Solution for Chapter 4

(compiled by Xinkai Wu, revised by Kip Thorne)

A.
Exercise 4.1 Pressure Measuring Device [by Kip Thorne]
We could choose our pressure measuring device (pressure meter) to exchange
only volume and not heat with its reservoir, but that leads to some tricky
subtleties as we see in Exercise 4.8; so it is better (more straightforward) to let
the meter exchange both volume and heat with its reservoir. This is analogous
to the meter that Kip discussed in class on Monday, which exchanged particles
and heat.
Then the probability for the meter to be in a (quantum) state with vol-
ume V and energy

E (probability measured either using an ensemble of
meter&reservoir systems or via the fraction of the time the meter spends in a
given state) is given by the meters Gibbs distribution:
= K exp[(

E PV )/kT] ;
or equivalently
ln = (

E PV )/kT + constant .
Here K is the normalization constant and P and T are the reservoirs pressure
and temperature.
Suppose that the reservoir plus meter has total energy

E
o
, volume

V
o
, and
number of particles N
o
; and the meter plus reservoir is a closed system. Then
by energy and volume conservation, when the meter has volume V and energy
E, the reservoir has volume V
r
= V
o
V and energy E
r
= E
o
E. Denote
by N
r
(

E
r
, V
r
) the total number of states of the reservoir that have energy E
r
,
volume V
r
and particle number N
r
, and similarly N
m
(E, V ) for the meter.
(The meter might not be made of the same kind of particles as the reservoir; it
might, for example, just be a mass-spring system that can be heated; the only
variables we care about for it are energy and volume.) Then the probability
is proportional to the following product of numbers of states for the reservoir
and meter:
N
r
(E
o
E, V
o
V, N
o
) N
m
(E, V ) .
Because the meter is tiny compared to the reservoir, when we take the logarithm
of this , the changes in the meter term, as energy and volume ow back and
forth, are negligible compared to the changes in the reservoir term, so
ln = ln[N
r
(E
o
E, V
o
V, N
o
)] + constant .
This logarithm of the number of reservoir states is proportional to the reservoir
entropy, so
ln = S
r
(E
o
E, V
o
V, N
o
)/k + constant .
1
Expanding to rst order in the meter energy and volume we obtain
ln =
_
S
r
E
r
_
Vr,Nr
E
_
S
r
V
r
_
Er,Nr
V .
Comparing with our previous expression for ln we see that the pressure and
temperature that appear in the meters Gibbs distribution are related to the
reservoirs entropy by the usual reservoir thermodynamic relations
1
T
=
_
S
r
E
r
_
Vr,Nr
,
P = T
_
S
r
V
r
_

Er,Nr
=
_

E
r
S
r
_
Nr,Vr
_
S
r
V
r
_

Er,Nr
=
_

E
r
V
r
_
Nr,Sr
.
In the last step we have used an obvious analog of Eq. (4) of Box 4.1.
B
Exercise 4.3 Enthalpy Representation of Thermodynamics [by Xinkai Wu]
(a) Combining H =

E + PV with the rst law d

E = TdS PdV + dN one
immediately nds dH = V dP +TdS+ dN. Having this expression we can com-
pute V, T, by dierentiating H w.r.t. P, S, N, respectively, namely H serves
as the fundamental potential in this case.
(b) In the rst law derived in part (a), theres dS, dN but no dV . Thus the only
extensive quantities exchanged between the system and the bath are volume and
energy; and the energy exchange is xed completely by the volume exchange,
d

E = PdV , i.e., it is entirely mechanical energy exchange. There is no ex-
change of heat and no exchange of particles. A physical situation which could
produce this ensemble is: the system is put in a chamber, with a frictionless
piston separating the system and the bath, where the walls and the piston are
impermeable and heat-insulating.
(c) The equations of state read o from the enthalpy rst law and the associated
Maxwell relations are:
V =
_
H
P
_
S,N
, T =
_
H
S
_
P,N
, =
_
H
N
_
P,S

2
H
PS
=
_
V
S
_
P,N
=
_
T
P
_
S,N
=

2
H
SP

2
H
SN
=
_
T
N
_
P,S
=
_

S
_
P,N
=

2
H
NS

2
H
NP
=
_

P
_
S,N
=
_
V
N
_
P,S
=

2
H
PN
(d) adding up small subsystems, we get
H = TS + N
2
(e) In Exercise 2.4 it is shown that the inertial mass per unit volume is

inert
ji
= T
00

ji
+ T
ji
for isotropic system
=
ji
+ P
ji
= ( + P)
ij
.
Thus the total inertial mass is isotropic and is given by V +PV =

E+PV = H.
(f) To create the sample, we need

E, the samples energy. To inject the sample
into the system, we need to perform an amount of work PV on the system.
Thus the total energy required to create the sample and perform the injection
is

E + PV , which is just the samples enthalpy. Thus, enthalpy has the
physical interpretation of energy of injection at xed volume V.
Exercise 4.2 Energy Representation for a Nonrelativistic Monatomic Gas [by
Xinkai Wu]
(a) To derive eqn. (4.18), one only has to substitute the fundamental potential
E as given by eqn. (4.15) into eqn. (4.16) and perform the partial dierentia-
tion. We assume this is trivial for everyone and omit the details here.
(b) Again, to verify eqn. (4.17) is simple dierentiation and we omit the details
here.
(c) Expressing exp
_
2S
3kN

5
3
_
in terms of (T, N, V ) using the second equation
of (4.18), then substituting the resulting expression into the rst equation of
(4.18), one readily gets P =
N
V
kT.
C
Exercise 4.4 Latent Heat and the Clausius-Clapeyron Equation [by Xinkai Wu]
(a) For xed temperature and pressure, the change in the Gibbs potential is
dG =
a
dN
a
+
b
dN
b
= (
a

b
)dN
a
where weve used the fact that dN
b
= dN
a
.
By the minimum principle for the Gibbs potential, N
a
will decrease when

a
>
b
, and increase when
a
<
b
. As a consequence, if phases a and b are in
equilibrium with each other, one must have
a
=
b
(subtracting m
H2O
c
2
from
both sides one nds
a
=
b
.) The Gibbs potential G is a function of (T, P, N),
so the chemical potential =
_
G
N
_
T,P
is also a function of (T, P, N). However,
is an intensive quantity and consequently cant depend on the extensive quan-
tity N, namely is a unique function of T and P. The above reasoning is valid
for any phase a. In a two-phase region, we must have
a
(T, P) =
b
(T, P), which
is one equation for two variables T and P, giving a line in the T P plane. In
a three-phase region,
a
(T, P) =
b
(T, P), and
b
(T, P) =
c
(T, P), which are
two equations for the two variables T and P, giving a point in the T P plane.
(b) The melting curve P = P(T) is determined by
ice
(T, P) =
water
(T, P).
Then taking temperature derivative of the above identity along the melting
3
curve gives
_

ice
T
_
P
+
_

ice
P
_
T
_
dP
dT
_
melt
=
_

water
T
_
P
+
_

water
P
_
T
_
dP
dT
_
melt

_
dP
dT
_
melt
=
__

water
T
_
P

ice
T
_
P
_
/
__

ice
P
_
T

water
P
_
T
_
Using the Maxwell relations that one reads o of dG = SdT + V dP + dN,
we nd
_

I
T
_
P
=
_
S
I
N
I
_
T,P
= m
H2O
s
I
_

I
P
_
T
=
_
V
I
N
I
_
T,P
= m
H2O
/
I
where I denotes either phase(ice or water), s
I
,
I
are the entropy per unit mass
and density for phase I, respectively. Thus
_
dP
dT
_
melt
=
m
H2O
(s
water
s
ice
)
m
H
2
O
ice

m
H
2
O
water
= (s
water
s
ice
)

ice

water

ice

water
=
q
melt
T
_

ice

water

ice

water
_
(c) In equilibrium we still have
water
(T, P
water
) =
vapor
(T, P
vapor
). However,
now P
water
is no longer equal to P
vapor
. Instead, P
water
= P
total
= P
vapor
+P
gas
.
So we have
water
(T, P
total
) =
vapor
(T, P
vapor
). Fixing T and dierentiating
the above identity w.r.t. P
total
, we have
_

water
P
total
_
T
=
_

vapor
P
vapor
_
T
_
dP
vapor
dP
total
_
T

_
dP
vapor
dP
total
_
T
=
_

water
P
total
_
T
/
_

vapor
P
vapor
_
T
=
m
H2O

water
/
m
H2O

vapor
=
vapor
/
water
Exercise 4.5 Electron-Positron Equilibrium at Low Temperatures [by Xinkai
Wu]
(a) The reaction equation e

+ p e

+ p + e

+ e
+
gives
e
+
p
=
2
e
+
p
+
e
+, which implies
e
+
e
+ = 0, i.e.
e
=
e
+.
(b) In what follows we shall use

to denote
e
, and
+
to denote
e
+. The
distribution function(density in phase space) for positrons and electrons (both
4
having g
s
= 2) is given by
N

=
2
h
3

=
2
h
3
1
e
(

E )/kT
+ 1

E mc
2
+p
2
/2m
N

=
2
h
3
1
e
(
p
2
2m
+mc
2
)/kT
+ 1
The density in coordinate space n

is given by
_
d
3
pN

. We know that n

>
n
+
because while positrons and electrons are created in pairs we also have
ionization electrons from hydrogen atoms. Thus we must have

>
+
. This
inequality combined with

+
+
= 0 gives

> 0,
+
< 0.
(c) In the dilute-gas regime,

exp
_
(
p
2
2m
+ mc
2

)/kT
_
. Its trivial to
perform the momentum-space integral and nd
n

=
2
h
3
(2mkT)
3/2
exp
_

mc
2
kT
_
<< 1 means (

mc
2
) << kT (see Chapter 2), i.e.
n

<<
2
h
3
(2mkT)
3/2
= 4.8 10
27
_
T
10
8
K
_
3/2
cm
3
The hydrogen mass density is n

m
p
(because m
e
<< m
p
and also electron-
positron pair generation is negligible.) Thus the dilute-gas region is given by
<< 8 10
3
_
T
10
8
K
_
3/2
g cm
3
(d) Dene x e
/kT
, then e
+/kT
= 1/x. And using the expression for n

we
found in part (c), we get
n = n

n
+
=
2
h
3
(2mkT)
3/2
e
mc
2
/kT
(x 1/x)
define y
1
4
n
3
e
mc
2
/kT
, where
h

2mkT
2y = (x 1/x)
x = y + (1 + y
2
)
1/2
Thus
n
+
n
=
1/x
x 1/x
=
1
2y
_
y + (1 + y
2
)
1/2

(e) [this part by Kip Thorne] The amount of phase space available to the ioniza-
tion electrons, per ionization electron, is of order V
phase
(2mkT)
3/2
mp

. If
one pair forms per ionization electron (our criterion for signicant pair forma-
tion), then each positron will also have available to it the phase volume V
phase
.
5
When this V
phase
is very large (i.e. when is very small, at xed T 10
8
K),
then the positron is almost completely unaware of the presence of the electrons,
so the probability of its meeting an electron and annihilating is very small. This
means that, although pair production occurs only very rarely (because of the
very few number of photons with sucient energy to produce a pair), once it
has occured, on average the positron lives for a very long time. It is this balanc-
ing of the long life, due to the huge phase space per particle, against the rarity
of pair production, that enables signicant pairs to form at T as low as 10
8
K.
D
Exercise 4.7 Fluctuations and Phase Transitions in a Van der Waals Gas [by
Xinkai Wu]
(a)
P =
_
E
V
_
S,N
=
h
2
2m
_
V
N
b
_
5/3
exp
_
2S
3kN

5
3
_

aN
2
V
2
T =
_
E
S
_
N,V
=
h
2
2mk
_
V
N
b
_
2/3
exp
_
2S
3kN

5
3
_
use the second equation to eliminate S in favor of N, V, T
P(N, V, T) =
kT
V/N b

a
(V/N)
2
=
kT
v b

a
v
2
where v V/N
while for an ideal gas
P(N, V, T) =
NkT
V
=
kT
v
One needs to solve the equation
_
P
v
_
T
= 0 for v. One nds that for T < T
crit
,
there are two solutions v
1
, v
2
(namely theres a local maximum and a local
minimum); when T = T
crit
, v
1
and v
2
merge into a single v
crit
; and when
T > T
crit
, theres no solution. Now lets determine T
crit
and v
crit
. At T
crit
, v
crit
,
we have
0 =
_
P
v
_
T
=
kT
(v b)
2
+
2a
v
3
0 =
_

2
P

2
v
_
T
=
2kT
(v b)
3

6a
v
4
Combining the above two equations, we nd T
crit
=
8a
27bk
=
8
27
T
0
, and v
crit
= 3b.
(b) For the system to be stable against volume uctuations under isother-
mal conditions, we must have
_
P
v
_
T
< 0 (the arguement for this is: sup-
pose initially our system is in equilibrium under the external pressure, namely
P
system
= P
external
. If by statistical uctuation the systems volume decreases
by a small amount v > 0, P
system
will correspondingly change by P
system
=
v
_
P
v
_
T
. If
_
P
v
_
T
> 0, then P
system
< 0, namely P
system
will drop below
P
external
and the system will keep collapsing. A similar reasoning can be car-
ried out when the volume increases a little bit by uctuation. Thus we conclude
6
_
P
v
_
T
> 0 would lead to instability against volume uctuations.)
So we can look at the slope of the curves in Fig. 4.7(a) and conclude that: when
T > T
crit
, the gas is stable all along the curve; when T = T
crit
, the gas is stable
except at the one particular point v = v
crit
; when T < T
crit
, the gas is stable in
two regions (and these are the two dierent phases that are both stable against
volume uctuations) b < v < v
min
and v > v
max
, where v
min
and v
max
are the
locations of the local minimum and the local maximum, respectively, and the
gas is unstable in the region v
min
< v < v
max
.
(c) By the principle of minimum Gibbs potential, the chemical potential at
point A must be equal to that at point B (see Exercise 4.4). Recall that be-
ing on the same isothermal curve A and B have the same temperature. Thus
(T, P
A
) = (T, P
B
) gives P
A
= P
B
, i.e. the straight line from A to B is hori-
zontal.
We have the following dierential equation
dG = d(N) = SdT + V dP + dN
Nd = SdT + V dP
d = sdT + vdP, where s S/N, v V/N
Along the isothermal curve, dT = 0, so d = vdP. Thus
|
B
|
A
=
_
B
A
d =
_
B
A
vdP
Easily seen,
_
B
A
vdP is just the dierence between the areas of the two stippled
regions. Thus |
A
= |
B
tells us that these two areas are equal.
(d) The fractional volume uctuation is given by

V
V
=

kT
V
=

kT

_
P
V
_
T,N
V
2
;
see Eq. (4.80) and the paragraph that follows it. We see that small values of

_
P
V
_
T,N
correspond to large volume uctuations. Thus near the critical
point (with T
crit
=
8a
27kb
, V
crit
= 3Nb, P
crit
=
a
27b
2
),
_
P
V
_
T,N
is very small
and the gas exhibit huge volume uctuations. (For T < T
crit
, near the the local
minimum and maximum on the P V plot, we also have small
_
P
V
_
T,N
.
However, as argued in part (b) and (c), these regions are not physical and in
general dont exist in nature. Also,
_
P
V
_
T,N
gets small when V gets large.
But
V
/V actually doesnt get huge, because
_
P
V
_
T,N
NkT/V
2
for large
V , and
V
/V remains 1/

N.)
Exercise 4.8 Fluctuations of Systems in Contact with a Volume Bath [by Xinkai
Wu]
(a) No matter whether the ensemble is in equilibrium or not, the (very gen-
eral) energy conservation law always holds. Since theres no heat or particle
7
exchange between the system and the bath, the change in the systems energy
is completely due to the work performed on the system by the bath, namely
d

E = P
b
dV . (Note that since the bath is huge compared to the system, the
baths pressure P
b
remains xed, so

E = P
b
V for nite volume change.)
This implies that H =

E + P
b
V = 0, i.e. the systems enthalpy is con-
served.
(b) Since the system is very small compared to the bath, the bath is always
in equilibrium. Then we can use the rst law for the bath

E
b
= T
b
S
b

P
b
V
b
+
b
N
b
. Since theres no particle exchange, N
b
= 0. Also by energy
and volume conservation,

E
b
=

E = P
b
V = P
b
V
b
. These combined
with the rst law tells us T
b
S
b
= 0, i.e. S
b
= 0. So we conclude that inter-
action with a system cannot change the baths entropy.
(c) The bath and the system form a closed supersystem with a total entropy
S
total
= S
b
+ S. This supersystem, like any closed system, must evolve toward
increasing S
total
. This combined with the fact that S
b
doesnt change (proved
in part (b)) tells us S always increases. When the supersystem reaches statisti-
cal equilibrium, it is in the microcanonical ensemble with
total
= e
S
total
/k
=
e
(S
b
S)/k
= const. This, combined with S
b
= const, gives us = e
S/k
=
const for the regions of phase space that have enthalpy H in the small range
H.
(d) For uctuations away from equilibrium, the probability e
S
total
/k
e
S/k
.
Note that the enthalpy H is a function of (P, S, N) (see Exercise 4.3). Inverting
this relation we get S = S(H, P, N). So we still have equations (4.65), (4.66)
but with E replaced by H, V replaced by P.
8