If $E_i$ is a $\mathbb R$-vector space, then the vector-valued trace $\operatorname{tr}_{E_1}:(E_2\otimes E_1^\ast)\otimes(E_1\otimes E_0)\to E_1\otimes E_0$ (or tensor contraction) is the linearization of the (surjective) bilinear operator \begin{equation}\begin{split}(E_2\otimes E_1^\ast)&\times(E_1\otimes E_0)\:&\to E_1\otimes E_0,\\(x_2\otimes\varphi_1)&\times(x_1\otimes x_0)&\mapsto \langle\varphi_1,x_1\rangle_{E_1}x_1\otimes x_0\end{split}\tag1\end{equation} Assume for simplicity that $E_i$ is finite-dimensional$^1$. If $E_0$ is replaced by $E_0^\ast$, then it's easy to see that $$\operatorname{tr}_{E_1}(A_2A_1)=A_2A_1\;\;\;\text{for all }A_i\in\mathcal L(E_{i-1},E_i)\tag2.$$

Is there a generalization of the kind \begin{equation}\begin{split}\bigotimes_{i=1}^kE_i\otimes E_{i-1}^\ast&\cong\mathcal L\left(\bigotimes_{i=1}^kE_{i-1},\bigotimes_{i=1}^kE_i\right)\\&\cong\bigotimes_{i=1}^k\mathcal L(E_{i-1},E_i)\ni\bigotimes_{i=1}^kA_i\mapsto A_k\cdots A_1\end{split}\tag3\end{equation} of this trace notion?

$^1$ This is clearly not necessary. We just need to replace the space $\mathcal L(E_{i-1},E_i)$ by its subset of finite rank operators.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.