I'm stuck on some identity for an exponentiated tensor product of operators, $e^{\hat A\otimes\hat B}$.

I'm learning weak measurement, reading the review by Kofman et al (2012, DOI: 10.1016/j.physrep.2012.07.001).

They study a discrete system S and a continuous meter M (In their Section 2.2). S and M get entangled in a pre-measurement via a brief interaction Hamiltonian: $$ \hat{H}=\hat{A}\otimes\hat{p} $$

where $\hat{A}$ is a discrete operator on S (with eigenstates $|{a_i}\rangle$), and $\hat{p}=-i\partial_q$ is a continuous operator M (on the variable q). During pre-measurement, states in $S\otimes M$ then evolve per the unitary operator $$ \hat{U}=e^{-i\hat{H}}= e^{-i\hat{A}\otimes\hat{p}} $$

If S starts in $|a_0\rangle, $ (an eigenstate of $\hat{A}$ with eigenvalue $a_0$) and M in $|m^{(0)}\rangle = \psi_0(q)$, the initial product state is $|a_0\rangle\otimes|m^{(0)}\rangle \equiv |a_om^{(0)}\rangle $. By unitary evolution we get (Kofman's equation 2.13): $$ |s_om^{(0)}\rangle \rightarrow e^{-i\hat{A}\otimes\hat{p}} |a_0\rangle\psi_0(q) = |a_0\rangle e^{-ia_0\hat{p}}\psi_0(q) $$

Now, that last equality has me confused.

So I'll spell it out with $\hat{A} = a_0\hat\sigma_z = a_0 \left(\begin{smallmatrix}1&0\\0&-1\end{smallmatrix}\right)$, and $|a_0\rangle = \left(\begin{smallmatrix}1\\0\end{smallmatrix}\right)$, and $\hat\sigma_z\otimes-i\partial_q = \left(\begin{smallmatrix}1&0\\0&-1\end{smallmatrix}\right)(-i\partial_q) $, for the left-hand side:

$$ e^{-a_0\left(\begin{smallmatrix}1&0\\0&-1\end{smallmatrix}\right)\partial_q} (\left(\begin{smallmatrix}1\\0\end{smallmatrix}\right)\psi_0(q)) $$

Now, I believe that the exponent can be viewed as three commuting operators, so that we could write it $$ e^{a_0}e^{-i\left(\begin{smallmatrix}1&0\\0&-1\end{smallmatrix}\right)}e^{-i\partial_q} (\left(\begin{smallmatrix}1\\0\end{smallmatrix}\right)\psi_0(q)) $$

[Edit: Per Oleg's answer, the above belief is wrong.]

Also, I believe that $\hat{A}$ only operates on $|a\rangle,$ and $\hat{p}$ only on $\psi_0(q)$. Then:

$$ \rightarrow e^{a_0} (\left(\begin{smallmatrix}e^{-i}&0\\0&e^{i}\end{smallmatrix}\right)\left(\begin{smallmatrix}1\\0\end{smallmatrix}\right)) (e^{-\partial_q}\psi_0(q)) = e^{a_0}\left(\begin{smallmatrix}e^{-i}\\0 \end{smallmatrix}\right)e^{-\partial_q}\psi_0(q) $$

$$ = e^{a_0+i}|a_0\rangle e^{-\partial_q}\psi_0(q) = |a_0\rangle e^{a_0-i-i\partial_q}\psi_0(q)$$

Which is not at all like the right-hand side in Kofman's equation 2.13(above)! They get a multiplication by $a_0$; I get an addition by $a_0-i$.


1 Answer 1


Here how it comes out: just Taylor expand the exponent, apply it, then sum it back $$e^{A\otimes B}\vert a_0\rangle\vert \psi_B\rangle= \sum_{n=0,1...}\frac{1}{n!} A^n\otimes B^n\vert a_0\rangle\vert \psi_B\rangle =\sum_{n=0,1...}\frac{1}{n!} a_0^nI\otimes B^n \vert a_0\rangle\vert \psi_B\rangle =e^{a_0I\otimes B}\vert a_0\rangle\vert \psi_B\rangle$$

In your calculation you said

Now, I believe that the exponent can be viewed as three commuting operators

followed by calculation that states something like

$$e^{A\otimes B}=e^{A\otimes I}e^{I\otimes B}$$

which is not true even when $A$ and $B$ are scalars, let alone operators.

  • $\begingroup$ #oleg, very nice, I got it now! Do you know of a resource reviewing this kind of operator math (for physicists)? $\endgroup$
    – Travis Lee
    Commented Aug 19, 2019 at 7:06
  • 1
    $\begingroup$ @Travis Lee Unfortunately I don't know of a good source for exactly this kind of tricks. I had to learn it the hard way like you: read papers, get stuck, then slowly figure it out. Here is s a book I liked (doi.org/10.1017/CBO9781139031103) where the QM math is treated with care but it is aimed at quantum information people. $\endgroup$
    – oleg
    Commented Aug 19, 2019 at 15:58

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.