To turn the probability of the projection over the Hilbert space $\mathcal H_A \otimes \mathcal H_B$ into the POVM probabilty over $\mathcal H_A$ we we use this equality: $$tr_Atr_B(ρΠ_i)=tr_A(ρ_Atr_B(ρ_BΠ_i))$$

My syllabus states that $\rho=\rho_A\otimes \rho_B$ but says nothing about the projector $\Pi_i$, as far as I know this is just an operator that can't necessarily be split into a tensor product and works on the full Hilbert space.

How does one obtain this equality?

What does $tr_B(\rho_B \Pi_i)$ even mean? As $\rho_B$ and $\Pi_i$ are matrices of different sizes.

What I have tried:
If I do assume that $\Pi_i=\Pi_A\otimes \Pi_B$ I get: $$ \begin{align} tr_Atr_B(ρΠ_i)&= tr_Atr_B((\rho_A \otimes \rho_B)\cdot(\Pi_A\otimes \Pi_B))\\ &=tr_Atr_B((\rho_A\cdot \Pi_A)\otimes(\rho_B\cdot \Pi_B)\\ &=tr_A((\rho_A\cdot \Pi_A) tr_B(\rho_B\cdot \Pi_B) \end{align} $$ Which is almost correct but now there is this $\Pi_A$ term. I'm not sure if I can move this $\Pi_A$ back into the $tr_B$ somehow. This is also all assuming I can even decompose the projector as a tensor product.


1 Answer 1


You can use the identity $$ \mathrm{tr}_{B}\left( (Y_A \otimes \mathbb{I}_{B}) X_{AB} \right) = Y_A \mathrm{tr}_B(X_{AB}) $$ where $X_{AB}$ can be any operator acting on the joint system $AB$. You can prove this quite quickly by writing the operators in the standard basis $Y_A = \sum_{i,j} y_{ij} |i \rangle \langle j |$ and $X_{AB} = \sum_{a,b,c,d} x_{abcd} |a \rangle \langle b| \otimes |c \rangle \langle d|$ and showing the two expressions are equal.

Using that expression you have $$ \begin{aligned} \mathrm{tr}_A (\mathrm{tr}_B ( \rho \Pi)) &= \mathrm{tr}_A(\mathrm{tr}_B((\rho_A \otimes \rho_B) \Pi_{AB})) \\ &= \mathrm{tr}_A(\mathrm{tr}_B((\rho_A \otimes \mathbb{I}_B)(\mathbb{I}_A \otimes \rho_B) \Pi_{AB})) \\ &= \mathrm{tr}_A(\rho_A\mathrm{tr}_B((\mathbb{I}_A\otimes \rho_B)\Pi_{AB})) \end{aligned} $$ which is what you wanted. Note that $\rho_B \Pi_{AB}$ is usually shorthand for $(\mathbb{I}_A \otimes \rho_B) \Pi_{AB}$.


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.