PHZ 5941

Condensed Matter I
Problem Set 1 — Solution
1.1 Problem 1.1, A&M, Pg. 38.
The Drude model assumes the probability an electron suffers a collision per unit time is
1/τ where τ is the “relaxation time.” This means that during an infinitesimal time interval
dt the probability that a collision occurs is dt/τ and the probability that a collision doesn’t
occur is 1 − dt/τ .
(a) To determine the probability that an electron picked at random did not suffer any
collisions during the preceding t seconds we can divide the time interval t into N smaller
intervals ∆t = t/N . For small ∆t the probability of no collision during the preceding t
seconds is then simply the product of the probabilities that no collision occurred during
each of these ∆t intervals,
N
∆t

p(t) ' 1 − . (1)
τ
This can then be made exact by taking the N → ∞ limit, with the result
N
t

p(t) = lim 1− = e−t/τ . (2)
N →∞ Nτ
This exponential distribution is characteristic of a Poisson process.
(b) To determine the probability that the time interval between two successive collisions of a
given electron is between t and t+dt we again divide the time interval from the first collision
(which we can take to be at time t = 0) to time t into N smaller intervals of ∆t = t/N . The
probability of no collision occurring from time 0 to time t and then a collision occurring in
the next time interval dt is thus
N
t dt dt

lim 1− = e−t/tau . (3)
N →∞ Nτ t τ
(c) From (a) we know that if an electron is picked at random then the mean time back to
the last collision (or up to the next collision) is
Z ∞ dt
hti = te−t/τ = τ. (4)
0 τ
(d) Likewise, from (b) we know that the mean time between collisions is given by the same
integral,
Z ∞ dt
te−t/τ = τ. (5)
0 τ
(e) The results of (c) and (d) appear to pose a paradox. From (c) it follows that if we choose
an electron at random and ask what the mean time interval is between the last collision the
electron suffered and the next collision it will suffer, the answer is 2τ . But from (d) we see
that the mean time between collisions is not 2τ but τ .
The resolution of this “paradox” can most easily be seen by considering the time line of
a single electron, shown below (X’s mark collisions).

Collisions

t
a b

From this figure it should be clear that if we pick a particular point in time at random, we
are more likely to choose it to be between two collisions which are separated by a large time
interval (e.g. point a) than one between two collisions separated by a small time interval
(e.g. point b). This will “bias” the average time interval between collisions to be larger than
τ . It is therefore not surprising that when an electron is chosen at random the mean time
between the last collision and the next collision for that electron is 2τ and not τ .
We can see this more concretely if we compute the probability distribution for T , the time
between the between the next and last collisions averaged over all electrons. This probability
can be written as a convolution of p(t) = e−t/τ ,
Z T dt1 dT T
P (T )dT = p(t1 ) p(T − t1 ) = 2 e−T /τ dT. (6)
0 τ τ τ
For this distribution we indeed find that
Z ∞ T −T /τ
hT i = T e dT = 2τ. (7)
0 τ2

1.2 Problem 1.2, A&M, Pg. 41

(a) Immediately after a collision the energy of an electron is
1
Ei = m~v 2 , (8)
2

2
where ~v is a random velocity vector (drawn from the appropriate thermal distribution).
After a time t during which the electron is under the influence of a force F~ = −eE,
~ and

if there are no collisions, the energy of this electron is

 2
1 ~
eE
Ef = m ~v − t . (9)
2 m

The average energy gained by an electron after time t is then

 2  +
~ ~ (eEt)2  1 e2 E 2 2
* + *
1  eE 1 1  eEt
∆E = m ~v − t − m~v 2 = m − · ~v + = t, (10)
2 m 2 2 m m2 2 m

where we have used the fact that h~v i = 0.

It follows that the average energy lost to the ions by an electron after a collision at time
t is

1 e2 E 2 2
∆E = t. (11)
2 m

(b) If we now average over all time intervals between collisions, t, we find that the average
energy lost to the ions per electron per collision is
Z ∞ dt 1 (eE)2 2 e2 E 2 t2
h∆Ei = e−t/τ t = . (12)
0 τ 2 m m

Let p be the energy dissipated per unit volume per unit time. To find p we need only
multiply h∆E by the number density of electrons n and divide by the mean time between
collisions τ . The result is

h∆Ein ne2 τ 2
p= = E = σE 2 . (13)
τ m

It then follows that the total power dissipated by a wire of length L and cross sectional
area A (and hence volume LA) is

LA L Lρ
P = LAp = LAσE 2 = (σE)2 = (Aj)2 = (Aj)2 = RI 2 , (14)
σ Aσ A

where we have used the fact that j = σE, I = Aj and R = Lρ/A = L/(Aσ).

3
1.3 Problem 1.4, A&M Pg. 26
(a) The equation of motion for the average momentum of the electrons in the presence of
~ and B
uniform E ~ fields is
!
1 ~ + p~ × B
p~˙ = − p~ − e E ~ . (15)
τ mc

~ = B ẑ and a circularly polarized AC E,

Here we consider the case of constant B ~

~ ± = E± (ω)(x̂ ± iŷ)e−iωt .
E (16)

Since (15) is linear, we expect the solution to have the form

p~± = p± (ω)(x̂ ± iŷ)e−iωt . (17)

Note that from (17) it can readily be shown that

p~± × ~z = ∓i~p± . (18)

Plugging everything into (15) then yields

1 B
 
−iωp± (ω) = − p± (ω) − e E± (ω) ∓ i p± (ω) . (19)
τ mc
Solving for p± we then find
−eE± (ω)
p± (ω) = , (20)
−iω ∓ iωc + 1/τ
where ωc = eB/mc is the cyclotron frequency. Recall that the current density is given by

~j = −ne~v = − ne p~, (21)

m
and so, for the present problem, we can define

~j± = j± (ω)(x̂ ± ŷ)e−iωt , (22)

with j± (ω) = −(ne/m)p± . Thus we have the desired result

ne2 τ /m
j± (ω) = E± (ω). (23)
1 − i(ω ∓ ωc )τ
(b) We now consider the case

~ ± = E± (x̂ ± iŷ)ei(kz−ωt)
E (24)

4
and

~j± = j± (x̂ ± iŷ)ei(kz−ωt) . (25)

Even though the fields are no longer uniform, provided kl  1, where l is the mean free
path, the local conductivity relation (23) should still be valid.
We derive the wave equation in the usual way, first taking the curl of Faraday’s law

~ × (∇
∇ ~ = −1 ∂ ∇
~ × E) ~ × B.
~ (26)
c ∂t
~ is transverse (so ∇
Then, using the fact for the present case E ~ × (∇
~ × E)
~ = −∇2 E) and

using Ampere’s law to eliminate ∇~ × B,

~ we have

1 ∂2E~ 4π ∂~j
~ =−
−∇2 E − . (27)
c2 ∂t2 c2 ∂t
~ ± and using the Drude result (23) for ~j± we obtain
Plugging in our Ansatz for E
ω2
k 2 E± (ω) = ± (ω)E± (ω), (28)
c2
where
4πi ne2 τ /m
± (ω) = 1 + (29)
ω 1 − i(ω ∓ ωc )τ
is the complex dielectric function, which can also be written
ωp2 1
± (ω) = 1 − , (30)
ω ω ∓ ωc + i/τ
where
!1/2
4πne2
ωp = (31)
m
is the plasma frequency.
The condition for propagating solutions to Maxwell’s equations is then that there exist
solutions to the equation

k 2 c2 = ± (ω)ω 2 (32)

for which k is real (and hence ± must be positive).

(c) For circularly polarized light with positive chirality (E+ ) the dielectric function is
ωp2 1
+ (ω) = 1 − (33)
ω ω − ωc + i/τ

5
If we assume that ωp  ωc , and ωc τ  1, then we see for ω > ωp

ωp2
+ (ω) ' 1 − 2 (34)
ω

and so propagating solutions exist (as is the case for B = 0).

For ω < ωc the dielectric function + again becomes positive and real (once |ω−ωc |  1/τ ,
and so propagating solutions exist for this frequency range as well.
(d) In the limit ω  ωc (and assuming ωc τ  1 and ωp  ωc ) we are justified in making
the approximation

ωp2
+ (ω) ' (35)
ωωc

for which we have propagating solutions whose dispersion is determined by the equation

ωp2
k 2 c2 = ω 2 + (ω) ' ω , (36)
ωc

which yields the desired result

k 2 c2
!
ω = ωc . (37)
ωp2

For B = 10kG the cyclotron frequency is ωc ' 1.8 × 101 1s−1 (see Eq. 1.22 in A& M). A
typical value for the plasma frequency in metals is ωp ' 1017 s−1 (see Eq. 1.40). Thus, for
k ' 2πcm−1 we have

ω ∼ 0.7 s−1 . (38)