Comprehensive Examination

Department of Physics and Astronomy

Stony Brook University

January 2019

General Instructions:

Three problems are given. If you take this exam as a placement exam, you must work on all
three problems. If you take the exam as a qualifying exam, you must work on two problems
(if you work on all three problems, only the two problems with the highest scores will be
counted).

Each problem counts for 20 points, and the solution should typically take approximately
one hour.

Use one exam book for each problem, and label it carefully with the problem topic and num-
ber and your ID number.

Write your ID number (not your name!) on each exam booklet.

You may use, one sheet (front and back side) of handwritten notes and, with the proc-
tor’s approval, a foreign-language dictionary. No other materials may be used.

1
Classical Mechanics 1

Linked rods in a plane

Two massless rods lie in the x ´ y plane (neglect gravity). Each rod has length `, and
each rod has mass M at one end (points A and B below). The massless end of the first rod
is hinged to a fixed pivot O, while its massive end is hinged at the point A to massless end of
the second rod as shown below. The hinges at O and A are constructed so that the second
rod can swing past past the first without obstruction.
y
B

A
O
x
(a) (7 points) Write down the Lagrangian of the system. Determine and interpret any
integrals of motion.

(b) At time t “ 0 both rods are aligned along the x axis. The mass at A moves with
velocity v0 , while the mass at B moves with velocity 2v0 ` ∆v, with ∆v ! v0 .

i (3 points) Qualitatively describe the subsequent motion of the system.

ii (10 points) Determine the angles of the masses relative to the x axis as functions
of time.

2
Solution:

(a) First we find the Lagrangian of the system. The generalized coordinates are φ1 and φ2 .
The coordinates of a point in A are

rA “ p` cospφ1 q, ` sinpφ1 qq , (1)

while the coordinates of B are

rB “ xA ` `pcospφ1 ` φ2 q, sinpφ1 ` φ2 qq . (2)

Here and below we will write φ3 “ φ1 ` φ2

Thus, the velocities are

vA “`p´ sinpφ1 q, cospφ1 qqφ9 1 , (3)

and

vB “vA ` `p´ sinpφ3 q, cospφ3 qqpφ9 3 q , (4)

respectively. The kinetic energy is 21 mvA2 ` 12 mvB2 yielding after some algebra
ˆ ˙
1 2 92 1 2 92 1 2 92 2 9 9
T “ m` φ1 ` m` φ1 ` m` φ3 , `m` cospφ3 ´ φ1 qφ1 φ3 , (5)
2 2 2
ˆ ˙
1 2 92 1 2 92 1 2 9 2 2
“ m` φ1 ` m` φ1 ` m` pφ1 ` φ9 2 q ` m` cospφ2 qφ9 1 pφ9 1 ` φ9 2 q , (6)
2 2 2
1 1
“ m`2 φ9 21 p3 ` 2 cospφ2 qq ` m`2 p1 ` cos φ2 qφ9 1 φ9 2 ` m`2 φ9 22 . (7)
2 2

We have two integrals of motion. The first is the total energy of the system.

h “ p1 φ9 1 ` p2 φ9 2 ´ L . (8)

We have

p1 “m`2 p3 ` 2 cospφ2 qqφ9 1 ` m`2 p1 ` cos φ2 qφ9 2 , (9)

p2 “m`2 p1 ` cospφ2 qqφ9 1 ` m`2 φ9 2 , (10)

The Hamiltonian function reduces to simply to the kinetic energy

1 1
h “ T “ m`2 φ9 21 p3 ` 2 cospφ2 qq ` m`2 p1 ` cos φ2 qφ9 1 φ9 2 ` m`2 φ9 22 , (11)
2 2
as one could expect.
The second is associated with the cyclic coordinate φ1 , which is the total angular mo-
mentum:
BL
J“ “ p1 . (12)
B φ9 1

3
 (i) It is perhaps a good place to use some intuition. If the total angular momentum is
large compared to the internal angular momentum. There will be a strong centrifugal
force which will tend to align rod B with rod A. The resulting sytem will exhibit small
oscillations around this configuration, while the center of mass will move at a constant
rate around the circle.
(ii) We should expand near φ2 = 0. φ9 1 is not small, but φ9 2 and φ2 are small. The Lagrangian
in this case is
1 1
L “ m`2 φ9 21 p5 ´ φ22 q ` 2m`2 φ9 1 φ9 2 ` m`2 φ9 22 . (13)
2 2
The angular momentum is
J “ 5m`2 φ9 1 ` 2m`2 φ9 2 . (14)
The equation of motion for φ2 is
d
pm`2 φ9 2 ` 2m`2 φ9 1 q “ ´m`2 φ9 21 φ2 . (15)
dt
On the RHS, which is already small, we can approximate
J
φ9 1 “ , (16)
5m`2
so the RHS can be written
J2
´ m`2 φ9 21 φ2 “´ φ2 . (17)
25m`2
The LHS can also be approximated. Since
dJ
“ 0, (18)
dt
we have
5m`2 φ:1 ` 2m`2 φ:2 “ 0 . (19)
So the LHS reads
4
m`2 φ:2 ` 2m`2 φ:1 “ m`2 φ:2 p1 ´ q . (20)
5
Then the equation of motion for φ2 is

1 2: J2
m` φ2 “ ´ φ2 . (21)
5 25m`2
Thus the motion is periodic with frequency

J2
ω02 “ . (22)
5pm`2 q2

Since
J » 5m`v , (23)

4
we have
5v 2
ω02 “ . (24)
`2
Putting together the results we have

φ2 “A sinpω0 tq ` B cospωtq (25)

J 29
φ9 1 “ ´ φ2 (26)
5m`2 5
The second equation is integrated to yield
Jt 2
φ1 “ 2
´ φ2 ` const (27)
5m` 5
At time t “ 0, φ1 “ φ2 “ 0, so our solutions can be written
C
φ2 ptq “ sinpω0 tq (28)
ω0
2
φ1 ptq “Dt ´ φ2 (29)
5
Then in order to match the initial condition we need
v0 ∆v
φ9 1 “ φ9 2 “ (30)
` `
So finally
∆v
φ2 ptq “ sinpω0 tq (31)
`ω0
ˆ ˙
v0 2∆v 2
φ1 ptq “ ` t ´ φ2 (32)
` 5` 5

5
asses
m1 = 2m and m2 = m, are suspended in a uniform gravita-
by identical massless springs with spring constant k. Assume
Classical Mechanics 2
ical motion occurs, and let z1 and z2 denote the vertical dis- k
the masses from their equilibrium positions.
2m
Coupled pendulums z1
the Lagrangian and find the resulting equations of motion.
k
mass m suspended (in a uniform gravitational
the subsequent Two identical
motion givenpendulums each have a zpoint
initial conditions 2 = h, and
field) on a massless rigid rod of length ` hanging from a frictionless pivot. The pendulums
ż2 = 0 at time t = 0.
are constrained to swing in the x ´ z plane, and are mounted on a block of mass M “ 2m
m
which is free to slide, without friction, in the x-direction. z2
ndula
pendula each have a point
nded (in a uniform grav-
) on a massless rigid rod
nging from a frictionless
ndula are constrained to
iform gravitational field) z
ne, and are mounted on a
M = 2m which is free to x
friction, in the x-direction.
(a) (10 points) Construct a Lagrangian of the system assuming small oscillations: (i) find
thethe
the Lagrangian of resulting normal
system. modes and
Simplify frequencies,
as much (ii) interpret any zero modes, and (iii) qual-
as possible.
itatively sketch the oscillation pattern for each mode.
he Lagrangian further, assuming small oscillations of the pendula. Then find
(b) (6and
ing normal modes points) base (the block with mass 2m) is pushed in the x-direction by
Now thefrequencies.
associated
an external force F ptq “ P0 δptq. Determine thee subsequent motion of the system,
assuming that the impulse P0 is so small that the subsequent oscillations may be
treated in a harmonic approximation.

(c) (4 points) Now consider a time dependent force pushing the base in the x-direction:

F ptq ” F0 e´|t|{τ , (33)

where τ ą 0 is some characteristic time scale. Assuming that the system is at rest for
t Ñ ´8, determine the total work done on the system by the time dependent force as
t Ñ `8.

6
Solution:

(a) First we coordinate the system. Choosing X to be the bottom of the center of mass of
the base, and φ1 and φ2 to be the angles. The coordinates of the first mass is

x1 “X ` ` sinpφ1 q » X ` `φ1 ` Opφ31 q , (34)

φ2
y1 “`p1 ´ cospφ1 q » ` , (35)
2
with an analogous expression for the second mass. Here we have chose y “ 0 to be the
resting point of the two masses.
The kinetic energy
1 1 1
T “ M X9 2 ` mpX9 ` `φ9 1 q2 ` mpX9 ` `φ9 2 q2 . (36)
2 2 2
The potential energy is just
U “ mgy1 ` mgy2 . (37)
So the total Lagrangian is
1 1 1
L “ mX9 2 ` mpX9 ` `φ9 1 q2 ` mpX9 ` `φ9 2 q2 ` mg`2 pφ21 ` φ22 q . (38)
2 2 2
First we note that there is a zero mode associated with the x translation of the system
BL
PX “ “ 4mX9 ` m`φ9 1 ` m`φ9 2 . (39)
B X9
This is clearly the momentum of the center of mass.
To find the equation of motion of the remaining coordinates it is useful to eliminate X
(which is cyclic) and to focus on φ1 and φ2 . This can be done using the Routh procedure.
The Routhian, ´R ” Leff , is
Leff “ L ´ pX X9 , (40)
and gives an effective Lagrangian for the remaining coordinates and a Hamilton equation of
motion for X.
In addition, since only the sum pφ1 ` φ2 q{2 “ φs couples to the center of mass, it is
convenient to change coordinates to φs and φ∆ “ pφ2 ´ φ1 q{2. This is also motivated by
symmetry considerations. We note the following

φ21 ` φ22 “2φ2s ` 2φ2∆ , (41)

φ9 21 ` φ9 22 “2φ9 2s ` 2φ9 2∆ . (42)

We also note that the velocity of the center of mass is

PX “4mX9 ` 2m`φ9 s . (43)

(44)

7
 The effective Lagrangian evaluates to
1 ´ ¯ 1 1
Leff “ m`2 φ9 21 ` φ9 22 ´ mg`2 pφ21 ` φ22 q ´ p4mqX9 2 , (45)
2 2 2
pPx ´ 2m`φ9 s q2
“m`2 φ9 2s ´ mg`φ2s ` m`2 φ9 2∆ ´ mg`φ2∆ ´ . (46)
2p4mq

Expanding the factor

pPX ´ 2m`φ9 s q2 PX2 2Px p2m`φ9 s q p2mq2 `2 φ9 2s

´ “´ ` ´ , (47)
2p4mq 2p4mq 8m 8m
PX2 1 dPX 9 1 2 9 2
“´ ´ `φ ´ m` φs . (48)
2p4mq 2 dt 2

Recognizing that PX is constant, the effective Lagrangian is

1
Leff “ m`2 φ9 2s ´ mg`φ2s ` m`2 φ9 2∆ ´ mg`φ2∆ ` const . (49)
2
From here the normal modes are simple. The first EOM

m`2 φ:s ` 2mg`φs “ 0 , (50)

yielding a resonance frequency (squared) of

g
ω12 “ 2 . (51)
`
The second normal mode associated with φ∆ is a simple pendulum
g
ω22 “ . (52)
`
(b) Now we add a force. After integrating the force on the block from t “ ´ to t “ `, the
momentum of the base is
p2mqX9 “ P0 . (53)
The momentum of the balls is zero

mx9 1 “ mX9 ` m`φ9 1 “0 , (54)

mx9 2 “ mX9 ` m`φ9 2 “0 . (55)

This means that just after the impulse

PX “P0 , (56)
P0
φ9 s “ ´ , (57)
2m`
φ9 ∆ “0 . (58)

8
 The subsequent oscillator motion of φs is just given by a simple harmonic oscillator result

P0 sinpω1 tq
φs ptq “ ´ , (59)
2m` ω1
with initial conditions described above. Similarly PX “ P0 “ const and φ∆ “ 0.
We can find Xptq from the Eq. (43) for the center of mass and our result for φs ptq

P0 “ 4mX9 ´ P0 cospω1 tq , (60)

Thus,
P0
X9 “ p1 ` cospω1 tqq , (61)
4m
which is easily integrated to find
P0 P0
Xptq “ t` sinpω1 tq . (62)
4m 4mω1
(c) Let us find the motion for an arbitrary force F ptq. In general
ż8
Xptq “ dt0 GR pt ´ t0 qF pt0 q . (63)
´8

One should realize that the Green function is the response to an impulsive force δpt ´ t0 q,
which was already found in part (b). We can use this to write
1 1
GR pt ´ t0 q “ pt ´ t0 q ` sinpω1 pt ´ t0 qq . (64)
4m 4mω1
Its derivative is
1 1
Bt GR pt ´ t0 q “ ` cospω1 pt ´ t0 qq . (65)
4m 4m
The work done is ż
9
W “ dtF ptqXptq , (66)

leading to
ż ż
W “ dt dt0 F ptqBt GR pt ´ t0 qF pt0 q . (67)

The Green function derivative consists of a constant, 1{4m, and cos term
ˆż ˙2 ż ż
1 1
W “ dtF ptq ` dt dt0 F ptq cospω1 pt ´ t0 qq F pt0 q . (68)
4m 4m

Writing the cospω1 pt ´ t0 qq “ peiω1 pt´t0 q ` e´iω1 pt´t0 q q{2 and performing the integrals, we find

1 1
W “ |F̃ p0q|2 ` |F̃ pω1 q|2 , (69)
4m 4m

9
where ż
F̃ pωq “ dteiωt F ptq , (70)

is the Fourier transform of F ptq. Thus the work is only sensitive to the spectral density of
the force at the resonant frequencies of the system, ω “ 0 and ω “ ω1 .
For the current case the Fourier transform is
2F0 τ
F̃ pωq “ (71)
1 ` pωτ q2

10
Classical Mechanics 3

Scattering between two particles

An incoming particle (particle A) of mass m and velocity ~v and collides with a free
particle (particle B) also of mass m, which is initially at rest in the laboratory frame. The
two particles interact with a repulsive 1{r2 potential
h
U prq “ , h ą 0,
r2
where ~r “ ~rA ´ ~rB is the relative coordinate of the two particles.
Find the angular distribution (the differential cross-section) of the deflected particles,
dσpθq{dθ, where θ is the angle between ~v and particle A’s velocity after the collision. Assume
that the process is repeated many times with random impact parameter b, distributed so
that the number of incident particles per transverse area is constant. In steps:

(a) (3 points) Write down the Lagrangian of the two particles. Introduce the center-of-
mass and relative coordinates of the system, and show that the equation of motion of
the relative coordinate ~r is that of a single (effective) particle in the potential U prq.
Express the energy E and angular momentum ` of the relative motion with respect to
the origin of ~r in terms of the “laboratory” frame quantities m, v and b.

(b) (6 points) Find the trajectory rpφq, where the φ is the polar angle, of the effective
particle with energy E and angular momentum ` moving in the potential U prq. Follow
the convention that φ “ 0 corresponds to the distance of closest approach as shown
below.

(c) (6 points) Find the differential cross-section dσpχq{dχ in the center-of-mass frame,
where χ is the scattering angle in this frame.

φ=0

r(φ) χ
b
~r = 0

Figure 1: The scattering angle χ in the center-of-mass frame.

(d) (5 points) Convert dσpχq{dχ into the initial “laboratory” frame to find dσpθq{dθ.

11
Solution

(a) Introducing the center-of-mass and relative coordinates: ~rc.m. “ p~r1 ` ~r2 q{2, and ~r “
~r1 ´ ~r2 , we see that the initial relative velocity is ~r9 “ ~v . The reduced mass of the relative
motion is m{2. Therefore, the energy and the angular momentum of the relative motion are:
mv 2 mvρ
E“ , l“ .
4 2

(b) To find the trajectory, we evaluate the standard integral for the polar angle φ for motion
in the central potential U prq:
żr ż 1{r
l dr{r2 du
φ“ ? “ ´
m rE ´ ph ` l2 {mq{r2 s1{2 rpmE{l2 q ´ p1 ` mh{l2 qu2 s1{2
1 ´ 1{r ¯
“ arccos ‰1{2 .
p1 ` mh{l2 q1{2
“
E{ph ` l2 {mq
The integration constant here was taken to be zero, the choice that makes φ “ 0 the peri-
center. Finally, the trajectory is given by the following equation:
1 ´ E ¯1{2
cos p1 ` mh{l2 q1{2 φ .
“ ‰
“ 2
r h ` l {m

(c) This equation shows that the angle 2φ0 between the incoming and outgoing directions
of the particle scattered by the 1{r2 potential is:
π
2φ0 “ .
p1 ` mh{l2 q1{2
This means that the impact parameter ρ “ 2l{mv depends on the scattering angle χ “ π´2φ0
as
ph{Eq1{2 pπ ´ χq
ρ“ “ ‰1{2 .
2πχ ´ χ2
Hence,
dρ ph{Eq1{2 π 2
“ ´“ ‰3{2 ,
dχ 2πχ ´ χ2
and the differential cross section dσpχq{dχ is:
dσpχq ρ ˇˇ dρ ˇˇ h π 2 pπ ´ χq 4h π 2 pπ ´ χq
“ ˇ ˇ“ “ , χ P r0, πs .
dχ sin χ dχ E sin χ χ2 p2π ´ χq2 mv 2 sin χ χ2 p2π ´ χq2

(d) To convert this result into the laboratory frame, one needs to find the velocity of the
scattered particle ~r91 “ ~r9c.m. ` ~r9 {2. Since ~r9c.m. “ ~v {2, we see from this relation that the
scattering angles in the center-of-mass and the laboratory frames are related simply as

θ “ χ{2 .

12
Equating the number of scattered particles in the two frames,

dσ dσpθq
sin χdχ “ sin θdθ ,
dχ dθ
and using the relation between the scattering angles, we find finally:

dσpθq dσ ˇˇ h π 2 pπ ´ 2θq
“ 4 cos θ ˇ “ , θ P r0, π{2s .
dθ dχ χ“2θ 2mv 2 sin θ θ2 pπ ´ θq2

Similarly, one can show that the angular distribution of the particles that initially were at
rest is the same.

13
Electromagnetism 1

Radiation from an undulator

a electron of charge q, mass m, and velocity v, propagating in

Consider an ultra-relativistic
the z direction with γ”1{ 1 ´ v 2 {c2 " 1. In the lab frame F0 , the electron experiences a
weak external sinusoidal magnetic field directed along the x axis:

Bpzq “ B0 sinpkzq x̂ . (1)

The electron is only slightly deflected from its straight line motion as it propagates in the
magnetic field.
In this problem we will compute the average energy radiated per unit time by the undu-
lating electron in two ways. In parts paq, pbq, pcq we will work in a frame F moving (with the
electron) at constant speed v in the z direction relative to the lab. In part pdq we will work
directly in the lab frame F0 .

(a) (4 points) Explicitly determine the external electromagnetic field in the moving frame
F by making a Lorentz transformation. Compute the instantaneous Poynting vector
in F (both magnitude and direction). Show that the transformed fields are equivalent
to a plane wave. What is the wavelength and amplitude associated with these fields?

(b) (6 points) Determine the average energy radiated per unit time by the electron in frame
F.

(c) (4 points) Show that the energy radiated per unit time by an accelerating charged
particle is invariant under Lorentz boosts in the z direction. What then is the average
energy radiated per unit time in the lab frame F0 by the undulating electron?
Hint: Boost the radiated energy and momentum in a time interval ∆t from the insta-
neous rest frame of the accelerating particle to a frame moving in the z direction.

(d) (6 points) The relativistic Larmor formula for the total energy W radiated per retarded
time T is ˆ 2 ˙
dW 2 q ` 6 2 4 2
˘
“ γ ak ` γ aK . (2)
dT 3 4πc3
Here ak is the acceleration of the electron parallel to its velocity, and aK is the accel-
eration perpendicular to its velocity.
Working in the lab frame, use Eq. (2) with the appropriate kinematic approximations
to determine the average energy lost per time by the relativistic electron. Compare
the result to that of part pcq.

14
Solution

(a) We use the transformation rule

F µν pXq “ Lµρ Lνσ F ρσ pXq , (3)

to determine the field strength in the electron’s frame. The only non-zero component of F ρσ
is F yz “ ´F zy “ Bx , which follows from the relation, F ij “ ijk Bk . The matrix pLqµν “ Lµν
takes the form ¨ ˛
γ 0 0 ´γβ
˚ 0 1 0 0 ‹
L“˚ ˝ 0
‹, (4)
0 1 0 ‚
´γβ 0 0 γ
and thus
F µν “ Lµy Lνz F yz ´ Lµz Lνy F yz . (5)
Looking at this formula, one of the indices in F µν needs to be y. Thus, the non-zero
components are E y “ F 0y and B x “ F yz :

E y pXq “ F 0y “ ´ L0spz Lyy F yz “ γβBx pXq , (6)

x yz y z yz
B pXq “ F “Ly Lz F “ γBx pXq . (7)

We also need to express the coordinates of the lab frame in terms of the electrons coordinates,
X “ pL´1 qX . This yields

ct “γpctq ` γβz , (8)

z “γβpctq ` γz . (9)

Thus, since Bx pXq “ B0 sinpkzq we find

Bx pXq “ B0 sin pkγvt ` kγzq , (10)

and the fields in the ultra relativistic limit β » 1 read

E “ γB0 sin pkγ pz ` ctqq ŷ , (11)

B “ γB0 sin pkγ pz ` ctq x̂ . (12)

We note the similarity between these fields and a plane wave of light with wavenumber
kγ (wavelength λ“2π{kγ) moving in the negative z direction with amplitude γB0 . The
Poynting vector S “ cE ˆ B is

S “ ´cγ 2 B02 sin2 pkγ pz ` ctqq ẑ , (13)

and is proportional to γ 2 .

15
(b) In frame F the electric field of the incident plane wave is

Ept, xq “ Ee´iωt`ikz ŷ , (14)

where it is conventionally understood that we are to take the real part of this expression.
From part paq, the amplitude of the electric field is E “ γB0 .
For an electron at z “ 0, the acceleration is given by Newton’s law
qE ´iωt
aptq “ e ẑ ” aω e´iωt . (15)
m
The radiated power is
2 q 2 aptq2
P “ , (16)
3 4π c3
and thus the time averaged radiated power reads
1 q 2 |aω |2
P “ . (17)
3 4π c3
Using the result from paq that E “ γB0
ˆ ˙ 4
1 q
P “ γ 2 B02 . (18)
12π m2 c3

(c) In the rest frame, the energy radiated in a time ∆t is ∆P 0 “ ∆E{c. The momentum
radiated in ∆t is zero, ∆P i “ 0. Over this time interval the change in the electron’s four
vector ∆X µ is p∆X 0 , ∆X i q “ pc∆t, 0q. We can now boost these four vectors in another
frame.
Boosting the momentum in Z direction we find

∆P 0 “γ∆P 0 , (19)
∆P z “ ´ γβ∆P 0 . (20)

Boosting the coordinates yields

∆X 0 “γ∆X 0 , (21)
∆X z “ ´ γβ∆X 0 . (22)

Thus, in the new frame the energy radiated per time is

∆P 0 γ∆P 0 ∆P 0
“ “ , (23)
∆X 0 γ∆X 0 ∆X 0
establishing the invariance of this quantity.
Since the total power is Lorentz invariant it may be computed in the rest frame of the
electron, and thus the frame F result of pbq holds in the lab frame F0
ˆ ˙ 4
1 q
P “ γ 2 B02 . (24)
12π m2 c3

16
(d) In the lab frame we can compute the acceleration of the electron using Newton’s law

dp v
“q ˆB, (25)
dt c
where
p “ γmv , (26)
is the relativistic momentum. Here the particle is only scarcely deflected from its straight line
motion. Thus, in a first approximation v » c t ẑ, and the magnetic field then causes small
deflections in the ẑ ˆ x̂ “ ŷ direction. The acceleration in the y direction is approximately
dv y qv q
ayK “ » B0 sinpkvtq » B0 sinpkctq . (27)
dt mγc mγ
The energy radiated is determined only by the transverse acceleration aK

q2
ˆ ˙ ˆ 2 ˙
dW 2 4 q 2
“ γ B 2 sin pkctq (28)
dT 3 4πc3 γ2 0
1 q4 2 2
“ γ B0 (29)
12π m2 c3
in agreement with part pdq.

17
Electromagnetism 2

A current loop and a sphere

A circular loop of radius R, lying flat in the xy-plane with center at the origin, carries a
uniform current I. A sphere of radius a ! R and permeability µ is placed at a height h " a
on the z axis above the xy-plane.

~ inside and outside the sphere.

a. (6 points) Determine the magnetic field B

b. (4 points) Determine the force between the sphere and the ring. Give a qualitative ex-
planation for the direction of the force for both paramagnetic (µ ą µ0 ) and diamagnetic
(µ ă µ0 ) materials.

c. (5 points) Determine the angular distribution of the force per area on the surface of
the sphere to leading order in a{R.

d. (3 points) Integrate the force per area in pcq to find the total force and compare the
result to pbq. Do they agree? Why or why not? Explain.

e. (2 points) Now place the sphere in the center of the ring at zero height, h “ 0. Is
this configuration stable or unstable configuration? Explain. Consider both paramagnetic
and diamagnetic materials.

18
Solution

a. The magnetic field created by the loop at a heigth h is

¿ ~ ˆ R̂h
I dl I2πR2
~ “
B “ ẑ “ B0 phqẑ (30)
Rh2 Rh3
C

with Rh2 “ R2 ` h2 . Throughout the coordinate system will be centered in the sphere. The
corresponding applied magnetic potential B ~ 0 “ ´∇ϕ
~ 0 is ϕ0 “ ´B0 phqz{µ0 . When a sphere
of radius a ! R is positioned at height h, we will assume that the outside potential around
the sphere ϕą asymptotes ϕ0 . Let ϕă be the potential inside the sphere. Since both satisfy
∇2 ϕă,ą “ 0, then for the lowest partial wave (dipole)

A
ϕą “ cos θ ` ϕ0 ϕă “ Cr cos θ (31)
r2
subject to the continuity equations at r “ a,
Bϕą Bϕă
φą “ φă µ0 “µ (32)
Br Br
which gives
ˆ ˙ ˆ ˙
´3 ~ ă“ 3µ ~0
ϕă “ B0 z Bă “ ´µ∇ϕ B (33)
µ ` 2µ0 µ ` 2µ0
~ ą “ ´µ0 ∇ϕ
and similarly for B ~ ą.

b. The magnetization follows as

´ ¯
~ ~ ~ ~
Bă “ µ0 H ` M “ µH (34)
and therefore

~ “ 3pµ ´ µ0 q B
M ~ 0 ” αB
~0 (35)
µ0 pµ ` 2µ0 q

~ 0 is
b. The energy of the sphere in the external field B

4πa3 ~
ˆ ˙
E “´ M ¨B ~0 (36)
3
The binary force between the loop and the sphere is along the z-axis by symmetry

19
 BE 18pµ ´ µ0 q 4πha3 2
Fz “ ´ “ B0 phq (37)
Bh µ0 pµ ` 2µ0 q 3Rh2
It is attractive for µ0 ă µ and repulsive for µ0 ą µ.

d. To determine the angular distribution of the force around the sphere, we note that
the surface current density is

~ă´H
~σM “ r̂ ˆ pH ~ ąq “ M
~ ˆ r̂ (38)
~ă´B
since r̂ ˆ pB ~ ą q. The pressure on the sphere or force per unit area is

~
dF
f~ ” “ ~σM ˆ B~ 0 “ αppB
~ 0 ˆ r̂q ˆ B
~ 0 q “ αB 2 phqpr̂ ´ ẑ pr̂ ¨ ẑqq
0 (39)
dA
which clearly shows that fz “ 0. The total force is
ż
~
F “ αB0 phq a2 dΩ pr̂ ´ ẑ pr̂ ¨ ẑqq
2
(40)

with manifest Fz “ 0. This is expected from the Lorentz character of the force on the surface
currents on the sphere.

e. The expression of Fz in (37) shows that although Fz “ 0 for h “ 0, a small devia-

tion up results in a repulsive force for µ0 ą µ and an attractive force for µ0 ă µ. The former
configuration is unstable, while the latter is stable.

20
Electromagnetism 3

A current sheet in an ohmic medium

Consider an infinite sheet lying in the zy plane carrying surface current1 , K “ K0 e´iωt ẑ.
The current is driven by an external source which sustains the current’s amplitude and
frequency.

K(t)

z
y
x

(a) (4 points) First consider the this current-carrying sheet in vacuum. Determine the
magnetic and electric fields to lowest (non-trivial) order in the frequency. Sketch the
amplitude of the electric field as a function of x for both positive and negative values
of x. Do your results for the electromagnetic fields hold everywhere in space? Explain.
(b) (8 points) Now place the same current sheet into an ohmic medium of conductivity σ
with2 σ " ω. Determine the (real) electric and magnetic fields everywhere in space.
Sketch the amplitude of the electric field as a function of x for both positive and
negative values of x.
Show how your results for the fields follow from the Maxwell equations and their
boundary conditions. Assume that3  “ µ “ 1.
(c) (8 points) Determine the total energy dissipated per time by the induced electric fields
in the ohmic medium. Show that it equals the work done per time by the external
source maintaining the surface current.
1
The units of K0 are amps/meter in SI units.
2
In SI units this reads σ{0 " ω.
3
In SI units this reads  “ 0 and µ “ µ0 .

21
 K(t)

K(t)

B(t) E z (t, x)
B(t) h

Figure 1: (a) Amperian loop used to determine the magnetic field. (b) Loop used to deter-
mine the electric field using Faraday’s law.

Solution:

(a) Drawing an Amperian loop of length ` as shown in Fig. 1(a) we find

¿
I
B ¨ d` “ , (41)
c
Kptq
2Bptq` “ `. (42)
c
Reinserting the direction of the magnetic field we find
#
K0 ´iωt
e ŷ xą0
Bptq “ 2cK0 ´iωt . (43)
´ 2c e ŷ x ă 0

One can also immediately check that the boundary conditions are satisfied:
Kptq Kptq
n ˆ pB2 ´ B1 q “ , or B y |x“0` ´ B y |x“0´ “ . (44)
c c

To determine the induced electric field we use Faraday’s Law in integral form
¿ ż
1
E ¨ d` “ ´ Bt B ¨ da . (45)
c
Drawing a surface as shown in Fig. 1(b) for x ą 0 we find
1
E z ptqh “ ` Bt B y ptqhx , (46)
c
22
 0.4 1

0.35
0.8

|Eω| (σ/ω)1/2/ [K0/2c ]

0.3
|Eω| / (K0/2c)

0.25 0.6
0.2
0.4
0.15
0.1
0.2
0.05
0 0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -3 -2 -1 0 1 2 3
xω/c x/δ

Figure 2: (a) The electric field in vacuum from part (a), Eq. (49). (b) The electric field in
the ohmic medium, Eq. (64). In case (b) (in contrast to (a)) the work done by the external
source is non-zero because the induced electric field does not vanish at x “ 0.

where we have recognized that the normal to the surface is in the negative y direction for a
loop as drawn in Fig. 1(b), and thus B ¨ da ă 0. For x ą 0 the electric field reduces to
´iω K0 ´iωt
E z ptq “ e x x ą 0. (47)
c 2c
A similar loop for x ă 0 gives
´iω K0 ´iωt
E z ptq “ e p´xq x ă 0, (48)
c 2c
leading to our final result
„ 
´iωK0 ´iωt
Ept, xq “Re e |x| ẑ , (49)
2c2
ω|x| K0
“´ sinpωtq ẑ . (50)
c 2c
A graph of the electric field (amplitude) in the z direction is shown in Fig. 2(a). The
amplitude of the electric field grows with x, and when
ωx
„ 1, (51)
c
the quasi-static approximation used here is no longer valid. At this point the electric field
becomes comparable to the magnetic field.

(b) When the sheet is in an ohmic medium, we first derive the appropriate magnetic diffusion
equation for magnetic field in conducting media. Writing j “ σE we find
σ
∇ ˆ p∇ ˆ Bq “ ∇ˆE. (52)
c
23
Using
∇ ˆ p∇ ˆ Bq “ ∇p∇ ¨ Bq ´ ∇2 B , (53)
and Faraday’s Law
1
∇ ˆ E “ ´ Bt B , (54)
c
we find the magnetic diffusion equation
σ
∇2 B “ Bt B . (55)
c2
In this problem we are motivated by the right hand rule to try a solution of the form

Bpt, xq “ B0 eikx´iωt ŷ . (56)

Substituting this ansatz into the diffusion equation, Eq. (55), and solving for k we find two
possible values of k c
σω 1
k˘ ” ˘ i 2 “ ˘ p1 ` iq . (57)
c δ
In the last step, we used the traditional definition of the skin depth
c
2c2
δ” . (58)
σω
The solution must decay as x Ñ ˘8 and should have a symmetric character. Thus, the
solution to the right and left of the current sheet takes the form
#
B0 eip1`iqx{δ e´iωt ŷ xą0
B “ Re . (59)
´B0 e´ip1`iqx{δ e´iωt ŷ x ă 0

The coefficient B0 can be found from the boundary conditions, Eq. (44),

K0
B0 “ , (60)
2c
and the solution therefore takes the form
K0 ´|x|{δ
B“˘ e cosp|x|{δ ´ ωtqŷ . (61)
2c
Here the positive sign is for x ą 0 and the negative sign is for x ă 0.
The electric field is found from Ampere’s Law together with the constitutive relation
j “ σE:
σE
∇ˆB “ . (62)
c
Since B is the y direction
K0 ik` x´iωt
p∇ ˆ Bqz “ Bx By “ ik` e , (63)
2c

24
leading to our final result for x ą 0
„ 
z cp1 ´ iq K0 ik` x´iωt
E pt, xq “ ´ Re e , (64)
σδ 2c
„? 
2c K0 ´x{δ ix{δ´iωt´iπ{4
“ ´ Re e e , (65)
σδ 2c
c ˆ ˙
ω K0 ´x{δ
“´ e cospx{δ ´ ωt ´ π{4q . (66)
σ 2c
In the last step we recognized that
? c
2c ω
“ . (67)
σδ σ
The solution for x ă 0 is follows by the relacements x Ñ |x|. A graph of the amplitude,
|Eω |9e´|x|{δ is shown in Fig. 2(b).

(c) The work done per area per time can be found in at least three ways:
1. By calculating the work done by the battery, ´K ¨ E. The work done by the battery
on the currents is, K ¨ Ebatt . The battery must supply an additional field Ebatt “ ´E
to counter balance the induced field E, and maintain the current K.
2. By calculating the Poynting flux flowing out of the sheet.
ş ş
3. By integrating the dissipation rate over the volume: dV j ¨ E “ σ dV E 2
They should all agree. All of these calculations involve time averages of harmonic quantities,
which are calculated as follows: if Aptq “ RerAω e´iωt s and Bptq “ RerBω e´iωt s then
1
AptqBptq “ RerAω Bω˚ s . (68)
2
The first method directly evaluates the work done by the battery
dW ˇ
“ ´ K ¨ E ˇx“0 (69)
dt dA
1
“ ´ RerKω pEωz q˚ s (70)
2 c
1 ω K0˚ p1 ´ iq
“ ` RerK0 ? s (71)
2 σ 2c 2
c
|K0 |2 ω
“ (72)
4c 2σ
The Poynting flux methods is evaluated similarly. There are two sides to the sheet leading
to a factor of two
dW
“2c E ˆ B ¨ x̂ , (73)
dt dA „ˆ ˙ 
1 Kω
“ ´ 2c Re pEωz q˚ , (74)
2 2c

25
which clearly agrees with Eq. (70). (The minus in passing from the first to the second line
follows from ẑ ˆ ŷ “ ´x̂.) We should expect agreement between these two methods since

´∇¨S “j¨E, (75)

Integrating ∇ ¨ S over a Gaussian “pillbox” shows directly that the Poynting flux out of the
sheet equals the work done by the battery, ´K ¨ E.
Finally, we integrate the ohmic dissipation over the volume. Here we evaluate the con-
tribution from x ą 0 and then multiply by a factor of two to account for x ă 0:
ż8
dW
“2σ dxE 2 , (76)
dt dA 0
1 |K0 |2 ω ´2x{δ
ż8 ˆ ˙
“2σ dx e , (77)
0 2 4c2 σ
|K0 |2
“ω 2
δ, (78)
c 8c
ω |K0 |2
“ . (79)
2σ 4c
All three methods thus give the same answer.

26
Quantum Mechanics 1

Resonant tunneling and state metastability

Your goal in this problem is to study and relate two basic quantum properties of a one-
dimensional non-relativistic particle of mass m, moving in the system of two similar, symmetric
potential barriers U(x) of width scale w, separated by a much larger distance a >> w – see the left panel
in the figure below. Each barrier has a tunneling transparency T  1 (which is some smooth function of
the particle’s energy E), defined as shown on the right panel of the figure, where I and I’ are the
probability currents of the incident and transferred monochromatic de Broglie waves.

U x 
U x  U x  a 
a I I'  TI

E
x x
w w

A (4 points). Calculate the (similarly defined) transparency T’ of the two-barrier system for a
monochromatic de Broglie wave incident from afar, as a function of a, T, and the particle’s energy E.

B (2 points). Sketch the calculated transparency as a function of a, and prove that its largest
value, reached at certain resonance values ares, equals 1, and hence may be larger than T. Give a physical
interpretation of this fact.

C (4 points). For the case of a very low barrier transparency, T << 1, and the distance a close to
one of its resonance values ares, calculate the energy width E of the transparency resonance. (For
certainty, use the standard “FWHM” definition of the width as the energy interval within that the
transparency is larger than ½.)

D (7 points). Now consider the situation when the particle is initially placed between the barriers
(again with a very low transparency T << 1), in its lowest-energy state, which is metastable due to the
tunneling through the barriers. Calculate the law of the time decay of the probability to find the particle
between the barriers, in the lowest nonvanishing approximation in T.

E (3 points). Compare the characteristic time  of the metastable state’s decay to the energy
width E calculated in Task C, and comment. Explain why such “energy-time uncertainty relation” is
less general than the canonical uncertainty relations - such as those between Cartesian components of
the generalized coordinates and the corresponding momenta.

Hint: Working on Task D, it may be helpful to represent the standing de Broglie wave describing
the metastable state as a sum of two waves traveling in opposite directions.
Solution:

A (4 points): Outside of the barriers, i.e. in any region where U(x) = 0, the stationary 1D Schrödinger
equation has the simple form
 2 d 2
  E  0 ,
2m dx 2
and hence is satisfied with any linear superposition of two de Broglie waves exp{ikx}, where the wave
number k is determined by the particle’s energy E:
2k 2
E.
2m
Due to the Schrödinger equation’s linearity for any x, for a single barrier located at x = 0, with
the wave incident from the left, the coefficients in such superpositions at x < 0 and x > 0 may be selected
as shown in Fig. below.

U x 
Aeikx tAe ikx

rAe ikx

0 x

Here t and r the (generally, complex) transmission and reflection coefficients, which may depend on the
energy of the incident wave, but not its amplitude A. Since the corresponding probability currents I are
proportional to the squares of the moduli of these coefficients, they are related by the probability
conservation law as
 tA  A , giving r  1  t  1  T .
2 2 2 2 2
rA (1)

For the system of two barriers, we may take the superpositions, in three regions outside the
barriers, in the form shown in Fig. below.

Ae ikx U x  B e ikx U x  a  Ce ikx

B e ikx

0 a x
Now requiring that the waves incident upon, transmitted through, and reflected from each barrier satisfy
the same relations as for a single barrier, we get the following relations between the complex amplitudes
of these waves:
B  tA  rB , B e ika  rB e ika , Ce ika  tB e ika .
This system of linear equations yields, in particular,
 t2A
C ,
1  r 2 exp2ika
so that using Eq. (1), the transparency of the system may be represented as

T' 
IC

C
2


t2

2
t

 2 2
T2
, (2)
IA A 1  r 2 exp2ika 1  r exp2ika   
2 2
1  1  T  exp2i ka   
2

where the real phase   arg r may be also a (smooth) function of k, and hence of the energy E.

B (2 points). The easiest way to analyze T’ as a function of a is to consider the denominator of

the last form of Eq. (2) as the square of the length of the 2D vector that is the difference of two vectors:
one of length 1, directed along some fixed axis, and the second vector of length (1 – T), turned by the
angle 2(ka + ) – see Fig. below.

1  T  difference
vector
2ka   
0 1
The change of a, which affects only this angle, leads to the mutual rotation of the two vectors
with the period (ka) = 2, and as a result, to oscillations of the denominator in Eq. (2), and hence of the
transparency T’, with the same period. This is the effect of resonant tunneling, physically arising from
the interference of the de Broglie waves reflected from two potential barriers. The largest value of T’ is
reached when the two vector are exactly aligned, i.e. when
ka    k n a    n , (3)
where n is an integer, so that the vector of their difference has the length T, and Eq. (2) yields T’max = 1.
This value may be indeed larger than the transparency T of a single barrier, because due to the
constructive interference of the reflected waves, the internal wave amplitudes B are resonantly
increased, resulting in the corresponding increase of the transmitted wave’s amplitude C.
Figure below shows plots of Eq. (2) for several values of the parameter T. The smaller is T, the
sharper is the resonance peak, i.e. the smaller is its FWHM (Full Width on Half-Maximum), (ka).

T  0 .9
0.8

0.6
T' ½
0.4 T  0 .6  ka 

0.2

T  0 .1
0
0 1 2 3
ka    / 
 C (4 points). If the single barrier’s transparency is very low, T << 1, the two vectors discussed
above have very close lengths, so that the transparency decreases to ½ when the angle 2(ka + ) between
them is increased to a small amount equal to approximately T – see the vector diagram below.

1  T 
2ka     T 2T
0 1 T
The distance between these two values is
 2ka     2ka     2ka     2T ,
Since T (and hence r and t, including   arg r) are relatively slow functions of k, we may use this
expression even if this change of the phase shift 2(ka + ) is achieved using a small change E of
particle’s energy, and hence of the wave vector, at constant a  ares:
T dE 2  2kn
k  , so that E  k  kk  T. (4)
a dk m ma
D (7 points). If T = 0, the particle placed between the barriers has genuine stationary states
described by standing waves,
 n  x   B sin k n x  const  ,
with the wave numbers given by Eq. (3). If T is nonvanishing but very small, this solution still may be
used as a reasonable approximation, but due to the particle “leakage” (tunneling) through the barriers,
the wavefunction has to be normalized not to unity, but to the probability W of finding the particle
between the barriers at this particular instant, which may be less than 1:

 dx  B  sin k x    dx  W .
2 2 2
n
inside inside

In the limit w << a, the exact interval of this integration are not important, and may be taken to be [0, a].
Since, according to Eq. (3), this interval contains an integer number of de Broglie half-waves, the
integral equals a/2, and we get
2 a 2 2W
B W, i.e. B  .
2 a
As suggested in the Hint, this solution may be represented in the same form as in Task A:
1/ 2
W B
 n  x   B exp ik n x  B exp ik n x, with B  B    .
2  2a 
Each of these waves, hitting the corresponding barrier, induces beyond it an outgoing wave with the
amplitude C = tB and hence with the probability current
k n k n k n W
I   I  
2 2
C tB T . (5)
m m m 2a
These leakage currents gradually reduce the probability W. The law of this reduction may be found from
the general probability conservation law, in our case taking the form1
dW
 I  I  0.
dt
With the result (5), this law gives a simple differential equation
dW W 1 k n
 , with  T, (6)
dt   ma
with the elementary solution W(t) = W(0)exp{-t/}, so that the constant  given by Eq. (6) is the
metastable state’s lifetime.

E (3 points). Comparing Eqs. (4) and (6), we see that their relation is remarkably simple:

E  , i.e. E    . (7)

This “energy-time uncertainty relation” is much more general than the analyzed situation, and is valid,
in particular, for a broad class of metastable states. However, it is still less general than canonical
uncertainty relation, in particular the Heisenberg’s relation

r j p j  . (8)
2
Indeed, in quantum mechanics, the Cartesian coordinates rj of a particle, the Cartesian
components pj of its momentum, and the energy E are regular observables, represented by operators. In
contract, the time is treated as a c-number argument, and is not represented by an operator, so that Eq.
(7) cannot be derived using such general assumptions as Eq. (8). Thus the time-energy uncertainty
relation should be applied with caution.

***

Finally, note that this the simple model discussed in this problem is a convenient platform for
discussion of several conceptual issues of quantum mechanics, including the relation between closed
(and hence energy-conserving) and open (and hence dissipative) systems, in particular their time
reversibility and irreversibility, and also of the so-called “wave-packet reduction” at quantum
measurements – see, e.g., Sec. 2.5 in https://commons.library.stonybrook.edu/egp/4/ .

1
If the system is initially in one of its higher metastable states, with kn > (kn)min, the perturbation caused by the
barrier transparency may also cause quantum transitions to lowest-energy states, thus affecting the probability
conservation law. This is why this problem asks only the explore the lowest-energy state.
Quantum Mechanics 2

Orbital motion perturbations

A spinless point particle of mass m is confined to a square-shape two-dimensional region of size
LL:
y

0 L x

A (2 points). Write down the stationary wave functions and the corresponding
eigenenergies of the particle.

B (5 points). Now consider there is a “point defect” inside the region, whose potential can
be modeled as
 
U = L2 (r − r0 ) ,

where r0 = x0 , y0  is the location of the defect. Treating the potential as a perturbation, calculate
the first-order corrections to the energies of the ground state and the first excited state. Find out
the locations of the defect (inside the square), at which the 1st excited state remains degenerate.

C (7 points). Now consider a moving “defect”, which oscillates along the x-direction at the
center of the square:
 L L
r0 =  + l sin t ,  ,
2 2
where l << L. Find out the selection rule for the particle excitation from the ground state to an
arbitrary excited state. Calculate the time-dependent probability for the charge to be in the first
excited state, for  << fi, where fi is the energy difference between the initial and final states,
using the time-dependent perturbation theory.

D (6 points). At  << fi, the time evolution of the system obeys the adiabatic theorem.
Use this fact to calculate the time-dependent transition probability from the ground state to the first
excited state. Compare your result with what you got in Task C.
Solution:
a) The wave functions are standing waves which vanish at the edges of the region:
2 𝑛𝑥 𝜋𝑥 𝑛𝑦 𝜋𝑦
Ψ𝑛𝑥 ,𝑛𝑦 = |n𝑥 , 𝑛𝑦 > = sin ( ) sin ( )
𝐿 𝐿 𝐿
where nx, ny = 1, 2, 3, ….. And the corresponding energies are:
𝜋 2 ℏ2 2
E𝑛𝑥 ,𝑛𝑦 = (𝑛 + 𝑛𝑦2 )
2𝑚𝐿2 𝑥
Ground state: (𝑛𝑥 , 𝑛𝑦 ) = (1,1), not degenerate.

First excited state, (𝑛𝑥 , 𝑛𝑦 ) = (1,2), (2,1), denegeracy=2.

b) Ground state is not degenerate, so the 1st order energy correction is:
(1)
ΔE11 = 〈1,1|𝑉′|1,1〉
4 𝜋𝑥 𝜋𝑦
= 2
∫ 𝑑𝑥𝑑𝑦 sin2 ( ) sin2 ( ) 𝛿(𝑥 − 𝑥0 )𝛿(𝑦 − 𝑦0 )𝜆𝐿2
𝐿 𝐿 𝐿
𝜋𝑥0 𝜋𝑦0
= 4𝜆sin2 ( ) sin2 ( )
𝐿 𝐿
For the first excited states: |1,2 > and |2,1 >
𝜋𝑥0 2𝜋𝑦0
〈1,2|𝑉′|1,2〉 = 4𝜆 sin2 ( ) sin2 ( )≡𝑎
𝐿 𝐿
𝜋𝑥0 2𝜋𝑥0 2𝜋𝑦0 𝜋𝑦0
〈1,2|𝑉′|2,1〉 = 4𝜆 sin ( ) sin ( ) sin ( ) sin ( )≡𝑏
𝐿 𝐿 𝐿 𝐿
2𝜋𝑥0 𝜋𝑥0 𝜋𝑦0 2𝜋𝑦0
〈2,1|𝑉′|1,2〉 = 4𝜆 sin ( ) sin ( ) sin ( ) sin ( )=𝑏
𝐿 𝐿 𝐿 𝐿
2𝜋𝑥0 𝜋𝑦0
〈2,1|𝑉′|2,1〉 = 4𝜆 sin2 ( ) sin2 ( )≡𝑐
𝐿 𝐿
𝑎 𝑏 𝑢1 (1) 𝑢1
Solving: ( ) (𝑢 ) = 𝐸2 (𝑢 )
𝑏 𝑐 2 2

(1) 𝑎+𝑐±√(𝑎−𝑐)2 +4𝑏2

We get: ∆𝐸2 =
2

𝑢1 2𝑏
=
𝑢2 𝑎 + 𝑐 ± √(𝑎 − 𝑐)2 + 4𝑏 2
𝑢1 1 𝛼
Hence: (𝑢 ) = ( )
2 1
2 √1+𝛼

For the 1st excited state to remain degenerate, (𝑎 − 𝑐)2 + 4𝑏 2 = 0, so 𝑎 = 𝑐 and 𝑏 = 0

𝐿 𝐿
Hence: 𝑥0 = 0, 2 , 𝐿; and 𝑦0 = 0, 2 , 𝐿
 (locations which do not break degeneracy for the 1st excited state)

b) First we can write down the time-dependent defect potential:

𝐿
𝑉 ′ = 𝜆𝐿2 𝛿(𝑥 − 𝑙 ∙ 𝑠𝑖𝑛(𝜔𝑡))𝛿 (𝑦 − )
2
A transition from the ground state (1,1) to (nx, ny):
𝑡 𝑛𝑦 𝜋𝑦
−𝑖 4 𝜋𝑥 𝜋𝑦 𝑛𝑥 𝜋𝑥
𝑑(𝑡) = ∫ 𝑑𝑥𝑑𝑦 ∫ 2 sin ( ) sin ( ) sin ( ) sin ( ) 𝜆𝐿2 𝛿(𝑥 − 𝑙sin(𝜔𝑡′))𝛿 (𝑦
ℏ 0 𝐿 𝐿 𝐿 𝐿 𝐿
𝐿 ′
− ) 𝑒 𝑖𝜔𝑓𝑖 𝑡 𝑑𝑡′
2
−4𝜆𝑖 𝑡 𝜋𝑙sin(𝜔𝑡 ′ ) 𝜋 𝑛𝑥 𝜋𝑙sin(𝜔𝑡 ′ ) 𝑛𝑦 𝜋 ′
= ℏ
∫0
sin ( 𝐿
) sin ( 2
) sin ( 𝐿
) sin ( 2 ) 𝑒 𝑖𝜔𝑓𝑖𝑡 𝑑𝑡′

𝐸𝑓 −𝐸𝑖 𝜋2 ℏ2 𝑛𝑥2 +𝑛𝑦2 −2

Here 𝜔𝑓𝑖 = . And between the ground state and excited state (nx, ny), 𝜔𝑓𝑖 = 2𝑚𝐿2 .
ℏ ℏ

For d(t) to be non-zero, ny = 1,3,5…

Since 𝑙 ≪ 𝐿, assuming the variables inside sine are very small, the above can be approximated by:

−4𝜆𝑖 𝑛𝑦 𝜋 𝑡 𝜋𝑙sin(𝜔𝑡 ′ ) 𝑛𝑥 𝜋𝑙sin(𝜔𝑡 ′ ) 𝑖𝜔 𝑡 ′

𝑑(𝑡) = sin ( )∫ 𝑒 𝑓𝑖 𝑑𝑡′
ℏ 2 0 𝐿 𝐿

−4𝜆𝑖 𝜋𝑙 2 𝑛𝑦 𝜋 𝑡 2 ′
= 𝑛𝑥 ( ) sin ( ) ∫ sin (𝜔𝑡 ′ )𝑒 𝑖𝜔𝑓𝑖𝑡 𝑑𝑡′
ℏ 𝐿 2 0

−4𝜆𝑖 𝜋𝑙 2 𝑛𝑦 𝜋 𝑡 1 𝑖𝜔𝑡′ 2
′
= 𝑛𝑥 ( ) sin ( ) ∫ [ (𝑒 − 𝑒 −𝑖𝜔𝑡′ )] 𝑒 𝑖𝜔𝑓𝑖 𝑡 𝑑𝑡′
ℏ 𝐿 2 0 2𝑖

𝜆𝑖 𝜋𝑙 2 𝑛𝑦 𝜋 𝑡 𝑖(𝜔 +2𝜔)𝑡′
= 𝑛𝑥 ( ) sin ( ) ∫ [𝑒 𝑓𝑖 + 𝑒 𝑖(𝜔𝑓𝑖−2𝜔)𝑡′ − 2𝑒 −𝑖𝜔𝑓𝑖 𝑡′ ] 𝑑𝑡′
ℏ 𝐿 2 0
′
𝜆𝑖 𝜋𝑙 2 𝑛𝑦 𝜋 𝑒 𝑖(𝜔𝑓𝑖+2𝜔)𝑡′ − 1 𝑒 𝑖(𝜔𝑓𝑖−2𝜔)𝑡′ − 1 2(𝑒 𝑖𝜔𝑓𝑖 𝑡 − 1)
= 𝑛𝑥 ( ) sin ( )[ + − ]
ℏ 𝐿 2 𝜔𝑓𝑖 + 2𝜔 𝜔𝑓𝑖 − 2𝜔 𝜔𝑓𝑖

For 𝜔 ≪ 𝜔𝑓𝑖 , this can be approximated by:

 ′
𝜆𝑖 𝜋𝑙 2 𝑛𝑦 𝜋 𝑒 𝑖(𝜔𝑓𝑖+2𝜔)𝑡′ − 1 𝑒 𝑖(𝜔𝑓𝑖−2𝜔)𝑡′ − 1 2(𝑒 𝑖𝜔𝑓𝑖𝑡 − 1)
𝑑(𝑡) = 𝑛𝑥 ( ) sin ( )[ + − ]
ℏ 𝐿 2 𝜔𝑓𝑖 𝜔𝑓𝑖 𝜔𝑓𝑖
′
𝜆𝑖 𝜋𝑙 2 𝑛𝑦 𝜋 𝑒 𝑖(𝜔𝑓𝑖+2𝜔)𝑡′ + 𝑒 𝑖(𝜔𝑓𝑖−2𝜔)𝑡′ − 2𝑒 𝑖𝜔𝑓𝑖𝑡
= 𝑛𝑥 ( ) sin ( )[ ]
ℏ 𝐿 2 𝜔𝑓𝑖

Hence the transition probability is:

𝜆 2 2 𝜋𝑙 4 2 𝑛𝑦 𝜋 (6 + 2 cos(4𝜔𝑡) − 8cos(2𝜔𝑡))
|𝑑(𝑡)|2 = ( ) 𝑛𝑥 ( ) sin ( ) 2
ℏ 𝐿 2 𝜔𝑓𝑖

𝑛𝑥2 4𝑚2 𝜆2 𝑙 4 2 𝑛𝑦 𝜋
= 2 4
sin ( ) sin4 (𝜔𝑡)
2 2
(𝑛 + 𝑛 − 2) ℏ 2
𝑥 𝑦

For |1,1> to |2,1> transition:

16𝑚2 𝜆2 𝑙 4 4
|𝑑(𝑡)|2 = sin (𝜔𝑡)
9ℏ4

d). Based on adiabatic theorem, the system starting from ground state would remain in ground state while
been very slowly perturbed. Under the slow perturbation, at any given moment, the ground state is (from
time-independent perturbation method):
(0) (0)
⟨𝑛𝑥 , 𝑛𝑦 |𝑉′|1, 1 ⟩
|1,1⟩ = |1, 1(0) ⟩ + ∑𝑛𝑥 ,𝑛𝑦 |𝑛𝑥 , 𝑛𝑦 (0) ⟩ (0) (0)
𝐸1,1 −𝐸𝑛𝑥 ,𝑛𝑦

The transition to the 1st excited state is thus described by the projection of the 1st excited state on the
perturbed ground state:

(0)
⟨2, 1(0)|𝑉′|1, 1(0) ⟩
⟨2, 1 |1,1⟩ = (0) (0)
𝐸1,1 − 𝐸2,1

2𝜋𝑥 𝜋𝑦 𝜋𝑥 𝜋𝑦 𝐿
∫ sin ( 𝐿 ) sin ( 𝐿 ) sin ( 𝐿 ) sin ( 𝐿 ) 𝜆𝐿2 𝛿(𝑥 − 𝑙sin(𝜔𝑡′))𝛿 (𝑦 − 2) 𝑑𝑥𝑑𝑦
= (0) (0)
𝐸1,1 − 𝐸2,1

2𝜋𝑙sin(𝜔𝑡) 𝜋𝑙sin(𝜔𝑡)
sin ( ) sin ( )
2 𝐿 𝐿
= 𝜆𝐿 (0) (0)
𝐸1,1 − 𝐸2,1

2(
𝜋𝑙 2
𝐿 ) sin(𝜔𝑡) = 4𝑚𝜆𝑙 sin2 (𝜔𝑡)
2
2𝑚𝐿
≈ 2𝜆𝐿2
𝜋 2 ℏ2 3 3ℏ2
And ⟨1, 2(0)|1,1⟩ = 0 obviously, because ⟨1, 2(0)|𝑉′|1, 1(0) ⟩ = 0 from the selection rule in c).
So the probably for the system to transit from ground state to the 1st excited state is:

2 16𝑚2 𝜆2 𝑙 4 4
|⟨2, 1(0)|1,1⟩| = sin (𝜔𝑡)
9ℏ4
This is the same result as derived from the time-dependent perturbation method.
Quantum Mechanics 3

Two fermions on a ring

Consider two spin-1{2 fermions of mass m, with coordinates x1 and x2 , confined to move on
a circle of circumference L and interacting through a spin-dependent potential

V “ ´uδpx1 ´ x2 q~s1 ¨ ~s2 , u ą 0,

where ~s “ tsx , sy , sz u is the operator of the spin 1/2 (in units of ~), so that the Hamiltonian
H of the two-electron system (in the standard notations) is:
1 2
H“ pp ` p22 q ` V .
2m 1
The circle is threaded by an infinitely-long solenoid which carriers a magnetic flux Φ, so that
the electron momenta are
~ B
pj “ ´ eA ,
i Bxj
where A is the vector potential produced by the magnetic flux, which can be taken to be
constant along the circle, and e is the electron’s charge.

(a) (4 points) What is the relation between A and Φ? Use this relation to write down the
single-particle eigenstates ψn pxq and eigenenergies En of one electron on the circle in
terms of φ and other parameters in the problem.

(b) (3 points) Derive the boundary conditions for the orbital part ψpx1 , x2 q of the two-
electron wavefunction at x1 “ x2 , if electron spins are in the triplet state.
ptq
(c) (4 points) Assuming that |Φ{ph{eq| ă 12 , find the ground-state energy E0 and the
ptq
orbital part ψ0 px1 , x2 q of the corresponding two-electron state.

(d) (4 points) Derive the same boundary conditions as in part (b) for electron spins in the
singlet state.
psq
(e) (5 points) For Φ “ 0, determine the ground-state energy E0 and the orbital part
psq
ψ0 px1 , x2 q of the ground-state two-electron wavefunction in the singlet spin state in
the limit of strong potential u Ñ 8.

37
Solution

(a) By Stokes’ theorem

ż ż ż
Φ“ ~ “
d~sB ~“
d~s∇ ˆ A d~lA
~ “ LA.

Since the wavefunctions on a circle should be periodic with the period L, the single-particle
eigenstates are
1 2π
ψn pxq “ ? eikn x , kn “ n , n “ 0, ˘1, ˘2, ... .
L L
Acting on these wavefunctions with the kinetic energy operator p2 {2m, and using the relation
between A and Φ, we get the eigenenergies that correspond to these eigenstates
p~kn ´ eAq2 phn ´ eLAq2 h2 pn ´ φq2
En “ “ “ .
2m 2mL2 2mL2
Here and below φ ” Φ{ph{eq.
(b) Any one of the triplet spin states is symmetric with respect to interchange of the two
spins:
1
|S “ 1y “ t| ÒÒy ; | ÓÓy ; ? r| ÒÓy ` | ÓÒys u .
2
Since the total wavefunction of the two electrons should be antisymmetric with respect to
interchange of the electrons, the symmetric nature of the spin part of the wavefunction makes
the orbitalˇ part ψpx1 , x2 q antisymmetric in the triplet states. This means that in this case,
ψpx1 , x2 qˇ “ 0, and ψpx1 , x2 qδpx1 ´ x2 q ” 0. Therefore, the interaction V does not have
ˇ
x1 “x2
any effect on the wavefunction, and electrons in the triplet state behave as non-interacting,
i.e. ψpx1 , x2 q is continuous together with its first derivatives at x1 “ x2 as at all other points.

(c) Electrons in the triplet state behave as non-interacting. Since in the triplet state, the two
electrons effectively have the same direction of spin, they can not occupy the same orbital
state. The lowest energy of the two-electron state is reached then, when the two electrons
occupy the two lowest-energy single-particle states. This means that for φ P r0, 1{2s,

ptq h2 p1 ´ 2φ ` 2φ2 q ptq 1

E0 “ E0 ` E1 “ , ψ0 px1 , x2 q “ ? pei2pix1 {L ´ ei2pix2 {L q.
2mL2 2L
For φ P r´1{2, 0s,

ptq h2 p1 ` 2φ ` 2φ2 q ptq 1

E0 “ E0 ` E´1 “ , ψ0 px1 , x2 q “ ? pe´i2pix1 {L ´ e´i2pix2 {L q.
2mL2 2L

(d) For electrons in the singlet spin state,

1
|S “ 0y “ ? p| ÒÓy ´ | ÓÒyq ,
2

38
the spin part of the wavefunction is antisymmetric, and the orbital part should be symmetric.
in this case, the δ-functional potential affects the wavefunction. To find the magnitude of
the potential, one needs to evaluate the magnitude of the spin part ~s1 ¨ ~s2 of the interaction
V . As usual, making use of the operator S ~ “ ~s1 ` ~s2 of the total spin, we have

~ 2 ´ ~s 2 ´ ~s 2 s{2 “ 1 S
~s1 ¨ ~s2 “ rS ~2 ´ 3 ,
1 2
2 4
In the singlet state |S “ 0y, this gives:
3
~s1 ¨ ~s2 |S “ 0y “ ´ |S “ 0y ,
4
i.e., the interaction V reduces to V “ p3u{4qδpx1 ´ x2 q, and its effect on ψpx1 , x2 q can be
described in the same way as for the standard single-particle δ-function potential. Integrating
the spatial Schrödinger equation over infinitesimal interval of x1 around the point x1 “ x2 ,
one obtains that the wavefunction is continuous at x1 “ x2 with discontinuous first derivative:
Bψ ˇˇ Bψ ˇˇ
ˇ ´ ˇ “ p3mu{2~2 qψpx1 “ x2 q .
Bx1 x1 “x2 `0 Bx1 x1 “x2 ´0
Similarly for x2 :
Bψ ˇˇ Bψ ˇˇ
ˇ ´ ˇ “ p3mu{2~2 qψpx1 “ x2 q .
Bx2 x2 “x1 `0 Bx2 x2 “x1 ´0
The fact that the wavefunction ψpx1 , x2 q is symmetric in the singlet state, implies that
Bψ ˇˇ Bψ ˇˇ
ˇ “ ˇ .
Bx1 x1 “x2 ´0 Bx2 x2 “x1 ´0
This equality transforms the first relation above into the sought boundary condition:
` Bψ Bψ ˘ˇˇ
´ ˇ “ p3mu{2~2 qψpx1 “ x2 q . (80)
Bx1 Bx2 x1 “x2 `0

(e) The boundary condition (80) together with the fact that momenta are finite, means
that ψpx1 “ x2 q “ 0 in the limit u Ñ 8. One can immediately see that such symmetric
wavefunction ψ psq px1 , x2 q vanishing at coincident coordinates can be constructed out of the
antisymmetric wavefunction of noninteracting electrons ψ ptq px1 , x2 q,
1
ψ ptq px1 , x2 q “ ? rψn1 px1 qψn2 px2 q ´ ψn1 px2 qψn2 px1 qs
2
by the sign change:
ψ psq px1 , x2 q “ ψ ptq px1 , x2 q sgn px1 ´ x2 q . (81)
Such a sign change implies however, that as the function of one coordinate, e.g. x1 , ψ psq
aquires an extra phase π when increasing past x2 . As a result, the single-particle wavefunc-
tions on a circle that could be used to build up ψ ptq and then ψ psq should be antiperiodic,
changing the quantization of momenta:
eikL “ ´1 , k “ π~p2n ` 1q{L.

39
In this case, the lowest-energy state corresponds to n1,2 “ 0, ´1, giving the ground state
energy
psq h2
E0 “ .
4mL2
The wavefunction that corresponds to this state is
?
psq 2 `π ˘
ψ0 px1 , x2 q “ sin |x1 ´ x2 | .
L L

40
Statistical Mechanics 1

Ideal gas in gravitational potential

Consider an ideal gas of N indistinguishable molecules of mass m in a cylindrical volume

V “ Ah with base area A and height h.

(a) (5 points) Calculate the canonical partition function of the

ş8 ideal gas Z g?pT, V q, including
2
the effect of gravity. You may find useful the integral 0 dt t2 e´t “ π{4.

(b) (4 points) Using the partition function, calculate the internal energy of the system.
Then, determine the heat capacity in two limits, T " mgh and T ! mgh. You may
find the Stirling approximation N ! « pN {eqN helpful.

(c) (2 points) Write down the condition for when the effect of gravity is negligible com-
pared to the thermal energy. What is the internal energy in that case? Neglecting
internal degrees of freedom, estimate at what temperature the thermal energy for oxy-
gen molecules O2 (16 O) at 1m height is of the order of the potential energy.

Now assume that the top of the cylinder is a piston of mass M that can
move vertically (change the height h), and that the gravity effects on the
gas itself are negligible, so that the gas volume can change but the pressure
is constant.

(d) (2 points) Write down the Hamiltonian of the combined piston + gas system H g`p ,
and express the potential energy of the piston in terms of the pressure and the volume
of the gas.

(e) (4 points) Find the canonical partition function Z g`p of the system with the piston.
One way to do this is to use the result for the partition function Z g pT, V q from part (a)
(now neglecting the potential energy of the gas). First, show that Z g`p is proportional
to its Laplace transform
ż8
g`p
Z pT, P q “ B dV e´αV Z g pT, V q (82)
0

for some B and αş8that you have to determine. Then compute Z g`p as above (you may
!
use the integral 0 dx xN e´bx “ bNN`1 ).

(f) (3 points) Compute the variance of the piston height h and express it in terms of
particle number N , piston mass M , and temperature T .
Hint: one way to do this is to use the pressure-dependent partition function Z g`p pT, P q.

41
Solution:

(a)The partition function of the system is given by

ˆ ˙N
1 ` p1q ˘N
g N p1q
Z “ z » z , (83)
N! e

with ż
1 p1q
z p1q
“ d3 p d3 x e´βH , (84)
p2π~q3
where
p2
H p1q “ ` mgy , (85)
2m
is the Hamiltonian for one particle with momentum p at height y, and β “ 1{T .
Evaluating the integral in z p1q ,
ż8 żh
p1q 4π 2 ´β 2mp2
z “ 3
dp p e Ae´βmgy dy
p2π~q 0 0
ˆ ˙3{2 ż 8 żh
4π 2m 2 ´t2
“ 3
dt t e Ae´βmgy dy
p2π~q β 0 0
ˆ ˙3{2 ? ż h
4π 2m π
“ Ae´βmgy dy
p2π~q3 β 4 0
ˆ ˙3{2 „ h
1 2πm A ´βmgy
“ ´ e
p2π~q3 β βmg 0
ˆ ˙3{2
1 2πm A ` ˘
“ 3
1 ´ e´βmgh ,
p2π~q β βmg

one gets the partition function of the system

« ˙3{2 ffN „
˘ N
ˆ 
N 1 2πm A ` ´βmgh
˘ N A ` ´βmgh
Z“ 1´e “ 1´e , (86)
e p2π~q3 β βmg e λ3T βmg
b
2π~2 β
where λT “ m
is the thermal wavelength.

(b) The internal energy of the system is given by

# « ˆ ˙3{2 ff+
B B N 2πm A ` ˘
U “´ ln Z “ ´ N ln 1 ´ e´βmgh
Bβ Bβ e βp2π~q2 βmg
3 2π~2
„ ˆ ˙
B 3 ´βmgh A
“´ N ln N ´ 1 ´ ln β ´ ln β ` lnp1 ´ e q ` ln ´ ln
Bβ 2 mg 2 m
5N mgh
“ `N
2β 1 ´ eβmgh

42
(c) Gravity becomes relevant when ε “ mgh
kT
„ 1, i.e., the potential energy is on the order of
the thermal energy at temperature T . For ε ! 1, we can neglect gravity. To get the internal
energy in this case, we expand the exponential:
5N mgh 5N mgh
U “ `N » `N
2β 1´e βmgh 2β 1 ´ 1 ´ βmgh ´ 12 pβmghq2 ` Opε3 q
5N mgh
ôU “ ´N
2β βmghp1 ` 12 βmghq
ˆ ˆ ˙˙
3 mgh
ôU “ N kT 1 ` O ,
2 kT
which in this limit is the well known internal energy of an ideal gas of non-interacting
particles.
The temperature at which ε becomes one for h “ 1 m, m “ 32u “ 5.312 ¨ 10´26 kg, is
mgh 5.312 ¨ 10´26 9.8 m2 kg s´2
T « “ “ 0.038 K (87)
k 1.38 ¨ 10´23 m2 kg s´2 K´1
Normal room temperature is 273.15 K, so the effect of gravity is negligible for oxygen at
room temperature, according to this simplified estimate.

(d) The pressure on the system, exerted on the gas by the piston is P “ M g{A. The
potential energy of the piston is M gh “ AhP “ V P . With that, the Hamiltonian of the
combined sytem is
H g`p “ H g ` k 2 {2M ` P V , (88)
where H g “ N
ř piq
i H is the Hamiltonian of the entire gas, and k is the momentum of the
piston in the z direction.

(e) With the Hamiltonian H g`p above, the canonical partition function becomes
ż ż
1 dhdk
g`p
Z pT, P q “ dω g
expr´βpH g ` k 2 {2M ` P V qs , (89)
N! 2π~
ş
where dω g is the phase space integral for the whole gas, assuming a constant volume
V “ Ah. One can first formally integrate the gas degrees of freedom,
ż „ ż  ż
g`p dhdk ´βpk2 {2M `P V q 1 g ´βH g dV dk ´βpk2 {2M `P V q g
Z pT, P q “ e dω e “ e Z pT, V q ,
2π~ N! 2π~A
(90)
where Z g pT, V q is the gas partition function and dV “ A dh. Integrating out the piston
momentum k, one gets
d ż
g`p 1 M
Z pT, P q “ 2
dV e´βP V Z g pV, T q . (91)
A 2π~ β

Thus, the partition function for constant pressure is thebLaplace transform of that at constant
volume, as given in part (e) with α “ βP and B “ A1 2π~ M
2β .

43
 In the high-temperature limit β Ñ 0, the partition function of the gas Z g simplifies to
«ˆ ˙3{2 ffN
1 M A 1 ´3N N
Z g pT, V q « 2
p1 ´ 1 ` βmghq “ λ V (92)
N! 2π~ β βmg N!

and its Laplace transform is given by

ż ˆ ˙N `1
g`p ´αV 1 ´3N N 1 N! 1 T
Z pT, P q “ B dV e λT V “ BλT´3N ¨ “ (93)
N! N! α N `1 AΛT λ3N
T P
b
2π~2 β
where ΛT “ M
is the thermal wavelength of the piston.

(f ) The variance of the piston height is essentially the variance of the volume of the gas at
constant pressure, xpδhq2 y “ A12 xpδV q2 y, and can be found from the equipartition theorem
for the oscillating
` ` ˘ ˘ pistonT 2 supported by the ideal gas with the isothermal compressibility
κT “ V1 ´ BV BP T
“ N P 2 . It is more fun, however, to use the machinery of the pressure-
dependent partition function developed“ in parts (e,d).
‰ The volume and its fluctuation can
g`p
be found as partial derivatives of ´ ln Z pT, P q with respect to the pressure,
˜ ` ˘¸
B ´ T ln Z g`p pT, P q
ˆ g`p ˙ ˆ ˙
1 1 BZ 1 B ln Z g`p
xV y “ ´ “´ “ , (94)
β Z g`p BP T β BP T BP
T
ˆ 2 g`p ˙ ˆ ˆ g`p ˙ ˙2
1 1 B Z 1 1 BZ
xpδV q2 y “ xV 2 y ´ xV y2 “ 2 g`p 2
´ ´ g`p
β Z BP T βZ BP T
˜ ` ˘ ¸
ˆ 2 g`p
˙ 2 g`p
1 B ln Z B ´ T ln Z pT, P q
“ 2 2
“ ´T , (95)
β BP T BP 2
T
` ˘
and thus one can identify ´T ln Z g`p pT, P q with the Gibbs potential. Using the expression
for Z g`p from above, one gets
˜ “ ` ˘ ‰¸
ˆ 2 2
T2
g`p
˙
B ln Z B pN ` 1q ´ ln P ` f pT q 1
xpδV q2 y “ T 2 2
“ T 2
2
“ pN ` 1q 2
« V2,
BP T BP P N
T
(96)
and the piston height fluctuation is
δV 1 V h 1 T
δh “ «? “? “? . (97)
A NA N N Mg

44
Statistical Mechanics 2

Thermodynamics of a polymer molecule

Consider a single polymer molecule that consists of N " 1 connected elementary identical
links. These links may be either “folded” (zero length) or extended (to length a) in the same
direction, and have the same intrinsic energy in both states. One end of the molecule is
fixed, and tension f may be applied to the other end.

(a) (3 points) Find the average length L0 of the molecule when no tension is applied to
the molecule (f “ 0). How does L0 depend on temperature? What is the variance of
the length xpδLq2 y (assuming constant temperature)?
Hint: Throughout the problem, assume that the change in the length of the molecule
is small, ∆L ! L0 .

(b) (4 points) The molecule is “stretched” to fixed length L “ L0 ` ∆L. How many micro-
scopic states of the molecule correspond to this state, and what is the corresponding
entropy? You may use the Stirling approximation ln n! “ pn ln n ´ nq.

(c) (3 points) Calculate the tension f required to stretch the molecule in part(b) to length
L “ L0 ` ∆L, and its elasticity kT at constant temperature T .

(d) (3 points) Now assume that the heat capacity of the unstretched molecule CpL0 q ” C0
is independent of temperature. Calculate the heat capacity Cf for a molecule stretched
by a constant force f .

(e) (3 points) What is the adiabatic elasticity kS if the molecule?

(f) (4 points) Suggest a design for a heat engine based on the temperature dependence of
the elasticity kT pT q and calculate its efficiency. Draw the engine’s operating cycle in
the pL, T q plane and show the direction of the cycle.

45
Solution

This problem can be solved using either microcanonical or canonical (Gibbs) distribution.
The solution here takes the microcanonical approach, which somewhat more complicated bit
is equally valid, but has the advantage of emphasizing how the tension in the rubber band
appears from the entropy if its disordered links.

(a) Since both states of each link have the same energy, their probabilities are the same and
equal to 12 and independent of the temperature. The average length then is
`1 1 ˘ N
L0 “ xLyf “0 “ N ¨ ¨0` ¨a “a
2 2 2
and also does not depend on the temperature. The number of extended links is given by the
binomial distribution, so its variance
? pδN` q2 “ N ¨ 1 ¨ p1 ´ 12 q “ 14 N gives the fluctuation
of the length δL “ a δN` “ 21 a N . Alternatively, one can consider the fluctuation of the
length of one (i-th) link,
1 a 1
pδli q2 “ xp∆li q2 y “ xli2 y ´ xli y2 “ a2 ´ p q2 “ a2 , (98)
2 2 4
which gives the same answer for N independent links pδLq2 “ N pδli q2 “ 14 a2 N .

(b) If the length of the molecule is known, then the precise number of extended N` and
folded N´ links can be determined:
N` ∆L ˘ N` ∆L ˘
L “ L0 ` ∆L “ aN` ô N` “ 1` , N´ “ N ´ N` “ 1´
2 L0 2 L0
The number of microscopic states, or the statistical weight Γ, corresponding to the molecule
having length L, is given by the number of all combinations of stretched N` and folded N´
links out of the total N “ N` ` N´ links,
N!
ΓpLq “ ,
N` !N´ !
from which it is easy to compute the entropy using the Stirling formula ln n! « pn ln n ´ nq:

Sd pLq “ ln ΓpLq « N ln N ´ N ´ pN` ln N` ´ N` q ´ pN´ ln N´ ´ N´ q

N` N´
“ ´N` ln ´ N´ ln
N N
N !´ ∆L ¯ ” 1 ´ ∆L ¯ı ´ ∆L ¯ ” 1 ´ ∆L ¯ı)
“´ 1` ln 1` ` 1´ ln 1´
2 L0 2 L0 L0 2 L0
” 1 ∆L
´ ¯ 2 ı
« N ln 2 ´
2 L0

In the last line, the expansion lnp1 ` xq “ x ´ 12 x2 ` Opx3 q was used. The first term is
the entropy of a molecule without load (at f “ 0), in which any link can be in two equally

46
probable states, whereas the second reflects the decrease in the entropy as the molecule is
stretched (or compressed) and more links become extended (or folded) than N {2.
The fluctuation of the length can also be easily found from the Gaussian probability
distribution ppLq9ΓpLq “ eSd pLq :
” N ´ ∆L ¯2 ı L2 a L0 1 ?
ppLq9 exp ´ ùñ xp∆Lq2 y “ 0 , or δL “ x∆Lq2 y “ ? “ a N .
2 L0 N N 2

(c) If the molecule is stretched at constant temperature, the change in the free energy is
T “const
dF “ ´S dT ` f dL “ f dL (99)

(note that when the molecule is stretched the external force performs work δWext “ f dL).
Alternatively, the free energy for the constant temperature also can be found from the number
of microscopic states above,
ZpT, Lq ΓpT, Lq 1 ´ ∆L ¯2
F pT, Lq ´ F pT, L “ L0 q “ ´T ln “ ´T ln « TN
ZpT, L0 q ΓpT, L0 q 2 L0
where the difference in the Gibbs partition functions is given only by the disorder of the
links because the energies of links in both states are the same. Finally, the tension in the
molecule is ˆ ˙
BF ∆L 2T ∆L
f“ “ TN 2 “ ,
BL T L0 a L0
and the isothermal elasticity is
ˆ ˙
Bf 2T
kT “ “ .
BL T aL0
Note that this elasticity is different from the elasticity of a thermally insulated molecule kS
because the work done by the external force f may produce heat and change the temperature
of the molecule, leading to an additional change in the tension f .

(d) First, note that the internal energy of the molecule is independent of its length L “
L0 ` ∆L because there is no energy difference between the different states of the links.
Therefore, all the internal energy is given by its “thermal” energy and does not depend on
the elongation ∆L,
EpT, Lq “ C0 T
neither does its heat capacity at constant length CL pT, Lq “ C0 “ const. However, the heat
capacity at constant tension f is larger than CL “ C0 because of the work performed by the
molecule as it shrinks with increasing temperature:
ˆ ˙ ˆ ˙
BE BL
Cf pT, f q “ ´f ,
BT f BT f

(the minus sign is due to the relation T dS “ dE ´ f dL where f dL is the work performed
by the external force). Since the internal energy is independent of L, it is also independent

47
 ` ˘
of f , therefore BE BT f
“ C0 . The length of the molecule can be found from the equation in
(c): ∆L “ aL 2T
0
f , and

aL0 2 L0 ´ ∆L ¯2 ´ ∆L ¯2
Cf pT, f q “ C0 ` f “ C 0 ` 2 “ C 0 ` N ą CL “ C0
2T 2 a L0 L0
(e) One elegant solution is to compute the ratio of the adiabatic to the isothermic elasticities
using the chain rule for determinants,
kS Bpf, Sq BpL, T q Bpf, Sq BpL, T q Cf N ´ ∆L ¯2
“ ¨ “ ¨ “ “1`
kT BpL, Sq Bpf, T q Bpf, T q BpL, Sq CL C0 L0

` Bf ˘the expression for kT from above. Alternatively, one can evaluate the partial deriva-
and use
tive BL S by changing the variables pL, Sq Ñ pL, T q and using the molecule’s “equation of
state” f “ 2T ∆L
a L0
:
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
Bf Bf Bf BT 2 ∆L BT
kS “ “ ` ¨ “ kT `
BL S BL T BT L BL S a L0 BL S
` ˘
The derivative BT BL S
may be evaluated using the 1st law of thermodynamics and dE “ C0 dT :
ˆ ˙
2T ∆L BT 2T ∆L
dE ´ f dL “ C0 dT ´ dL “ T dS “ 0 ñ “
a L0 BL S aC0 L0
and, finally,
4T ´ ∆L ¯2 ” 2L0 ´ ∆L ¯2 ı ” N ´ ∆L ¯2 ı
kS “ kT ` “ kT 1 ` “ k T 1 ` ,
a2 C0 L0 aC0 L0 C0 L0
consistent with the previuos answer.

(f )
The simplest solution is to implement the Carnot cycle
which has efficiency η “ 1 ´ T1 {T2 shown on the right.
Note that the entropy S2 ą S1 because the shorter
(less stretched) molecule has larger entropy. Thus, the
molecule has to be cooled (S1 ´S2 ă 0) as it is stretched
at T1 and heated (S2 ´ S1 ą 0) as it is released at tem-
perature T2 .

To draw the diagram in pL, T q axes, we need to find how the temperature changes when the
molecule is stretched adiabatically. For S “ const ô dS “ 0, we have
2T ∆L “ p∆Lq2 ‰
0 “ T dS “ dE ´ f dL ô C0 dT “ dL ô T “ T0 exp
a L0 aC0 L0
and the dependency of the temperature on the elongation is exponential (for ∆L ! L0 ).

48
Statistical Mechanics 3

The thermoelectric effect

A (3 points). For a gas of similar non-relativistic particles, write:

- the Liouville theorem,
- the Boltzmann transport equation in its general form, and
- the Boltzmann equation in the relaxation-time approximation.1

B (2 points). Spell out the last equation for a free, isotropic Fermi gas of particles with electric
charge q, in the presence of a uniform, time-independent external electric field E.

C (4 points). Solve the obtained equation in the linear approximation in the weak applied field E,
and use the result to express the densities of the electric current (je) and of the heat flow (jh) as integrals
over the single-particle energy .

Hint 1: The heat flow density in a gas with the single-particle probability distribution w(r, p) may be
calculated as

j h       v wd 3 p ,
where  its chemical potential, and v is the particle’s velocity.

D (2 points). Give a physical interpretation of the formula given in Hint 1.

E (2 points). Use the first result of Task C to obtain an explicit expression for the Ohmic
conductivity  (defined as the coefficient in the differential form je = E of the Ohm law) via the gas
density n, particle's charge q and mass m, and the relaxation time , for arbitrary temperature.

F (3 points). For a degenerate Fermi gas, use the second result of Task C and the Sommerfeld
expansion formula to calculate the so-called Peltier coefficient in the linear relation jh = je.

Hint 2: The Sommerfeld expansion may be represented in several forms; for this problem, the most
useful of them is
 f     2 2 d 2  

 
0   
   d       T   , at T   ,
6 d 2

where () is a smooth function equal to zero at  = 0, and f() is the Fermi-Dirac distribution.

G (4 points). Suggest a simple method to measure the coefficient  experimentally.

Hint 3: Think about a loop made of two different conducting materials, one of them with a known .

1
The approximation is sometimes called the BGK model.
Solution

A (3 points). The general form of the Liouville theorem is

dw w d  w w 
  q j  p j   0 ,
dt t j 1  q j p j 

where w({qj}, {pj}, t) is the probability density to find the particle in the (2d + 1)-dimensional space of
its d generalized coordinates qj, d generalized momenta pj, and time t. For a non-relativistic gas of 3D
particles (d = 3, {q1, q2, q3} = r), each under the effect of an external force F(r, t) but otherwise free to
move, the equation takes the form
w
 v rw  F  pw  0,
t
where v is the particle's velocity, and r and p are the del operators in the coordinate and momentum
spaces.
The Boltzmann transport equation differs from the Liouville theorem by addition of the so-called
scattering integral, giving an approximate account of relatively rare particle scattering events, for
example in the form
w
t
 
 v   r w  F   p w   d 3 p' p' p w(r, p' , t )  pp' w(r, p, t ) ,

where  are the rates of the corresponding scattering events.

This equation is most frequently used in a simpler form further by assuming that the scattering
integral is proportional to the local deviation,
~  ww ,
w 0

of the probability density from its value w0 in equilibrium, so that equation takes the form
w ~
w
 v rw  F  pw   ,
t 
where  is a phenomenological constant, called the relaxation time. (Physically, this is the average time
a particle moves between two consequent scattering events.)

B (2 points). For a free, isotropic Fermi gas

g
w0  f  , (1)
2 3
where g = 2s + 1 is the spin degeneracy, and f() is the Fermi-Dirac distribution

f    (   ) / T
1
,
e 1
 = p2/2m is the single-particle energy,  is the chemical potential, and T is temperature in energy units.
For particles with electric charge q in an electric field E, in the absence of other forces, we may write F
= qE. If E is time- and space-independent, so should be w, so that the Boltzmann equation, in the
relaxation-time approximation, reduces to
~
~   w .
qE   p w0  w (2)


C (4 points). If the applied electric field is relatively weak (qEv << T, ), it may cause only a
weak deviation of the probability density from its equilibrium value (1), so that in the linear
approximation in E we may neglect w ~ in comparison with w on the left-hand side of Eq. (2), and it
0
may be readily solved:
w~  q E   w .
p 0

Since, according to Eq. (1), w0 depends on p only via the energy  = p2/2m, we may write pw0 =
(w0/)p. But the jth Cartesian component of this vector,
  p12  p12  p12 p j
  
p     vj,
j
p j p j 2m m

so that the gradient p is just the particle's velocity v, and we may write

~   q E  v  - w0  .
w
  
Let us first use this expression for the calculation of the electric current density
je   qvwd 3 p  q  vw0  w
~d 3 p .

Since in the equilibrium state (with w = w0), the net current density has to equal zero, the integral of the
first term in the parentheses has to vanish. For the integral of the second term, we get
 w  gq 2  f    3
je  q 2  v E  v   0  d 3 p  3 
v E  v    d p.
   2    
Due to axial symmetry of this expression, the vector j is directed along the applied field,
gq 2 E  f    3
je  v
2
cos 2    d p,
2  3
  
where Here  is the angle between the field E and the vector p = mv. Transferring to the polar
coordinates, and performing an easy angular integration, we get
gq 2 E 2  
 f    gq E 4 
 f   
2
je   d  sin d cos θ  p dp v     p dp v 2 
2 2 2 2
.
2  
3
0 0 0   2  3 3 0   
As requested in the assignment, the remaining, one-dimensional integral may be readily recast to that
over particle energies  = p2/2m: p2 = 2m, so that p = (2m)1/2, dp = (m/2)1/2, and v2 = 2/m:
 f   
 8m 

gq 2 E 4 3 1/ 2 
je     d . (3)
2  3 3 0  
Now according to the expression for the heat flow density jh given in Hint 1, it may be obtained
from Eq. (3) for je by the replacement q  ( - ), immediately giving

 8m       f   d .


gq E 4 3 1/ 2
jh  (4)
2 3 3 0   

D (2 points). By definition, the chemical potential  of a system of independent particles is just

an average energy of a particle, and a deviation ( - ) of the particle’s energy  from this average may
be interpreted as its thermal energy.
Another possible explanation is that for a uniform gas of N particles, the factor ( - ) in the
expression for jh may be represented as (E – G)/N, where E is the total energy, and G the Gibbs energy
of the system. But according to the basic thermodynamics,
E  G  TS  PV ,
where S is entropy, P the pressure, and V the volume of the gas. The full differential of this difference,
d E  G   TdS  SdT  PdV  VdP .
so that in the absence of mechanical work dW = -PdV and of changes of pressure and temperature (as in
our current problem), the differential is reduced to just TdS, i.e. to the elementary heat dQ.

E (2 points). We may take the integral in Eq. (3) by parts. Since the function under it vanishes at
both ends of the integration interval (at  = 0, due to the factor 3/2, and at   +, due to the
exponentially decreasing Fermi distribution function f), the result is

   2gq 
 
gq 2 4 4

2
  f   d 8m 3 8m 1 / 2  f   3  1 / 2 d
1/ 2

2  3 3 0
3
3 0
2

q 2 gm 3 / 2
   f   d .
1/ 2

m 2 2  3 0

But the factor following the multiplication sign is just the well-known expression for the density n  N/V
of the gas in thermal equilibrium. Indeed, using Eq. (1), and the substitutions listed just before Eq. (3),
we get
 
g g gm 3 / 2
n   w0 d p  f  d p  4  p dp f    d f   .
2 3  
3 3 2 1/ 2
(5)
2 3 0 2 2  3 0

Hence in the relaxation-time approximation, the Ohmic conductivity may be expressed by the
Drude formula,
q 2
 n, (6)
m
at arbitrary temperature.

F (3 points). Applying the Sommerfeld expansion formula, given in Hint 2, with () = 3/2( - )
 5/2 - 3/2, i.e. with () = 0 and d2()/d2= = 31/2, to Eq. (4), for the ratio jh/E   we get
gq 4  2T 2
  38m  , for T   .
1/ 2
(7)
2  3 6
3

This expression may be further simplified taking into account that for the degenerate Fermi gas, at T <<
(T)  (0)  F, the Fermi distribution becomes very simple,
1, for    F ,
f    
0, for  F   ,
enabling a simple integration in Eq. (5):
F
gm 3 / 2 gm 3 / 2 2 3 / 2
n   d  F .
1/ 2

2 2  3 0 2 2  3 3
Plugging this expression into Eq. (6) for , and the result into Eq. (7), we get simply
2 T2
 , for T   F .
2q  F
We see that the phenomenological relaxation time  has canceled, implying that the last result is
more robust to violations of the relaxation-time approximation.

G (4 points). Figure below shows the traditional (and apparently the easiest) way to measure the
Peltier coefficient.

( 1   2 ) I

2I

I  jA

1 I


2I

( 1   2 ) I
An external voltage source drives a certain dc current I = jeA (where A is the conductors’ cross-
section area), the same in the whole loop. However, if the materials 1 and 2 are different, the power P 
jhA of the voltage-induced heat flow2 may be different in two parts of the loop. Indeed, if the whole
system is kept at the same temperature, the integration of the relation jh = je over the cross-sections of
each part yields
P1, 2   1, 2 A1, 2 j1, 2   1, 2 I 1, 2   1, 2 I .
This equality means that in order to sustain a constant temperature, the following power difference,
P   1   2 I ,
has to be extracted from one junction of the two materials (in the Fig. above, shown on the top), and
inserted into the counterpart junction.
This is the thermoelectric (or Peltier) effect, may be used, in particular, for the measurement of
the coefficient  of one of the materials, provided that its value in the counterpart material is known.
Another, much more common, application of this effect is for thermoelectric cooling. Indeed, if a
constant temperature is not maintained, one of the junctions is heated (in excess of the bulk, Joule
heating), while the latter one is cooled. Such Peltier refrigerators, which require neither moving parts
nor fluids, are very convenient for modest (by a few tens C) cooling of relatively small components of
various systems - from sensitive radiation detectors on mobile platforms (including spacecraft), all the
way to cold drinks in vending machines.

2
Note that we are discussing the heat transferred through the conductors, not the additional Joule heat generated
in them by the current. (The latter heat, with the power density jeE = E2, is proportional to the square of the
applied voltage V, rather than proportional to it as the Peltier heat.)