78 4. TOP
an integer number equal to or greater than 0. For both of our two synthesis strategies, the formula
gives a zero overhead, expressing that both methods are ideally efficient.
4.8 FURTHER SYNTHESIS
For the synthesis of classical reversible circuits, after the primal synthesis (Section 2.3) and the
dual synthesis (Section 2.4), Section 2.6 presented a “refined synthesis.” Is such a third synthesis
method also possible for quantum circuits? Unfortunately, the answer is “no.”
Indeed, for classical computing, a controlled 2 2 matrix can only be either a controlled
IDENTITY gate or a controlled NOT gate. As a controlled IDENTITY gate, in fact, it is performing
no action at all; it can be deleted from any circuit. Hence, for classical computing, a controlled
2 2 matrix is a controlled NOT gate. For quantum computing, a controlled 2 2 gate can be
any of the 1
4
matrices of U(2). erefore, a controlled U(2) gate, acting on the first qubit, is
represented by a matrix like (2.3), however, with the 2
w1
blocks, either
1 0
0 1
or
0 1
1 0
replaced by blocks from U(2). For w D 3, this looks like
0
B
B
B
B
B
B
B
B
B
B
B
@
a
1
b
1
a
2
b
2
a
3
b
3
a
4
b
4
c
1
d
1
c
2
d
2
c
3
d
3
c
4
d
4
1
C
C
C
C
C
C
C
C
C
C
C
A
; (4.14)
where each block
a
j
b
j
c
j
d
j
is some U(2) matrix. Hence, each controlled U(2) has 2
w1
times
four degrees of freedom, thus 2
wC1
degrees of freedom. A cascade of 2w 1 controlled U(2)
gates thus has a total of .2w 1/2
wC1
parameters. With n D 2
w
, this means 2nŒ2 log
2
.n/ 1
degrees of freedom, i.e., too few to synthesize an arbitrary member of the n
2
-dimensional group
U(n).
We thus can conclude that a “refined synthesis” method is impossible for quantum com-
puting. Fortunately, such a method is not necessary, as both the “primal synthesis” method and
the “dual synthesis” method are already optimal in the quantum case.
We finaly remark that, both in quantum circuit (4.5) and in quantum circuit (4.10), we
can change the order of the blocks, as mentioned in Section 2.8 for classical circuits. We also
can vary the order of the quantum wires. And, finally, we can reduce the number of controlling
qubits.