Grover’s algorithm

Raffaele Solcà
Assistant: Dr. Alejandro Daleo
17.03.2008
 Contents
1. Introduction
• Grover’s algorithm
• Elements of a quantum computer
2. The algorithm
• Abstraction of the problem
• Design of the algorithm
• Example with n=2
• Analysis of the steps of the algorithm
• Implementation with local transformations

3. Geometrical interpretation
• Geometrical understanding of the algorithm
• Number of steps necessary to find the solution
• Probability of failure
4. The case with t solutions
• Number of steps and probability of failure
2
1. Introduction

What is Grover’s algorithm?

• An Algorithm based on quantum computation.
• It provides a solution of the search problem.

Simple example of the search problem

• We have a phonebook and we are looking for a
person who has a specific phone number.
• With a classical algorithm we need O( N ) steps to find
the person.

• With Grover’s algorithm we need only O( N ) steps.

3
1. Introduction

A quantum computer operates on qubits…

• State of a qubit: ϕ =α 0 +β 1

… using quantum gates

• Unitary transformation
• They are reversible
• Can act on single or many qubits

Measurement of qubits
• The wavefunction collapses in one of the states.
ϕ = ∑ α i i → j with probability α j
2

4
1. Introduction

Some gates we will need:

Hadamard Gate
• Generate superposition
• Defined as: 1 1 1
H=  
2 1 − 1

Walsh-Hadamard Gate
• Generalization of H for many qubits
• Equivalent to apply H on each qubit

γ 1 ⊗L ⊗ γ n H γ 1 ⊗L ⊗ H γ n
γ i ∈ {0,1}

• Matrix : Wij =
1
(− 1)i ⋅ j 5
N
1. Introduction

Oracle
• Is a quantum mechanical operator.
• It is defined as

• If the second register is initialized with

1
(0 − 1 )
2
it simply swaps the sign of the amplitude of the state
with f ( x ) = 1
6
2. The algorithm

The abstracted problem

• Define a function f : A → {0,1}
A is a set with N = 2 n elements
There is only an element s ∈ A with f (s ) = 1

• We are looking for the element x for which f (x ) = 1

7
2. The algorithm

Design of the algorithm

• The steps of the algorithm:

i. Initialization of first register in the
superposition state:
 1 1 1 1 
 , , ,L, 
 N N N N

( )
ii. Repeat the following operation O N times:
a. Apply the oracle
1
b. Apply the diffusion matrix Dij = −δ ij + 2
N
iii. Measure the resulting state of first register.
8
2. The algorithm

Example with N=4 elements (n=2 qubits),

we assume that the solution is the state
2 = 10

• We will see that we only need to apply step ii. once.

• We need to know the number of iterations needed

because we will see that the probability does not
increase monotonically with them.

9
 2. The algorithm

Apply the steps of Grover’s algorithm on

the example.

i. Initial state 1 1 1 1
 , , , 
2 2 2 2

1 1 1 1
ii. a. Application of the oracle  , ,− , 
2 2 2 2
ii. b. Application of D
1
Dij = −δ ij + 2 1 1 1 1 1 1 1 1
N −  , ,− ,  + 2 , , ,  = (0,0,1,0 )
2 2 2 2 4 4 4 4

iii.Measurement 2 = 10
with probability p =1
10
2. The algorithm

Analysis of the steps.

i. Initialization of first register in the superposition state:
 1 1 1 1 
 , , ,L, 
 N N N N

To reach this superposition we can initialize the first

register in the state (1,0, K ,0 ) and apply the Walsh-
Hadamard gate.

• For instance we apply the Walsh-Hadamard gate to the

n=2 initialized state ψ 0 = 0 0

ψ =W ψ0 =
1
( 0 + 1 )⊗ ( 0 + 1 ) = ∑ i
1 1 3
2 2 2 i =0
11
2. The algorithm

Analysis of the steps.

( )
ii. Repeat the following operation O N times:
a. Apply the oracle
1
b. Apply the diffusion matrix Dij = −δ ij + 2
N
a. The application of the oracle changes the sign of the
amplitude of the component for which f ( x ) = 1

b. We prove that D is unitary

• An example of an application of D

12
 2. The algorithm

Implementation with local transformations

• A local transformation acts on single states.
• We implement the algorithm with local transformations
because these transformations can be implemented with
elementary quantum gates.

• We prove that D can be represented as a product of

local transformations:
1
Dij = −δ ij + 2 D = W (R1 − I )W
N where
R1, 00 = 2 and R1,ij = 0 for j or i ≠ 0

• W , (R1 − I ) are local transformations.

13

We have seen

• The search problem

• A classical solution
• A quantum mechanical solution: Grover’s algorithm
• How it is defined
• How it works: Example with n=2

We will see

• Geometrical interpretation
• An application of step ii.
• The number of iterations we need to find the
right answer
• The case with t solution

14
3. Geometrical interpretation

Why can we do a simple geometrical representation?

• All amplitudes are real
• The amplitudes of the non-searched states are equal
⇒ We can represent all the states of the algorithm as a
linear combination of :
• the state of the searched element s
• the linear combination of the other states

Why do we make a geometrical representation?

• Easy understanding of the algorithm
• Easy determination of the number of steps we have
to apply 15
3. Geometrical interpretation

Observation
• Define θ as sin (θ ) =
1
N

• Observe that the application of the oracle is a

reflection around the horizontal axis

16
3. Geometrical interpretation

Application of step ii. on a generic state

as in figure

17
3. Geometrical interpretation

The angle between the states before and

after the application of step ii.:

18
3. Geometrical interpretation

Determination of the steps needed to

approximate the solution

19
3. Geometrical interpretation

Determination of the number of

applications of step ii.

• After m applications of step ii. we have the state

• The probability of finding the searched state is

• If we want to have a probability of about 1 we have

to choose
(2m + 1)θ ≈ π ⇔ m ≈ π − 1
2 4θ 2
Clearly we need an integer number of iterations m
π 
⇒ m≈  20
 4θ 
3. Geometrical interpretation
 1 

For small θ (big N) θ ≈ sin θ = 
 N

π
• We can approximate m ≈  N 
4 

Probability of failure

• The probability of failure is given by

pu = cos 2 ((2m + 1)θ )
π 
= sin 2  − (2m + 1) θ  < sin 2 θ =
1
2  N

21
3. Geometrical interpretation

Example with n=2

• By definition: sin (θ ) =
1
⇒ θ = 30°
2

⇒ Only one application

22
3. Geometrical interpretation

Probability of success as a function of n

after applications of step ii.

23
4. The case with t solutions

Analyze the problem with t solutions

sin (θ ) =
t
• We define:
N

• After m applications of step ii. we have the state

ψ m = sin ((2m + 1)θ ) s + cos((2m + 1)θ ) u

π 
• As for one solution we need m =   steps
 4θ 
with a probability of failure
t
pu < sin 2 θ =
N 24
 Summary

The search problem

• The classical solution
• A quantum mechanical solution: Grover’s algorithm

Grover’s algorithm
• How it is defined
• How it works: Example with n=2

Geometrical interpretation
• The effect of the application of step ii.
• How many step we have to apply
• What is the probability of failure / of finding the correct answer

The case with t solutions

• Drawback: in order to have high probability of success we have to
know the number t of solutions and t << N.

25