CS259Q Fall2024 Problem Set 3: Solutions

Problem 2. Optimal Measurement

|ϕ⟩ and |ψ⟩ are arbitrary states prepared in equal superposition. Differentiating between states is done through their
angular difference from our measurement basis (this is what determines our probability). Bob’s optimal general
strategy is as follows:
• Since we know both states, apply a unitary such that we end up in the following 2 dimensional configuration
w.r.t our measurement basis (without loss of generality):
s⊥ |ϕ⟩

|ψ⟩

|s⟩
With the angle between the two states, |ϕ⟩ and |ψ⟩, is θ. Also assume the angle between |ψ⟩ and |s⟩ is α.

• Then proceed to measure the system in the s basis. Our protocol guesses |ψ⟩ if we measure s and |ϕ⟩ if we
measure s⊥ so the probability of succeeding with guessing right is a function of the closeness of the respective
states to the measurement basis.

P (Successful guessing) = P (Successful guessing, |ψ⟩) + P (Successful guessing, |ϕ⟩)

= P (Successful guessing | |ψ⟩)P (|ψ⟩) + P (Successful guessing | |ϕ⟩)P (|ϕ⟩)
= P (measure s | |ψ⟩) ∗ 1/2 + P (measure s⊥ | |ϕ⟩) ∗ 1/2
(Equal probability of both states)
cos2 α + sin2 (α + θ)
=
2
2 + cos 2α − cos (2α + 2θ)
=
4

Note that the last term is maximised at α = π/4 − θ/2 which you can check using a simple first derivative test.
This gives the probability of success = 1+sin
2
θ
when the two vectors are symmetrical around the π/4 angle.
You can also check this probability with a few cases. If the angle between the two states is π2 then they are
fully distinguishable if measured in its own basis as they are orthonormal (sin( π2 ) = 1). If the angle between the
two states is 0 then the probability of guessing which one it is 0 as they are indistinguishable (sin(0) = 0). If
the angle between the two states is π4 then you should be able to guess which one it is more than half the time
(sin ( π4 ) ≈ 0.707 =⇒ P ≈ 0.854).

Problem 3. SWAP Test

1
Carrying out the successive application of operator in the circuit, we get that the global state is:
|Ψ⟩ = |0⟩ |ϕ⟩ |ψ⟩ (initial state)
→ √12 |0⟩ |ϕ⟩ |ψ⟩ + √12 |1⟩ |ϕ⟩ |ψ⟩ (applying H ⊗ I ⊗ I)
(1)
→ √12 |0⟩ |ϕ⟩ |ψ⟩ + √12 |1⟩ |ψ⟩ |ϕ⟩ (applying C-SWAP)
→ 12 |0⟩ (|ϕ⟩ |ψ⟩ + |ψ⟩ |ϕ⟩) + 12 |1⟩ (|ϕ⟩ |ψ⟩ − |ψ⟩ |ϕ⟩) (applying H ⊗ I ⊗ I)
The probability that the first qubit is measured to be zero is:
2
P (Q1 : 0) = ∥(|0⟩⟨0| ⊗ I) |Ψ⟩∥ (2)
1 2
= ∥|ϕ⟩ |ψ⟩ + |ψ⟩ |ϕ⟩∥ (3)
4
1
= (⟨ϕ| ⟨ψ| + ⟨ψ| ⟨ϕ|) (|ϕ⟩ |ψ⟩ + |ψ⟩ |ϕ⟩) (4)
4
1
= (2 ⟨ϕ|ϕ⟩ ⟨ψ|ψ⟩ + 2 ⟨ϕ|ψ⟩ ⟨ψ|ϕ⟩) (5)
4
2
1 + |⟨ϕ|ψ⟩|
= , (6)
2
where we recall that states are normalized (i.e. ⟨ϕ|ϕ⟩ = ⟨ψ|ψ⟩ = 1), and the complex
 conjugate
 property of Hilbert
∗ 2
space inner products ⟨ϕ|ψ⟩ = ⟨ψ|ϕ⟩ . Naturally, P (Q1 : 1) = 1 − P (Q1 : 0) = 1 − |⟨ϕ|ψ⟩| /2. The more similar
the states, the higher the likelihood that the outcome will be 0 — therefore this circuit can serve as a test of state
collinearity.

Problem 4. One Qubit Universality

Rz (θ1 )HRz (θ2 )HRz (θ3 ) =

 θ2 θ2 θ θ

−i 2 i 2
e−i 2 +ei 2
eiθ3 ( e 2 2−e 2 )
iθ2 iθ2
!
1+e iθ3 1−e 2
e ( 2 ) θ 2
= ei 2 
 
2 
iθ2 iθ2
eiθ1 ( 1−e2 ) ei(θ1 +θ2 ) ( 1+e2 )  θ
−i 2
θ
i 2
θ
−i 2
θ
i 2

eiθ1 ( e 2 2−e 2 ) ei(θ1 +θ3 ) ( e 2 2+e 2 )
π
cos θ22 −ei(θ3 + 2 ) sin θ22
 
θ
i 22
=e π
ei(θ1 − 2 ) sin θ22 ei(θ1 +θ3 ) cos θ22
On the other hand, we will argue that any unitary U can be written in the form below with a global phase factor
cos θ2 −eiϕ2 sin θ2
 
iϕ
U =e
eiϕ1 sin θ2 ei(ϕ1 +ϕ2 ) cos θ2
This is because, first observe that the first column of U is a 1 qubit state space vector and hence can be represented
with a Bloch sphere format (by potentially taking a global phase factor from U).
The rows also need to be normalized hence the second entry of the first row should also be similar to the second
entry of the first column (with potentially a different phase), and finally the second entry of the second row (and
column) is set uniquely based on the values of the other entries to keep the rows and columns orthonormal.
Therefore, any arbitrary unitary can be represented in the form above, as it represents an arbitrary basis change
(the first column is an arbitrary qubit state on the Bloch sphere, the second column is an arbitary state orthogonal
to the first). Hence for any arbitrary unitary in the form of above determined by θ, ϕ, ϕ1 , ϕ2 , we can set θ1 = ϕ1 + π2 ,
θ2
π
θ2 = θ, and θ3 = ϕ2 − 2 to acheive U by applying Rz (θ1 )HRz (θ2 )HRz (θ3 ) up to a global phase of ei(ϕ− 2 )
.

Problem 5. the Solovay-Kitaev algorithm

3
(a) Since at each stage ϵi+1 = O(ϵi2 ) to get from ϵ0 to ϵn ≤ ϵ we need n (the number of steps) to be
n = log 32 (logϵ0 (ϵ)) + O(1)

2
Since in each step Li+1 = 5Li , after running n = log 32 (logϵ0 (ϵ)) + O(1) We will have the number of gats:

log 3 (logϵ0 (ϵ))+O(1)

Ln = 5 2 L0

since ϵ0 ≤ ϵ ≤ 1, logϵ0 (ϵ) = O(ln 1ϵ ), therefore we have:

( ) +O(1)
ln O(ln 1 )
ϵ ln 5
1 1
ln 3
) = O(ln3.97 )
ln 5 3
Ln = L0 (e ) 2 = O(ln ln 2
ϵ ϵ
(b) Based on Neuman series :
+∞
X
A−1 = (I + ∆A )−1 = A† = (−∆A )i = I − ∆A + ∆2A + O(δ 3 )
i=0

where (O(δ 3 )) is a matrix of norm O(δ 3 ). Therefore we have:

ABA† B † = (I + ∆A )(I + ∆B )(I − ∆A + ∆2A + O(δ 3 ))(I − ∆B + ∆2B + O(δ 3 )) =

I + ∆A + ∆B − ∆A − ∆B + ∆2A + ∆2B − ∆2A − ∆2B + 2∆A ∆B − ∆A ∆B − ∆B ∆A + O(δ 3 ) =

I + ∆A ∆B − ∆B ∆A + O(δ 3 )
Note that (∆A ∆B ) and (∆B ∆A ) have norm less than ||∆A ||||∆B || ≤ δ 2 - the norm operator is submultiplicative.
Because of submultiplicativity of the norm operator, any term that is the multiplication of more than two ∆A or
∆B has norm O(δ 3 ) so we have not written it in this multiplication and have written O(δ 3 ) which is a matrix with
norm O(δ 3 ).
We can do similar analysis with other terms in the expression above. Using norm subadditivity, we can see that

||ABA† B † − I|| ≤ ||∆A ∆B − ∆B ∆A + O(δ 3 )|| = O(δ 2 )

as required.

(c) From above we have

ABA† B † = I + ∆A ∆B − ∆B ∆A + O(δ 3 )
So, the answer is ∆A ∆B − ∆B ∆A

3
(d) Let à = I + ∆A + εA and B̃ = I + ∆B + εB where εA and εB are matrices with norm O(ϵi ).
Based on Neuman series:
+∞
X
Ã−1 = (I + ∆A + εA )−1 = (−∆A − εA )i = I − ∆A − εA + (∆A + εA )2 + O(ϵi + δ)3
i=0

Therefore we have:

ÃB̃ Æ B̃ † = (I +∆A +εA )(I +∆B +εB )(I −∆A −εA +(∆A +εA )2 +O(ϵi +δ)3 )(I −∆B −εB +(∆B +εB )2 +O(ϵi +δ)3 ) =

I + ∆A + ∆B − ∆A − ∆B + εA + εB − εA − εB + (∆A + εA )2 + (∆B + εB )2 − ∆2A − ∆2B + ∆A ∆B − ∆B ∆A +

∆A εB +εA ∆B −∆A εA −εA ∆A −∆A εB −εA ∆B −∆B εA −εB ∆A −∆B εB −εB ∆B +∆A εB +εA ∆B +O(ϵ2i )+O(ϵi +δ)3 =

I + ∆A ∆B − ∆B ∆A + ∆A εB − εB ∆A + εA ∆B − ∆B εA + O(ϵ2i ) + O(ϵi + δ)3

(e) Open ended discussion and analysis of the original paper’s algorithm using results from the previous parts.