I apologize if this is a dumb question but I have really thought about this a while and I can’t understand it. I have tried to prove this using the power series of the exponential function but I did not get anywhere. I really don't understand why the first term is there. So can someone help me understand why the following is true: \begin{equation} e^{-H}\partial_j e^{H} = \partial_j + \partial_j H\tag{1} \end{equation}

Where $H$ is an arbitrary time independent Hamiltonian that can be taken to be a scalar field for our purposes.

  • 5
    $\begingroup$ That's a very strange identity. Where is it from? $\endgroup$
    – ACuriousMind
    Feb 1 at 18:29
  • $\begingroup$ @TobiasFünke it’s the partial derivative with respect to the jth coordinate. So $\partial_j = \partial/\partial x_j$. $\endgroup$ Feb 1 at 18:51
  • $\begingroup$ @ACuriousMind a professor said this was the case in lecture but did not show it. $\endgroup$ Feb 1 at 18:51
  • $\begingroup$ The formula $e^{A}Be^{-A} = e^{[A, \cdot]} B$ comes to mind (for operators $A$ and $B$ on a Hilbert space. With this you can try to work out, how the commutators have to look for the right hand side to be correct. $\endgroup$ Feb 1 at 18:52
  • $\begingroup$ @SebastianRiese But the time derivative is not an operator on the Hilbert space. That being said, I don't know what $\partial_j$ means, even after the comment of OP, i.e. whether or not $j$ could also mean the time coordinate. $\endgroup$ Feb 1 at 18:54

2 Answers 2


It's just Leibnitz rule. For any $\psi(x)$ we have $$ e^{-H(x)}\partial_x e^{H(x)} \psi(x) = e^{-H(x)} (\partial_x e^{H(x)})\psi(x) + e^{-H(x)} e^{H(x)}\partial_x \psi\\ = e^{-H(x)}e^{H(x)} \partial_x H\psi(x)+ \partial_x \psi\\ = (\partial_x H )\psi(x) + \partial_x \psi\\ =\left[(\partial_x H) +\partial_x\right] \psi\\ \equiv (H'+\partial_x)\psi $$

  • $\begingroup$ I was being very dumb, this is probably what the professor meant. Thank you $\endgroup$ Feb 1 at 18:56
  • $\begingroup$ With the $\partial_x$ in the last line not acting on the $\psi(x)$ to the right? If I read this expression I would expect that it acts on the $H(x)$ and the $\psi(x)$. $\endgroup$ Feb 1 at 18:57
  • 1
    $\begingroup$ @Sebastian Riese Yes. I'll edit to make clear. $\endgroup$
    – mike stone
    Feb 1 at 19:08
  • 2
    $\begingroup$ Imagine not using align $\endgroup$
    – Kyle Kanos
    Feb 1 at 19:25
  • 7
    $\begingroup$ This derivation is only correct under the assumption that $H(x)$ commutes with $\partial_xH(x)$, for instance when $H(x)$ only depends on the coordinate operator. $\endgroup$ Feb 1 at 21:29
  1. OP's identity$^1$ $$\begin{align} e^{-H}\partial e^H ~\equiv~&e^{-[H,\cdot]}\partial\cr ~\equiv~& \partial+[\partial, H] +\frac{1}{2}[[\partial, H],H] +\frac{1}{6}[[[\partial, H],H],H] +\ldots\cr ~\stackrel{?}{=}~&\partial+[\partial, H]\end{align}\tag{1}$$ is not true in general.

  2. Example: If $H=ax\partial$, then the left-hand side $e^{-H}\partial e^H=e^{a}\partial$ is different from the right-hand side $\partial +[\partial,H]=(1+a)\partial$.

  3. OP's identity (1) becomes true if we additionally assume that $H$ commutes with $[\partial,H]$.

$^1$ Note that the $\partial\equiv\frac{\partial}{\partial x}$ operator notation is ambiguous, cf. e.g. this Phys.SE post. In particular, we interpret OP's notation $\partial H$ on the right-hand side to effectively mean $[\partial,H]$, cf. the Leibniz rule. OP's identity (1) is in general wrong with any interpretation.


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.