1

Ph 219A/CS 219A

Exercises
Due: Friday 15 October 2021

1.1 How far apart are two quantum states?

Consider two quantum states described by density operators ρ and
ρ̃ in an N -dimensional Hilbert space, and consider the complete or-
thogonal measurement {Ea , a = 1, 2, 3, . . . N }, where the Ea ’s are one-
dimensional projectors satisfying
N
X
Ea = I . (1)
a=1

When the measurement is performed, outcome a occurs with proba-

bility pa = tr ρEa if the state is ρ and with probability p̃a = tr ρ̃Ea if
the state is ρ̃.
The L1 distance between the two probability distributions is defined
as
N
1X
d(p, p̃) ≡ kp − p̃k1 ≡ |pa − p̃a | ; (2)
2
a=1

this distance is zero if the two distributions are identical, and attains
its maximum value one if the two distributions have support on disjoint
sets.

a) Show that
N
1X
d(p, p̃) ≤ |λi | (3)
2
i=1

where the λi ’s are the eigenvalues of the Hermitian operator ρ− ρ̃.

Hint: Working in the basis in which ρ − ρ̃ is diagonal, find an
expression for |pa − p̃a |, and then find an upper bound on |pa − p̃a |.
Finally, use the completeness property eq. (1) to bound d(p, p̃).
b) Find a choice for the orthogonal projector {Ea } that saturates the
upper bound eq. (3).
 2

Define a distance d(ρ, ρ̃) between density operators as the maximal L1

distance between the corresponding probability distributions that can
be achieved by any orthogonal measurement. From the results of (a)
and (b), we have found that
N
1X
d(ρ, ρ̃) = |λi | . (4)
2
i=1

c) The L1 norm kAk1 of an operator A is defined as

h i
kAk1 ≡ tr (AA† )1/2 . (5)

How can the distance d(ρ, ρ̃) be expressed as the L1 norm of an

operator?

Now suppose that the states ρ and ρ̃ are pure states ρ = |ψihψ| and
ρ̃ = |ψ̃ihψ̃|. If we adopt a suitable basis in the space spanned by the
two vectors, and appropriate phase conventions, then these vectors can
be expressed as
   
cos θ/2 sin θ/2
|ψi = , |ψ̃i = , (6)
sin θ/2 cos θ/2

where hψ|ψ̃i = sin θ.

d) Express the distance d(ρ, ρ̃) in terms of the angle θ.

e) Express k |ψi − |ψ̃ik2 (where k · k denotes the Hilbert space norm)
in terms of θ, and by comparing with the result of (d), derive the
bound
d(|ψihψ|, |ψ̃ihψ̃|) ≤ k |ψi − |ψ̃ik . (7)

f ) Bob thinks that the norm k |ψi − |ψ̃ik should be a good measure of
the distinguishability of the pure quantum states ρ and ρ̃. Explain
why Bob is wrong. Hint: Remember that quantum states are
rays.

1.2 Which state did Alice make?

Consider a game in which Alice prepares one of two possible states:
either ρ1 with a priori probability p1 , or ρ2 with a priori probability
 3

p2 = 1 − p1 . Bob is to perform a measurement and on the basis of the

outcome to guess which state Alice prepared. If Bob’s guess is right,
he wins; if he guesses wrong, Alice wins.
In this exercise you will find Bob’s best strategy, and determine his
optimal probability of error.
Let’s suppose (for now) that Bob performs a POVM with two possible
outcomes, corresponding to the two nonnegative Hermitian operators
E1 and E2 = I − E1 . If Bob’s outcome is E1 , he guesses that Alice’s
state was ρ1 , and if it is E2 , he guesses ρ2 . Then the probability that
Bob guesses wrong is

perror = p1 tr (ρ1 E2 ) + p2 tr (ρ2 E1 ) . (8)

a) Show that X
perror = p1 + λi hi|E1 |ii , (9)
i
where {|ii} denotes the orthonormal basis of eigenstates of the
Hermitian operator p2 ρ2 −p1 ρ1 , and the λi ’s are the corresponding
eigenvalues.
b) Bob’s best strategy is to perform the two-outcome POVM that
minimizes this error probability. Find the nonnegative operator
E1 that minimizes perror , and show that error probability when
Bob performs this optimal two-outcome POVM is
X
(perror )optimal = p1 + λi . (10)
neg
P
where neg denotes the sum over all of the negative eigenvalues
of p2 ρ2 − p1 ρ1 .
c) It is convenient to express this optimal error probability in terms
of the L1 norm of the operator p2 ρ2 − p1 ρ1 ,
X X
kp2 ρ2 − p1 ρ1 k1 = tr |p2 ρ2 − p1 ρ1 | = λi − λi , (11)
pos neg

the difference between the sum of positive eigenvalues and the

sum of negative eigenvalues. Use the property tr (p2 ρ2 − p1 ρ1 ) =
p2 − p1 to show that
1 1
(perror )optimal = − kp2 ρ2 − p1 ρ1 k1 . (12)
2 2
 4

Check whether the answer makes sense in the case where ρ1 = ρ2

and in the case where ρ1 and ρ2 have support on orthogonal
subspaces.
d) Now suppose that Alice decides at random (with p1 = p2 = 1/2)
to prepare one of two pure states |ψ1 i, |ψ2 i of a single qubit, with

|hψ1 |ψ2 i| = sin(2α) , 0 ≤ α ≤ π/4 . (13)

With a suitable choice of basis, the two states can be expressed

as    
cos α sin α
|ψ1 i = , |ψ2 i = . (14)
sin α cos α
Find Bob’s optimal two-outcome measurement, and compute the
optimal error probability.
e) Bob wonders whether he can find a better strategy if his POVM
{Ei } has more than two possible outcomes. Let p(a|i) denote the
probability that state a was prepared, given that the measure-
ment outcome was i; it can be computed using the relations

pi p(1|i) = p1 p(i|1) = p1 tr ρ1 Ei ,
pi p(2|i) = p2 p(i|2) = p2 tr ρ2 Ei ; (15)

here p(i|a) denotes the probability that Bob finds measurement

outcome i if Alice prepared the state ρa , and pi denotes the prob-
ability that Bob finds measurement outcome i, averaged over
Alice’s choice of state. For each outcome i, Bob will make his
decision according to which of the two quantities

p(1|i) , p(2|i) (16)

is the larger; the probability that he makes a mistake is the

smaller of these two quantities. This probability of error, given
that Bob obtains outcome i, can be written as
1 1
perror (i) = min (p(1|i), p(2|i)) = − |p(2|i) − p(1|i)| . (17)
2 2
Show that the probability of error, averaged over the measure-
ment outcomes, is
X 1 1X
perror = pi perror (i) = − |tr (p2 ρ2 − p1 ρ1 ) Ei | . (18)
2 2
i i
 5

f ) By expanding in terms of the basis of eigenstates of p2 ρ2 − p1 ρ1 ,

show that
1 1
perror ≥ − kp2 ρ2 − p1 ρ1 k1 . (19)
2 2
P
(Hint: Use the completeness property i Ei = I.) Since we
have already shown that this bound can be saturated with a two-
outcome POVM, the POVM with many outcomes is no better.

1.3 Eavesdropping and disturbance

Alice wants to send a message to Bob. Alice is equipped to prepare
either one of the two states |ui or |vi. These two states, in a suitable
basis, can be expressed as
   
cos α sin α
|ui = , |vi = , (20)
sin α cos α

where 0 < α < π/4. Suppose that Alice decides uniformly at random
to send either |ui or |vi to Bob, and Bob is to make a measurement
to determine what she sent. Since the two states are not orthogonal,
Bob cannot distinguish the states perfectly.

a) Bob realizes that he can’t expect to be able to identify Alice’s

qubit every time, so he settles for a procedure that is successful
only some of the time. He performs a POVM with three possible
outcomes: ¬u, ¬v, or DON’T KNOW. If he obtains the result
¬u, he is certain that |vi was sent, and if he obtains ¬v, he is
certain that |ui was sent. If the result is DON’T KNOW, then
his measurement is inconclusive. This POVM is defined by the
operators

E¬u = A(I − |uihu|) , E¬v = A(I − |vihv|) ,

EDK = (1 − 2A)I + A (|uihu| + |vihv|) , (21)

where A is a positive real number. How should Bob choose A

to minimize the probability of the outcome DK, and what is
this minimal DK probability (assuming that Alice chooses from
{|ui, |vi} equiprobably)? Hint: If A is too large, EDK will have
negative eigenvalues, and Eq.(21) will not be a POVM.
b) Eve also wants to know what Alice is sending to Bob. Hoping that
Alice and Bob won’t notice, she intercepts each qubit that Alice
 6

sends, by performing
 1
an
orthogonal measurement that projects
0 1


onto the basis 0 , 1 . If she obtains the outcome 0 , she
sends the state |ui on to Bob, and if she obtains the outcome 01 ,

she sends |vi on to Bob. Therefore each time Bob’s POVM has a
conclusive outcome, Eve knows with certainty what that outcome
is. But Eve’s tampering causes detectable errors; sometimes Bob
obtains a “conclusive” outcome that actually differs from what
Alice sent. What is the probability of such an error, when Bob’s
outcome is conclusive?

1.4 What probability distributions are consistent with a mixed

state?
A density operator ρ, expressed in the orthonormal basis {|αi i} that
diagonalizes it, is X
ρ= pi |αi ihαi | . (22)
i
We would like to realize this density operator as an ensemble of pure
states {|ϕµ i}, where |ϕµ i is prepared with a specified probability qµ .
This preparation is possible if the |ϕµ i’s can be chosen so that
X
ρ= qµ |ϕµ ihϕµ | . (23)
µ

We say that a probability vector q (a vector whose components are

nonnegative real numbers that sum to 1) is majorized by a probability
vector p (denoted q ≺ p), if there exists a doubly stochastic matrix D
such that X
qµ = Dµi pi . (24)
i
A matrix is doubly
P stochastic
P if its entries are nonnegative real num-
bers such that µ Dµi = i Dµi = 1. That the columns sum to one
assures that D maps probability vectors to probability vectors (i.e.,
is stochastic). That the rows sum to one assures that D maps the
uniform distribution to itself. Applied repeatedly, D takes any input
distribution closer and closer to the uniform distribution (unless D is
a permutation, with one nonzero entry in each row and column). Thus
we can view majorization as a partial order on probability vectors such
that q ≺ p means that q is more nearly uniform than p (or equally close
to uniform, in the case where D is a permutation).
 7

Show that normalized pure states {|ϕµ i} exist such that eq. (23) is
satisfied if and only if q ≺ p, where p is the vector of eigenvalues of ρ.
Hint: Recall that, according to the Hughston-Jozsa-Wootters Theo-
rem, if eq. (22) and eq. (23) are both satisfied then there is a unitary
matrix Vµi such that
√ X√
qµ |ϕµ i = pi Vµi |αi i . (25)
i

You may also use (but need not prove) Horn’s Lemma: if q ≺ p,
then there exists a unitary (in fact, orthogonal) matrix Uµi such that
q = Dp and Dµi = |Uµi |2 .