QM1 Problem Set 1 solutions Mike Saelim

If you nd any errors with these solutions, please email me at [email protected].

n
c
n
A
n
.
The rest is a straightforward evaluation, using the property A|a = a|a:
a|f(A)|a

n
c
n
a|A
n
|a

n
c
n
a|a
n
|a

n
c
n
a
n
a|a

= f(a

)
aa
.
(b) Since the operator can be written as a diagonal matrix of its eigenvalues, A =

a
a|aa|,
the determinant is simply the product of all these eigenvalues.
The trace of the logarithm of the operator is
Tr ln A =

a
a| ln A|a =

a
ln a,
so taking the exponential of that gives
exp(Tr ln A) = exp(

a
ln a) =

a
a = det A.
(c) e
iA
is unitary if e
iA
(e
iA
)

= 1, where 1 is the identity operator. To show this, lets decompose

a
f(a)|aa|, which can be reasoned out
from what we proved in part (a). (You can also compute it out directly by replacing A with its
spectral decomposition in the Taylor expansion of f(A).) So,
e
iA
=

a
e
ia
|aa|
(e
iA
)

a
(e
ia
|aa|)

a
e
ia
|aa|.
Putting these together,
e
iA
(e
iA
)

b
e
ia
e
ib
|aa|bb| =

b
e
ia
e
ib
|a
ab
b|
=

a
|aa| = 1.
Alternatively, given that we showed (e
iA
)

= e
iA
, we could just say that e
iA
(e
iA
)

= e
iA
e
iA
=
e
0
= 1, but we would have to specify that we can only combine the two exponentials naively
because their arguments commute. (e
X
e
Y
= e
X+Y
is not true if [X, Y ] = 0.)
1
2 I nd it easiest to start with the right-hand side and work to the left-hand side of this
equation. We can even use our friends the Taylor expansion and spectral decomposition again.
f(U
1
AU) =

n
c
n
(U
1
AU)
n
=

n
c
n
U
1
AUU
1
AUU
1
AU . . . U
1
AU
= U
1

n
c
n
A
n
U = U
1
f(A)U.
3 (a) Calculating e
i n
involves using the Taylor expansion of the exponential, which will
necessitate calculating ( n )
2
. One way to do this is by brute force. If we let n = (n
x
, n
y
, n
z
),
n =
_
n
z
n
x
in
y
n
x
+in
y
n
z
_
( n )
2
=
_
n
2
x
+n
2
y
+n
2
z
0
0 n
2
x
+n
2
y
+n
2
z
_
=
_
1 0
0 1
_
since n is a unit vector. Thus, factors of ( n )
n
will simply alternate between the identity matrix
and n .
The other way to do this is to cheat (sorta) and use the identity you prove in part (b).
( n )
2
= ( n n) 1 +i ( n n) = 1.
Either way,
e
i( n)
= 1 +i( n )
1
2
( n )
2

i
3!
( n )
3

3
+. . .
= 1 +i( n )
1
2
1
2

1
3!
i( n )
3

3
+. . .
= cos 1 +i( n ) sin
where the bold 1 signies the 2x2 identity matrix.
(b) The quickest way to prove this identity is to use two identities that dene the Pauli matrices:
[
i
,
j
] =

k
2i
ijk

k
{
i
,
j
} = 2
ij
1.
Then,
(a )(

b ) =

ij
a
i
b
j

j
=

ij
a
i
b
j
_
1
2
{
i
,
j
} +
1
2
[
i
,
j
]
_
=

ij
a
i
b
j
_

ij
1 +

k
i
ijk

k
_
=a

b 1 +i a

b.
4 (a) Again with the Taylor expansions:
de
xB
dx
=
d
dx
_
1 +xB +
1
2
x
2
B
2
+
1
3!
x
3
B
3
+. . .
_
= B +xB
2
+
1
2
x
2
B
3
+. . . = Be
xB
.
2
Or, this can also be done with the denition of the derivative, extended to functions of operators:
de
xB
dx
= lim
0
1

[e
(x+)B
e
xB
]
= e
xB
lim
0
1

[e
B
1]
= e
xB
lim
0
1

_
B +
1
2

2
B
2
+. . .
_
= Be
xB
but one must be careful to specify that you can split e
(x+)B
into e
xB
e
B
only because xB and B
commute.
(b) This will require us to use the denition of the derivative, as well as a Taylor expansion of
the operators A(x) and B(x):
dAB
dx
= lim
0
1

[A(x +)B(x +) A(x)B(x)]

__
A(x) +
dA(x)
dx
+O(
2
)
__
B(x) +
dB(x)
dx
+O(
2
)
_
A(x)B(x)
_
= lim
0
1

_
A(x)B(x) +
dA(x)
dx
B(x) +A(x)
dB(x)
dx
+O(
2
) A(x)B(x)
_
= A

B +AB

.
(c) This relies on the identity in part (b). Since AA
1
= 1,
d
dx
(AA
1
) =
dA
dx
A
1
+A
dA
1
dx
= 0
=
dA
1
dx
= A
1
dA
dx
A
1
.
5 Photon Filters
The linear polarizer can be represented by a matrix that picks out the component of the photons
state vector that is parallel to the axis rotated by from the x-axis. In the x-y basis, it takes the
x and y components of the input state vector, and spits out the x and y components of the output
state vector. We can do this by rotating the state vector by , picking out its x-component, and
rotating the result back by :
P
xy

=
_
cos sin
sin cos
__
1 0
0 0
__
cos sin
sin cos
_
=
_
cos
2
sin cos
sin cos sin
2

_
.
What does the matrix look like if we have it take in and spit out R and L components instead?
One way is to transform P
xy

with the matrices that transform the components of the state vector.
The prompt gives us the transformation for the basis states:
_
|R
|L
_
= U
_
|x
|y
_
U =
1

2
_
1 i
1 i
_
.
3
However, the components of the state vector will actually transform with U

:
| =
_

x

y
_
_
|x
|y
_
=
_

x

y
_
U

U
_
|x
|y
_
=
_

R

L
_
_
|R
|L
_
=
_

R

L
_
=
_

x

y
_
U

=
_

L
_
= U

y
_
So, to transform P

into a matrix that transforms between R-L components, we use the U

matrix
that transforms x-y components into R-L components, and its inverse U
T
that transforms R-L
components into x-y components:
P
RL

= U

P
xy

U
T
=
1

2
_
1 i
1 i
__
cos
2
sin cos
sin cos sin
2

_
1

2
_
1 1
i i
_
=
1
2
_
1 e
2i
e
2i
1
_
.
Another way, which pretty much does the same thing but avoids the confusion over what matrix
to use, involves playing with bras and kets. Notice that P

= || where | = cos |x +sin |y

2
(|R + |L) + sin
i

2
(|R |L)
=
1

2
(e
i
|R +e
i
|L).
Using this |, P

gives the same result.

Q
xy
U
T
=
1
2
_
1 +i 1 i
1 i 1 +i
_
=
1

2
_
e
i/4
e
i/4
e
i/4
e
i/4
_
.
Now lets consider the combination QP
/4
Q. Plugging and chugging,
(QP
/4
Q)
xy
=
1
2
_
1 i
i 1
_
(QP
/4
Q)
RL
=
_
0 1
0 0
_
So, we can summarize the eects of this system in a table:
Input Output Transmission probability = |Output|
2
|x
1
2
|R
1
2
|y
i
2
|R
1
2
|R 0 0
|L |R 1
This lter blocks right-circularly polarized light, and turns left-circularly polarized light into right-
circularly polarized light.
4
Indeed, this ts with what we would expect QP
/4
Q to do. Weve set the fast and slow axes
of our quarter wave plates to be parallel to the x and y axes, so light linearly polarized parallel
to the x and y axes will be unaected by the rst quarter-wave plate: the plate cannot induce a
shift between two components if only one is present. Then, it is halved in intensity by the 45

linear polarizer, and turned into circularly-polarized light by the second quarter-wave plate. The
rst quarter-wave plate will also shift left-circularly polarized light into linearly polarized light
parallel to the axis of the linear polarizer, while it shifts right-circularly polarized light into linearly
polarized light perpendicular to that axis.
6 Sakurai 1.7
Since we can decompose any state into a linear combination of eigenstates of A, | =

a
c
a
|a,
lets consider these operators eects on eigenstates rst.
(a) For some eigenstate |a,

(Aa

)|a =

(a a

)|a
which equals 0 because of the term a

= a. Since this holds for all the eigenstates, the operator

=a

(Aa

)
(a

)
|a =

=a

(a a

)
(a

)
|a.
We have two possible cases: either a = a

and all the terms will be 1, or a = a

, and one term will

=a

(Aa

)
(a

)
= |a

|,
is the projection operator. Note that this is only possible if the states are nondegenerate, because
if two dierent states have the same eigenvalue, the product will have 0/0 for one of its terms.
(c) The null operator,

(S
z
a

) =
_
S
z
+
h
2
__
S
z

h
2
_
,
will return 0 for either state.
The projection operator for S
z

h
2
,

=h/2
(S
z
a

)
(
h
2
a

)
=
1
2

S
z
h
,
will return |h/2 when acting on |h/2 and 0 when acting on |h/2.
5