MASSACHUSETTS INSTITUTE OF TECHNOLOGY

MIT 2.111/8.411/6.898/18.435
Quantum Information Science I
October 28, 2010

Problem Set #7 Solution

(due in class, 04-Nov-10)

P1: (Continuous time Grover with noise) Consider a continuous time Grover algorithm on n qubits
employing two Hamiltonians, an oracle Hamiltonian, HO = EO |xihx|, and a driving Hamiltonian
HD = ED |ψi hψ|, where |xi is the target state and |ψi is state with all the qubits being |0i + |1i. Here,
we explore the impact of a simple noise model on this algorithm.

(a) Ideally, HD has the same strength as the oracle HO . However, we the strength of HD may
fluctuate.
√ Assuming ED = EO (1 + δ), calculate the probability of being in the target state at
time π 2n /(2EO ).
Answer:

H = HO + HD = EO |xihx| + ED |ψihψ|

 ⊗n
1 1 X
|ψi = √ (|0i + |1i) =√ |yi
2 2n y

where y are binary numbers from 0 to 2n − 1.

|ψi can be expanded as
√
1 2n − 1 0
|ψi = √ |xi + √ |x i
2n 2n
|x0 i is orthogonal to |xi.
Therefore, in the basis |xi and |x0 i, H can be written as
" √ n # " √ n #
1 2 −1 1 2 −1
 
1 0 n n 1 + (1 + δ) n (1 + δ) n
H = EO + ED √22n −1 2n2−1 = EO √ n2 n
2
0 0
2n 2n (1 + δ) 22n−1 (1 + δ) 2 2−1
n

Writing in terms of Pauli operators,

√ n
2+δ 1+δ δ 2 −1
H = EO ( I + ( n − )Z + (1 + δ) X) = EO (a0 I + a1 Z + a2 X)
2 2 2 2n
√
After time t = π 2n /(2EO ), the evolution operator is

exp(−iHt) = exp(−ia0 EO tI −ia1 EO tZ −ia2 EO tX) = exp(−ia0 EO t)exp(−it0 (a01 EO Z +a02 EO X))
p p p
where a01 = a1 / a21 + a22 , a02 = a2 / a21 + a22 , t0 = a21 + a22 t.
Hence,

exp(−iHt) = exp(−ia0 EO t)(cos (EO t0 )I − i sin (EO t0 )(a01 Z + a02 X))

Therefore, the state after evolution is

|ψ(t)i = exp(−iHt)|ψi
√ √
a0 + a02 2n − 1 a0 − a01 2n − 1 0
 
= exp(−ia0 EO t) cos (EO t0 )|ψi − i sin (EO t0 )( 1 √ |xi + 2 √ |x i)
2n 2n
 2

The probability of being in the target state is

p = |hx|ψ(t)i|2
√
0 a01 + a02 2n − 1 2
0
= | cos (EO t )hx|ψi − i sin (EO t ) √ |
2n
√
(a0 + a02 2n − 1)2
= cos2 (EO t0 )/2n + sin2 (EO t0 ) 1
2n
1+δ/2
p
To first order in δ, a21 + a22 = √
2n
. Therefore EO t0 = (1 + δ/2)π/2.

√ 2
a01 + a02 2n − 1 2n − 1 2

0 0 2 2
cos (EO t ) = −πδ/4 , sin (EO t ) = 1 − π δ /32 , √ =1− δ
2n 4

1 π2
p(δ) = 1 − δ 2 (2n − 1)( + n ) = 1 − λδ 2
4 2 16
π2
where λ = (2n − 1)( 14 + 2n 16 ).
(b) Suppose we run the algorithm on an imperfect quantum computer in which each qubit experiences
1
a δ that has a random value given by the probability distribution √2πσ exp(−δ 2 /(2σ 2 )).
Calculate how many times one needs to run the algorithm to know what the target state is with
probability 2/3, as a function of σ.
Answer:
The probability of getting the right answer each time is

Z∞
1
p(δ) √ exp(−δ 2 /(2σ 2 )) dx = 1 − λσ 2
2πσ
−∞

Therefore, after running the algorithm N times, the probability of NOT getting the target state
goes down exponentially as (λσ 2 )N . In order to succeed with probability 2/3, N ≥ log(λσ2 ) 31 .

P2: (Compositions of Hamiltonian operations) Consider a physical system with four energy levels
which are addressable, |0i, |1i, |2i, and |3i. You are provided with controls which turn one of two
Hamiltonians Hb , which couples {|0i, |2i}, and Ha , which couples {|2i, |3i} and {|0i, |1i}. Specifically,

0 α 0 0 0 0 β 0
   
α∗ 0 0 0 0 0 0 0
Ha (α) =  Hb (β) =  ∗ . (1)
0 0 0 α β 0 0 0
0 0 α∗ 0 0 0 0 0

Note that α and β are both complex numbers. These may be visualized as transitions between some
subset of energy levels, eg in the hyperfine levels of an atomic system:
|0ñ

|3ñ

Ha Hb

|1ñ

|2ñ

Your mission is to perform qubit rotations in the {|0i, |1i} qubit subspace, while leaving {|2i, |3i} alone.
 3

(a) Suppose H1 and H2 are Hamiltonians such that tr|H1 | ≤  and tr|H2 | ≤ . Prove that
e−iH1 e−iH2 eiH1 eiH2 = e−iHc + O(3 ), where Hc = i[H1 , H2 ] = i(H1 H2 − H2 H1 ).
Answer:
Up to second order in 

e−iH1 = I − iH1 − H12 /2 + O(3 ) , e−iH2 = I − iH2 − H22 /2 + O(3 )

eiH1 = I + iH1 − H12 /2 + O(3 ) , eiH2 = I + iH2 − H22 /2 + O(3 )

Retaining only up to second order in 

e−iH1 e−iH2 eiH1 eiH2 = (I − iH1 − H12 /2)(I − iH2 − H22 /2)(I + iH1 − H12 /2)(I + iH2 − H22 /2)
= I − H1 H2 + H2 H1 + O(3 )
= e−iHc + O(3 )
(b) Show our goal is possible in principle, by constructing a set of Hamiltonians from which H01 =
γ|0ih1| + γ ∗ |1ih0| can be generated. Specifically, compute Hc = i[Hb |β=1 , Hb |β=i ]/2 and Hd =
i[Ha |α=i , Hc ] and explain how to obtain H01 from this, by quantum simulation techniques.
Answer:

1 0 0 0
 
0 0 0 0
Hc = i[Hb |β=1 , Hb |β=i ]/2 = 
0 0 −1 0
0 0 0 0

0 1 0 0
 
1 0 0 0
Hd = i[Ha |α=i , Hc ] = 
0 0 0 −1
0 0 −1 0

Define He = i[Ha , Hc ]

0 −iα 0 0
 
iα∗ 0 0 0
He = i[Ha , Hc ] = 
0 0 0 iα
0 0 −iα∗ 0

H01 can be obtained by

H01 = (Ha |α=γ + He |α=iγ )/2

Note that here we can take the sum of Ha |α=γ and He |α=iγ because they commute. eitH01 =
eitHa |α=γ /2 eitHe |α=iγ
h
/2
. i
θ
(c) Let Rx (θ) = exp −i 2 (|0ih1| + |1ih0|) be a rotation about the x̂ axis of the {|0i, |1i} qubit sub-
space (it acts as identity on the {|2i, |3i} subspace). Give a sequence of individual Hamiltonian
evolutions, eg U = eit1 Ha |α=1 eit2 Hb |β=i · · · , turning on Ha and Hb sequentially (only one Hamil-
tonian on at a time), with specified values of α, β, and pulse durations, such that U = Rx (θ)
exactly.
Answer:

0 0 1 0 0 0 −i 0
   
0 0 0 0 0 0 0 0
Hb |β=1 = = X02 , Hb |β=−i =  = Y02
1 0 0 0 i 0 0 0
0 0 0 0 0 0 0 0
 4

Composing X02 and Y02 we can get any operation on |0i and |2i. For example
 
eiθ/2 0 0 0
 0 1 0 0
eiY02 π/4 eiX02 θ/2 e−iY02 π/4 = eiZ02 θ/2 = 
 0 0 e−iθ/2 0 = e
 iθ/4
Uz (−θ/2)01 ⊕ e−iθ/4 Uz (θ/2)23
0 0 0 1

   
−iHa |α=−i π/4 1 1 −1 1 1 −1
Ha |α=−i = Y01 ⊕ Y23 , e =√ ⊕√
2 1 1 01 2 1 1 23

eiHa |α=−i π/4 eiZ02 θ/2 e−iHa |α=−i π/4 = eiθ/4 Ux (θ/2)01 ⊕ e−iθ/4 Ux (−θ/2)23 = U1

But

e−iHa |α=1 θ/4 = Ux (θ/2)01 ⊕ Ux (θ/2)23 = U2

Combining U1 and U2 we get

U1 U2 = eiθ/4 Ux (θ)01 ⊕ e−iθ/4 I23

Therefore, the total pulse sequence for applying an X operation on |0i and |1i is

Rx (θ) = eiHa |α=−i π/4 eiHb |β=−i π/4 eiHb |β=1 θ/2 e−iHb |β=−i π/4 e−iHa |α=−i π/4 e−iHa |α=1 θ/4
h i
(d) Do the same for Rz (θ) = exp −i θ2 (|0ih0| − |1ih1|) , such that you now have “pulse sequences”
for performing arbitrary operations on the {|0i, |1i} qubit.
Answer:
Rotation around Z axis can be obtained from rotaion around X axis as
Rz (θ) = eiHa |α=−i π/4 Rx (θ)eiHa |α=−i π/4
= eiθ/4 (eiY01 π/4 Ux (θ)01 e−iY01 π/4 ) ⊕ e−iθ/4 (eiY23 π/4 I23 e−iY23 π/4 )
= eiθ/4 Uz (θ)01 ⊕ e−iθ/4 I23