Physics 215 Solution Set 3 Winter 2018

1. Consider the one-dimensional problem of a particle moving in a delta-function potential:

V (x) = −Aδ(x) .

(a) Solve for the bound state energies and wave functions. Consider the cases A > 0 and
A < 0 separately.

The time-independent Schrödinger equation for this problem is

~2 d 2 ψ
− − Aδ(x)ψ(x) = Eψ(x) . (1)
2m dx2
Integrating this equation from x = −ǫ to x = ǫ, where ǫ > 0 is a positive infinitesimal
quantity, Z ǫ
~2  ′ ′

− ψ (ǫ) − ψ (−ǫ) − Aψ(0) = E ψ(x) dx .
2m −ǫ

where ψ ′ ≡ dψ/dx. We require that ψ(x) is continuous at x = 0. If this were not true
then dψ/dx would behave like a delta function near x = 0 and d2 ψ/dx2 would behave lie
the derivative of a delta function.1 Such behavior is not compatible with eq. (1). Hence, it
follows that Z ǫ Z ǫ
ψ(x) dx ≃ ψ(0) dx = 2ǫψ(0) ,
−ǫ −ǫ
which vanishes in the limit of ǫ → 0. We conclude that
2mA
lim ψ ′ (ǫ) − ψ ′ (−ǫ) = − 2 ψ(0) .
 
(2)
ǫ→0 ~
That is, dψ/dx is not continuous at x = 0; its discontinuity is determined by eq. (2).
To solve for the bound state energies, we solve eq. (1) for x 6= 0. In this case, we can set
δ(x) = 0. Normalizable energy eigenstates exist if and only if E < 0 and A > 0. That is, we
solve the differential equation,
~2 d 2 ψ
− + |E|ψ(x) = 0 , for x 6= 0,
2m dx2
where E = −|E|, under the assumption that limx→±∞ ψ(x) = 0. The solution to this
equation after imposing the boundary conditions at x = ±∞ is
(
Ne−αx , for x > 0,
ψ(x) =
Neαx , for x < 0,
1
Recall that dΘ(x)/dx = δ(x), where Θ(x) is the Heavyside step function.

1
where N is a normalization constant, and
r
2m|E|
α≡ . (3)
~2
Note that the sign of α (which is positive) is determined since limx→±∞ ψ(x) = 0.
The energy eigenvalues are determined by imposing eq. (2). Noting that N = ψ(0),
eq. (2) yields
2mAN
−2αN = − ,
~2
which implies that
mA
α= 2 . (4)
~
Note that we must require that A > 0 in order that bound state solutions exist, since A < 0
is incompatible with the requirement that α > 0. Finally, inserting eq. (3) into eq. (4) yields

mA2
E = −|E| = − .
2~2
That is, there is precisely one bound state. The corresponding energy eigenfunction is
 1/2
mA 2
ψE (x) = e−mA|x|/~ , (5)
~2
p
where we have determined that the normalization constant is N = mA/~2 by requiring
that Z ∞
|ψE (x)|2 dx = 1 .
−∞

The normalization condition yields

Z ∞ Z ∞ ∞
1 −2α|x| −2α|x| 1 −2α|x|  1 ~2
= e dx = 2 e dx = − e  = α = mA ,
|N|2 −∞ 0 α 0

after using eq. (4). The phase of N is conventionally taken to be unity, in which case
N = (mA/~2 )1/2 , as indicated in eq. (5).

(b) In the case of A > 0, consider a scattering process where the incident wave enters
from the left with E = ~2 k 2 /(2m) > 0 (where E is the energy eigenvalue of the Hamiltonian).
Determine the corresponding reflection coefficient R and the transmission coefficient T as
a function of k. Write the coefficients in terms of the dimensionless parameter b ≡ E/Eg
where −Eg is the ground state energy obtained in part (a), in the case of A > 0. What is
the behavior of T (b) as b → −1?

2
Consider a scattering process where the incident wave enters from the left. Then, the solution
to
~2 d 2 ψ
− − Eψ(x) = 0 , for x 6= 0,
2m dx2
where E > 0, is given by
(
Beikx + Ce−ikx , for x < 0.
ψ(x) = (6)
Deikx , for x > 0,
and r
2mE
k= , (7)
~2
Note that k > 0, under the assumption that no incident wave enters from the right. As
discussed in part (a), we demand that the wave function ψ(x) is continuous at x = 0, whereas
dψ/dx is discontinuous at x = 0 according to eq. (2). Noting that ψ(0) = B + C = D and
lim ψ ′ (−ǫ) = ik(B − C) , lim ψ ′ (ǫ) = ikD ,
ǫ→0 ǫ→0

which we make use of in eq. (2), we end up with two equations,

B +C = D,
  2mA
ik D − B + C = − 2 D ,
~
which can be rewritten in the following form,
D C
− = 1, (8)
B B
 
D 2imA C
1− 2 + = 1. (9)
B ~ k B
Adding eqs. (8) and (9), we immediately solve for D/B,
D ~2 k
= 2 . (10)
B ~ k − imA
Plugging back in to eq. (8) yields,
C imA
= 2 . (11)
B ~ k − imA

The corresponding reflection and transmission coefficients are

 2
C  m2 A2
R =   = 4 2 , (12)
B ~ k + m2 A2
 2
D ~4 k 2
T =   = 4 2 . (13)
B ~ k + m2 A2

3
It is convenient to rewrite eqs. (12) and (13) in terms of the the energy eigenvalue for the
scattering problem, E = ~2 k 2 /(2m) [cf. eq. (7)], and the parameter E0 ≡ −mA2 /(2~2 ),
which we recognize from part (a) is the ground state energy of the attractive delta function
potential. In terms of E and E0 , eqs. (12) and (13) yield

E0 E
R=− , T = .
E − E0 E − E0
Note that the transmission coefficient T (E) exhibits a pole at the ground state energy E0 of
the attractive delta function potential.

(c) In the case of A < 0, consider a scattering process where the incident wave enters from
the left with E = ~2 k 2 /(2m) > 0. Investigate the location of any poles of the transmission
amplitude in the complex k plane and the complex E plane, respectively. Explain your result
in light of the fact that the repulsive delta function potential possesses no bound states.

In part (b), the transmission amplitude for one-dimensional scattering was obtained in
eq. (10). In the case under consideration here, the only difference from part (b) is that
the sign of A is negative. From eq. (10), we see that the poles of the transmission amplitude
correspond to ~2 k − imA = 0. Solving for k yields
imA
k= . (14)
~2
where k ≡ (2mE/~2 )1/2 > 0. Since A < 0, it follows that Im k < 0, which means that
in the complex k-plane, the pole is located on the negative imaginary axis. Hence, in the
complex E-plane, the pole is located on the negative Re E-axis on the second Riemann
sheet. In contrast, poles corresponding to bound states reside on the negative Re E-axis
on the first Riemann sheet of the complex E-plane. We conclude that the repulsive delta
function potential supports no bound states, as expected.
Indeed, if we rewrite eq. (6) in the form,

 1 eikx + R e−ikx ,

for x < 0.
ψ(x) ∼ T T (15)
eikx ,

for x > 0,

and substitute eq. (14) into eq. (15), we see that the resulting wave function diverges expo-
nentially as x → ±∞. This implies that the resulting wave function cannot correspond to a
bound state.

4
2. A one-dimensional potential has the following form:

+∞ ,
 for x < 0,
V (x) = −V0 , for 0 < x < b,

0, for x > b,


where V0 and b are positive constants.

(a) Find V0 as a function of b such that there is just one bound state, of about zero
binding energy, for a particle of mass M.

To solve the bound state problem, we require that −V0 < E < 0. The relevant boundary
conditions are ψ(0) = 0, due to the infinitely high barrier at x = 0 and limx→∞ ψ(x) = 0.
The solution to the time-independent Schrödinger equation is
(
A sin(kx) , for 0 ≤ x < b ,
ψ(x) = −κx
Be for x > b,

where E = −|E| and

r r
2M(V0 − |E|) 2M|E|
k= , κ= , (16)
~2 ~2
with k > 0 and κ > 0.2
Continuity of ψ and dψ/dx at x = b yields two equations,

A sin(kb) = Be−κb ,
Ak cos(kb) = −Bκe−κb .

Diving these two equations, we obtain

1 1
tan(kb) = − . (17)
k κ
Note that eq. (16) implies that
2MV0
k 2 + κ2 = . (18)
~2
We can also rewrite eqs. (17) and (18) as

kb cot(kb) = −κb , (19)

2MV0 b2
(kb)2 + (κb)2 = . (20)
~2
2
We require κ > 0 so that ψ(x) → 0 as x → ∞. The requirement of k > 0 is conventional, since the sign
of k can be absorbed into the definition of the constant A.

5
Eqs. (19) and (20) can be solved graphically by looking for intersection of the function
y = −x cot x with the circle of radius r = (2MV0 b2 /~2 )1/2 in the first quadrant of the x–y
plane (where x = kb and y = κb).

κb

bound
state

kb
π
2

It is clear from the figure above that if V0 is less than some critical value, then the radius
of the circle, r < 21 π, in which case there are no intersections in the first quadrant, and
therefore no bound states. The critical value of of V0 is obtained by setting r = 21 π. That is,

2MV0 b2 π2
= ,
~2 4
which yields
π 2 ~2
V0 = . (21)
8Mb2

(b) Applying this crude model to the deuteron (a bound state of a proton and a neutron),
evaluate V0 in MeV, assuming b = 1.3 × 10−13 cm and M = 12 mp (where mp is the proton
mass).

The rest mass of the proton is mp c2 = 938.272 MeV. Therefore, eq. (21) yields
π 2 ~2 π 2 (6.5817 × 10−22 MeV sec)2 (3 × 1010 cm/sec)2
V0 = = ≃ 60.7 MeV .
8Mb2 4(1.3 × 10−13 cm)2 (938.272 MeV)

6
 (c) Why did I set M = 21 mp rather than M = mp in part (b)?

Applying the Schrödinger equation to the bound state of a proton and a neutron, we proceed
as in classical mechanics by converting the two-body problem into a one-body problem by
separating off the center-of-mass motion. The corresponding mass employed in the one-
body problem is the reduced mass, M = mp mn /(mp + mn ). Since mp ≃ mn , it is a good
approximation to take M = 12 mp .

REMARK: We have obtained here an estimate of the mean value of the nuclear potential
in light of the fact that the deuteron is a weakly bound system with a bound state energy
of Eb ≃ −2.2 MeV. Since Eb ≪ V0 (which was estimated in part (b) to be ∼ 60 MeV), it
follows that our approximation of Eb ≃ 0 [corresponding to the critical value of V0 obtained
in eq. (21)] is justified.

3. Consider a one-dimensional quantum mechanical problem with a time-independent Hamil-

tonian, H. The time evolution operator, evaluated in the coordinate basis, also known as
the propagator, is given by,
Z ∞
′ −iHt/~ ′
G(x, t ; x , 0) = hx| e |x i = dp hx |pi hp| e−iHt/~ |x′ i , (22)
−∞

where we have taken the initial time to be t0 = 0.

(a) The free particle Hamiltonian is given by H = P 2 /(2m). Evaluate the free particle
propagator by explicitly performing the p-integration using eq. (22).

Plugging H = P 2/(2m) into eq. (22), and using the fact that P |pi = p |pi,3
Z ∞
ip2 t
 
′ ′
G(x, t ; x , 0) = dp hx |pi hp |x i exp − .
−∞ 2m~
Using hx |pi = (2π~)−1/2 eipx/~ and hp |x′ i = hx |pi∗ , it follows that
Z ∞
p2 t
  
′ 1 i ′
G(x, t ; x , 0) = exp p(x − x ) − dp . (23)
2π~ −∞ ~ 2m

In order to perform the integral given in eq. (23), we “complete the square,” by rewriting
the argument of the exponent in eq. (23) as,
 2 2
p(x − x′ ) m(x − x′ ) im(x − x′ )2
 
pt it
−i − =− p− + .
2m~ ~ 2m~ t 2~t
3
Since P is self-adjoint, we also have hp| P = hp| p (since the eigenvalues of a self-adjoint operator are
real), and hp| f (P ) = hp| f (p), for any function f (P ) expressed as a Taylor series in P .

7
Plugging this last result back into eq. (23) yields,
Z ∞ ( 2 )
′ 2 ′
 
1 im(x − x ) it m(x − x )
G(x, t ; x′ , 0) = exp exp − p− dp .
2π~ 2~t −∞ 2m~ t

We now change integration variables by defining p′ ≡ p−m(x−x′ )/t. The limits of integration
do not change and we are left with,
Z ∞
im(x − x′ )2 itp′ 2
  
′ 1
G(x, t ; x , 0) = exp exp − dp′ .
2π~ 2~t −∞ 2m~
The integral can now be evaluated by making use of,
Z ∞ Z ∞ r
−iap2 −iap2 −iπ/4 π
e dp = 2 e dp = e , for a > 0, (24)
−∞ 0 a
which was derived in the class handout entitled, A Gaussian integral with a purely imaginary
argument. Identifying a ≡ t/(2m~), and noting that a > 0 if we take t > 0 corresponding to
propagation forward in time, we end up with,
im(x − x′ )2
 m 1/2  
′ −iπ/4
G(x, t ; x , 0) = e exp . (25)
2π~t 2~t
Finally, if we interpret e−iπ/4 = i−1/2 , we can rewrite eq. (25) as,
im(x − x′ )2
 m 1/2  
′
G(x, t ; x , 0) = exp . (26)
2πi~t 2~t

(b) For the one-dimensional harmonic oscillator, where H = P 2 /(2m) + 12 mω 2 X 2 , can

you evaluate the propagator by explicitly performing the p-integration in eq. (22)? Why is
this calculation doomed to failure?

If we try to compute the propagator of the one-dimensional harmonic oscillator using the
same technique as the one employed in part (a), then we would have to evaluate
it P 2
  
2 2
hp| exp − 1
+ mω X |x′ i . (27)
~ 2m 2
If we could write
it P 2 2 2 2
      
2 2 ? itP itmω X
exp − + 1 mω X = exp − exp − , (28)
~ 2m 2 2m~ 2~
then it would follow that
it P 2 it p2
     
2 2 ′ ? 2 ′2
hp| exp − 1
+ 2 mω X |x i = exp − 1
+ 2 mω x hp |x′ i , (29)
~ 2m ~ 2m
after employing the eigenvalue equations for P and X.

8
 However, eqs. (28) and (29) are both false. In particular, eq. (28)
 would
 be correct if and
2 2
only if P and X were commuting operators. But, we know that X, P = i~I, so that
 2 2
X , P = X X, P 2 + X, P 2 X
   
       
= XP X, P + X X, P P + P X, P X + X, P P X
= 2i~(XP + P X) . (30)
Thus, X 2 , P 6= 0 (nor is it a c-number). Hence, to evaluate the matrix exponential on the
2
 

left hand side of eq. (28), one must employ the Zassenhaus formula for matrix exponentials,4
exp t(A + B) = etA etB exp − 12 t2 A, B exp 16 t3 2 B, [A, B] + A, [A, B]
        

··· ,
where the exponents of higher order in t involve nested commutators.
Consequently, the right hand side of eq. (28) must be corrected by multiplying by a
complicated product of exponentials. Hence, it is not possible to simply replace the operators
P and X with their corresponding eigenvalues in eq. (27).
In fact, one can surmount these difficulties by a clever manipulation of eq. (27). The first
step is to divide up the interval from x to x′ into N equal segments. In the limit of infinitely
large N, each segment is infinitesimally small. By inserting N complete sets of position
eigenstates and corresponding momentum eigenstates and then formally taking N → ∞,
one can develop the so-called path integral representation of the propagator.
The resulting path integral can then be evaluated exactly in the case of the harmonic
oscillator. Since we do not have time in this course to pursue this formalism, we shall seek
another technique for computing the propagator of the one-dimensional harmonic oscillator.
This is the goal of the next two parts of this problem.

NOTE ADDED:

Perhaps I am being a little too hasty in stating that a direct evaluation of eq. (27)
is doomed to failure. Indeed, a direct evaluation has been carried out in F. A. Barone,
H. Boschi-Filho and C. Farina, Three methods for calculating the Feynman propagator, Amer-
ican Journal of Physics 71, 483 (2003). In this paper, the authors provide three different
methods of evaluation. The first method corresponds to the method you will employ in parts
(c) and (d) of this problem. The third method employs the path integral technique alluded
to above. But the second method makes use of the remarkable fact that the exponential in
eq. (27) can be expressed as the following product of exponentials,
it P 2
  
2 2
exp − 1
+ mω X = eiωt/2 exp(−αX 2 ) exp(−βP 2 ) exp(ωtP X/~) exp(βP 2) exp(αX 2) ,
~ 2m 2
(31)
4
The Zassenhaus formula for matrix exponentials is sometimes referred to as the dual of the Baker-
A B
Campbell Hausdorff formula. The latter provides a formula for
 e e in terms of a single matrix exponential,
1

whose argument is an infinite series of terms, A+ B + 2 A, B + · · · , where succeeding terms consist of nested
commutators. For further details, see, e.g., Fernando Casas, Ander Nurua and Mladen Nadinic, Efficient
computation of the Zassenhaus formula, Comp. Phys. Commun. 183, 2386 (2012).

9
where
mω 1
α≡ , β≡− .
2~ 4m~ω
Using eq. (31), it is now possible to evaluate eq. (27), with the help of the relation,
hp′ | exp(ωtP X/~ |pi = e−iωt δ(p′ − e−iωt p) . (32)
(Can you prove the above result?) As a challenge to the reader, try evaluating eq. (27) using
the results of eqs. (31) and (32) and then derive the propagator for the harmonic oscillator
using eq. (22).

(c) Show that G(x, t; x′ , 0) = hx, t | x′ , 0i, where the |x, ti are basis states in the Heisenberg
representation. Deduce the following differential equation for G,
∂G
i~ = hx, t| H |x′ , 0i ,
∂t
where the boundary condition at t = 0 is G(x, 0; x′ , 0) = δ(x − x′ ).

Recall that for basis states in the Heisenberg representation,

|x , ti = eiHt/~ |xi , (33)
assuming that the Hamiltonian H is time-independent. It immediately follows that |x , ti
satisfies the wrong-sign Schrödinger equation,
∂
i~ |x , ti = −H |x , ti . (34)
∂t
For the corresponding bras, we take the adjoint of eqs. (33) and (34) to obtain,
hx , t| = hx| e−iHt/~ , (35)
and
∂
hx , t| = hx , t| H ,
i~ (36)
∂t
where we have used the fact that H is self-adjoint.
Eq. (50) then yields,
G(x, t; x′ , 0) ≡ hx| e−iHt/~ |x′ i = hx , t | x′ , 0i ,
after using eqs. (33) and (35). Finally, using eq. (36), it immediately follows that
∂G
i~ = hx, t| H |x′ , 0i . (37)
∂t
The relevant boundary condition at t = 0 is
G(x, 0; x′ , 0) = hx , 0 | x′ , 0i = hx | x′ i = δ(x − x′ ) . (38)

10
 (d) Evaluate the propagator for the one-dimensional harmonic oscillator by employing
the following steps. First, by using the Heisenberg equations of motion, express P ≡ P (0)
in terms of X(t) and X ≡ X(0). Then solve the differential equation obtained in part (c),
subject to the boundary condition at t = 0.

The one-dimensional harmonic oscillator Hamiltonian is given by

P2
H= + 1 mω 2 X 2 . (39)
2m 2
To solve for the harmonic oscillator propagator, we shall solve the differential equation given
by eq. (37). Since H is time-independent, H(t) = H(0), and we may employ eq. (39), where
P ≡ P (0) and X ≡ X(0). Our strategy is to rewrite P (0) in terms of the Heisenberg picture
operators X(t) and X ≡ X(0). To accomplish this, we make use of the Heisenberg equations
of motion,
dP (t) i  dX(t) i 
= − P (t) , H(t) , = − X(t) , H(t) .
dt ~ dt ~
The commutators are easily evaluated:
P 2 (t) 1
 
+ 2 mω X(t) = −i~mω 2 X(t) ,
2
 
P (t) , H(t) = P (t) ,
2m
P 2 (t) 1
 
  2 P (t)
X(t) , H(t) = X(t) , + 2 mω X(t) = i~ .
2m m
That is, the Heisenberg equations of motion are given by,
dP (t) dX(t) P (t)
= −mω 2 X(t) , = .
dt dt m
Taking the derivative of dX(t)/dt with respect to time and using the equation for dP (t)/dt
yields
d2 X(t)
= −ω 2 X(t) .
dt
The solution to this equation is
P
X(t) = sin ωt + X cos ωt , (40)
mω
where P ≡ P (0) and X ≡ X(0).
We can now express P in terms of X(t) and X,
mω  
P = X(t) − X cos ωt . (41)
sin ωt
We would like to insert this into eq. (39). In computing P 2, it is important to remember
that X(t) does not commute in general with X(0). In particular, in light of eq. (40),
 
  P sin ωt  i~ sin ωt
X , X(t) = X , sin ωt + X cos ωt = X , P] = .
mω mω mω

11
 Thus, squaring eq. (41) yields

m2 ω 2
 
2 2 2 2
 
P = X + X cos ωt − cos ωt X(t)X + XX(t)
sin2 ωt

m2 ω 2
 
2 2 2 i~ sin ωt cos ωt
= X + X cos ωt − 2X(t)X cos ωt − .
sin2 ωt mω

Inserting this result into eq. (39), we end up with

mω 2
 
2 2 2 i~ sin ωt cos ωt
H= X (t) + X cos ωt − 2X(t)X cos ωt − + 12 mω 2 X 2 . (42)
2 sin2 ωt mω

We can now evaluate hx , t| H |x′ , 0i by employing the eigenvalue equations,

X(t) |x , ti = x |x , ti , X |x , 0i = X(0) |x , 0i = x |x , 0i .

It then follows that

mω 2
   
′ 2 ′2 2 ′ i~ sin ωt cos ωt 1 2 ′2
hx , t| H |x , 0i = x + x cos ωt − 2xx cos ωt − + 2 mω x hx , t | x , 0i
2 sin2 ωt mω

mω 2  2
 
′2 ′
 i~ω cos ωt
= x + x − 2xx cos ωt − hx , t | x , 0i
2 sin2 ωt 2 sin ωt

Plugging this result into eq. (37) yields the following differential equation,

−imω 2  2
 
∂G ′2 ′
 ω cos ωt
= x + x − 2xx cos ωt − G, (43)
∂t 2~ sin2 ωt 2 sin ωt

subject to the boundary condition specified by eq. (38).

Obtaining the solution to eq. (43) is straightforward. Integrating over t, and making use
of
dt 1 cos ωt 1 cos ωt 1
Z Z Z
2 dt = − cot ωt , 2 dt = − , dt = ln | sin ωt| ,
sin ωt ω sin ωt ω sin ωt sin ωt ω
we end up with
( )
imω (x2 + x′ 2 ) cos ωt − 2xx′
 
G(x, t ; x′ , 0) = g(x, x′ ) exp − 12 ln | sin ωt| ,
2~ sin ωt

where g(x, x′ ) is a function of x and x′ to be determined. We can rewrite G(x, t ; x′ , 0) in

the following form,
(  2 )
′2 ′
imω (x + x ) cos ωt − 2xx
G(x, t ; x′ , 0) = g(x, x′ )| sin ωt|−1/2 exp . (44)
2~ sin ωt

12
 We now impose the boundary condition specified by eq. (38). However, we cannot just
set t = 0 in eq. (44), since the resulting expression is not well defined. Instead, we shall
consider the limit as t → 0+ and impose the condition that5
lim+ G(x, t ; x′ , 0) = δ(x − x′ ) .
t→0

Since cos ωt ≃ 1 and sin ωt ≃ ωt as t → 0, it follows that

im(x − x′ )2
 
′ ′ −1/2
δ(x − x ) = lim+ g(x, x )(ωt) exp . (45)
t→0 2~t
Eq. (45) should be compared with the following representation of the delta function6
(x − x′ )2
 
′ 1
δ(x − x ) = lim exp − , (46)
∆→0 (π∆2 )1/2 ∆2
If we define ∆2 ≡ 2i~t/m, then we can therefore identify7
 mω 1/2
g(x, x′ ) = . (47)
2πi~
Plugging this result back into eq. (44), we end up with,
(  2 )
′2 ′
 mω −1/2 imω (x + x ) cos ωt − 2xx
G(x, t ; x′ , 0) = exp , (48)
2πi~ sin ωt 2~ sin ωt
assuming that 0 < t < π/ω, in which case we can drop the absolute value signs from the
factor | sin ωt| that appears in the prefactor of the exponential in eq. (44).8

(e) Check that in the limit of ω → 0, the result of part (d) reduces to the free particle
propagator obtained in part (a).

If one evaluates eq. (48) in the limit of ω → 0, the end result is,
im(x − x′ )2
 m 1/2  
′
G0 (x, t ; x , 0) = exp , (49)
2πi~t 2~t
which is the expression for the free-particle propagator previously obtained in eq. (26).
5
The t → 0 limit should be taken such that t = 0 is approached from the positive side (this is the meaning
of the symbol t → 0+ ), since one is typically interested in propagation forward in time, i.e. t > 0.
6
See e.g., eq. (1.10.19) of Ramamurti Shankar, Principles of Quantum Mechanics, 2nd edition, p. 61.
7
Strictly speaking, these steps are valid only for Re ∆2 > 0. When ambiguities of this nature arise in
quantum mechanics, the safest procedure is to add an infinitesimal real part to ∆2 in such a way that
convergence is assured. At the end of the calculation, this infinitesimal quantity can be taken to zero.
8
The result given in eq. (48) is ambiguous, since it is not clear how to interpret i−1/2 in the prefactor.
One can show that the naive choice, i−1/2 = e−iπ/4 , is correct only for 0 < t < π/ω. This ambiguity is
associated with the fact that sin ωt = 0 at t = π/ω which means that the propagator diverges at this point
(called the first caustic). For values of t > π/ω, an additional phase factor arises that cannot be fixed by the
limiting procedure of eqs. (45)–(47). A simple way of deriving this phase factor is given in Nora S. Thornber
and Edwin F. Taylor, American Journal of Physics, 66, 1022 (1998). You can also find references there to
the original literature and references to the few textbooks that address this phase ambiguity.

13
4. The partition function is defined by,

Z(β) ≡ Tr e−βH ,

where H is a time-independent Hamiltonian operator and β is a real positive parameter.

(a) In the case of a one-dimensional quantum mechanical problem, show that

Z
Z(β) = G(x, −i~β ; x, 0) dx ,

where the propagator is defined by,

G(x, t; x′ , 0) ≡ hx| e−iHt/~ |x′ i . (50)

Evaluating the trace in a coordinate basis,

Z Z 
−βH 3 −βH 3

Z(β) ≡ Tr e = d x h~ x| e |~
xi = d x G(~
x, t; ~
x, 0) ,
t=−i~β

which yields the result quoted in eq. (50).

(b) Show that the ground state energy E0 is given by:

1 ∂Z
E0 = lim − .
β→∞ Z ∂β

One can also compute Z(β) by evaluating the trace with respect to a basis of energy eigen-
states. Assuming that the basis states are normalized to unity, hEm |En i = δmn , where
H |En i = En |En i, it follows that
X X X
Z(β) ≡ Tr e−βH = hEn | e−βH |En i = e−βEn hEn |En i = e−βEn , (51)
n n n

where I have adopted a convention in which Em ≤ En for m < n. Henceforth, we will assume
that the ground state energy E0 is non-degenerate. Hence, eq. (51) yields
X
En e−βEn
1 ∂Z E0 eβE0 + E1 eβE1 + E2 eβE2 + · · ·
− = nX =
Z ∂β e−βEn eβE0 + eβE1 + eβE2 + · · ·
n

E0 + E1 eβ(E1 −E0 ) + E2 eβ(E2 −E0 ) + · · ·

= ,
1 + eβ(E1 −E0 ) + eβ(E2 −E0 ) + · · ·

14
where E0 is the (non-degenerate) ground state energy as previously noted. This means that
En − E0 > 0 for n = 1, 2, 3, . . .. Taking the limit as β → ∞, it follows that e−β(En −E0 ) → 0
for all n = 1, 2, 3, . . .. Consequently,
1 ∂Z
E0 = lim − . (52)
β→∞ Z ∂β

REMARK: Using eq. (51), one can immediately establish the following result,
Z ∞ X 1
e−βE Z(β)dβ = .
0 n
E + En

However, in the case of the harmonic oscillator, both sides of this relation are divergent.
Thus, this relation does not provide a method for computing the energy levels of the harmonic
oscillator. Nevertheless, all is not lost as we now examine in parts (c) and (d) of this problem.

(c) Using the results of part (b) of this problem and part (d) of Problem 3, compute the
ground state energy of the one-dimensional harmonic oscillator.

One can now evaluate Z(β) as follows. First, we set x′ = x and t = −i~β in the expression
for G(x, t ; x′ , 0) given in eq. (48). Using sin(iz) = i sinh z and cos(iz) = cosh z, we obtain,9
−1/2
mωx2 (cosh β~ω − 1)
  
mω
G(x, −i~β ; x, 0) = exp − . (53)
2π~ sinh β~ω ~ sinh β~ω
It follows that
∞  1/2
1 2
Z
Z(β) = dx G(x, −i~β ; x, 0) = .
−∞ 2 cosh β~ω − 1
In light of the identity, sinh2 (z/2) = 21 (cosh z − 1), we can rewrite Z(β) as
1
Z(β) = 1

, (54)
2 sinh 2
β~ω
which yields,
1 ∂Z
= 12 ~ω coth 1


− 2
β~ω .
Z ∂β
Finally, we use eq. (52) to compute the ground state energy.
1 ∂Z
E0 = lim − = 12 ~ω ,
β→∞ Z ∂β
as expected for the ground state energy of the harmonic oscillator.
9
In contrast to eq. (48), no phase ambiguities arise in the derivation of eq. (53). Indeed, since β > 0, the
prefactor is equal to the positive inverse square root of a positive quantity.

15
 (d) Consider the the propagator for a one-dimensional quantum system governed by
a time-independent Hamiltonian with only discrete (bound state) energy levels, {En }, for
n = 0, 1, 2, 3, . . .. Using eq. (50), show that the full energy spectrum of H can be determined
from10
Z ∞ X ∞
X
−iHt/~ −iHt/~
Tr e = dx G(x, t ; x, 0) = hEn | e |En i = e−iEn t/~ . (55)
−∞ n n=0

Strictly speaking, the sums on the right-hand side of eq. (57) are not convergent. However,
one can give mathematical meaning to these sums by extending the time parameter to the
complex plane.
In particular, let t = −i~β (where β is a positive real parameter). Using eq. (57), obtain
all the energy eigenvalues of the one-dimensional harmonic oscillator.

In analogy with part (a) of this problem, we can write Tr e−iHt/~ in two different ways. With
respect to the coordinate basis,
Z ∞ Z ∞
−iHt/~ −iHt/~
Tr e = dx hx| e |xi = G(x, t ; x, 0) .
−∞ −∞

where we have made use of eq. (50). Alternatively, we can compute the trace with respect
to the energy eigenstate basis,
X X
Tr e−iHt/~ = hEn | e−iHt/~ |En i = e−iEn t/~ ,
n n

under the assumption that the energy eigenstates are normalized to unity. Hence, eq. (55)
is established. If we now put t = −i~β, then we obtain
X
Z(β) ≡ Tr e−βH = e−βEn , (56)
n

which is a result that was previously obtained in part (b) [cf. eq. (51)].
In part (c) we derived an explicit formula for Z(β) [cf. eq. (54)]. We can rewrite this
result in the following way:
∞ ∞
1 1 e−β~ω/2 −β~ω/2
X
−βn~ω
X
−β(n+ 1 )~ω
Z(β) = 1

= β~ω/2 −β~ω/2
= −β~ω
= e e = e 2 .
2 sinh 2
β~ω e −e 1−e n=0 n=0

Comparing with eq. (56), we conclude that the energy levels of the harmonic oscillator are

En = ~ω(n + 21 ) .
10
In eq. (57), the trace is expressed as a diagonal sum of matrix elements by employing two different basis
choices (the coordinate basis and the energy basis, respectively). Of course, the trace is a basis-independent
quantity, so one may choose any orthonormal basis to compute it.

16
5. Consider a particle in one dimension trapped between two impenetrable walls at x = 0
and x = L.

(a) Determine the bound state energy levels, En , of the particle. (Here, n labels the
possible energy eigenvalues: n = 1 is the ground state, n = 2 is the first excited state, etc.).

We solve the time-independent Schrödinger equation,

~2 d2 ψ(x)
Hψ(x) = − = Eψ(x) , (57)
2m dx2
subject to the boundary conditions, ψ(0) = ψ(L) = 0, which are a consequence of the two
impenetrable walls at x = 0 and x = L. As usual, we define
r
2mE
k= ,
~2
where k > 0. Then, the solution to eq. (57) is
(
A sin kx , for 0 ≤ x ≤ L,
ψ(x) =
0, for x > L and x < −L,

after imposing the boundary condition ψ(0) = 0. We next impose the boundary condition
ψ(L) = 0, which yields,

kL = nπ , for any positive integer n. (58)

Note that since k > 0, n cannot be negative, and n = 0 is not allowed since this would
correspond to ψ(x) = 0, which is not an eigenfunction of H. Eq. (58) then yields,
~2 k 2 π 2 ~2 n2
En = = . (59)
2m 2mL2

The normalization constant A is easily determined by setting

Z L
1 2 nπx
 
2
= sin dx = 12 L ,
|A| 0 L
p
which yields A = 2/L after setting the arbitrary phase factor to unity. It is convenient to
write the expressions for the energy eigenstates with the help of the Heavyside step function.
Then, the eigenfunction corresponding the the nth energy level is given by
r
2  nπx 
ψn (x) = sin [Θ(x) − Θ(x − L)] , (60)
L L
where Θ(x) is the Heavyside step function, and the quantity Θ(x) − Θ(x − L) in eq. (60)
ensures that the wave function is zero beyond the two impenetrable walls at x = 0 and
x = L.

17
 (b) Suppose that at time t = 0, the state of the particle is given by the wave function

ψ(x, t = 0) = Ax(L − x) [Θ(x) − Θ(x − L)] ,

where Θ(x) is the Heavyside step function and A is a normalization constant.

If an energy measurement is performed at time t = 0, what is the probability that
the particle will be observed to be in the ground state? Find an exact expression for the
probability that the particle will be observed to be in a state of energy En (for any positive
integer n).

The normalization constant is determined in the usual way,

Z L
1 2 2 2 1 1 1

L5
= x (L − x) dx = L 3
−2+5 = .
|A|2 0 30

Hence, after setting the the arbitrary phase factor to unity,

r
30
A= .
L2
Thus, r
30
ψ(x, t = 0) = x(L − x) [Θ(x) − Θ(x − L)] , (61)
L2
Next, we expand ψ(x, t = 0) in terms of energy eigenstates,
∞
X
ψ(x, 0) = cn ψn (x) , (62)
n=1

where the energy eigenstates (which are a complete set of states), are given by eq. (60).
Using the fact that the {ψn (x)} are orthonormal,
Z L
ψn∗ (x)ψm (x) dx = δnm ,
0

we can multiply eq. (62) on the left by ψn∗ (x) and integrate to project out the coefficients cn .
Thus,
L  1/2  1/2 Z L
30 2  πnx 
Z
cn = ψn∗ (x)ψ(x, 0) dx = x(L − x) sin dx .
0 L5 L 0 L

after the change of variables, y = πx/L,

√ Z
2 15 π  y
cn = y 1− sin(ny)dy .
π2 0 π

18
Employing the integrals,
Z π
π
y sin(ny)dy = − (−1)n ,
0 n
π
π2 2
Z
y 2 sin(ny)dy = − (−1)n + 3 (−1)n − 1 ,

0 n n
we end up with √
4 15 
cn = 3 3 1 − (−1)n .

π n
We conclude that the probability of finding the particle in the state with energy En , denoted
by Pn , is given by
240 
Pn = |cn |2 = 6 6 1 − (−1)n ]2 .
π n
The ground state corresponds to n = 1. Thus, the probability that the particle will be
observed to be in the ground state is
960
P1 = = 0.9986 .
π6
This means that ψ(x, 0) is a pretty good approximation to the ground state energy eigen-
function!

(c) Evaluate the expectation value of the Hamiltonian with respect to the wave function
given in eq. (61). What is the average value of the energy at time t = 0?

The average energy is given by the expectation value of the Hamiltonian. Thus, in light of
eq. (57),
Z L
~2 d2 ψ(x)
hEi = hψ| H |ψi = − ψ ∗ (x) dx .
2m 0 dx2
Plugging in eq. (61),
30 L ~2 d 2 30~2 L 5~2
Z   Z
hEi = 5 x(L − x) − x(L − x)dx = x(L − x)dx = . (63)
L 0 2m dx2 mL5 0 mL2

 WARNING :

The calculation presented above is too naive, although the end result turns out to be
correct. Note that starting from eq. (61), it follows that
r  
dψ 30
= (L − 2x) [Θ(x) − Θ(x − L)] + x(L − x) [δ(x) − δ(x − L)] , (64)
dx L2

19
after using δ(x) = dΘ(x)/dx. Luckily, the second term on the right hand side of eq. (64)
does not contribute, since xδ(x) = 0 and (L − x)δ(x − L) = 0. Hence,
r  
dψ 30
= (L − 2x) [Θ(x) − Θ(x − L)] .
dx L2
However, when we take the second derivative,
r 
d2 ψ

30
= −2 [Θ(x) − Θ(x − L)] + (L − 2x) [δ(x) − δ(x − L)] ,
dx2 L2
we cannot immediately discard the terms proportional to the delta functions. Indeed, (L −
2x)δ(x) = Lδ(x) and (L − 2x)δ(x − L) = −Lδ(x − L), Hence,
r 
d2 ψ

30
= −2 [Θ(x) − Θ(x − L)] + L [δ(x) + δ(x − L)] ,
dx2 L2

When we compute hEi more carefully, we obtain

Z L
~2 d2 ψ(x)
hEi = − ψ ∗ (x) dx
2m 0 dx2

15~2 ∞
Z  
=− x(L − x) [Θ(x) − Θ(x − L)] −2 [Θ(x) − Θ(x − L)] + L [δ(x) + δ(x − L)] .
mL5 −∞
Since an overall factor of x(L − x) appears in the integrand, we can use it to kill the delta
function terms, i.e., x(L − x) [δ(x) + δ(x − L)] = 0. It then follows that
30~2 L 5~2
Z
hEi = x(L − x)dx = .
mL5 0 mL2

The reason I mention all this is that keeping track of the terms that arise by taking
multiple derivatives of the Heavyside step function will sometimes matter. In particular,
if you were to calculate hE 2 i = hψ| H 2 |ψi, the naive calculation of the integrand would
give zero, but the correct calculation would yield terms proportional to the derivative of the
delta function, which contributes to the final result. Indeed, the correct calculation yields
hE 2 i = 30~4 /(m2 L4 ). (Try it!)

(d) The expectation value of H can be computed by a different method than the one
used in part (c). First, expand ψ(x, 0) as a linear combination of energy eigenstates, and
then show that the expectation value of H can be expressed as an infinite sum. Using this
technique, obtain an expression for the average value of the energy at time t = 0, and then
employ the result obtained in part (c) to determine the value of the sum,
∞
X 1
.
n=0
(2n + 1)4

20
We can also compute hEi by inserting a complete set of energy eigenstates,

X X ∞
X
hEi = hψ| H |ψi = hψ| H |En i hEn |ψi = En hψ|En i hEn |ψi = En |cn |2 , (65)
n n n=1

after making use of eq. (62) and identifying cn = hEn |ψi.

In part (b), we computed

240   960 ,

for n odd,
6 6
|cn | = 6 6 1 − (−1) ] = π n
2 n 2
π n  0, for n even.


Inserting this result into eq. (65) and making use of eq. (59), we end up with

480~2 X 1
hEi = 4 ,
π mL2 n>0 n4
n odd

which we can rewrite as ∞

480~2 X 1
hEi = 4 . (66)
π mL2 n=0 (2n + 1)4
Comparing eqs. (63) and (66), we conclude that
∞
X 1 π4
= . (67)
n=0
(2n + 1)4 96

ADDED NOTE:

Sums such as the one given in eq. (67) are related to the famous Riemann zeta function,11
∞ ∞
X 1 1 X 1
ζ(z) = z
= 1−z
(−1)n+1 z ,
n=1
n 1−2 n=1
n

where the first sum is valid for Re z > 1 and the second sum is valid for Re z > 0 (assuming
that z 6= 1 where the Riemann zeta function diverges). It then follows that
X 1 ∞ ∞
1X 1 1X 1
z
= z
+ (−1)n+1 z = (1 − 2−z )ζ(z) , for Re z > 1.
n>0 n 2 n=1 n 2 n=1 n
n odd

11
See e.g., I.S. Gradshteyn and I.M. Ryzhik, Table of Integral, Series and Products, 8th edition (Academic
Press, Waltham, MA, 2015) pp. 1046–1049.

21
It is well known that ζ(4) = π 2 /90. Hence,
X 1 π4
4
= .
n>0 n 96
n odd

in agreement with eq. (67).

(e) After preparing the state given by eq. (61) at time t = 0, suppose that instead of
performing an energy measurement at time t = 0, I wait a while and then make the first
energy measurement at a later time t > 0. Do any of the results obtained in parts (b) and (c)
change? Explain.

The results of parts (b) and (c) do not change as a function of time. To prove this, we first
note that eq. (60) is equivalent to
X
|ψ(0)i = cn |Ei ,
n

when evaluated with respect to the coordinate basis. Multiplying on the left by hEm | and
using hEm |En i = δmn , it immediately follows that cn = hEn |ψ(0)i. Next, we recall that
X X
|ψ(t)i = U(t) |ψ(0)i = U(t) |En i hEn |ψ(0)i = cn e−iEn t/~ |En i (68)
n n


where we have used the fact that the time evolution is given by U(t) = exp −iHt/~ for a
time-independent Hamiltonian, and H |En i = En |En i. It immediately follows that
∞
X
ψ(x, t) = cn e−iEn t/~ ψn (x) , (69)
n=1

after evaluating eq. (68) with respect to the coordinate basis. From eq. (69), we conclude that
the probability of a state with wave function ψ(x, t) to be observed in an energy eigenstate
with energy En at time t is given by
|cn e−iEn t/~ |2 = |cn |2 ,
which is indeed time-independent.
To see that hEi is also time-independent, we make use of Ehrenfest’s equation,
 
d i
 ∂Ω
hΩi = [H , Ω] + .
dt ~ ∂t
Setting ω = H, it follows that
d
hEi = 0 ,
dt
for any time-independent Hamiltonian. This is the quantum mechanical statement of energy
conservation.

22
6. Consider a periodic potential in one-dimension which satisfies V (x + ℓ) = V (x).

(a) Show that the translation operator T = exp(−iℓP/~) commutes with the Hamilto-
nian:
P2
H= + V (x) . (70)
2m

We first note that [P 2 , T ] = 0, since P commutes with any function of P . Thus, we focus
on [T , V (X)]. Consider the action of this commutator on an arbitrary basis state |xi,


[T , V (X)] |xi = T V (X) − V (X)T |xi = V (x) − V (X) T |xi , (71)

where we have used V (X) |xi = V (x) |xi, where X |xi = x |xi. Using the result of problem
5(c) on Problem Set 1 (with P = ~K), it follows that

T |xi = exp(−iℓP/~) |xi = |x + ℓi . (72)

Since V (X) |x + ℓi = V (x + ℓ) |x + ℓi, it then follows from eqs. (71) and (72) that


[T , V (X)] |xi = V (x) − V (x + ℓ) |x + ℓi = 0 , (73)

where we have used the periodicity of V (x) in the final step above. Since eq. (73) is true for
an arbitrary basis state |xi, it follows that [T , V (X)] = 0.

(b) We may choose the energy eigenstates to be simultaneous eigenstates of the transla-
tion operator T . Show that the general form of such eigenfunctions is:

ψ(x) = exp(ipx/~)up (x) .

where up (x+ℓ) = up (x). That is, the eigenfunctions are plane waves modulated by a function
with the periodicity of the potential.
Consider the eigenvalue problem for T ,

T |ψi = t |ψi . (74)

To determine the possible values of t, examine eq. (74) in the p-representation. Since

hp| T |ψi = hp| exp(−iℓP/~) |ψi = e−iℓp/~ hp|ψi ,

we can conclude that

t = e−iℓp/~ . (75)

Next, we consider the eigenvalue equation for T in the x-representation. Using eqs. (74)
and (75),
hx| T |ψi = e−iℓp/~ hx|ψi . (76)

23
In light of eq. (72), T |xi = |x + ℓi. It follows that hx| T = hx − ℓ|. To verify this last
assertion,12 consider
hx| T |x′ i = hx|x′ + ℓi = δ(x − x′ − ℓ) ,
when we act with T to the right. If we act with T to the left we will get the same result if
hx| T = hx − ℓ|, namely,

hx| T |x′ i = hx − ℓ|x′ i = δ(x − x′ − ℓ) ,

Consequently, eq. (76) yields

hx − ℓ|ψi = e−iℓp/~ hx|ψi .

which we can rewrite as

ψ(x − ℓ) = e−iℓp/~ ψ(x) . (77)

Although ψ(x) is not a periodic function, it is straightforward to check that e−ipx/~ ψ(x)
is periodic, since the equation,

e−i(x−ℓ)p/~ ψ(x − ℓ) = e−ipx/~ ψ(x) ,

is equivalent to eq. (77). Thus, we introduce the notation,

up (x) ≡ e−ipx/~ ψ(x) .

Then,
ψ(x) = eipx/~ up (x) ,
where up (x + ℓ) = up (x). This result is known as Bloch’s Theorem. The description of ψ(x)
as a plane wave modulated by a function that possesses the periodicity of the potential V (x)
is called a Bloch wave.

12
Another way to derive hx| T = hx − ℓ| is as follows. The action of the operator
T on
 the ket |xi
† 
is T |xi ≡ |T xi. Likewise, the action of the operator
 † T on the bra hx| is hx| T ≡ T x . Since
 P is
† T x = T † |xi = |x − ℓi and the adjoint of T † x is
hermitian,

†  it follows that T = exp(iℓP/~). Thus

T x = hx| T = hx − ℓ|.

24
 APPENDIX: An alternative solution to Problem 6

Since V (x + ℓ) = V (x), we can expand V (X) in a Fourier series,

X∞
V (X) = cn exp(2πinX/ℓ) .
n=−∞

To compute T V (X), where T = exp(−iℓP/~), we need to calculate the product of two

exponentials. Here we make use of the identities,
eA eB = exp A + B + 12 [A, B] ,


(78)

eB eA = exp A + B − 21 [A, B] ,


   
under the assumption that A, [A, B] = B, [A, B] = 0. Using [X , P ] = i~I,
 
X iℓP 2πinX
T V (X) = cn exp − + − iπn
n
~ ℓ
 
X iℓP 2πin(X − ℓ)
= cn exp − + + iπn
n
~ ℓ

= V (X − ℓ)T = V (X)T .
Hence, we have demonstrated that [T , V (X)] = 0.
It then follows that [H , T ] = 0. Thus, we can choose |ψi = |E, ti to be a simultaneous
eigenstate of H and T . That is, H |E, ti = E |E, ti and T |E, ti = t |E, ti. Consequently,
hx| HT |E, ti = tE hx|E, ti , hx| T H |E, ti = E hx − ℓ|E, ti ,
after noting [cf. footnote 12] that hx| T = hx − ℓ|. Since HT = T H, it follows that
ψ(x − ℓ) = tψ(x) , (79)
where ψ(x) ≡ hx|E, ti. Finally, we compute t by evaluating hp| T |E, ti in two different ways,
hp| T |E, ti = t hp|E, ti ,
and
hp| T |E, ti = hp| exp(−iℓP/~) |E, ti = exp(−iℓp/~) hp|E, ti .
Subtracting the last two equations yields t = e−iℓp/~ . Plugging this back into eq. (79) yields
ψ(x − ℓ) = e−iℓp/~ ψ(x) , (80)
which reproduces eq. (77). The final step is the same as in our previous solution. Namely,
we introduce,
up (x) ≡ e−ipx/~ ψ(x) .
and show that eq. (80) implies that up (x − ℓ) = up (x). Hence, we conclude as before that
ψ(x) = eipx/~ up (x) ,
where up (x + ℓ) = up (x).

25