Lecture 5

Quantum Fourier transform and Phase Estimation

Classical Fourier transform
Classical Fourier transform
Fourier transform converts the signal/data from time domain to frequency domain
N−1
1
xje 2πjki/N
N ∑
yk =
j=0 where yk = y0, … . yN−1 & xk = x0, … . xN−1
Quantum Fourier transform
Phase Superposed state in the
computational basis
1
exp(2πizb/2n) | b⟩
2n ∑
| z⟩ →
b
Any binary or decimal number is converted into a superposed state. ’n’ represents
number of qubits.
For a 3 qubit system, 2 -> 010, 7 -> 111, etc

Example of a 2 qubit system (n=2):

2 2 2
| 00⟩ → [ | 00⟩ + e 2π0i/(2 ) | 01⟩ + e 2π0i2/(2 ) | 10⟩ + e 2π0i3/(2 ) | 11⟩]
2πi/(22) 2πi2/(22) 2πi3/(22)
| 01⟩ → [ | 00⟩ + e | 01⟩ + e | 10⟩ + e | 11⟩]
2πi2/(22) 2πi4/(22) 2πi6/(22)
| 10⟩ → [ | 00⟩ + e | 01⟩ + e | 10⟩ + e | 11⟩]
| 11⟩ → [ | 00⟩ + e 3πi/(2) | 01⟩ + e πi4 | 10⟩ + e πi6 | 11⟩]
i(ϕ = π/2)
1 1 1 1
1 1 i −1 −i 1(ϕ = 0)
QFT = −1(ϕ = π)
2 1 −1 1 −1
1 −i −1 i
−i(ϕ = 3π/2)
Circuit for QFT
| x1⟩ H
OR
| x2⟩ Z H

Let us see if the circuit performs QFT

SWAP I⊗H CTRL − Z

1 0 0 0 1 1 0 0 1 0 0 0 1 1 1 1
0 0 1 0 1 1 −1 0 0 1 0 1 0 0 1 1 i −1 −i
0 1 0 0 (2) 0 0 1 1 2 0 0 1 0 2 1 −1 1 −1
0 0 0 1 0 0 1 −1 0 0 0 i 1 −i −1 i
Geometrical Demonstration of QFT- What happens if the input state is a superposed
state !!

Let us consider a 3 qubit system:

1 0
0 0
0 0
0 0
| 0⟩ = | 000⟩ = ......... . | 7⟩ = | 111⟩ =
0 0
0 0
0 0
0 1
After QFT, amplitudes remain the same as we normalise and hence the measurement yields
any of the states with equal probability.
However, the phase changes in steps as can be depicted easily by the following figures:
1 - qubit system ( step size: 1/2 revolution)

| 0⟩ | 1⟩
 2 - qubit system ( step size: 1/4 revolution)

| 0⟩
| 0⟩ | 0⟩

and so on……

Applying QFT, hence we get…

 and so on…

In the case of a superposed state as an input, QFT is applied on each state of

superposed state independently and added. For example,

QFT8( | 0⟩ + | 4⟩) = QFT8( | 0⟩) + QFT8( | 4⟩)

QFT of a superposed state
PHASE ESTIMATION

Suppose an unitary operation is performed on a qubit |u> such that U | u⟩ → e 2πiϕ | u⟩

Phase estimation procedure helps us in estimating the phase ϕ

Quantum Phase estimate procedure has two registers:

1. First register contains t qubits initially in the state |0>.
2. t depends on the number of digits of accuracy one desires to have for the
phase and the probability of success of measurement we wish to have.

Second register begins with |u> and contains as many qubits as are necessary to store |u>
QFT is applied in two phases. The first phase consists of application of H gate followed
by control-U operation as seen below:
 |0⟩ + e |1⟩ |0⟩ + e |1⟩ . . . |0⟩ + e |1⟩
2t/2
t
2# −1
1
PHASE ESTIMATION: = e2πiϕk |k⟩ . (5.20)
2t/2 k=0

We omit the second register from this description, since it stays in the state |u⟩ throughout
the computation.
Stage I of the circuit
⎧
|0⟩ +
t−1 ϕ)
⎪
⎪
⎪ |0⟩ H ··· • e2πi(2 | 1⟩
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
First register ⎨
|0⟩ + e2πi(2
2 ϕ)
t qubits |0⟩ H • ··· |1⟩
⎪
⎪
|0⟩ + e2πi(2
⎪
⎪ 1 ϕ)
⎪
⎪
⎪
⎪ |0⟩ H • ··· |1⟩
⎪
⎪
|0⟩ + e2πi(2
⎪ 0 ϕ)
⎩ |0⟩ ···
⎪
H • |1⟩
⎪

⎧
⎪
⎨
Second register ⎩ |u⟩ U2
0
U2
1
U2
2
··· U2
t−1
|u⟩

√
Figure 5.2. The first stage of the phase estimation procedure. Normalization factors of 1/ 2 have been omitted, on
the right. Remember !!!

Exercise 5.7: Additional insight into the circuit in Figure 5.2 may be obtained ( by0 e iα)
| ψe⟩ 1 0

showing, as you should now do, that the effect of the = sequence of controlled-U
operations like that in Figure
U 5.2 is to take the state |j⟩|u⟩ to |j⟩U j
|u⟩. (Note
that this does not depend on |u⟩ being an eigenstate of U .)
The second stage of phase estimation is to apply the inverse quantum Fourier transform
PHASE ESTIMATION: Stage II of the circuit

To apply inverse Fourier Transform to the qubits in the first register from stage I.
Measurement is performed to the qubits from the first register to obtain an estimate of
the phase Phase estimation 223

❴✤✤ ❴ ❴ ❴ ❴ ❴ ! ❴ ❴ ✤✤
✤✤ ✙ ✤✤
✤✤ ✙✙ ✙ ✤
✤❴ ❴ ❴ ❴ ❴✙ ✙ ❴ ❴ ❴ ✤✤

Figure 5.3. Schematic of the overall phase estimation procedure. The top t qubits (the ‘/’ denotes a bundle of
How does this work !!
wires, as usual) are the first register, and the bottom qubits are the second register, numbering as many as required
to perform U . |u⟩ is an eigenstate of!U with e2πiϕ . The output of the measurement is an
" eigenvalue
1
#$
approximation to ϕ accurate to t − log 2 + 2ϵ bits, with probability of success at least 1 − ϵ.
Lets go back to QFT once and rewrite it in terms of binary fractions.
ϕ =the
Summarizing, 2πphase
× ωestimation algorithm allows one to estimate the phase ϕ of an
eigenvalue of a unitary operator−k
U , given the corresponding eigenvector |u⟩. An essential
∑ k
= 2πof thisbprocedure
feature at the heart 2 is the ability of the inverse Fourier transform to
k=1
perform the transformation
= 2π × (0.b1b22−1
...) t
1 %
e2πiϕj |j⟩|u⟩ → |ϕ̃⟩|u⟩ , (5.22)
2t/2
In this representation, 1/2=0.1, 1/8=0.001,
j=0
3/8=0.011
where |ϕ̃⟩ denotes a state which is a good estimator for ϕ when measured.

5.2.1 Performance and requirements

The above analysis applies to the ideal case, where ϕ can be written exactly with a t
bit binary expansion. What happens when this is not the case? It turns out that the
 based upon the quantum Fourier transform. As an incidental bonus we obtain the classical
fast Fourier transform, in the exercises!
The equivalence
Rewriting the QFT inofterms
the product representation
of Binary fractions, (5.4) and the definition (5.2) follows
from some elementary algebra:
n
2# −1
1 2πijk/2n
|j⟩ → e |k⟩ (5.5)
2n/2 k=0
1 1
1
$ %n −l
&
2πij kl 2
# #
= ... e l=1 |k1 . . . kn ⟩ (5.6)
2n/2 k1 =0 kn =0
1 1 '
n
1 # #
2πijkl 2−l
= ... e |kl ⟩ (5.7)
2n/2 k1 =0 kn =0 l=1
n 1
( )
1 ' #
2πijkl 2−l
= e |kl ⟩ (5.8)
2n/2 l=1 kl =0
n *
1 −l
+
|0⟩ + e2πij2 |1⟩
'
= (5.9)
2n/2 l=1
! "! " ! "
|0⟩ + e2πi0.jn |1⟩ |0⟩ + e2πi0.jn−1 jn |1⟩ · · · |0⟩ + e2πi0.j1 j2 ···jn |1⟩
= .(5.10)
2n/2
The product representation (5.4) makes it easy to derive an efficient circuit for the
quantum
RememberFourier transform.ofSuch
the example a circuit
2 qubits !!! is shown in Figure 5.1. The gate Rk denotes
the unitary transformation 2 2 2
| 10⟩ → [ |(00⟩ + e 2πi2/(2 ) )
| 01⟩ + e 2πi4/(2 ) | 10⟩ + e 2πi6/(2 ) | 11⟩]
Rewriting the QFT in terms of Binary fractions,

Following the binary fractions notation, let us take an example and show that both
the notations give same value
1
| 10⟩ = ( | 0⟩ + e 2πi0.0 | 1⟩)( | 0⟩ + e 2πi0.10 | 1⟩)
2
1
= ( | 0⟩ + | 1⟩)( | 0⟩ + e iπ | 1⟩)
2 i(ϕ = π/2)
= | 00⟩ + e iπ | 01⟩ + | 10⟩ + e iπ | 11⟩
−1(ϕ = π) 1(ϕ = 0)

−i(ϕ = 3π/2)
 The second stage of phase estimation is to apply the inverse quantum Fourier transform
on the first register. This is obtained by reversing the circuit for the quantum Fourier
transform in the previous section (Exercise 5.5), and can be done in Θ(t2 ) steps. The
Phase estimation third again
and final!!stage of phase estimation is to read out the state of the first register by
doing a measurement in the computational basis. We will show that this provides a pretty
Let the phase togoodbeestimate
estimated
of ϕ. An is overall
represented
schematic as ϕ algorithm
of the = 0.ϕ1 . .is. ϕ
shown
t in Figure 5.3.
To sharpen our intuition as to why phase estimation works, suppose ϕ may be ex-
The output after theexactly
pressed first in
stage
t bits, of
as ϕPhase
= 0.ϕ1 . estimation
. . ϕt . Then thecircuit gives
state (5.20) us from
resulting below after
the first
rewriting in terms
stageof of binary fractions:
phase estimation may be rewritten
1 ! 2πi0.ϕt
"!
2πi0.ϕt−1 ϕt
" !
2πi0.ϕ1 ϕ2 ···ϕt
"
|0⟩ + e |1⟩ |0⟩ + e |1⟩ . . . |0⟩ + e |1⟩ . (5.21)
2 t/2

The second stage of phase estimation is to apply the inverse quantum Fourier transform.
After the second Butstage,
comparing I.e.,the
applying inversewith
previous equation QFT,the we getform
product the for
phase as | ϕtransform,
the Fourier 1ϕ2ϕ3 . . . ϕt⟩
Equation (5.4), we see that the output state from the second stage is the product state
A measurement|ϕwill 1 . . .give
ϕt ⟩. Aus the phase
measurement in phi exactly basis therefore gives us ϕ exactly!
the computational