2
$\begingroup$

It is known that for n-qubit matrices X, Y $\in \mathbb{C}^{2^{n}\times 2^{n}}$ (and Pauli matrices $I, X, Y, Z$) such that $$ X = \sum_{M \in \{I, X, Y, Z\}^{n}} x_{M}M_{1}\otimes ... \otimes M_{n} $$ and $$ Y = \sum_{M \in \{I, X, Y, Z\}^{n}} y_{M}M_{1}\otimes ... \otimes M_{n} $$ their trace is $$ Tr(XY) = 2^{n} \sum_{M \in \{I, X, Y, Z\}^{n}}x_{M} y_{M} $$ and we can thus use the trace to calculate the coefficients $x_{M}, y_{M}$, but how do we know that is true? I understand that for all tensor products where $M \neq I$ the trace is zero but where exactly does the $2^{n}$ come from, and why is there still a summation in the trace if the tensor product only yields a non-zero trace for one scenario?

$\endgroup$

2 Answers 2

2
$\begingroup$

Letting $\sigma_j$ for $j \in \{I, X, Y, Z\}$ denote a Pauli matrix, its easy to verify the identity \begin{equation} \text{Tr}(\sigma_i \sigma_j) = 2 \delta_i^j \tag{1} \end{equation}

since $\sigma_i \sigma_j$ give the identity when $i=j$ but gives another non-identity Pauli operator (i.e. traceless) when $i \neq j$. Now its straightforward to show \begin{align} \text{Tr}(XY) &= \text{Tr}\left(\sum_{\mathbf{i}\in\{I,X,Y,Z\}^n} x_\mathbf{i} \sigma_{i_1} \otimes \cdots \otimes \sigma_{i_n} \sum_{\mathbf{j}\in\{I,X,Y,Z\}^n} y_\mathbf{j} \sigma_{j_1} \otimes \cdots \otimes \sigma_{j_n}\right)\tag{2} \\&= \sum_{\mathbf{i},\mathbf{j}\in\{I,X,Y,Z\}^n} x_\mathbf{i} y_\mathbf{j} \text{Tr}\left(\sigma_{i_1}\sigma_{j_1} \otimes \cdots \otimes \sigma_{i_n} \sigma_{j_n}\right)\tag{3} \\&= \sum_{\mathbf{i},\mathbf{j}\in\{I,X,Y,Z\}^n} x_\mathbf{i} y_\mathbf{j} \prod_{k=1}^n\text{Tr}\left(\sigma_{i_k}\sigma_{j_k}\right)\tag{4} \\&= \sum_{\mathbf{i},\mathbf{j}\in\{I,X,Y,Z\}^n} x_\mathbf{i} y_\mathbf{j} \left(2^n \prod_{k=1}^n \delta_{i_k}^{j_k} \right)\tag{5} \\&= 2^n \sum_{\mathbf{i},\mathbf{j}\in\{I,X,Y,Z\}^n} x_\mathbf{i} y_\mathbf{j} \delta_\mathbf{j}^\mathbf{i}\tag{6} \\&= 2^n \sum_{\mathbf{i}\in\{I,X,Y,Z\}^n} x_\mathbf{i} y_\mathbf{i} \tag{7} \end{align}

And so the $2^n$ comes from $\text{Tr}(\sigma_I)=2$ and the fact that every pair of terms between $X$ and $Y$ is either orthogonal or multiplies to $I_{2^n} := I \otimes \cdots \otimes I$ (with trace $2^n$). The sum remains because there are actually $4^n$ different pairs whose product is $I_{2^n}$. For example with three qubits, multiplying $x_{III} I_8$ with $y_{III} I_8$ results in an operator with nonzero trace, but so too does multiplying $x_{IXY} (I\otimes X \otimes Y)$ and $y_{IXY}(I\otimes X \otimes Y$), and so on.

$\endgroup$
0
$\begingroup$

where exactly does the $2^n$ come from

If you have a term in the product where all the Paulis are the same from both the $X$ and the $Y$, then you're left with the $2^n\times 2^n$ identity matrix. This has trace $2^n$.

why is there still a summation in the trace if the tensor product only yields a non-zero trace for one scenario?

When you multiply $X$ and $Y$, there are two summations (better to use different summation indices to make it super-clear!). The fact that the trace identifies equal terms accounts for one of the summations, leaving your answer with one summation.

$\endgroup$

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.