This preview shows page 1. Sign up to view the full content.
Unformatted text preview: as one of the swap bits, with c = 0 as the other swap bit. In the end, we
measure the third register where c went in, and this will be a ∧ b Now look
what happens: if a = b = 1, then b and c swap, so that the third register reads
b = 1: true! If a is one but b is 0, then b and c swap, but the third register is
b = 0: false. And clearly, if a = 0 then so to will read the third register. 4.2. SIMULATING CLASSICAL CIRCUITS 39 Figure 4.1: Note that the input and output have the same number of qubits
in the reversible quantum circuit. If we construct the corresponding reversible circuit (lets call it RC), we
have a small problem. The CSWAP gates end up converting input scratch
bits to garbage. Why is this a problem? We have our output f (x), don’t
we. This seems like it should be good enough. But in fact it is not. All the
junk that gets made in the inbetween steps can be entangled with the output
qubits. Thus if it gets measured, it will screw alter our function. Furthermore,
there is a principle which states that any unmeasured, thrown away qubits are
just as good as measured. This is called the principle of deferred measurement,
and it means that junk qubits are no good.
So how do we restore the scratch bits to 0 on output? We use the fact that
RC is a reversible circuit. We use the CSWAP gates, for example, to produce
the output f (x). We can then copy the output onto some scratch qubits,
which we will keep as our output. We then use the reverse of our circuit RC
on the input, f (x), and the junk to turn it all back to 0’s and x. But because
we copied f (x) in the middle, we keep it at the end.
The sequence of steps for the overall circuit is
C copy y (C ) − 1 (x, 0k , 0m , 0k , 1) −→ (x, y, garbagex , 0k , 1) −→ (x, y, garbagex , y, 1) −→ (x, 0k , 0m , y, 1) . 40 CHAPTER 4. FOURIER SAMPLING & SIMON’S ALGORITHM ˆ
Overall, this gives us a clean reversible circuit C corresponding to C . 4.3 Fourier Sampling Consider a quantum circuit acting on n qubits, which applies a Hadamard
gate to each qubit. i.e. the circuit applies the unitary transformation H ⊗n ,
or H tensored with itself n times.
Another way to deﬁne this unitary transformation H2n is as the 2n × 2n
matrix in which the (x, y ) entry is 2−n/2 (−1)x·y .
n
Applying the Hadamard transform (or the Fourier transform over Z2 ) to
n states
the state of all zeros gives an equal superposition over all 2
1
H2n 0 · · · 0 = √
2n x∈ { 0 ,1 } n  x . In general, applying the Hadamard transform to the computational basis
state u yields:
1
H 2n  u = √
(−1)u·x x
2n x∈{0,1}n
We
deﬁne the Fourier sampling problem as follows: Input an n qubit state
⊗n φ and measure the resulting state
φ =
x∈{0,1}n αx x. Compute H
ˆ
ˆ2
y αy y to output y with probability αy  .
Fourier sampling is probably the most fundamental primitive we use in
quantum algorithms (where in place of the Hadamard t...
View
Full
Document
 Fall '13

Click to edit the document details