Global phase uc

[adjs]

Summary

QuantumCircuit.uc does not preserve global phase. This PR adds the gate $P = e^{-i\pi/4}$ (see fig. 3 of arXiv:quant-ph/0410066).

Details and comments

[Cryoris]

If it doesn't break anything I think we should just always add the correct global phase. It looks more like a bug to me and not a feature we want to preserve 🙂

[adjs]

Hi @Cryoris , it breaks test_isometry_inverse in test/python/circuit/test_isometry.py
I didn't check the reason.

[Cryoris]

That's odd, nothing should work based off the global phase not being conserved. I think we should try get to the bottom of this and then hopefully remove the argument.

[adjs]

I converted the isometry into a gate (instead of Instruction). With this modification the isometry.inverse works as expected.

The change from instruction to gate will not break anyone's code because a gate is also an instruction and isometry is a unitary operation. Let me know if this is the best strategy.

[ewinston]

could this be something like inv_gate for clarity?

[ewinston]

I converted the isometry into a gate (instead of Instruction). With this modification the isometry.inverse works as expected.

The change from instruction to gate will not break anyone's code because a gate is also an instruction and isometry is a unitary operation. Let me know if this is the best strategy.

Since generally speaking an isometry is not always unitary could you keep this as an instruction?

[adjs]

Thanks for the comment @ewinston. I reverted isometry to an instruction.

I thought that the isometry should be a gate, because its implementation contains only unitary parts.

[ewinston]

Thanks for the update @adjs.

[adjs]

I made a commit that messed up this PR. Then I'll open another one in place of this one.

[adjs]

I'm closing this PR in favour of PR #8231