3
$\begingroup$

Definitions

  • $\mathcal{C}$ is the $n$-qubit Clifford group.
  • $\mathcal{P}_C$ is the group of $n$-qubit permutation matrices in the $n$-qubit Clifford group. This group is generated by all CNOTs and Pauli $X$ strings. Note that for all $n$, $\mathcal{P}_C$ is a strict subgroup of $\mathcal{C}$.
  • $\mathcal{P}$ is the group of all $n$-qubit unitary permutation matrices. There are $2^n!$ such binary unitary matrices. Note that when $n\ge 3$ this group contains many non-Clifford elements. And so for $n \ge 3$, $\mathcal{P}_C$ is a strict subgroup of $\mathcal{P}$

Equivalence relation 1: $P_1 \equiv_1 P_2$ iff $P_1 = Q_L P_2 Q_R$ for some $Q_L,Q_R \in \mathcal{P}_C$. Here $P_1,P_2 \in \mathcal{P}$.

Equivalence relation 2: $P_1 \equiv_2 P_2$ iff $P_1 = C_L P_2 C_R$ for some $C_L,C_R \in \mathcal{C}$. Here $P_1,P_2 \in \mathcal{P}$.

Question

Are these equivalence relations equivalent?

Since $\mathcal{P}_C \subset \mathcal{C}$, $P_1 \equiv_1 P_2 \implies P_1 \equiv_2 P_2$. So the question boils down to does $P_1 \equiv_2 P_2 \implies P_1 \equiv_1 P_2$?

$\endgroup$
5
  • $\begingroup$ for the second equivalence it might be easier to work with $C_L,C_R$ being in a smaller subgroup of $\cal C$. I think the normalizer of $\cal P$ in the unitary group is the generalized permutation group so it might enough to restrict to this subgroup. A (maby?) quick check would be to count double cosets in each case for small $n=1,2..$ and see if they match. $\endgroup$
    – unknown
    Commented Jun 4 at 16:20
  • $\begingroup$ Does this answer help at all? $\endgroup$
    – xzkxyz
    Commented Jun 4 at 19:08
  • $\begingroup$ @xzkxyz For fixed $n$ all three groups I defined above are finite. This fact can be used to exhaustively check the equivalence relations above for small $n$, but I don't know how to use this fact more generally. Please let me know if I'm missing something. $\endgroup$ Commented Jun 5 at 4:30
  • $\begingroup$ @unknown You are correct that the normalizer of $\mathcal{P}$ in the unitary group is the generalized permutation group. I'm also interested in the case where $C_R = C_L^\dagger$. Please post an answer if you have insight into this case. The equivalence relations above are equal for $n=1, 2$. In those cases all permutations are in the same equivalence classes by equivalence relation 1 (and therefore by relation 2 as well). $\endgroup$ Commented Jun 5 at 4:33
  • 1
    $\begingroup$ to me it seems likely that the two are the same; I don't have a proof...just some possible directions to look : ${\cal P_C}$ (which has order $|GL(n,2)|*2^n$) covers all the permutations in ${\cal C}$...now to add generalized permutations (aka monomial matrices) you'd have to throw in all pauli $Z$ strings and see if these break the equivalence. $\endgroup$
    – unknown
    Commented Jun 6 at 22:19

1 Answer 1

1
$\begingroup$

Here is a partial answer based on the suggestion by @unknown.

I'll introduce an intermediate equivalence relation:

Equivalence relation 1.5: $P_1 \equiv_{1.5} P_2$ iff $P_1 = G_L P_2 G_R$ for some $G_L,G_R \in \mathcal{G}_C$.

Here $P_1,P_2 \in \mathcal{P}$ as before and $G_L,G_R$ are generalized permutations (or monomial matrices) in the Clifford group. This group is generated by $\mathcal{G}_C = \mathcal{P}_C \cup \mathcal{D}_C$. $\mathcal{D}_C$ is the diagonal Clifford group. $\mathcal{G}_C$ is a strict subgroup of the Clifford group since no Hadamard gates are contained in the group.

As before, since Since $\mathcal{P}_C \subset \mathcal{G}_C \subset \mathcal{C}$, it must be the case that $P_1 \equiv_1 P_2 \implies P_1 \equiv_{1.5} P_2 \implies P_1 \equiv_2 P_2$.

Now, we will look at the conditions for $G_L P_1 G_R \in \mathcal{P}$. This is a necessary condition for any $P_1 \equiv_{1.5} P_2$.

First, note that any element of $G\in\mathcal{G}_C$ can be written as a product $AD$ where $A\in \mathcal{P}_C$ and $D$ is a diagonal Clifford gate. This is also true for generalized permutations with the full permutation group and any diagonal matrix group. Then, $$G_L P_1 G_R = A_L D_L P_1 A_R D_R = A_L P_1 (P_1^{-1} D_L P_1) A_R D_R = [A_L P_1 A_R] [A_R^{-1} P_1^{-1} D_L P_1 A_R D_R].$$

Since a diagonal matrix conjugated by a permutation matrix is always a diagonal matrix we see that the terms in the first square bracket are all permutations and therefore the product is a permutation. And since $A_R^{-1} P_1^{-1} D_L P_1 A_R$ is a diagonal matrix, we see that the product of all terms in the second square bracket is a diagonal matrix. For $G_L P_1 G_R$ to be a permutation, the diagonal part must be Identity since this is the only element in the union of diagonal gates and permutations.

So, $A_R^{-1} P_1^{-1} D_L P_1 A_R D_R = I \implies D_R = A_R^{-1} P_1^{-1}D_L^{-1} P_1 A_R$.

Finally, we can substitute this condition into our equivalence relation: $$G_L P_1 G_R = P_2 \implies A_L D_L P_1 A_R D_R = P_2 \implies A_L D_L P_1 A_R (A_R^{-1} P_1^{-1}D_L^{-1} P_1 A_R) = P_2 \implies A_L P_1 A_R = P_2.$$

This proves that $P_1 \equiv_{1.5} P_2 \implies P_1 \equiv_{1} P_2$.

And from our earlier discussion can see that $P_1 \equiv_1 P_2 \iff P_1 \equiv_{1.5} P_2$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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