VB
.
VA
Figure 3: In this modified Carnot cycle, the upper isotherm is itself broken into two stages.
AA is an adiabatic free expansion, and A B is a quasistatic isotherm. Stages BC, CD, and
DA are all quasistatic.
"RTB
= 1.04 ! 106 Pa
VB
PB =
(b) We can compute the area under the curve to get the work in this process.
B
VB
V
#RT A
dV ! #RTB (1 ! C )
V
VB
VA
"
' .V +
V $
= /RTB &ln,, B )) ( 1 + C # = 2.64 ! 10 3 J
VB "
% - VA *
C
dQ
dE + PdV
(c) "S = ! dS = !
=!
T
T
B
B
B
where E =
PV
, take it into (1), we get:
" !1
PdV + VdP P
$
dV
%R dP
+ dV = !
%R
+
($ " 1)T
T
$ "1
V
$ "1 P
B
B
#S = !
=
'V
)
(R ln%% C
) !1
& VB
Problem 2. Solution:
$
"" = !40.3 J / K
#
(1)
First of all, lets consider the work during this period, i.e. S ABCD
B
D
PAV A#
PAVD#
= ! # dV + ! # dV + PA (V A " VD )
A V
C V
!
1
PA (V A " VD ) "
PAVB1"! (V A! " VD! )
! "1
! "1
Secondly, in this period, the ideal gas absorb heat during D->A, therefore:
A
A
A
' PV $
)
"" + PA (V A ! VD ) =
Q = ( dQ = ( dE + PdV = ( d %%
PA (V A ! VD )
) !1#
) !1
D
D
D &
Finally, we get:
W
1 r 1!" ( s " ! 1)
#=
= 1!
, where s and r have been defined in the problem.
Q
"
s !1
'V
Otto cycle has the efficiency of ) = 1 ! %% min
& Vmax
$
""
#
( !1
If r=18, s=2, ! Diesel = 81% < ! Otta = 85% , however an optimized Otto Cycle has
the parameter of r=8, ! max ofOtta = 75% , which is less than the efficiency in Diesel
cycle.
Problem 3. Solution:
(a) The unit of S is J/K, V is m 3 and E is J, therefore, a should has the unit of:
m6 K 5 / J 4 .
aS 5
5aS 4
2aS 5
(b) E = 2 2 " dE = 2 2 dS ! 3 2 dV
V N
V N
V N
(1)
T=
(2)
5aS 4
2aS 5
P
=
,
V 2N 2
V 3N 2
&V #
P$ !
%N"
1/ 2
5 5 / 4 a 1 / 4 = 2T 5 / 4
PV 1 / 2 = C = const.
Therefore, the work in this isotherm process is:
C
W = ! PdV = ! 1 / 2 dV = 2C (V f1 / 2 " Vi1 / 2 )
V
(d) Always be aware of the difference between isotherm process and adiabatic
process!
First, lets rewrite the equation of state as:
16
P 4V 2 a =
N 2T 5
3125
We can then get the differential relation:
16 2 4
4 P 3V 2 adP + 2 P 4VadV =
N T dT
(1)
625
As adiabatic process, dE = ! PdV
(2)
aS 5
(VN )1 / 2 5 / 4
Where E = 2 2 = 5 / 4 1 / 4 T , and P is determined by the equation of state
V N
5 a
as above.
The (2) can give us the relation of dV and dT, dT = !
Invoking this relation to (1), we can get:
(
1 & 'V #
1
=
$
!
V % 'P " dQ =0 3P
Problem 4. Solution:
!"
( )E %
T =&
# = 0
kB
' )S $V
!S
(N%
)
& # exp(
Nk B
'V $
2T
dV
V
( )E %
(N%
P = *&
# = !" 0 & #
' )V $ S
'V $
1+!
exp(
!S
)
Nk B
Therefore, PV = NkT = !RT . ! This system obeys the ideal gas law.
(b) As for adiabatic process, dE = dQ ! PdV = ! PdV = !
Therefore, dE =
Nk B T
dV
V
Nk B T
by using the relation in (a).
!
Nk B
dT , invoking this expression into (1):
!
Nk B
Nk T
dT = ! B dV ! TV " = const.
"
V
1 ( )E %
R
&
# =
" ' )T $V !
1+!
' (V $
c P = cV + p%
R
" =
!
& (T # p
Problem 5. Solution:
(a)
The volume relation should be the same as the normal Carot Cycle. i.e.
dQ
! T = 0 , where only A->B and C->D contribute dQ. Thus,
(1)
nR(T2 ! T1 )
" !1
& VC
% VD
DA: Q DA = 0 ; W DA = !
#
&V #
!! = 'nRT1 ln$$ B !!
"
% VA "
nR(T2 ! T1 )
" !1
&V #
&V #
W = W AA' + W A'B + WBC + WCD + WDA = nRT2 ln$$ B !! ' nRT1 ln$$ B !!
% V A' "
% VA "
By using the defined parameter x and r, we can get:
W = nR(T2 ! T1 ) ln r ! nRT2 ln (1 + (r ! 1) x )
(c)
r 1"T1 / T2 " 1
! 0.26
r "1
&
#
&
#
VB
r
!! = nRT2 ln$$
!! , therefore
(d) Q A'B = nRT2 ln$$
V
+
x
(
V
'
V
)
1
+
x
(
r
'
1
)
%
"
B
A "
% A
"=
T
T
W
ln r
= 1! 1
< 1! 1
Q A' B
T2 ln r ! ln(1 + (r ! 1) x)
T2
T
p
V,N
S
V
p,N
T
V
p,N
S
p
=1.
V,N
(4) Express V
T
V
p & 'T #
V
& 'T #
!! =
, $$
! =
% 'V " p , N nR % 'p "V , N nR
Finally we get:
! (T , S , N )
=1
! ( p ,V , N )
Problem 2. Solution:
!2
=C
& 'V #
($
! = 2CT
% 'T ")
& 'V #
( C) = CV + p$
! = CV + 2 pCT = CV + 2 R
% 'T ")
(b) ! (T , p ) = Te
) ln T +
p / p0
= Const
p
RT
1
R
RT & 'V #
= ln T +
= ln Const. ) +
(
$
! =0
p0
p0V
T p0V p0V 2 % 'T "*
pV
& 'V #
( C) = CV + p$
! = CV + R + 0
T
% 'T ")
3
(c) # (T , p ) = p V = const. ! VT
"3 / 2
=C
3
R
2
Problem 3. Solution:
(a) The extensivity of S imposes a relation of:
S (!E , !V , !N ) = !S ( E ,V , N )
So we can easily get the constraint:
# + " +! =1
(b) Usually the stability of a system depends on the second order differential of a given function,
e.g. potential energy, entropy.
As a standard procedure, we should evaluate
!2S
, in which xi are dimensionless
!xi !x j
& '2S #
! should
$ 'x 'x !
% i j"
be non-positive matrix. In other words, the eigenvalues of the following matrix should be
non-positive.
*)
*( #
& * (* ' 1)
$
!
) ( ) ' 1)
)( !
$ *)
$ *(
)(
( (( ' 1) !"
%
The final answer is $ , # , " ! [0,1]
Problem 4. Solution:
(*)
"p
C
( )S %
( )S %
dS = &
dV
# dT + &
# dV = V dT +
T
!T
' )T $V
' )V $T
"
) *V &
1
) *V &
$$ dp = V" p dT # V! T dp $ dp = p dT #
dV = '
dV
$ dT + ''
!T
V! T
( *T % p
( *p %T
Invoking them into (*), we can get:
1 # ! pT
) *T &
V'
$ =
( *V % H ! + C p" T # T! 2
p
p
V
Problem 5. Solution:
& Vf
% Vi
(a) For isothermal process of ideal gas with fixed N, (F = 'T(S = ' nRT ln$$
#
!! , where
"
!S have obtained in the Problem 1. Or we can directly get the result from the expression:
&V
dF = " SdT " pdV + dN = " pdV ! (F = 'nRT ln$$ f
% Vi
(b) d" = ! SdT ! pdV ! Nd = ! pdV = dF
&V
* () = 'nRT ln$$ f
% Vi
#
!!
"
#
!!
"
(1) The latent heat of vaporization of hexane (C6 H14 ) is = 30.8 kJ/mol. The boiling point
at p = 1.00 atm is T = 68.9 C. Assuming that is roughly constant over this part of the
vaporization curve, what will be boiling point be at a pressure p = 0.50 atm?
SO2
2 SO3 .
What is the maximum number of phases of this system that can coexist at any point
in thermodynamic state space?
(3) A small college dormitory contains five different rooms. Each room houses two persons. Ten students live in the dorm: four physics majors and six chemistry majors.
(a) Suppose the dormitory rules forbid physics and chemistry majors from sharing a
room. How many possible distinct room assignments are there? Note that the rooms
are themselves distinct.
(b) If the rules are relaxed and physics and chemistry majors are allowed to room together, how many possible distinct room assignments are there?
(c) Generalize to the case of M + N rooms, 2M physics majors, and 2N chemistry majors. What is the numbers of possible room assignments W1 (M, N ) assuming the
rule from part (a) is enforced? What is the number of possible room assignments
W2 (M, N ) assuming the rule is relaxed? Writing M = x (M + N ), what is the
change in dimensionless entropy ln W2 ln W1 upon relaxing the rule in the limit
where M + N is large and x is finite?
l
pl
dp l dT
dp
=
2
p R T2
dT TS RT
Suppose l = const. , we can get:
ln
p
l 1 1
= ( ) T = 321.4 K = 48.4 o C
p0
R T T0
Problem 2. Solution:
is equilvalent to
, i.e. one chemical reaction results in one equation which will lead to degree
of freedom decrease by one. Therefore if we have reaction, the degree will decrease by
. As a result, we can just simply revise the existed relation:
d coex = 2 + to d coex = 2 +
(b) d = 2 + , if
= 4 and = 2 , max = 4
Problem 3. Solution:
2
(a) Select two rooms for chemistry C 52 , chemistry students assigned in this two rooms C 4 , the
remained six physics students are assigned in three rooms: C 62 C 42 . There are totally
C 52 C 42 C 62 C 42 = 5400
(b) If we dont distinguish majors, the problem is equilvalent to: How should we assign 10 people
to 5 rooms? Therefore, we can easily write down the answer as: C102 C 82 C 62 C 42 = 113400 .
(c) According to the analysis above, if there are M+N rooms corresponding (a) assignment rule:
( M + N )! (2 M )! (2 N )!
M ! N! 2 M
2N
W2 = C 22( M + N ) C 22( M + N ) 2 L C 22 =
(2 M + 2 N )!
2M +N
Using the formula ln( N !) N (ln N 1) where N is large enough, we can get:
ln W2 ln W1 = M ln
M +N
M +N
+ N ln
M
N
(1) Consider a modified version of the Kac ring model where each spin now exists in one
of three states: A, B, or C. The flippers rotate the internal states cyclically: ABCA.
(a) What is the Poincare recurrence time for this system? Hint: the answer depends on
whether or not the total number of flippers is a multiple of 3.
SOLUTION: If the number of flippers Nf is a multiple of 3, then each spin will have
made an integer number of complete cyclic changes ABCA after one complete
passage around the ring. The recurrence time is then N , where N is the number of
sites. If the number of flippers Nf is not a multiple of 3, then the recurrence time is
simply 3N .
(b) Simulate the system numerically. Choose a ring size on the order of N = 10, 000
and investigate a few flipper densities: x = 0.001, x = 0.01, x = 0.1, x = 0.99.
Remember that the flippers are located randomly at the start, but do not move as the
spins evolve. Starting from a configuration where all the spins are in the A state, plot
the probabilities pA (t), pB (t), and pC (t) versus the discrete time coordinate t, with t
ranging from 0 to the recurrence time. If you can, for each value of x, plot the three
probabilities in different colors or line characteristics (e.g. solid, dotted, dashed) on
the same graph.
SOLUTION: See figs. 1, 2, 3.
(c) Lets call at = pA (t), etc. Explain in words why the Stosszahlansatz results in the equations
at+1 = (1 x) at + x ct
bt+1 = (1 x) bt + x at
ct+1 = (1 x) ct + x bt .
1x
0
x
Q= x
1x
0 .
0
x
1x
P
(d) Show that the total probability is conserved by a Markov process if i Qij = 1 and
verify this is the case for the equations in (c).
P
P
SOLUTION: The total probability is i Pi . Assuming i Qij = 1, we have
X
Pi (t + 1) =
XX
i
Qij Pj (t) =
XX
j
Qij Pj (t) =
Pj (t)
1
3
b b
t
t
,
a
t+1
b
t+1
1
3
ct ct
1
3
a
= R t ,
bt
and find the 2 2 matrix R. Note that this is not a Markov process in A and B,
since total probability for the A and B states is not itself conserved. Show that the
eigenvalues of R form a complex conjugate pair. Find the amplitude and phase of
these eigenvalues. Show that the amplitude never exceeds unity.
SOLUTION: Substituting at = a
t + 31 , etc. into the Markov process and eliminating
ct = a
t + bt , we obtain
R=
1 2x x
x
1x
= 2 (2 3x) + (1 3x + 3x2 ) .
3
2
x.
2
(x) =
ln 1 3x + 3x2
(x) = tan
3x
.
2 3x
Since x(1 x) achieves its maximum volume on the unit interval x [0, 1] at x = 21 ,
where x(1 x) = 41 , we see that 12 |(x)| 1, hence 0 (x) ln 2. We plot (x)
and (x) in fig. 3.
If you managed to get this far, then youve done all that was asked. However, one
can go farther and analytically solve the equations for the Markov chain. In so doing,
we will discuss the linear algebraic aspects of the problem.
The matrix R is real but not symmetric. For such a matrix, the characteristic polyno
mial satisfies P () = P ( ), hence if is a root of P ( = 0), which is to say is an
eigenvalue, then so is . Accordingly, the eigenvalues of a real asymmetric matrix
are either real or come in complex conjugate pairs. We can decompose such a matrix
R as a sum over its eigenvectors,
X
Rij =
i j ,
where
X
Rij j = i
i Rij = j .
X
i i = .
=
i
One can check that the following assignment of eigenvectors is valid for our R(x)
matrix:
~ =
~ =
1
3
1
ei/3
ei/6 1 ei/3
~ =
~ =
1
3
1
ei/3
ei/6 1 ei/3 .
a
~t = t .
bt
Suppose we begin in a state where at=0 = 1 and bt=0 = ct=0 = 0. Then we have
a
t=0 = 23 and bt=0 = 13 , hence
+2/3
~
C =
.
1/3
2
3
2
3
et/ cos t
et/ sin t
with ct = a
t + bt .
(f) The fact that the eigenvalues of R are complex means that the probabilities should
oscillate as they decay to their equilibrium values pA = pB = pC = 31 . Can you see this
in your simulations?
SOLUTION: Yes! The oscillation is particularly clear in the lower panel of fig. 1.
Figure 1: Simulation of three state Kac ring model with initial conditions at=0 = 0.7, bt=0 =
0.2, ct=0 = 0.1. Note the oscillations as equilibrium is approached.
(2) Create your own pixelated image to iterate under the cat map. Dont do anything extravagant something with less than 25 black pixels should be fine. Choose a denominator
k which is minimally acceptable to convey your image. Then iterate the pixel coordinates
Figure 2: Simulation of three state Kac ring model with initial conditions at=0 = 0.7, bt=0 =
0.2, ct=0 = 0.1.
under the cat map. Show how your image gets messed up after a few iterations of the map,
but is nevertheless recurrent. Youll need to write a computer code to do this problem.
SOLUTION: Thomas Tran has kindly agreed to make his solution public. His iterated cat
map, acting on an image of a blobfish, can be viewed at
https://physics-forums.ucsd.edu/showthread.php?t=73
Ive uploaded Thomas image files to the Physics 140 web page and included a link from
the homework page:
http://physics.ucsd.edu/students/courses/fall2008/physics140/CATMAP/catmap.html
(3) Find (T, p) for the nonrelativistic ideal gas in d dimensions, and for the ultrarelativistic ideal gas in d dimensions.
Figure 3: Phase angle and relaxation time for the Markov process derived via the
Stosszahlansatz from the three state Kac ring model.
SOLUTION: We have
"
m
E
SNR (E, V, N ) = 21 dN kB ln
2
d~ N
"
1/d
d (d)
SUR (E, V, N ) = dN kB ln
d~c
V
N
2/d #
V
N
1/d #
1
dE + dV dN .
T
T
T
This says
=
T
S
N
E,V
m
1
2 dkB ln d~2
E
N
1/d
dkB ln [d (d)]
d~c
2/d
V
N
E
N
+ 1+
1/d
V
N
d
2
kB
+ (1 + d)kB
(NR)
(UR) .
Figure 4: Iterated cat map acting on a pixelated image of a blobfish. (Credit: Thomas Tran)
We need to eliminate E/N and V /N . To this end, we write
N
1
2 dkB E (NR)
S
1
=
=
T
E V,N
dkB N
(UR)
E
7
and
p
=
T
Thus,
S
V
E,N
kB
kB
N
V
(NR)
N
V
(UR)
m
d
1
2/d (k T )1+ 2d
p
+
1
+
dk
ln
kB
B
B
2
2
2
2~
=
[d (d)]1/d
1+
1/d
dkB ln
p
(kB T ) d + (1 + d)kB
~c
Thus, we have
(T, p) =
1
k
T
ln
p
1
+
d
ln
T
+
C
(NR)
2
B
k T ln p (1 + d) ln T + C
B
(UR) ,
(NR)
(UR) .
(1) Consider a monatomic ideal gas, represented within the grand canonical ensemble.
Show that the probability of finding the system to have N atoms is given by the Poisson
distribution,
1 hN i
e
hN iN .
PN =
N!
P N
Solution : The grand partition function is =
N z ZN , where z = exp() is the
fugacity.
The
ordinary
canonical
partition
function
is ZN = (V /3T )N /N !, where T =
p
2~2 /mkB T is the thermal wavelength. Thus, = exp(zV /3T ), and the average number
of particles is
zV
ln = 3 .
hN i = z
z
T
The probability that there are N particles in the system is clearly
PN =
z N ZN
1 hN i
=
e
hN iN .
N!
(2) Derive the grand canonical distribution when there are several types of particles present.
Solution : Let Z(N1 , N2 , . . . , NK ) be the OCE partition function for K species of particles
when there are Nj particles of species j. Then introduce K chemical potentials j (or K
fugacities zj = exp(j )) and write
X
X
eN1 1 eNK K Z(N1 , . . . , NK ) .
(T, V, 1 , . . . , K ) =
N1
NK
(3) An ideal paramagnet is described by the model in 3.11 of the notes, i.e.
= H
H
0
N
X
j ,
j=1
where each j = 1. Suppose the system starts off at a temperature T = 10 mK and a field
H = 20 T. The field is then lowered adiabatically to H = 1 T. What is the final temperature of
the system?
Solution : This problem may be solved by dimensional analysis. Clearly the entropy S
is a function of T and H, and the only dimensionally correct possibility is S(T, H, N ) =
N kB f (0 H/kB T ). We conclude that during an adiabatic process that the ration H/T is
constant. Thus,
H
T
Tf = Ti f = i = 500 nK .
Hi
20
Explicitly, we have
X
Z=
o H/kB T
=1
!N
= 2 cosh
0 H
kB T
(1)
H
F = kB T ln Z = N kB T ln 2 N kB T ln cosh 0
kB T
The entropy is
F
S=
T
= N kB
H,N
H
ln 2 + ln cosh 0
kB T
0 H
kB T
)
0 H
.
sinh
kB T
This is of the required form, S(T, H, N ) = N kB f (0 H/kB T ), with f (x) = ln(2 cosh x) +
x sinh x.
(4) Consider a nonrelativistic ideal gas. From dimensional analysis, we conclude that
|p|k = Ck (mkB T )k/2 .
R 2+k p2 /2mk T
B
dp p
e
|p|k = 0 R
2
dp p2 ep /2mkB T
mkB T , we have
Ck =
R
2
dx x2+k ex /2
0
R
dx x2 ex2 /2
Ck =
R
1
k
2k/2 dy y 2 + 2 ey
0
R
0
=2
dx y 2 ey
k/2
3
2 kB T .
3+k
2
.
23
8
,
2
F .
2
Solution : We have
CV = T
2F
.
T 2
= kB 2
.
T
(2)
Then
2
1
2
F
CV =
kB
kB
2
2 F
3 F
2 F
+ 2
= kB
= kB
= kB 2
2 ln Z
2
F = kB 2
.
2
(6) Consider a three state system with energy levels at = 0, = , and = W , with
0 W . Compute the free energy for such a system, f (T ). Derive an expression for
the heat capacity c(T ). You may find the results from problem (5) useful. Plot the specific
heat c(T ) versus kB T / for W = , W = 2, and W = 6.
Solution : The partition function is
= Tr e H = 1 + e + eW .
Thus,
f (T ) = kB T ln = kB T ln 1 + e/kB T + eW/kB T .
2
ln
2
2 e + W 2 eW + (W )2 e(+W )
.
= kB 2
2
1 + e + eW
c(T ) = kB 2
(3)
(1) Calculate, under equilibrium conditions, the number of photons in a cavity of volume
1 m3 at a temperature of T = 293 K. Compare this number with the number of ideal gas
molecules under the same conditions.
3
From the notes, eqn. 4.82, we have n(T ) = 20.405 T [K] cm3 . With a volume V =
106 cm3 , we have N = 5.13 1014 . For a classical gas at T = 293 K and p = 1 atm, we have
N=
(1)
(2) Thanksgiving turkey typically cooks at a temperature of 350 F. Calculate the total
2 V (kB T )4
= 3.78 105 J .
15 (~c)3
(2)
For air, which is a diatomic ideal gas, we have E = 25 pV . What do we take for p? If we
assume that oven door is closed at an initial temperature of 63 F which is 300 K, then with
a final temperature of 350 F = 450 K, we have an increase in the absolute temperature by
50%, hence a corresponding pressure increase of 50%. So we set p = 32 atm and we have
E=
5
2
(3)
(3) Let L denote the number of single particle energy levels and N the total number of
particles for a given system. Find the number of possible N -particle states (L, N ) for
each of the following situations:
(a) Distinguishable particles with L = 3 and N = 3.
D (3, 3) = 33 = 27.
(b) Bosons with
L = 3 and N = 3.
BE (3, 3) = 53 = 10.
(4)
1/3
1
k
k1/2
=
.
=
4/3
2 d/dk
30 3/2
32 0
(5)
32
7
3
e/kB T
7
3
(6)
(kB T )7/3
4/3
(7)
E
T
AkB
=
32
10
3
7
3
kB T
0
4/3
(8)
(b) Assuming the particles are S = 0 bosons, is there an Bose condensation transition? If
yes, compute the condensation temperature Tc (n) as a function of the particle density. If
no, compute the low-temperature behavior of the chemical potential (n, T ).
The following integral may be useful:
Z s1
X
u
du
ns (s) (s) ,
=
(s)
eu 1
n=1
0
where (s) is the gamma function and (s) is the Riemann zeta-function.
The condition for Bose-Einstein condensation is
Z
n = d g()
0
1
e/kB Tc
1
=
32
1
4
3
Thus,
Tc = 0
kB
32 n
34 43
!3
4
3
kB Tc
0
1/3
(9)
(10)
(5) Recall how we derived the GCE probability distribution based on the maximization of
the entropy S under the constraint of fixed average energy E and particle number N .
1
if one
(a) Show that one obtains the Bose-Einstein distribution n = e( ) 1
extremizes the entropy function
X
n ln n (1 + n ) ln(1 + n )
S = kB
S = kB
X n
ln
n
1
+
n
(11)
We also have
N =
(12)
n .
(13)
E =
We then write
S = S N
and compute
S =
Setting
X
"
X
n E
n N E
n
kB ln
1 + n
1
,
T
we recover the Bose-Einstein distribution,
+ N + E n = 0 .
N =
E =
n =
(14)
1
e( )/kB T
,
T
(15)
(16)
(17)
1
(b) Show that one obtains the Fermi-Dirac distribution n = e( ) + 1
if one extremizes the entropy function
X
n ln n + (1 n ) ln(1 n )
S = kB
X n
n
ln
1
(18)
(19)
n .
(20)
E =
We then write
S = S N
and compute
S =
Setting
X
"
X
n E
n N E
n
kB ln
1 n
1
,
T
we recover the Bose-Einstein distribution,
+ N + E n = 0 .
N =
E =
n =
(21)
1
e( )/kB T
+1
,
T
(22)
(23)
(24)
(6) Hydrogen (H2 ) freezes at 14 K and boils at 20 K under atmospheric pressure. The
density of liquid hydrogen is 70 kg/m3 . Hydrogen molecules are bosons. No evidence has
been found for Bose-Einstein condensation of hydrogen. Why not?
If we treat the H2 molecules as bosons, and we ignore the rotational freedom, which is
appropriate at temperatures below rot = 85.4 K, we have
2~2
Tc =
mkB
3
2
2/3
= 6.1 K .
(25)
Thus, the critical temperature for ideal gas Bose-Einstein condensation is significantly below the freezing temperature for H2 . The freezing transition into a regular solid preempts
any BEC phenomena.
(1) For each of the following situations, explain clearly and fully why it is or is not thermodynamically possible.
(a) Energy function E(S, V, N ) = a S V N with a constant. [6 points]
No! E(S, V, N ) = 3 E(S, V, N ) is homogeneous of degree 3 not extensive.
(b) Equation of state V = a N p T with a constant. [6 points]
No! The isothermal compressibility T = V1 V
p T = 1/p is negative, which
violates T > S > 0.
(c) A system where V
T p,N < 0 over some range of T and p. [6 points]
Yes! Many systems, such as water, contract upon a temperature increase over some
range of temperature.
(d) The phase diagram for a single component system depicted in fig. 1 (left panel). (You
only need know that a superfluid is a distinct thermodynamic phase.) [6 points]
dp
s
No! This one is tricky. From the Clapeyron equation, we have dT
.
= v
coex
Nernsts law says that the entropy of both the solid and superfluid phases must vanish at T = 0. Therefore all coexistence curves which intersect the pressure axis at
T = 0 must do so with zero slope.
(e) The phase diagram for a single component system in fig. 1 (right panel). (You only
need know that BCC, HCP, and FCC solids are distinct phases.) [6 points]
No! The Gibbs phase rule d = 2+ gives the dimension of thermodynamic space
over which distinct phases among species can coexist. For = 1 we have 3,
since d 0. So four phase coexistence with a single component is impossible.
(f) E(S, V, N ) = a N 2 V 1 exp(S/N b) with a and b constant. [6 points]
Yes! E is properly extensive and convex. One can derive E = pV = N bT , which is
the ideal gas law with kB replaced by b.
(g) 15 Joules of heat energy are required to raise the temperature of a system by T =
1 C at constant volume. 10 Joules of heat energy are required to raise the temperature of the same system by T = 1 F at constant pressure.
[6 points]
Yes! The heat capacity at constant volume is CV = dQ
dT V = 15 J/K. The heat cadQ
5
pacity at constant pressure is Cp = dT p = 10 J/ 9 K = 18 J/K. Stability requires
Cp > CV , which is satisfied.
(h) A heat engine operating between reservoirs at temperatures T1 = 400 K and T2 =
600 K. During each cycle, the engine does work W = 300 J and the entropy of the
upper reservoir decreases by 2.00 J/K. [8 points]
Yes! The only possible obstacle here is whether the engines efficiency is greater than
W
that of the corresponding Carnot cycle, for which C = 1 T1 = 13 . We have = Q
2
2
Q
and S2 = T 2 . Thus, = W/ T2 (S2 ) = 300 J/ (600 K)(2.00 J/K) = 41 < C .
2
E
S
aS 4
.
NV 2
V,N
4aS 3
.
NV 2
= 32 a
.
3
4
8
3
12
p
N V
8a S
N
Eliminating the
S4
E
N
S,V
aS 4
.
N 2V 2
= 2 2
=
=
.
p
N V
2aS 4
2N
64ap3
Note that we used the equation of state to eliminate the ratio V /N in terms of the
other intensive variables T and p. Thus,
(T, p) =
T4
.
64ap2
(d) Suppose the volume is isothermally expanded by a factor of eight (V 8V at constant T ). Then the temperature is isobarically increased by a factor of two (T 2T
at constant p). By what factor does the entropy change? Be sure to indicate whether
S increases or decreases. [10 points]
From part (a) we have S(T, V, N ) = (N V 2 T /4a)1/3 , hence isothermal expansion by
a factor of eight leads to a quadrupling (82/3 ) of the entropy. We next need S(T, p, N ),
which we obtain by eliminating V using part (b):
64a3 S 9 N 2 V 6
16aS
T3
=
2 8 =
.
2
3
6
p
N V
4a S
N
(3) Consider the equilibrium between a single component gas (assumed ideal and diatomic) and a solution of it in a liquid or solid solvent. Let G denote the chemical potential
of the gas molecules in the gas, and S their chemical potential in solution.
(a) What are the conditions for thermal, mechanical, and chemical equilibrium of the
gas and solute? [10 points]
Thermal equilibrium means T is constant throughout. Mechanical equilibrium means
p/T is constant throughout. Chemical equilibrium means /T is constant throughout, where is the chemical potential of the solute.
(b) Let (T, p) be the Gibbs free energy of a single solvent molecule in the liquid, i.e.
without the entropy of mixing term. If the concentration of solute is x, what is the
chemical potential S (T, p), i.e. when the entropy of mixing term is included? You
may assume x 1. [10 points]
Let N0 be the number of solvent molecules and NS the number of solute molecules
in the liquid. The entropy of mixing is
"
#
NS
N0
+ NS ln
Smix = kB N0 ln
N0 + NS
N0 + NS
NS
kB NS ln
.
eN0
T Smix
NS
= (T, p) + kB T ln x .
S (T, p) = (T, p)
(c) For an ideal gas, one has (T, p) = kB T (p) ( 21 f + 1) ln T , where f is the familiar
number of relevant degrees of freedom per molecule and (p) is a function of p
alone. Determine (p). [10 points]
We have, using a Maxwell relation deriving from the exactness of dG,
kB T
V
V
=
.
=
=
p T,N
N T,p N
p
This says
1
(p) =
p
p
(p) = ln p + 0 ,
where 0 is a constant.
(d) In liquids, the dependence of (T, p) on p typically is weak. Assuming (T, p) =
(T ), find an expression for x(T, p) in equilibrium. [10 points]
Applying the condition of equilibrium between solute and vapor, we have
0 + ln p
Exponentiating, we have
1
2f
(T )
+ 1 ln T =
+ ln x .
kB T
1
where A = exp(0 ) is a constant which also makes the units work out.
(e) Suppose we further neglect the temperature dependence of and write simply (T ) =
0 , where 0 is a positive constant. Sketch x(T, p) versus temperature at constant
pressure. Be sure to identify any relevant features, such as maxima, minima, vanishings, etc. [10 points]
1
The T ( 2 f +1) factor diverges as T 0 and vanishes as T . For (T ) = 0
a constant, the quantity e0 /kB T vanishes essentially at T = 0, and overwhelms
the power law divergence of the prefactor. As T , e0 /kB T 1. Thus, there
is a maximum concentration, which, after differentiating ln x, is found to occur at
kB T = 0 /( 21 f + 1) = 72 0 . See fig. 2 for a plot.
(1)
where S takes values 1, 0, or +1 (three possibilities) and takes values 1 but not 0 (two
possibilities).
(a) Find the partition function Z(T, H). [10 points]
(b) Find the magnetization M (T, H). [10 points]
(c) Find the zero field magnetic susceptibility, (T ) = (M/H)H=0 . [10 points]
(d) Provide a physical interpretation of your result for (T ) in the limits J 0 and
J . [5 points]
Z = Tr e h =
XX
eE(S,)/kB T .
(2)
It helps to make a little table of the energy values: From this we obtain
E
= +1
= 1
S = +1
J 20 H
J
S=0
0 H
0 H
S = 1
J
J + 20 H
J/kB T
Z(T, H) = 2 e
20 H
cosh
kB T
H
+ 2 cosh 0
kB T
+ 2 eJ/kB T
(3)
The magnetization is
kB T Z
Z H
T
2 H
H
20 eJ/kB T sinh k 0T + 0 sinh k 0 T
B
B
=
.
20 H
0 H
J/k
T
J/k
BT
e B cosh k T + cosh k T + e
M =
F
H
(4)
To find the zero field susceptibility, we first expand M to lowest order in the field H:
M (T, H) =
20
4 eJ/kB T + 1
H + O(H2 ) .
kB T 1 + eJ/kB T + eJ/kB T
(5)
Thus,
(T ) =
4 eJ/kB T + 1
20
M
.
=
H H=0 kB T 1 + eJ/kB T + eJ/kB T
(6)
(T ) =
420 /kB T
520 /3kB T
J 0
20 e|J|/kB T /kB T
J .
(7)
In the case J , the spins are ferromagnetically locked, and the only allowed configurations are | S, i = | 1, 1 i and | S, i = | 1, 1 i. The magnetization in these states
is M = 20 , respectively, hence the system acts as a single spin- 12 object with effective
magnetic moment
0 = 20 . The Curie susceptibility is then
20 /kB T = 420 /kB T .
When J = 0, the S and spins are independent. The spin gives a contribution 20 /kB T to
the susceptibility. The S spins contribution is 220 /3kB T , because only 23 of the states have
a moment; the S = 0 state doesnt contribute to the susceptibility.
When J , the spins are locked antiferromagnetically, meaning the only allowed
configurations are | S, i = | 1, 1 i and | S, i = | 1, 1 i, neither of which has a net
magnetic moment. Thus, = 0. If we consider large but finite |J|, then higher energy
states contribute, with exponentially small Boltzmann weights, leading to an exponentially
suppressed susceptibility, as we find by explicit calculation.
(b) If the excitations obey photon statistics (i.e. = 0), find CV (T ). [10 points]
(c) If the excitations obey Bose-Einstein statistics, show that the system undergoes BoseEinstein condensation. Find the critical temperature Tc (n), where n is the number
density of the excitations. [10 points]
Solution : We have
g() =
g
k2
3
3
= 2 k5/3 = 2 A9/4 5/4 () .
2
2 d/dk
8 A
8
(8)
e/kB T
(9)
=
Here, weve used the result
3V
8 2
13
4
13
4
(kB T )13/4
A9/4
(10)
Z
t1
dt t
= () () .
e 1
(11)
We therefore have
CV (T ) =
E
T
3V
= 2
8
17
4
13
4
kB T 9/4
.
A
(12)
(13)
1
z 1 e/kB T 1
1
e/kB Tc
3
= 2
8
1
Thus,
A
Tc =
kB
8 2 n
3( 49 ) ( 94 )
9
4
9
4
4/9
kB Tc 9/4
.
A
(14)
(15)
p2 c2 + m2 c4 .
d p p2 c2 +m2 c4
2A
p2 c2 +m2 c4
(T, A) = A
dp
p
e
=
e
h2
h2
0
A
2
Z
mc 2
2
d sinh() cosh() emc cosh()
~
0
Z
A mc 2
2
=
dx x emc x
2 ~
1
2
kB T
kB T
2
A mc
1
+
emc /kB T .
=
2 ~
mc2
mc2
(16)
(17)
Note that
Z
Z
d
d e
1
1
x
x
dx x e
=
dx e
=
=
1+
e .
d
d
(18)
= N kB T ln
+ N kB T
F (T, A, N ) = kB T ln Z(T, A, N ) = kB T ln
N!
N
"
#
mc 2
A
2
(19)
= N mc + kB T N kB T ln
2N ~
kB T
kB T
N kB T ln
N kB T ln 1 +
.
mc2
mc2
Thus, the entropy is
kB T
kB T
F
+ N kB ln 1 +
= N kB ln
S=
T A,N
mc2
mc2
"
#
kB T
mc 2
A
+ N kB
.
+ N kB ln
mc2 + kB T
2N ~
Accordingly, the specific heat is
1
S
mc2 kB2 T
kB2 T
c (T ) =
T
+
= kB +
2 .
2
N
T A,N
mc + kB T
mc2 + kB T
4
(20)
(21)
= S2 H S ,
h
0
1 M
N H H=0 .
(d) Find the zero field entropy S(T, H = 0, N ). (Hint : Take H 0 first.)
[5 points]
Solution : The partition function is Z = N , where is the single particle partition function,
H
.
(1)
= Tr e h = 1 + 2 e/kB T cosh 0
kB T
Thus,
(a)
"
/kB T
F = N kB T ln = N kB T ln 1 + 2 e
H
cosh 0
kB T
#
(2)
The magnetization is
(b)
0 H
2
sinh
0
kB T
k T Z
F
= B
=
M =
H
/k
T
H
Z
H
e B + 2 cosh k 0 T
(3)
To find the zero field susceptibility, we expand M to linear order in H, which entails expanding the numerator of M to first order in H and setting H = 0 in the denominator. We
then find
1
220
/k T
(4)
(c) (T ) =
kB T e B + 2
To find the entropy in zero field, it is convenient to set
H 0 first. The free energy is then
given by F (T, H = 0, N ) = N kB T ln 1 + 2 e/kB T , and therefore
(d)
S=
N
1
F
= N kB ln 1 + 2 e/kB T +
T
T
2 + e/kB T
(5)
(2) A classical gas consists of particles of two species: A and B. The dispersions for these
species are
p2
p2
A (p) =
,
B (p) =
.
2m
4m
In other words, mA = m and mB = 2m, and there is an additional energy offset
associated with the B species.
(a) Find the grand potential (T, V, A , B ).
[10 points]
(b) Find the number densities nA (T, A , B ) and nB (T, A , B ).
[5 points]
(c) If 2A B is an allowed reaction, what is the relation between nA and nB ?
(Hint : What is the relation between A and B ?)
[5 points]
(d) Suppose initially that nA = n and nB = 0. Find nA in equilibrium, as a function of T
and n and constants.
[5 points]
Solution : The grand partition function is a product of contributions from the A and B
species, and the grand potential is a sum:
(a)
(B +)/kB T
A /kB T
23/2 V kB T 3
= V kB T 3
T e
T e
Here, we have defined the thermal wavelength for the A species as T T,A =
For the B species, since the mass is twice as great, we have T,B = 21/2 T,A .
(6)
p
2~2 /mkB T .
1
A /kB T
= V 3
T e
V A
(7)
nB =
1
(B +)/kB T
.
= 23/2 V 3
T e
V B
(8)
If the reaction 2A B is allowed, then the chemical potentials of the A and B species are
related by B = 2A 2. We then have
nA 3T = e/kB T
(9)
nB 3T = 23/2 e(2+)/kB T
(10)
(b)
and
(b)
nB = 23/2 nA 3T
2
e/kB T
(11)
(12)
(13)
(d)
nA =
1 +
1 + 16 2 n3T e/kB T
8 2 3T e/kB T
(14)
(3) A branch of excitations for a three-dimensional system has a dispersion (k) = A |k|2/3 .
The excitations are bosonic and are not conserved; they therefore obey photon statistics.
(a) Find the single excitation density of states per unit volume, g(). You may assume
that there is no internal degeneracy for this excitation branch.
[10 points]
(b) Find the heat capacity CV (T, V ).
[5 points]
(c) Find the ratio E/pV .
[5 points]
(d) If the particles are bosons with number conservation, find the critical temperature Tc
for Bose-Einstein condensation.
[5 points]
k2
3
1
= 2 k7/3 .
2
2 d/dk
4 A
(15)
g() =
3 7/2
4 2 A9/2
(16)
3V
4 2
11
2
Thus,
(b)
CV =
E
T
e/kB T
(kB T )11/2
11
2
A9/2
3V
4 2
13
2
The pressure is
11
2
kB
(17)
kB T
A
9/2
Z
p = = kB T d g() ln 1 e/kB T
V
(18)
(19)
Z
3 7/2
= kB T d 2 9/2 ln 1 e/kB T
4 A
0
3 (k T )11/2
= 2 B 9/2
4
A
=
3V
4 2
9
2
Thus,
11
2
(kB T )11/2
A9/2
E
=
pV
(c)
Z
ds s7/2 ln 1 es
11
2
9
2
(20)
9
2
(21)
e/kB Tc
(22)
9
2
and therefore
(d)
A
Tc =
kB
9
2
kB T
A
9/2
4 2 n
3 29 92
(23)
!2/9
(24)
f () T = f () S , where
ZT
1
dt f (t)
f () T = lim
T T
0
Z
f () S = d () f ,
(25)
(26)
(27)
(e) A system with L = 6 single particle levels contains N = 3 fermions. How many
distinct many-body states are there? [3 points]
L
The general result for bosons is FD (L, N ) = N
, so we have = 63 = 20.
(f) Explain how the Maxwell-Boltzmann limit results, starting from the expression for
BE/FD (T, V, ). [3 points]
We have
X
BE/FD = kB T
ln 1 z e /kB T .
(28)
where the sum is over all energy eigenstates of the single particle Hamiltonian.
(29)
(g) For the Dieterici equation of state, p (1 bn) = nkB T exp(an/kB T ), find the second
virial coefficient B2 (T ). [3 points]
We must expand in powers of the density n:
ean/kB T
an
1 + bn + . . .
= nkB T 1
+ ...
1 bn
kB T
2
3
= nkB T + b kB T a n + O(n ) .
p = nkB T
(30)
and so we identify
p = nkB T 1 + B2 (T ) + B3 (T ) n2 + . . . ,
B2 (T ) = b
a
.
kB T
(31)
(32)
(h) Explain the difference between the Einstein and Debye models for the specific heat
of a solid. [4 points]
The Einstein model assumes a phonon density of states g() = CE ( 0 ), while for
the Debye model one has g() = CD 2 (D ), where CE,D are constants, and D
is a cutoff known as the Debye energy. At high temperatures, both models yield a
Dulong-Petit heat capacity of 3N kB , where N is the number of atoms in the solid. At
low temperatures, however, the Einstein model yields an exponentially suppressed
specific heat, while the specific heat of the Debye model obeys a T 3 power law.
(i) Who composed the song yerushalayim shel zahav? [50 quatloos extra credit]
The song was composed by Naomi Shemer in 1967. In 2005, it was revealed that it
was based in part on a Basque folk song.
(1) Consider a two-dimensional gas of fermions which obey the dispersion relation
(k) = 0 (kx2 + ky2 ) a2 + 12 (kx4 + ky4 ) a4 .
Sketch, on the same plot, the Fermi surfaces for F = 0.1 0 , F = 0 , and F = 10 0 .
Solution : It is convenient to adimensionalize, writing
x kx a
y ky a
.
0
(1)
(2)
In other words, we are interested in the level sets of the function (x, y) x2 +y 2 + 21 x4 + 21 y 4 .
When is small, we can ignore the quartic terms, and we have an isotropic dispersion, with
= x2 + y 2 . I.e. we can write x = 1/2 cos and y = 1/2 sin . The quartic terms give a
contribution of order 4 , which is vanishingly small compared with the quadratic term in
the 0 limit. When O(1), the quadratic and quartic terms in the dispersion are of
the same order of magnitude, and the continuous O(2) symmetry, namely the symmetry
under rotation by any angle, is replaced by a discrete symmetry group, which is the group
of the square, known as C4v in group theory parlance. This group has eight elements:
I , R , R2 , R3 , , R , R2 , R3
(3)
Explicitly, we can set y = 0 and solve the resulting quadratic equation in x2 to obtain the
maximum value of x, which we call a(). One finds
q
2
1 4
x
+
x
=
0
=
a
=
1 + 2 1 .
(4)
2
So long as x {a, a}, we can solve for y(x):
qp
1 + 2 2x2 x4 1 .
y(x) =
(5)
A sketch of the level sets, showing the evolution from an isotropic (i.e. circular) Fermi
surface at small , to surfaces with discrete symmetries, is shown in fig. 1.
Figure 1: Level sets of the function (x, y) = x2 +y 2 + 21 x4 + 12 y 4 for = ( 21 n)4 , with positive
integer n.
(2) Using the Sommerfeld expansion, compute the heat capacity for a two-dimensional
electron gas, to lowest nontrivial order in the temperature T .
Solution : In the notes, in section 4.7.6, we obtained the result
E
=
V
ZF
2
d g() +
(k T )2 g(F ) + O(T 4 ) .
6 B
(6)
This entails a heat capacity of CV,N = V 31 2 kB g(F ) kB T . The density of states at the
Fermi level, g(F ), is easily found to be
g(F ) =
Thus,
CV,N
d n
.
2 F
d 2
=N
k
6 B
kB T
F
(7)
(8)
(3) 3 He atoms consist of an odd number of fermions (two electrons, two protons, and one
neutron), and hence is itself a fermion. Consider a kilomole of 3 He atoms at standard
temperature and pressure (T = 293, K, p = 1 atm).
(a) What is the Fermi temperature of the gas?
(b) Calculate /kB T and exp(/kB T ).
(c) Find the average occupancy n() of a single particle state with energy 32 kB T .
Solution : Assuming the gas is essentially classical (this will be justified shortly), we find
the gas density using the ideal gas law:
n=
1.013 105 Pa
p
=
= 2.51 1025 m3 .
kB T
(1.38 1023 J/K)(293 K)
(9)
It is convenient to compute the rest energy of a 3 He atom. The mass is 3.016 amu (look it
up on Google), hence
m3 c2 = 3.016 (931.5 MeV) = 2.809 GeV .
(10)
For the conversion of amu to MeV/c2 , again try googling. Well then need ~c = 1973 eV A.
I remember 1973 because that was the summer I won third prize in an archery contest at
Camp Mehakeno. Thus,
F =
(3
n)
=
(3 2 2.51 1025 m3 )2/3
2m3 c2
2.809 109 eV
= 1.14 105 eV .
(11)
2~2
mkB T
1/2
2
2 (1973 eV A)
(2.809 109 eV) (86.2 106 eV/K) (293 K)
3
= ln z = 12.2
kB T
1/2
,
= 0.587 A
(12)
3 = 5.08 106 .
) (0.587 A)
(13)
(14)
To find the occupancy f ( ), we note = 23 (12.2) kB T = 13.7 kB T , in which
case
1
1
n() = ()/k T
= 13.7
= 1.12 106 .
(15)
e
+1
B
e
+1
(4) For ideal Fermi gases in d = 1, 2, and 3 dimensions, compute at T = 0 the average
energy per particle E/N in terms of the Fermi energy F .
Solution : The number of particles is
Z d
g d kFd
dk
(k
k)
=
V
,
N = gV
F
(2)d
(2)d d
(16)
where g is the internal degeneracy and d is the surface area of a sphere in d dimensions.
The total energy is
Z d
g d kFd
d k ~2 k2
~2 kF2
E = gV
(k
k)
=
V
.
(17)
F
(2)d 2m
(2)d d + 2 2m
Therefore,
E
d
=
.
N
d+2 F
(18)
(5) Obtain numerical estimates for the Fermi energy (in eV) and the Fermi temperature (in
K) for the following systems:
(a) conduction electrons in silver, lead, and aluminum
(b) nucleons in a heavy nucleus, such as 200 Hg
Solution : The Fermi energy for ballistic dispersion is given by
F =
~2
(3 2 n)2/3 ,
2m
(19)
where m is the effective mass, which one can assume is the electron mass m = 9.11
1028 g. The electron density is given by the number of valence electrons of the atom
3,
divided by the volume of the unit cell. A typical unit cell volume is on the order of 30 A
and if we assume one valence electron per atom we obtain a Fermi energy of F = 3.8 eV,
and hence a Fermi temperature of 3.8 eV/(86.2 106 eV/K) = 4.4 104 K. This sets the
overall scale. For detailed numbers, one can examine table 2.1 in Solid State Physics by
Ashcroft and Mermin. One finds
TF (Ag) = 6.38 104 K
(20)
Nuclear densities are of course much higher. In the literature one finds the relation R
A1/3 r0 , where R is the nuclear radius, A is the number of nucleons (i.e. the atomic mass
number), and r0 1.2 fm = 1.2 1015 m Under these conditions, the nuclear density is
on the order of n 3A/4R3 = 3/4r03 = 1.4 1044 m3 . With the mass of the proton
mp = 938 MeV/c2 we find F 30 MeV for the nucleus, corresponding to a temperature
of roughly TF 3.5 1011 K.
(6) Show that the chemical potential of a three-dimensional ideal nonrelativistic Fermi gas
is given by
#
"
2 kB T 2 4 kB T 4
+ ...
(n, T ) = F 1
12
F
80
F
and the average energy per particle is
#
"
E
5 2 kB T 2 4 kB T 4
3
+ ... ,
= 5 F 1 +
N
12
F
16
F
where 0 (n) is the Fermi energy at T = 0. Compute the heat capacity CV (T ) to terms of
order T 3 . How does the T 3 contribution to the electronic heat capacity compare with the
contribution from phonons?
Solution : From the Sommerfeld expansion we have
Z
Z
2
7 4
d () f ( ) = d () +
(kB T )2 () +
(k T )4 () + O(T 6 )
6
360 B
2
4
2
7 4
2 d
2 d
6
= 1+
(k T )
++
(k T )
+ O(T ) H() ,
6 B
d2
360 B
d4
(21)
(22)
where () = H (). Lets work this out to second order in T 2 for the case () = g(). The
integral then gives the overall density n. We write = F + and expand the RHS to
second order in . Thus yields
ZF
n = d g() + g(F ) + 21 g (F ) ()2 + . . .
(23)
2
2
(kB T )2 g (F ) +
(k T )2 g (F ) + . . .
6
6 B
+
7 4
(k T )4 g (F ) + . . . .
360 B
RHS of the first line above comes from expanding the integral in the first term on the RHS
of the previous equation to second order in . The subsequent lines come from the expansions of the second and third terms on the RHS of the previous equation, respectively.
We expand out to the necessary order in each case. From this equation we thus obtain
#
"
2
4
7 g (F )
1 g (F ) 3 g (F ) g (F )
2 g (F )
4
+ ...
= (kB T )
(k T )
+
6
g(F )
36 B
2 g(F )
g2 (F )
10 g(F )
(24)
2 (kB T )2
4 (kB T )4
+ O(T 6 ) .
( 2)(2 7)
6
F
360 3F
(25)
2 (kB T )2 4 (kB T )4
+ O(T 6 ) ,
12 F
80 3F
(26)
360
C+1
N
=
V
+2
1 + ( 1)
2
6
kB T
2
+ ( 3)( 2)( 1)
7 4
360
kB T
4
+ ...
(27)
)
(28)
45
Unfortunately were not quite done, since we now must expand in a power series in T ,
invoking our previous result. Working this out (by hand!), I obtain
)
(
4 kB T 4
+1
E
2 kB T 2
+ ( + 2)(2 7)
+ O(T 6 ) .
=
1 + ( + 2)
N
+2 F
6
F
120
F
(30)
1
5
1
1
1
Setting = 2 we have 6 ( + 2) = 12 and 120 ( + 2)(2 7) = 16 , as indicated in the
statement of the problem. Our formula holds for general , so we can find the result for
d = 2 by setting = 0.
The heat capacity is
E
CV,N =
T V,N
(
= N kB
2
( + 1)
3
In d = 3, with =
1
2,
kB T
F
4
+ ( + 1)(2 7)
30
kB T
F
3
+ O(T 5 )
(31)
3 4
the order T 3 term is 20
(kB T /F )3 N kB . The low temperature
(phonon)
phonon contribution is CV
The ratio is
12 4
3
5 (T /D ) N kB , where D
(elec)
CV
(phonon)
CV
1 D 3
.
=
16 TF
(1) Consider a model in which there are three possible states per site, which we can denote
by A, B, and V. The states A and B are for our purposes identical. The energies of A-A, A-B,
and B-B links are all identical and equal to W . The state V represents a vacancy, and any
link containing a vacancy, meaning A-V, B-V, or V-V, has energy 0.
(a) Suppose we write = +1 for A, = 1 for B, and = 0 for V. How would you
write a Hamiltonian for this system? Your result should be of the form
X
=
H
E(i , j ) .
hiji
Find a simple and explicit function E(, ) which yields the correct energy for each
possible bond configuration.
Solution : The quantity i2 is 1 if site i is in state A or B and is 0 in state V. Therefore
we have
X
=W
H
i2 j2 .
hiji
(b) Consider a triangle of three sites. Find the average total energy at temperature T .
There are 33 = 27 states for the triangle. You can just enumerate them all and find
the energies.
Solution : Of the 27 states, eight have zero vacancies each site has two possible
states A and B with energy E = 3W . There are 12 states with one vacancy, since
there are three possible locations for the vacancy and then four possibilities for the
remaining two sites (each can be either A or B). Each of these 12 single vacancy states
has energy E = W . There are 6 states with two vacancies and 1 state with three
vacancies, all of which have energy E = 0. The partition function is therefore
Z = 8 e3W + 12 eW + 7 .
Note that Z( = 0) = Tr 1 = 27 is the total number of microstates. The average
energy is then
24 e3W + 12 eW
1 Z
=
W
E=
Z
8 e3W + 12 eW + 7
(c) For a one-dimensional ring of N sites, find the 3 3 transfer matrix R. Find the free
energy per site F (T, N )/N and the ground state entropy per site S(T, N )/N in the
N limit for the cases W < 0 and W > 0. Interpret your results. The eigenvalue
equation for R factorizes, so you only have to solve a quadratic equation.
W 2 2
R = e
eW
eW
1
1
1 ,
1
where the row and column indices are A (1), B (2), and V (3), respectively. The partition
function on a ring of N sites is
N
N
Z = N
1 + 2 + 3 ,
where 1,2,3 are the three eigenvalues of R. Generally the eigenvalue equation for a 3 3
matrix is cubic, but we can see immediately that det R = 0 because the first two rows
are
identical. Thus, = 0 is a solution to the characteristic equation P () = det I R = 0,
and the cubic polynomial P () factors into the product of and a quadratic. The latter is
easily solved. One finds
P () = 3 (2x + 1) 2 + (2x 2) ,
where x = eW . The roots are = 0 and
= x +
1
2
x2 x +
9
4
Thus
+ (T 0 , W < 0) = 2 e|W |/kB T + e|W |/kB T + O(e2|W |/kB T .
F (T 0 , W < 0) = N |W | N kB T ln 2 + . . .
S(T 0 , W < 0) = N ln 2 .
Then
+ (T 0 , W > 0) = 2 + 23 eW/kB T + O(e2W/kB T .
F (T 0 , W > 0) = N kB T ln 2 13 N kB T eW/kB T + . . .
S(T 0 , W > 0) = N ln 2 .
Thus, the ground state entropies are the same, even though the allowed microstates are
very different. For W < 0, there are no vacancies. For W > 0, every link must contain at
least one vacancy.
(2) Consider a two-state Ising model on a one-dimensional ring, where each site supports
a spin n = 1. The Hamiltonian is
X 1 + n1 n 1 + n n+1
H = J
.
2
2
n
(a) Show explicitly that P =
1
2
1 + is zero if 6= and is unity if = .
Solution : Well this is pretty obvious. Since each Ising variable is 1, we have
= +1 when = and = 1 when 6= . Thus, P = 21 (1 + ) registers
a 1 when = and 0 when 6= .
is proportional to the number of consecutive triples of sites (n1 , n , n+
(b) Show that H
1) for which n1 = n = n+1 .
Solution : Clearly P P = 1 if and only if = = . So the above Hamiltonian counts the number of consecutive triples where n1 = n = n+1 .
(c) Calculate the partition function for a ring of N = 4 sites. There are 24 = 16 states of
the ring, so you can attack the problem by direct enumeration of all the possibilities.
Remember that the system is a ring.
Solution : On an N = 4 site ring, there are two states with energy E = 4J, corresponding to all spins and all spins . There are eight states with energy E = J,
corresponding to four configurations with one spin (and three spins) and four
configurations with one spin (and three spins). The remaining 42 = 6 configurations have two and two spins, and have energy E = 0. Thus,
Z = 2 e4J + 8 eJ + 6 .
Note that Z( = 0) = Tr 1 = 16.
(3) For the Mayer cluster expansion, write down all possible unlabeled connected subgraphs which contain four vertices. For your favorite of these animals, identify its symmetry factor s , and write down the corresponding expression for the cluster integral b .
For example, for the diagram with four vertices the symmetry factor is s = 8 and the
cluster integral is
Z
Z
Z
Z
1
d
d
d
d r1 d r2 d r3 ddr4 f (r12 ) f (r23 ) f (r34 ) f (r14 )
b =
8V
Z
Z
Z
1
d
d
d r1 d r2 ddr3 f (r12 ) f (r23 ) f (r1 ) f (r3 ) .
=
8
(Youll have to choose a favorite other than .) If youre really energetic, compute s and
b for all of the animals with four vertices.
(4) Explain physically, using pictures and/or drawing analogies to other common knowledge (of upper division physics majors) why the pair distribution function g(r) should
have the features it does in fig. 5.7 of the notes.
Solution : The radial distribution function,
g(r) =
1
X
(r xi + xj ) ,
2
Vn
i6=j
tells us how likely it is to find a pair of particles with separation r. For particles which
behave as hard spheres at distances larger than some diameter a, we therefore expect
g(r) 0 for r < a. Beyond r = a the distribution function exhibits a damped oscillation
on a length scale a, since the relative absence of pairs at a particular radial separation r
leads to a relative abundance of pairs at radial separation r + a. See the figures in chapter
5 of the notes.
(5) An ionic solution of dielectric constant and mean ionic density n fills a grounded
conducting sphere of radius R. A charge Q lies at the center of the sphere. Calculate the
ionic charge density as a function of the radial coordinate r, assuming Q/r kB T .
Solution : Debye-Huckel
4
4e
,
n+ n
ext
(r) .
r + D =
2
r r
1
r
cosh(D r)
sinh(D r)
+B
.
r
r
(1) Consider an Ising ferromagnet where the nearest neighbor exchange temperature is
JNN /kB = 50 K and the next nearest neighbor exchange temperature is JNNN /kB = 10 K.
What is the mean field transition temperature Tc if the lattice is:
(a) square
(b) honeycomb
(c) triangular
(d) simple cubic
(e) body centered cubic
Hint : As an intermediate step, you might want to show that the mean field transition
temperature is given by
kB TcMF = z1 JNN + z2 JNNN ,
where z1 and z2 are the number of nearest neighbors and next-nearest neighbors of a given
lattice site, respectively.
where JNN and JNNN are the nearest neighbor and next nearest neighbor exchange interaction energies. According to sketches in fig. 1, we have
(a) square lattice : z1 = 4 and z2 = 4. Thus, TcMF = 240 K.
(b) honeycomb lattice : z1 = 3 and z2 = 6. Thus, TcMF = 210 K.
(c) triangular lattice : z1 = 6 and z2 = 6. Thus, TcMF = 360 K.
(d) simple cubic lattice : z1 = 6 and z2 = 12. Thus, TcMF = 420 K.
(e) body-centered cubic lattice : z1 = 8 and z2 = 6. Thus, TcMF = 460 K.
X
hiji
Si Sj H
where Si 1 , 0 , +1 .
1
X
i
Si ,
Figure 1: Nearest neighbors (red circles), next nearest neighbors (light blue squares), and
some third nearest neighbors (green triangles) for five common lattices. (a) square, (b)
honeycomb, (c) triangular, (d) simple cubic, and (e) body centered cubic.
(a) Writing Si = m + Si and ignoring terms quadratic in the fluctuations, derive the
mean field Hamiltonian HMF .
Solution : We have
Si Sj = (m + Si )(m + Sj )
= m2 + m (Si + Sj ) + Si Sj
= m2 + m (Si + Sj ) + Si Sj .
We ignore the fluctuation term, resulting in the mean field Hamiltonian
X
Si .
HMF = 21 N zJm2 zJm + H
i
(b) Find the dimensionless mean field free energy density, f = FMF /N zJ, where z is
the lattice coordination number. You should define the dimensionless temperature
kB T /zJ and the dimensionless field h H/zJ.
Solution : The effective field is Heff = zJm + H. Note that
X
zJm + H
Heff S/kB T
.
e
= 1 + 2 cosh
kB T
S
(c) Find the self-consistency equation for m = hSi i and show that this agrees with the
condition f /m = 0.
Solution : Extremizing the free energy f (m) with respect to m, we obtain the mean
field equation:
2 sinh m+h
f
.
=0
=
m=
m
1 + 2 cosh m+h
P
m+h
(m+h)S/
2
sinh
S
e
.
m = PS (m+h)S/ =
1 + 2 cosh m+h
Se
+ ... .
= ln 3 + 12 1
3
36 3
3
(e) Find the critical temperature c .
Solution : The critical temperature is identified as the value of where the coefficient
of the m2 term in the free energy vanishes. Thus, c = 32 .
(f) Find m(c , h).
Solution : Setting = c = 23 , we extremize f (m) and obtain the equation
f (m, c , h) = 0 =
2h
m3
3
9c
3c
1/3
=
m(c , h) = 6 c2 h
X
hiji
i
j
1/3
8
3h
is a three-dimensional unit
find the mean field transition temperature TcMF . Here, each
i
vector, which can be parameterized using the usual polar and azimuthal angles:
= sin cos , sin sin , cos .
i
i
i
i
i
i
, ... ,
)=
Tr A(
1
N
Z Y
N
d
i=1
, ... ,
),
A(
1
N
where
di = sin i di di .
= m + , where m = h i. Youll want
Hint : Your mean field Ansatz will look like
i
i
i
to ignore terms in the Hamiltonian which are quadratic in fluctuations, i.e. i j . You
can, without loss of generality, assume m to lie in the z direction.
= m + and neglecting the fluctuations, we arrive at the mean
Solution : Writing
i
i
field Hamiltonian
X
,
Z=e
d zJm
e
4
!N
2
1
e
f (m, ) = 2 m ln
4
sinh(m/)
1 2
= 2 m ln
m/
m2
m4
1 2
= 2 m ln 1 + 2 +
+ ...
6
120 4
1 2
m4
m +
+ ... .
= 21 1
3
180 3
Setting the coefficient of the quadratic term to zero, we obtain c = 31 .
X
hiji
I(i , j ) H
i ,A ,
where on each site i there are four possible choices for i : i {A, B, C, D}. The interaction
matrix I(, ) is given in the following table:
I
A
B
C
D
A
+1
1
1
0
B
1
+1
0
1
C
1
0
+1
1
N (1 , . . . , N ) =
N
Y
(i )
D
0
1
1
+1
i=1
() = x ,A + y (,B + ,C + ,D ) .
What is the relationship between x and y? Henceforth use this relationship to eliminate y in terms of x.
Solution : The density matrix must be normalized, hence
Tr = x + 3y 1
y = 31 (1 x) .
= 12 zJ x2 + 3y 2 4xy 4y 2 H x
= 12 zJ x2 + 31 (1 x)2 43 x(1 x) 49 (1 x)2 H x
1
zJ 1 + 10 x 20 x2 H x .
= 18
= kB x ln x + 3 y ln y
(
= kB
1x
x ln x + (1 x) ln
3
)
x
1
1 + 10 x 20 x2 hx + x ln x + (1 x) ln
,
= 18
3
where h = H/zJ and = kB T /zJ are the dimensionless field and temperature, respectively. The mean field equation is obtained by extremizing f (x, , h):
3x
f
5
= 0 = 9 (1 4x) h + ln
.
x
1x
(e) What value x does x take when the system is disordered?
Solution : Clearly x = y = 41 in the disordered phase, since each state is then equally
probable. The global symmetry of the problem, which is Z4 , is then unbroken.
(f) Write x = x + 43 and expand the free energy to fourth order in . The factor 34
should generate manageable coefficients in the Taylor series expansion. You may
want to use a symbolic manipulator like Mathematica here.
Solution : We write x = 41 + 34 , in which case
3x
5
f = 9 (1 4x) h + ln
1x
2
3
5
3
3 + 47 4 + O(5 ) .
= ln 4 4 h + 2 12
(g) Sketch as a function of T for H = 0 and find Tc . Is the transition first order or second
order?
Solution : The transition in zero field is first order, but youd have had to read
ahead a little in the notes to understand this. The point is that whenever a Landau
expansion of the free energy has a cubic term, e.g. for
f () = f0 + 21 a 2
1
3
u 3 + 14 b 4 + . . . ,
1
2
a 2 13 u 3 + 41 b 4 = 0 .
1
2
a u + b 2 = 0
1
3
u + 41 b 2 = 0 .
Eliminating the quadratic term, we obtain = 3a/u at the first order transition, and
inserting this into either of the above equations
we obtain the relation u2 = 29 ab. For
5
our specific model, we have a = 3 12 , u = 3, and b = 7. Thus, the first order
transition occurs at a critical temperature
c =
35
76
Note that the sign of the quadratic term in f () is still positive at this point, and re5
. If there were no cubic term, we would
mains so down to a temperature = 12
expect a second order transition at this latter temperature, but as we see it is preempted by the first order transition.
(1) Find in the scientific literature two examples of phase transitions, one first order and
one second order. Describe the systems and identify the order parameter in each case. Reproduce a plot of the order parameter versus temperature showing how the transition is
discontinuous (first order) or continuous (second order). If the transition is first order, is a
latent heat reported? If second order, are any critical exponents reported? Some examples
of phase transitions, which might help you in your searches: liquid-gas transitions, Curie
transitions, order-disorder transitions, structural phase transitions, normal fluid to superfluid transitions (3 He and 4 He are two examples), metal-superconductor transitions, crystallization transitions, etc. An excellent search engine for scholarly publications is available
at scholar.google.com.
Solution : For an example of a system which exhibits both a second-order and a firstorder transition as a function of temperature, see fig. 1, which shows the cubic lattice
constant a(T ) versus temperature for the solid material C60 . The C60 molecule is one of a
class known as fullerenes, after the architect and futurist Buckminster Fuller, the inventor
of the geodesic dome. At a temperature TcII = 260 K, there is a discontinuity in the a(T )
curve, signaling a first order phase transition, This is akin to the density discontinuity
in water when it boils or freezes. For T > TcII = 260 K, the C60 molecules form a face
centered cubic lattice, while below this temperature the lattice is simple cubic. For T
[TcI , TcII ], with TcI = 90 K, the individual C60 molecules are in one of two orientational states.
There is long-ranged orientational order in this phase both orientational states are not
equally probable however as T approaches TcI from above, the distribution of molecular
orientations approaches a nearly fixed distribution and varies very little for T < TcI .
Figure 1: Temperature variation of the cubic lattice constant a(T ) of C60 as a function of
temperature, from W. I. F. David et al., Europhys. Lett. 18, 219 (1992).
(2) Consider a two-state Ising model, with an added dash of flavor from what you have
learned in your Physics 130 (Quantum Mechanics) sequence. You are invited to investigate
the transverse Ising model, whose Hamiltonian is written
X
X
x x
= 1
J
H
iz ,
H
ij i j
2
i,j
0 1
1 0 i
iz
1 0
.
0 1 i
You may find the material in 6.15 of the notes useful to study before attempting this problem.
(a) Using the trial density matrix,
i =
1
2
1
2
mx ix + 12 mz iz
f (, h, mx , mz ), where = kB T /J(0),
compute the mean field free energy F/N J(0)
and h = H/J(0). Hint: Work in an eigenbasis when computing Tr ( ln ).
Solution : We have Tr( x ) = mx and Tr( z ) = mz . The eigenvalues of are
1
2 1/2 . Thus,
2
2 (1 m), where m = (mx + mz )
"
#
1
+
m
1
m
1
m
1
+
m
.
ln
ln
+
f (, h, mx , mz ) = 21 m2x hmz +
2
2
2
2
(b) Derive the mean field equations for mx and mz .
Solution : Differentiating with respect to mx and mz yields
f
1+m
m
= 0 = mx + ln
x
mx
2
1m
m
f
1+m
m
= 0 = h + ln
z .
mz
2
1m
m
Note that we have used the result
m
m
=
m
m
where m is any component of the vector m.
(c) Show that there is always a solution with mx = 0, although it may not be the solution
with the lowest free energy. What is mz (, h) when mx = 0?
Solution : If we set mx = 0, the first mean field equation is satisfied. We then have
mz = m sgn(h), and the second mean field equation yields mz = tanh(h/). Thus, in
this phase we have
mx = 0
,
mz = tanh(h/) .
1+m
m = ln
,
2
1m
which is equivalent to m = tanh(m/). Plugging this into the second mean field
equation, we find mz = h. Thus, when mx 6= 0,
p
mz = h
,
mx = m2 h2
,
m = tanh(m/) .
Note that the length of the magnetization vector, m, is purely a function of the temperature in this phase and thus does not change as h is varied when is kept fixed.
What does change is the canting angle of m, which is = tan1 (h/m) with respect
to the z axis.
(e) Show that for 1 there is a curve h = h () below which mx 6= 0, and along which
mx vanishes. That is, sketch the mean field phase diagram in the (, h) plane. Is the
transition at h = h () first order or second order?
Solution : The two solutions coincide when m = h, hence
h = tanh(h/)
(h) =
ln
2h
.
1+h
1h
Inverting the above transcendental equation yields h (). The component mx , which
serves as the order parameter for this system, vanishes smoothly at = c (h). The
transition is therefore second order.
(f) Sketch, on the same plot, the behavior of mx (, h) and mz (, h) as functions of the
field h for fixed . Do this for = 0, = 21 , and = 1.
Solution : See fig. 2.
ddx
2
1
2 a ||
i
2 .
+ 14 b ||4 + 12 ||
is a complex-valued field, and both b and are positive. This theory is approHere (x)
priate for describing the transition to superfluidity. The order parameter is h(x)i.
Note
and (x),
where
that the free energy is a functional of the two independent fields (x)
is the complex conjugate of . Alternatively, one can consider F a functional of the real
and imaginary parts of .
Figure 2: Solution to the mean field equations for problem 2. Top panel: phase diagram.
The region within the thick blue line is a canted phase, where mx 6= 0 and mz = h > 0;
outside this region the moment is aligned along z and mx = 0 with mz = tanh(h/).
so that the free energy
(a) Show that one can rescale the field and the coordinates x
can be written in the form
Z
h
i
F = 0 ddx 12 ||2 + 14 ||4 + 21 ||2 ,
where and x are dimensionless, 0 has dimensions of energy, and where the sign
on the first term on the RHS is sgn (a). Find 0 and the relations between and
and x.
and between x
Solution : Taking the ratio of the second and first terms in the free energy density,
1/2
. Taking the ratio of the third to the first
we learn that has units of A |a|/b
1/2
= x to
terms yields a length scale = /|a|
. We therefore write = A and x
obtain the desired form of the free energy, with
1
0 = A2 d |a| = |a|2 2 d b1 2 d .
1
2
||2 12 2 .
equation.
(c) Consider a two-dimensional system (d = 2) and let a < 0 (i.e. T < Tc ). Consider
the case where (x) describe a vortex configuration: (x) = f (r) ei , where (r, ) are
two-dimensional polar coordinates. Find the ordinary differential equation for f (r)
which extremizes F .
Solution : In two dimensions,
2 =
1
1 2
2
+
+
.
r 2 r r r 2 2
+ f f3 = 0 .
dr 2 r dr r 2
1
4
1f
2 2
where R is the radius of the system, which we presume is confined to a disk. Consider a trial solution for f (r) of the form
f (r) =
r2
r
,
+ a2
F=
1
2
1
4
which, up to a constant, is the desired form of the free energy. It is a good exercise to
show that the Euler-Lagrange equations,
d (rF)
(rF)
=0
f
dr
f
results in the same ODE we obtained for f in part (c). We now insert the trial form
for f (r) into F . The resulting integrals are elementary, and we obtain
)
(
2
2 a2
4
R
R
a
+ 2 ln
+1 + 2
.
F (a, R) = 14 0 1 2
(R + a2 )2
a2
R + a2
Taking the limit R , we have
R2
F (a, R ) = 2 ln
a2
+ a2 .
We now extremize with respect to a, which yields a = 2. Note that the energy in
the vortex state is logarithmically infinite. In order to have a finite total free energy
(relative to the ground state), we need to introduce an antivortex somewhere in the
system. An antivortex has a phase winding which is opposite to that of the vortex,
i.e. = f ei . If the vortex and antivortex separation is r, the energy is
2
r
1
V (r) = 2 0 ln 2 + 1 .
a
This tends to V (r) = 0 ln(d/a) for d a and smoothly approaches V (0) = 0, since
when r = 0 the vortex and antivortex annihilate leaving the ground state condensate. Recall that two-dimensional point charges also interact via a logarithmic potential, according to Maxwells equations. Indeed, there is a rather extensive analogy
between the physics of two-dimensional models with O(2) symmetry and (2 + 1)dimensional electrodynamics.
(1) Consider the situation in 7.6.2 of the notes, where a temperature gradient is applied
at constant pressure. Show that there is no resulting particle current: j = 0.
Solution : We have
d3p ( h) 0
f (vT ) v
h3 kB T 2
Z
2
3/2
2
=
T (kB T )
d 1/2 e/kB T ( h)
3mkB T 2
j=
(1)
h i
2 (kB T )3/2
Thus,
since h = 52 kB T . Thus, j = 0.
Z
( 32 + s)
1
d s+ 2 e/kB T =
(kB T )s .
( 32 )
(2)
h2 i =
3
2
52 (kB T )2 = hi h ,
(3)
n3T
p2 /2mkB T
p2
1+
2mkB T
2
and investigate its relaxation to the equilibrium distribution f 0 (p) = n3T ep /2mkB T using
the Boltzmann equation in the relaxation time approximation, with no external forces. Find
f (r, p, t). Then find N (t) and E(t), the time-dependent values for the total particle number
and total energy. You may abbreviate N0 nV , where V is the system volume and N0
is the number of particles at equilibrium. Then, drawing upon your understanding of
collisional invariants, explain why your calculation is complete BS. What has gone wrong?
Solution : It is convenient to express everything in terms of the single particle energy
= p2 /2m:
f (p, t = 0) = n3T e 1 + = f 0 1 + ,
(4)
where = 1/kB T as usual. In the absence of temperature or velocity gradients, and with
no external forces, the Boltzmann equation takes the form
f
=
t
f
t
coll
(5)
So we have
f (p, t) = f 0 + f 0 et/ .
(6)
(7)
Z
( 3 + s)
h i = d PMB () s = 2 3
(kB T )s .
( 2 )
s
(8)
(9)
Thus, we find
N (t) = N0 1 + 23 et/
t/
E(t) = 32 N0 kB T 1 + 15
,
4 e
(10)
(11)
where N0 = n V . We see that neither particle number nor total energy are conserved,
which is clearly wrong. This is a defect of the relaxation time approximation. In general,
the collision integral must annihilate the five collisional invariants, which are total particle
number, total energy, and the three components of the total momentum. These quantities
are all conserved by the collisions. Every quantity other than these should relax on some
microscopic time scale, but the collisional invariants dont relax at all, and the relaxation
time approximation (RTA) is too crude to distinguish between the collisional invariants
and other quantities everything relaxes at the same rate within the RTA.
Incidentally, there is a minor modification to the relaxation time approximation collision
integral which results in number conservation, although it does not yield energy conservation. This is the so-called BGK collision integral, named for P. L. Bhatnagar, E. P. Gross,
and M. Krook, Phys. Rev. 94, 511 (1954). BGK took
f
t
coll
n0 (r)
(12)
where
d3p
f (r, p, t)
h3
Z 3
dp 0
n0 (r) =
f (r, p) .
h3
n(r, t) =
d3p
h3
f
t
coll
=0,
(13)
(14)
(15)
and there is local number conservation by collisions1 . The nice thing about the BTG collision integral is that it is still linear in f .
(3) Consider a classical gas of charged particles in the presence of a magnetic field B. The
Boltzmann equation is then given by
e
f
h 0
f v T
vB
=
kB T 2
mc
dv
f
t
coll
Making the relaxation time approximaConsider the case where T = T (x) and B = B z.
tion, show that a solution to the above equation exists in the form f = vA(), where A()
is a vector-valued function of (v) = 21 mv 2 which lies in the (x, y) plane. Find the energy
current j . Interpret your result physically.
Solution : Well use index notation and the Einstein summation convention for ease of
presentation. Recall that the curl is given by (A B) = A B . We write f =
v A (), and compute
f
A
= A + v
v
v
A
= A + v v
.
(16)
(17)
Thus,
vB
We then have
f
f
= v B
v
v
A
= v B A + v v
= v B A .
v A
h 0
e
f
v
T
=
v
B
A
kB T 2
mc
(18)
(19)
eB
( h) 0
n A =
f T ,
mc
kB T 2
(20)
The BGK collision integral also can be used in the presence of streaming terms.
(22)
(23)
(24)
(25)
1
1 + c2 2
B + C c = 0 ,
c2 2
1 + c2 2
C + A c = 0 ,
(26)
c
.
1 + c2 2
(27)
B=
C=
+ c2 2 n n c n
1
M
X
1 + c2 2
X .
(28)
d3p
2
v v A () =
3
h
3m
j =
d3p 2
A () ,
h3
(29)
. Next, we use
d3p 2
5 2 T
X =
k T
,
3
h
3m B x
(30)
5
kB2 T
2 2
T
+
n
(
nT
)
+
T
.
c
c
3m 1 + c2 2
(31)
2
3m
hence
2
3m
= z and T = T (x) x.
We see that the energy current j is flowing
We are given that n
and along y.
Why does heat flow along y?
It is because the particles
both along x
there is a Lorentz force in the y
are charged, and as they individually flow along x,
direction, so the energy flows along y as well.
and compute the average p4 (t) . You should assume that
(s1 ) (s2 ) (s3 ) (s4 ) = (s1 s2 ) (s3 s4 ) + (s1 s3 ) (s2 s4 ) + (s1 s4 ) (s2 s3 )
where (s) = (s). You may further assume that p(0) = 0.
Solution : Integrating the Langevin equation, we have
Zt
p(t) = dt1 e(tt1 ) (t1 ) .
(32)
Raising this to the fourth power and taking the average, we have
p (t) =
Zt
Zt
Zt
Zt
(tt1 )
(tt2 )
(tt3 )
dt1 e
dt2 e
dt3 e
dt4 e(tt4 ) (t1 ) (t2 ) (t3 ) (t4 )
0
Zt
Zt
2
32
1 e2t .
2
4
(33)
We have here used the fact that the three contributions to the average of the product of the
four s each contribute the same amount to hp4 (t)i. Recall = 2M kB T , where M is the
mass of the particle. Note that
2
p (t) = 3 p2 (t) .
(34)
(5) For the brave only! Due to quantum coherence effects in the backscattering from impurities, one-dimensional wires dont obey Ohms law (in the limit where the inelastic
mean free path is greater than the sample dimensions, which you may assume). Rather,
let R(L) = R(L)/(h/e2 ) be the dimensionless resistance of a quantum wire of length L, in
units of h/e2 = 25.813 k. Then the dimensionless resistance of a quantum wire of length
L + L is given by
R(L + L) = R(L) + R(L) + 2 R(L) R(L)
q
+ 2 cos R(L) 1 + R(L) R(L) 1 + R(L) ,
where is a random phase uniformly distributed over the interval [0, 2). Here,
R(L) =
L
,
2
Show that this equation may be solved in the limits R 1 and R 1, with
P (R, z) =
for R 1, and
1 R/z
e
z
Solution : From the composition rule for series quantum resistances, we derive the phase
averages
L
R = 1 + 2 R(L)
2
L
2 L 2
L
2
+ 2 R(L) 1 + R(L)
1+
(R) = 1 + 2 R(L)
2
2
2
L
= 2 R(L) 1 + R(L)
+ O (L)2 ,
2
(35)
(36)
(37)
2R + 1
2
F2 (R) =
2R(1 + R)
.
2
(38)
Note that 2F1 (R) = dF2 /dR, which allows us to write the Fokker-Planck equation as
R (1 + R) P
P
=
.
(39)
L
R
2
R
Defining the dimensionless length z = L/2, we have
P
P
=
R (1 + R)
.
z
R
R
(40)
(41)
which is satisfied by P (R, z) = z 1 exp(R/z). For this distribution one has hRi = z.
In the opposite limit, R 1, we have
P
2P
P
= R2
+ 2R
2
z
R
R
2P
P
=
+
,
2
(42)
(43)
P (R, z) dR = (4z)
exp
(ln R z)2
4z
(44)
d ln R .
(45)
hiji
where the first sum is over all nearest neighbor pairs and the second
sum is over all next
nearest neighbor pairs. The spin variables Si take values in the set 32 , 12 , + 21 , + 32 .
(a) Making the mean field Ansatz Si = m + (Si m), where m = hSi i is presumed
MF . You may denote z as the
independent of i, derive the mean field Hamiltonian H
1
number of nearest neighbors and z2 as the number of next nearest neighbors of any
site on the lattice.
[15 points]
Solution : As usual, we neglect fluctuations and obtain
X
m2 J(0)
m+H
MF = 1 N J(0)
Si ,
H
2
i
where
=
J(0)
J(R) = z1 J1 + z2 J2 .
m2
12 N J(0)
=e
"
#N
3(J (0) m + H)
J(0) m + H
2 cosh
+ 2 cosh
.
2kB T
2kB T
Thus,
F (m, T, H) =
2
1
2 N J(0) m N kB T
"
#
J(0) m + H
3(J (0) m + H)
ln 2 cosh
.
+2 cosh
2kB T
2kB T
sinh
2 cosh
m+h
+ 3 sinh 3(m+h)
2
2
3(m+h)
m+h
+
2
cosh
2
2
(g) What is the mean field result for |Si | ? Interpret your result in the and 0
limits. Hint : We dont neglect fluctuations from the same site.
[10 points]
Solution : We have
3
2
Note that as we have |S| 1, since all four states are equally probable
and two of them have |S| = 12 and the other two have |S| = 32 . As 0 the ground
state configurations are selected. These are two completely
polarized states, one with
Si = + 23 i and the other with Si = 32 i. Thus |S| 23 in this limit.
|S| =
e3(m+h)2/ +
3
2
hiji
where the first sum is over all links of a lattice of coordination number z. The spin variables
Si take values in the set {1 , 0 , 0 , +1}. Note that there are two distinct states, each with
Si = 0, and a total of four possible states on each site. Taking the trace for a single site
means we sum over the four independent states, one with S = +1, two with S = 0, and
one with S = 1.
(a) Making the mean field Ansatz Si = m + (Si m), where m = hSi i is presumed
MF .
independent of i, derive the mean field Hamiltonian H
[15 points]
Solution : As usual, we neglect fluctuations and obtain
X
MF = 1 N zJm2 (zJm + H)
Si .
H
2
i
12 N zJm2
=e
Thus,
"
2 + 2 cosh
"
zJm + H
kB T
#N
zJm + H
kB T
#
(c) Adimensionalize, writing = kB T /zJ and h = H/zJ. Find the dimensionless free
energy per site f = F/N zJ.
[15 points]
Solution : We have
"
#
m
+
h
f (m, , h) = 21 m2 ln 2 + 2 cosh
"
#
m
+
h
= 12 m2 2 ln 2 cosh
.
2
m+h
2
m(h) =
h
.
2 1
hS 2 i =
Note that as we have hS 2 i 21 , since all four states are equally probable and
two of them have S 2 = +1 and the other two have S 2 = 0. As 0 the ground
state configurations are selected. These are two completely polarized states, one with
Si = +1 i and the other with Si = 1 i. Thus hS 2 i 1 in this limit.
X
i,j
Jij Si Sj H
Si
where Si {2, 1, 0, 1, 2}. The system is on a simple cubic lattice, with nearest neighbor
coupling J1 /kB = 40 K and next-nearest neighbor coupling J2 /kB = 10 K.
(a) Find the mean field free energy per site f (, h, m), where = kB T /J(0), h = H/J(0),
Si .
(1)
J(0)/k
B = z1 J1 /kB + z2 J2 /kB = 360 K ,
(2)
since z1 = 6 and z2 = 12 on the simple cubic lattice. Well need this number in part
), taking the logarithm, and
(c). Computing the partition function ZMF = Tr exp( H
MF
m=
= 0. Thus,
m+h
+ 4 sinh 2m+2h
2m+2h
2 cosh m+h
+
2
cosh
2 sinh
1+
f
m
(4)
To find c , we set h = 0 and equate the slopes of the LHS and RHS of the above equation.
This yields
= 720 K
(c)
= 2 T = J(0)
(5)
c
To find the zero field susceptibility, we assume that m and h are both small and expand the
RHS of the self-consistency equation, yielding
m(h, ) =
(d)
2h
2
() =
2
2
(6)
When < c , we need to expand the RHS of the self-consistency equation to order m3 .
2
4
Equivalently, we can work from f , and using cosh x = 1 + x2 + x24 + . . . , we have
m2
m4
4m2 4m4
1 2
+ ... + 2 +
+ ...
f = 2 m ln 5 + 2 +
12 4
3 4
m2 17 m4
1 2
= ln 5 + 2 m ln 1 + 2 +
+ ...
60 4
2 2 13 m4
m +
+ ... ,
(7)
= ln 5 +
2
60 3
since ln(1 + x) = x
and obtain
(e)
x2
2
x3
3
m2 =
15
13
2 (2 )
60
13
(2 )
q
A = 2 15
13
(8)
In deriving the above result we have assumed c = 2 and worked only to lowest order
in the difference c .
(2) The Landau free energy of a crystalline magnet is given by the expression
f = 12 t m2x + m2y + 14 b1 m4x + m4y + 21 b2 m2x m2y ,
where the constants and b1 are both positive, and where t is the dimensionless reduced
temperature, t = (T Tc )/Tc .
(a) Rescale, so that f is of the form
f = 0 12 t 2x + 2y +
1
4
4x
4y
2 2x
2y
where mx,y = s x,y , where s is a scale factor. Find the appropriate scale factor and
find expressions for the energy scale 0 and the dimensionless parameter in terms
of , b1 , and b2 .
[5 points]
(b) For what values of is the free energy unbounded from below?
[5 points]
(a)
Note that
f=
1
4
2x
2y
b1 x,y
0 =
2
b1
b2
b1
2
1
t
x
1
2
2
+ 2 0 x y
2y
1
t
(9)
(10)
We need to make sure that the quartic term goes to positive infinity when the fields x,y
tend to infinity. Else the free energy will not be bounded from below and the model is
unphysical. Clearly the matrix in the first term
on the RHS has eigenvalues 1 and
1
corresponding (unnormalized) eigenvectors 1
. Since 2x,y cannot be negative, we only
need worry about the eigenvalue 1 + . This is negative for < 1. Thus,
(b)
1 is unphysical
(11)
f
= t + 2x + 2y x = 0
x
f
= t + 2y + 2x y = 0
y
(c2)
(12)
Clearly phase I with x = y = 0 is a solutionto these equations. In phase II, we set one of
the fields to zero, y = 0 and solve for x = t, which requires t < 0. A corresponding
solution exists if we exchange x y . In phase III, we solve
2
t
x
t
1
2x = 2y =
=
.
(13)
t
2y
1
1+
This phase also exists only for t < 0, and > 1 as well, which is required if the free
energy is to be bounded from below. Thus, we find
(d1)
fI = 0
(14)
(d2)
(d3)
(x,II , y,II ) = ( t , 0) or (0 , t) ,
(x,III , y,III ) =
t
1+
(1 , 1) or
t
1+
(1 , 1) ,
fII = 14 0 t2
fIII =
0 t2
2 (1 + )
(15)
(16)
To find the phase diagram, we note that phase I has the lowest free energy for t > 0. For
t < 0 we find
1
,
(17)
fIII fII = 41 0 t2
+1
which is negative for || < 1. Thus, the phase diagram is as depicted in fig. 1.
2
ecp/kB T
where the factor of 2 comes from summing over the two independent polarization states.
(a) Consider a photon gas (in three dimensions) slightly out of equilibrium, but in steady
state under the influence of a temperature gradient T . Write f = f 0 + f and
write the Boltzmann equation in the relaxation time approximation. Remember that
(b) What is the formal expression for the energy current, expressed as an integral of
something times the distribution f ?
[5 points]
(c) Compute the thermal conductivity . It is OK for your expression to involve dimensionless integrals.
[10 points]
Solution : We have
df 0 =
dT
2cp ecp
2cp ecp
d
=
.
cp
2
cp
2
(e 1)
(e 1) kB T 2
f 0
r
f
t
coll
(18)
1
f
2 c2 ecp/kB T
p T =
2
cp/k
T
B
(e
1)2 kB T
(19)
j (r) =
d3p 2
c p f (p, r)
h3
(20)
Integrating, we find
Z
p2 ecp/kB T
2c4
3
d
p
= 3
3h kB T 2
(ecp/kB T 1)2
Z
s4 es
8kB kB T 3
ds s
=
3c
c
(e 1)2
0
4kB
3 2 c
Z
kB T 3
s3
,
ds s
c
e 1
(21)
where we simplified the integrand somewhat using integration by parts. The integral may
be computed in closed form:
Z
In = ds
0
sn
= (n + 1) (n + 1)
es 1
I3 =
4
,
15
(22)
and therefore
(c)
2 kB
=
45 c
kB T
c
3
(23)
(4) At the surface of every metal a dipolar layer develops which lowers the potential energy for electrons inside the metal. Some electrons near the surface escape to the outside,
leaving a positively charged layer behind, while overall there is charge neutrality. The
situation is depicted in fig. 2. The electron density outside the metal is very low and
Maxwell-Boltzmann statistics are appropriate.
2 = 4 = 4en0 ee/kB T
(24)
d
dx ,
(25)
we have
d
dx
1 2
2
d
e/kB T
,
=
4n0 kB T e
dx
(26)
d
= (8n0 kB T )1/2 ee/2kB T .
dx
(27)
Note also the choice of sign here, due to the fact that the potential e for electrons must
increase with x. The boundary term at x = must vanish since n() = 0, which requires
ee()/kB T = 0.
Integrating once more, we have
e(x)/2kB T
2n0 e2
kB T
1/2
(x + a) ,
(28)
(x) =
x+a
2kB T
ln
e
a
with
a=
kB T
2n0 e2
1/2
(30)
n(x) = n0
a
x+a
2
(31)