Given the Hermitian Pauli group products $$ \Omega_{a,b}=\{\pm 1,\pm i\}_{a,b}\cdot \{I,X,Y,Z\}_{a,b}^{\otimes n} $$ composed of $n$ 2x2 pauli matrices $(I,X,Y,Z)$ in tensor product, such that they span a $2^n$ by $2^n$ matrix space, and $a,b=0,1,\cdots , 2^n-2, 2^n-1$, does there exist a reference for their matrix index representation $$ \Omega_{a,b} = \sum_{k=0}^{2^n-1}\sum_{l=0}^{2^n-1}A_{a,b,k,l}\left|k\right>\left<l\right| $$ to find the complex coefficients $A_{a,b,k,l}$, where I write them using bras and kets, and matrix multiplication $$ \Omega_{a,b}\Omega_{g,h}=f_{a,b,g,h,u,v}\Omega_{u,v} $$ to find the complex structure constants $f_{a,b,g,h,u,v}$?

I have seen references, such as ZX, that use $\Omega_{a,b}\approx Z^aX^b$ notation, but those representations are never scaled by complex values to ensure they are Hermitian, nor have I seen an explicit matrix index representation for them.

  • $\begingroup$ Why do you describe the reference to quantum error correction as 'ZX'? $\endgroup$ Oct 21 '19 at 16:45

If you write $Z^aX^b$, there's an implicit "add a phase $i$ to make it Hermitian if necessary", although I guess there are a couple of different conventions you might use the determine the sign used. So long as you're clear about the convention it doesn't really matter because you've got the extra $\pm 1$ freedom to add in to adjust for it.

As for an explicit matrix index representation, $$ A_{a,b,k,l}=i^{a\cdot b}(-1)^{a\cdot k}\delta_{b\oplus l=k} $$ where $\delta$ is the kronecker delta. Here I've chosen a slightly different convention for introducing the $i$ which makes the representation easier to write down: wherever there's a pair $ZX$, assume it's $iZX=Y$.

If you want to work out the effect of the product of two terms, you could do it long-hand, but I think it's easier to work it our with the Pauli operators first and translate it back: $$ Z^aX^bZ^gX^h=(-1)^{b\cdot g}Z^aZ^gX^bX^h=(-1)^{b\cdot g}Z^{a\oplus g}X^{b\oplus h}. $$

  • $\begingroup$ Thanks for the answer. Is there a reference for $A_{a,b,k,l}$ and $f_{a,b,k,l,u,v}$ or is it simply easy to derive and is common knowledge? Also, computing $f_{a,b,k,l,u,v}$ looks more complicated now, because the hamming weight has to be evaluated over modulo 4 instead of modulo 2 arithmetic, e.i. $(i)^{a\cdot b}$ versus $(-1)^{a\cdot b}$. If you could provide $f_{a,b,k,l,u,v}$ using $(i)^{a\cdot b}$, that would be welcomed. $\endgroup$ Oct 21 '19 at 13:30
  • $\begingroup$ @linuxfreebird it’s easy to derive. I just stared at it for about 30 secs and wrote down the answer. $\endgroup$
    – DaftWullie
    Oct 21 '19 at 16:41

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.