in the context of non-relativistic quantum mechanics I want to show that, for any $A$ and $B$ operators

$$e^{A}e^{B}=e^{A+B} $$

if and only if


I remember my professor told use about looking for a differential equation but I don't remember the details, and I want to be able to prove it. Brute force the series doesn't seem to be a good idea.

Any hint will be appreciated thanks.


5 Answers 5


There is no 'only if' because it is not true: \begin{align} e^{A+B} = e^A e^B \end{align} does not necessarily imply $[A,B] = 0$.

One can easily find an example of this using matrices. Here's one: \begin{align} A= \begin{pmatrix} 0 & 0 \\ 0 & 2\pi i \end{pmatrix}, B=\begin{pmatrix} 0 & 1 \\ 0 & 2 \pi i \end{pmatrix}. \end{align} $[A,B] \neq 0$ but $e^{A+ B} = e^A e^B = I$.

Edit: Let me help with the if part, using a differential equation as OP desires. Compute \begin{align} \frac{d}{dt}(e^{t(A+B)}e^{-tA}e^{-tB}), \end{align} and show that it is $0$ if $[A,B] = 0$.

That implies that $e^{t(A+B)}e^{-tA}e^{-tB}$ is independent of $t$. In particular, plugging in $t = 0$ gives $e^{t(A+B)}e^{-tA}e^{-tB} = I$ for all $t$. Then plug in $t = 1$ to get $e^{(A+B)}e^{-A}e^{-B} = I$.


  • $\begingroup$ I don't think this counterexample is correct: I get exp(A) = I, but I don't get exp(B) = I. $\endgroup$ Mar 21, 2013 at 3:34
  • $\begingroup$ @ColinMcFaul Actually I get the same as nervxxx. I tried wolfram alpha as well: http://www.wolframalpha.com/input/?i=MatrixExp+%28{{0%2C1}%2C{0%2C2+pi+i}}%29 $\endgroup$
    – Michael
    Mar 21, 2013 at 5:13
  • $\begingroup$ @ Colin McFaul then you are doing something wrong. wolframalpha.com/input/… $\endgroup$
    – nervxxx
    Mar 21, 2013 at 5:13
  • 2
    $\begingroup$ +1 because this seems to be true --- but on the other hand I think the OP is implicitly asking about Hermitian operators. $\endgroup$
    – N. Virgo
    Mar 21, 2013 at 5:35
  • 1
    $\begingroup$ The Taylor expansion of the exponential function shows $1+ 2\pi i\sum_{k \geq 1} (2 \pi i)^{k-1}/k! = 1$. $\endgroup$ Mar 24, 2013 at 17:24

Define the function $f(u)=e^{uA}e^{uB}$. If $A$ and $B$ commute, you can take the derivatives of the $f(u)$ as follows:

$$\frac{df(u)}{du}=\frac{de^{uA}}{du}e^{uB}+e^{uA}\frac{de^{uB}}{du}=Ae^{uA} e^{uB}+e^{uA}Be^{uB}$$

$$\frac{d^2f(u)}{du^2}=A^2e^{uA} e^{uB}+2Ae^{uA}Be^{uB}+e^{uA}B^2e^{uB}$$


Now take the Maclaurin series expansion:

$$f(u)=1+(A+B)u+\frac{1}{2!}(A^2+2AB+B^2)u^2+\ldots=\\ =1+(A+B)u+\frac{1}{2!}(A+B)^2 u^2+\frac{1}{3!}(A+B)^3 u^3+\ldots=e^{(A+B)u}$$

For $u=1$ you get the desired result:


Also check this:

Baker–Campbell–Hausdorff formula

Matrix exponential

  • $\begingroup$ This shows the if but not the only if... $\endgroup$
    – Fabian
    Mar 20, 2013 at 20:21

I've never heard of trying to find a differential equation to prove this; I've only done by brute forcing the series expansion. That isn't as bad as it sounds. On the one side of the equation you get a single infinite series with terms of $(A + B)^n$. The other side of the equation gives you two infinite series multiplied together, with terms of $A^m$ and $B^l$. Don't make any assumptions about $[A,B]$ yet, and expand everything out until you have all the terms that contain $AB$ and $BA$. You should find that the constant terms and the terms of order $A$ and $B$ all cancel out, which will leave you with your condition for the commutator.


Remember that you are working with operators.


$$e^{A}e^B =(1+\frac{A}{2}+\frac{AA}{3!}+...)(1+\frac{B}{2}+\frac{BB}{3!}+...)$$


$$e^{A+B} =(1+\frac{A+B}{2}+\frac{(A+B)^2}{3!}+...)$$.

Try matching the terms up using commutators.


The proof (due to Glauber, given in Messiah) is as follows: consider f(x)= exp(xA)+exp(xB)

differentiate it ---> df/dx= Aexp(xA)exp(xB)+exp(xA)exp(xB)B

                  df/dx=f(x)[exp(-xB)Aexp(xB)+B] here we have to take care of the order of operators
                  df/dx=f(x)[exp(1-xB)A(1+xB) + B]  expand exponential


It is easy to check that the solution to this first-order differential equation is one at x = 0 which is given by


for x=1 and [A,B] =0; you can get the required answer


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.