1
$\begingroup$

So far, the control bits of most control gates have an AND relationship. How to construct a quantum gate whose control bits are XOR with the fewest auxiliary qubits? And then generalize it, is it possible to realize multi-control quantum gates of arbitrary control logic with different relationships?

$\endgroup$

1 Answer 1

1
$\begingroup$

I think what you are asking for is a block unitary matrix. To be concrete I'll discuss a three-qubit case, the generalization is straightforward. Consider $$U=|00\rangle\langle00|\otimes U_{00}+|01\rangle\langle01|\otimes U_{01}+|10\rangle\langle10|\otimes U_{10}+|11\rangle\langle11|\otimes U_{11}$$ or, in matrix representation $$ U=\begin{pmatrix}U_{00}&0&0&0\\0&U_{01}&0&0\\0&0&U_{10}&0\\0&0&0&U_{11}\end{pmatrix} $$ where $U_{00}-U_{11}$ are $2\times2$ matrices.

So, the matrix $U_{00}$ is applied to the last qubit if the first two are in the state $|00\rangle$, $U_{01}$ is applied when the control qubits are in the state $|01\rangle$ etc. Using these block unitaries, you can realize your arbitrary control logic.

The case when $U_{00}=U_{10}=U_{01}=\mathbb{1}$ and $U_{11}=V$ is the standard controlled-$V$ on the last qubit.

$$ U=\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&V\end{pmatrix} $$

To insert $V$ at another place, say at $|01\rangle\langle01|$ you just need to swap $|01\rangle$ and $|11\rangle$. This is easy, since $|01\rangle=X_0 |11\rangle$. Explicitly, start with a standard $$C^2V=|11\rangle\langle11|\otimes V+(1-|11\rangle\langle11|)\otimes 1$$ and conjugate it with $X_0$ to get $$X_0C^2VX_0=|01\rangle\langle01|\otimes V+(1-|01\rangle\langle01|)\otimes 1=\begin{pmatrix}1&0&0&0\\0&V&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}$$

General block unitary can be constructed as a product of these simpler ones with only a single control state. This method uses no extra ancilla qubits. Note however that it may be very inefficient in terms of depth/gate counts.

$\endgroup$
1
  • $\begingroup$ It's really cool theory, and the whole thing looks very straightforward and concise. I don't know if there is any relevant literature for this theory. The direct construction of the U operator you discussed opened my mind, and I will also think about the idea in practice, which may be a useful research. Also Qiskit seems to be able to set the Uniray matrix directly, I will do a verification on it. $\endgroup$
    – R-X Zhao
    Jul 27 at 8:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.