3

This is essentially a follow-up on the very interesting answer given here

Is there a closure property for the entire Clifford hierarchy?

I'm interested in sufficient conditions to conclude that a given gate is (or is not) in the Clifford hierarchy. I would be interested in any sort of "quick checks" I can put a matrix through to check whether or not it is in the Clifford hierarchy.

Along those lines:

The paper

https://arxiv.org/pdf/1608.06596.pdf

shows that a diagonal matrix in the $ k $th level of the Clifford hierarchy must have all entries $ 2^k $ roots of unity.

Which monomial matrices are in the Clifford hierarchy?

1 Answers1

1

$ \mathcal{C}^{(1)} $ is the Pauli group and all elements of the Pauli group are monomial matrices. Thus the monomial matrices in the first level of the Clifford hierarchy are exactly all the Pauli matrices.

Lemma: Suppose that $ U,V $ are unitaries such that $$ U=VC $$ for some Clifford gate $ C $. Then $ U \in \mathcal{C}^{(k)} $ if and only if $ V \in \mathcal{C}^{(k)} $.

Proof. $ U \in \mathcal{C}^{(k)} $ iff $$ UPU^{-1} \in \mathcal{C}^{(k-1)} $$ for all $ P \in \mathcal{C}^{(1)} $ iff
$$ (VC)P(VC)^{-1}=VCPC^{-1}V^{-1} \in \mathcal{C}^{(k-1)} $$ for all $ P \in \mathcal{C}^{(1)} $ iff
$$ VP'V^{-1} \in \mathcal{C}^{(k-1)} $$ for all $ P' \in \mathcal{C}^{(1)} $ (recall conjugation by a Clifford gate is an automorphism of the Pauli group, by definition) which is true iff $ V \in \mathcal{C}^{(k)} $

Recall that every monomial matrix $ M $ can be written as $$ M=DP $$ for $ D $ a diagonal matrix and $ P $ a permutation matrix (to get $ P $ just replace all the nonzero entries of $ M $ by $ 1 $ and to get $ D $ just take all the nonzero entries row by row and list them down the the diagonal).

So any monomial matrix whose corresponding permutation matrix $ P $ is in the Clifford group and whose corresponding diagonal matrix $ D $ is in the Clifford hierarchy will be in the Clifford hierarchy . And the diagonal matrix $ D $ is in the Clifford hierarchy iff it has a certain form in terms of $ 2^k $ roots of unity see

https://arxiv.org/abs/1608.06596

unclear what kind of permutation matrices show up in the Clifford group/ Clifford hierarchy. For example Toffoli shows up in third level but not second level. Recall that Toffoli is $ CCX $. So this is an example of the general fact that $ C^kX $ is in the $ k+1 $ level of the Clifford hierarchy but not in the $ k $ level of the Clifford hierarchy.

  • For one and two qubits all permutations matrices are Clifford, but for three or more qubits there are non-Clifford permutations. An example of such a matrix is the non-Clifford Toffoli gate. – Jonas Anderson Nov 09 '22 at 21:07
  • @JonasAnderson You are so right. I wonder if anyone has ever tried to characterize which permutation matrices are in the Clifford group? Or more generally in the Clifford hierarchy? – Ian Gershon Teixeira Nov 09 '22 at 21:08
  • 1
    Pauli X strings and CNOTs generate the group of Clifford permutations (relative to the computational basis states). I have unpublished work on the subset (they do not form a group) of permutations in the Clifford Hierarchy. I list some rules they must follow but it seemed intractable generally. I exhaustively checked them all for n=3. I can send it to you if you like. – Jonas Anderson Nov 09 '22 at 21:17
  • @JonasAnderson That would be lovely! My email is igt@umd.edu – Ian Gershon Teixeira Nov 09 '22 at 23:03
  • @JonasAnderson So $ 2 $ qubits seems to be special in that all permutation matrices are Clifford. Is it also the case that all diagonal gates with entries a $ 2^k $ root of unity are in the $ k $ level of the $ 2 $ qubit Clifford hierarchy? – Ian Gershon Teixeira May 03 '23 at 19:01
  • @JonasAnderson are you sure that all permutation matrices in the $ 2 $ qubit Clifford group are generated by Pauli X strings and CNOTs? The group generated by $ IX,XI,CNOT_{1,2} ,CNOT_{2,1} $ seems to be only size $ 8 $. Do you mean to include $ SWAP $ also? – Ian Gershon Teixeira May 03 '23 at 19:46
  • 1 and 2 qubits are special in that all permutations are Clifford. This is not true for 3 or more qubits since Toffoli is a non-Clifford permutation gate. You can use $CNOT_{1,2} CNOT_{2,1} CNOT_{1,2} = SWAP$ to get the $SWAP$ gate. – Jonas Anderson May 04 '23 at 03:45
  • For 1 qubit diagonal gates with $2^k$ root of unity entries are in level $k$ of the Clifford Hierarchy. For 2 qubits this holds for some diagonal gates are such as $T\otimes T$, but the control-$T$ gate which only has $2^3$ root of unity entries is in the 4th level of the Clifford Hierarchy. If you compile a diagonal gate using only diagonal Pauli rotations, I believe the $2^k$ root of unity entries in these rotation gates will correspond to the correct level of the Clifford Hierarchy. – Jonas Anderson May 04 '23 at 03:52
  • 1
    oh ya I just made typo yesterday when I wrote down $ CNOT_{2,1} $ you're right I'm getting size $ 24 $ now. I'll have to think more about the 2 qubit diagonal gates thing, maybe I'll ask a separate question – Ian Gershon Teixeira May 04 '23 at 13:05