PHYS-4601 Homework 3 Due 29 Sept 2011

This homework is due in class on the due date. If you wish to turn it in ahead of time, you may
email a PDF or give a hardcopy to Dr. Frey.
Throughout this assignment, ignore time dependence of wavefunctions and suppress time in your
notation. That is, dont bother writing out that a wavefunction or state depends on time.
1. Boundary Conditions and Operators
Consider a particle in 1D conned to the line segment 0 < x < L (note that the Hamiltonian is
not specied). All wavefunctions must satisfy Dirichlet boundary conditions (0) = (L) = 0.
It is easy to see that functions with these boundary conditions and the usual inner product
| =
_
L
0
dx

(x)(x) (1)
form a Hilbert space.
(a) Check that the momentum operator p satises the Hermiticity condition p| = |p.
Solution: In terms of integrals,
p | = ih
_
L
0
dx
d

dx
.
Notice that the factor of i is +i rather than the usual i from the momentum operator
because we complex conjugated in going to the bra state. Now integrate that expression
by parts.
ih
_
L
0
dx
d

dx
= ih
_
L
0
dx

d
dx
+ih[

|
L
0
.
Fortunately, the last term on the right (the surface term) vanishes by the boundary
conditions. And, further, the integral we have remaining is precisely |p . This
shows that p satises the Hermiticity condition.
(b) Find the eigenstates and eigenvalues of the operator p
2
in this Hilbert space. Are any of
the eigenstates of p
2
also eigenstates of the momentum operator p? Explain.
Solution: First o, we are trying to nd solutions to the eigenvalue equation
h
2
d
2

dx
2
(x) = p
2
(x) ,
where p
2
is just a number eigenvalue. We know the solutions to this equation are com-
plex exponentials exp[ipx/h] or linear combinations thereof. To satisfy the boundary
condition at x = 0, we need to choose the linear combination that gives sin[px/h], and
to satisfy the boundary condition at x = L, we need p = n/L, where n is a positive
integer (n = 0 gives a wavefunction equal to zero, which is not an eigenfunction, and n
negative is equivalent to n positive, since the sine is an odd function). Therefore,
p
2
=
n
2

2
L
2
(eigenvalue) , (x) = Asin
_
nx
L
_
(eigenfunction) ,
where A is an normalization constant. We can also check if these are eigenfunctions of p
as follows:
p = ih
d
dx
= ihA
_
n
L
_
cos
_
nx
L
_
.
Thats not proportional to the wavefunction, so no eigenfunction of p
2
is an eigenfunction
of p. This seems a bit strange, since any eigenfunction of p must be an eigenfunction of
p
2
. (Think about that statement if its not obvious.) So whats happening?
(c) Presumably you have an apparent paradox. What is the resolution? Hint: Ask if p acting
on any wavefunction in this Hilbert space always gives another wavefunction in this Hilbert
space.
Solution: The question to ask is if p is actually even an operator on our Hilbert space.
Consider one of the eigenfunctions of p
2
. As we saw, p acting on this eigenfunction gives
us something proportional to cos [nx/L]. If you check, this function does not satisfy the
boundary conditions for our Hilbert space, since cos(0) = 1. An operator on a Hilbert
space takes a vector on the Hilbert space and returns another vector on Hilbert space,
so momentum is not an operator on this Hilbert space.
(d) Now change the boundary conditions to Neumann boundary conditions (d/dx = 0 at
x = 0, L). Does p satisfy the Hermiticity condition? Is p a linear operator on this new
Hilbert space?
Solution: Lets check the Hermiticity condition rst. As in part (a),
p | = |p +ih[

|
L
0
.
However, in this case, the boundary term does not necessarily vanish. You can see
this if you consider the two functions = 1 and = cos(x/L) (you can multiply by
normalization constants if you want). So p does not satisfy the Hermiticity condition.
Furthermore, p acting on the function cos(x/L), which satises the Neumann boundary
conditions, gives something proportional to sin(x/L), which does not. Once again, p is
not even an operator on the Hilbert space.
The lesson of this problem is to be careful with naive assumptions; boundary conditions can
have a nontrivial eect.
2. Probabilities and Densities
One of the postulates is that the probability (density) for measuring eigenvalue of some
observable for a system in state | is |||
2
, where | is the corresponding eigenstate of
the observable. For the rest of this problem, consider a state |. You may work in one
dimension. Hint: You will nd it useful to relate this abstract inner product to the usual one
on wavefunctions.
(a) Show that the probability density for measurements of position is the square of the absolute
value of the wavefunction, as expected.
Solution: Since the eigenstate of position is |x, were told the probability density is
|x||
2
. As we see in the class notes, x| = (x), the wavefunction. (This follows
Page 2
because the wavefunction of |x is (xx

) as a function of x

.) Therefore, the probability

density is |(x)|
2
, like weve always known.
(b) Show that the probability density to measure momentum p is given by the square of the
absolute value of the Fourier transform of the wavefunction (with appropriate normaliza-
tion).
Solution: We know from the class notes that the wavefunction for a momentum state
is
|p
p
(x) =
1

2h
e
ipx/h
.
Therefore,
p| =
_
dx

2h
e
ipx/h
(x) .
This is, up to normalization, the Fourier transform of (x), where p/h is the Fourier
conjugate variable. (Indeed, we see that the momentum-space wavefunction is the
Fourier transform.) Therefore, the probability density |p||
2
is the absolute square of
the Fourier transform of the usual position-space wavefunction.
(c) Finally, nd the probability of measuring energy E if the corresponding eigenstate has
wavefunction
E
(x).
Solution: Well, once again,
E| =
_
dx

E
(x)(x) .
That gives the rather ungainly formula for the probability of measuring energy E to be

_
dx

E
(x)(x)

2
.
However, it seems much simpler if we consider this is just the coecient of the basis state
|E in the expansion of | in the basis of energy eigenstates. Its just how much of the
state is lined up along the energy eigenstate (squared)!
3. Unitary Operators
Weve talked about Hermitian operators quite a bit. Unitary operators are another type of
operator that are quite important in quantum mechanics. By denition, a unitary operator U
satises U

= U
1
.
(a) First, show that if U| = | (ie, | is an eigenstate of U), then | is an eigenstate
of U
1
with eigenvalue 1/ Then use this fact to show that an eigenvalue of a unitary
operator U satises ||
2
= 1.
Solution: To start with, we are of course assuming that U is invertible. Then the
eigenvalue equation is
U| = | U
1
U| = U
1
| .
Page 3
If we recall that U
1
U = 1, we get the equation
U
1
| = (1/)| ,
which is the dening equation for eigenstates and eigenvalues of U
1
. Thats the rst
thing we wanted to show. But now consider that
|U

| = (U|)

| =

| .
(Actually, this really implies that U

| =

|, but we dont even need so much

information.) We also know that
|U

| = |U
1
| = (1/)| .
This tells us that 1/ =

, or ||
2
= 1.
(b) Show that U = exp[iA] is unitary if the operator A is Hermitian. We dene the exponential
of an operator by a power series
exp[iA]

n
1
n!
(iA)
n
= 1 +iA
1
2
A
2
+ (2)
Hint: You may want to show that (AB)

= B

.
Solution: In a notation that shows clearly where the operators are acting (at some
expense in precision), we recall that O

| = |O denes the adjoint operator.

But then
|A B = A

|B = B

|
by moving the operators sequentially, and
|AB = (AB)

|
by denition of the adjoint. Therefore, (AB)

= B

. This clearly implies that (A

n
)

=
(A

)
n
for any operator A, since the power of an operator just means repeated application
of it. Also, the adjoint of a sum of operators is the sum of the adjoints because of the
inner products linearity properties. Therefore,
exp[iA]

n
1
n!
(i
n
)

(A
n
)

n
1
n!
(i)
n
_
A

_
n
.
Since A is given to be Hermitian A

= A, we nd
exp[iA]

n
1
n!
(iA)
n
= exp[iA] .
Since weve dened the exponential of an operator with the same power series as the
exponential of a number, its clear that exp[iA] is the inverse of exp[iA] in that the two
exponentials multiply to give the identity.
(c) Using the expansion above, argue that the operator exp[ipa/h], where p is momentum and
Page 4
a is a constant, translates a wavefunction by a distance a. That is, show that
e
ipa/h
(x) = (x +a) . (3)
So that exponential carries out translations. We will nd that unitary operators often
represent transformations like this. Hint: Think about the wavefunctions Taylor series
around x.
Solution: Lets expand out both sides of equation (3). First, the l.h.s. is (using the
denition of the operator p)
e
ipa/h
(x) = exp
_
a
d
dx
_
(x) =
_

n
1
n!
a
n
d
n
dx
n
_
(x) =

n
1
n!
a
n
d
n

dx
n
(x) .
By the 0th derivative of (x), I mean the function (x) itself. However, this looks a lot
like the r.h.s. of (3) if we write out the Taylor expansion:
(x +a) = (x) +a
d
dx
(x) +
1
2
a
2
d
2

dx
2
(x) + =

n
1
n!
a
n
d
n

dx
n
(x) .
Thats just the same as the l.h.s., so weve shown equation (3). What this does is tell us
that the new state exp[ipa/h]| has a wavefunction at x equal to the wavefunction of
| some distance to the right.
4. REMOVED
5. Homework Comments
The following questions are ungraded, but your answers are greatly appreciated. This will be
the last time I ask this, but you can always feel free to comment.
(a) On a scale of 1 to 10, with 1 being very easy, 10 very dicult, and 5 the average of
homeworks from your physics classes last year, how dicult was this assignment?
(b) On a scale of 1 to 10, with 1 being very short, 10 very long, and 5 the average of homeworks
from your physics classes last year, how long was this assignment?
Solution: You dont really need my solution for this.
Page 5