Can someone help me with the following question?

Let $M$ be a general operator on the composite system $\mathcal{H}_A\otimes \mathcal{H}_B$ and let $O_A$ be an operator on $\mathcal{H}_A$. Using the definition of a partial trace that was given in class, together with the properties of the full trace, prove that $Tr_B(O_AM) = O_ATr_B(M)$.

I got stuck here:

$$M = \sum\limits_{\alpha}M^{(A)}_{\alpha}\otimes M^{(B)}_{\alpha}; O_AM = (O\otimes\mathbb{1}_B)M = \sum\limits_{\alpha}OM_\alpha^{(A)}\otimes M_\alpha^{(B)}\rightarrow$$ $$Tr_B(O_AM) = Tr_B\left(\sum\limits_{\alpha} OM_\alpha^{(A)}\otimes M_\alpha^{(B)}\right)=\sum\limits_\alpha Tr_B\left(OM_\alpha^{(A)}\otimes M_\alpha^{(B)}\right)$$$$=\sum\limits_\alpha OM_\alpha^{(A)}Tr\left(M_\alpha^{(B)}\right)$$

Editor's note: the $\mathbb{1}_B$ actually looked more like $\mathbb{R}$ except with a 1; however, I couldn't get the font to work properly in mathjax. Apologies.

  • 2
    $\begingroup$ Hi, welcome to QCSE. This clearly sounds like homework, which, while not necessarily disallowed at this site, is generally disfavored. I recommend you revise your question. Start by rewriting it in Latex, and then indicating clearly what you have done, and where you are stuck. $\endgroup$ Commented Apr 24, 2020 at 21:31

2 Answers 2


Firstly, just because an operator $M$ is defined on a composite space, it does not mean that the operator itself has a tensor product structure. You need $$ M=\sum_{i,j,k,l}M_{ij,kl}|ij\rangle\langle kl| $$ Now let $$ O=\sum_{i,k}O_{i,k}|i\rangle\langle k|. $$ With these in place, you just calculate the two sides of the equation and see if they're the same. (In a strict mathematical sense, this is not the best way to present the answer, but it's the best way to start to understand it before working out a sequence so that you can start from the left-hand side and finally arrive at the right-hand side.)

So, I can work out the product \begin{align*} \Phi=(O\otimes I)M&=\left(\sum_{i,k}O_{i,k}|i\rangle\langle k|\otimes\sum_j|j\rangle\langle j|\right)\left(\sum_{j,k,l,m}M_{kj,ml}|kj\rangle\langle ml|\right) \\ &=\left(\sum_{i,j,k}O_{i,k}|ij\rangle\langle kj|\right)\left(\sum_{j,k,l,m}M_{kj,ml}|kj\rangle\langle ml|\right) \\ &=\sum_{i,j,k,l,m}O_{i,k}M_{kj,ml}|ij\rangle\langle ml|. \end{align*} Taking the partial trace of this, I get $$ \text{Tr}_B\Phi=\sum_{i,j,k,m}O_{i,k}M_{kj,mj}|i\rangle\langle m| $$ (basically summing over indices $j=l$).

Now, I need to compare this to $$ O\cdot\text{Tr}_B(M)=\left(\sum_{i,k}O_{i,k}|i\rangle\langle k|\right)\left(\sum_{j,k,m}M_{kj,mj}|k\rangle\langle m|\right)=\sum_{i,j,k,m}O_{i,k}M_{kj,mj}|i\rangle\langle m|, $$ so the two are the same.

  • $\begingroup$ First of all thanks. Second i only have problem with :O⋅TrB(M) because you ignored the fact that it operates only on the A system and then it should be :(O⊗I) and that is where i am stuck $\endgroup$ Commented Apr 25, 2020 at 13:12
  • $\begingroup$ @Vladimirkozlov Made an edit to more explicitly show taking identity into account. $\endgroup$
    – DaftWullie
    Commented Apr 27, 2020 at 8:40

A way that's perhaps more compact (if less terse), we can write

$$ \mathrm{Tr}_B(O_AM)=\mathrm{Tr}_B((O_A\otimes \mathbb 1_B)M)=\sum_i (\mathbb 1_A\otimes \langle i|_B) (O_A\otimes \mathbb 1_B)M(\mathbb 1_A\otimes |i\rangle_B)=\\=\sum_i O_A(\mathbb 1_A\otimes \langle i|_B) M(\mathbb 1_A\otimes |i\rangle_B)=O_A\mathrm{Tr}_B(M)$$

The only nontrivial thing I used is $(\mathbb 1_A\otimes \langle i|_B) (O_A\otimes \mathbb 1_B)=O_A(\mathbb 1_A\otimes \langle i|_B)$, it's pretty clearly true, but to see it explicitely just write

$$(\mathbb 1_A\otimes \langle i|_B) (O_A\otimes \mathbb 1_B)= (\mathbb 1_A\otimes \langle i|_B) O_A\otimes\sum_j |j\rangle\langle j|=O_A\otimes\langle i|_B=O_A(\mathbb 1_A\otimes\langle i|_B) $$

  • $\begingroup$ It's even simpler: $(\mathbb 1_A\otimes \langle i|_B) (O_A\otimes \mathbb 1_B) = O_A\otimes \langle i|_B$ due to mixed-product rule. Also I wouldn't use the same symbol for $O_A$ and $O_A \otimes \mathbb 1_B$, it's confusing $\endgroup$
    – Danylo Y
    Commented Apr 28, 2020 at 10:40

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.