2

I initially wanted to prove that if a Lie group $G$ is abelian, then $\exp(X+Y) = \exp(X)\exp(Y)$ for any $X, Y \in \frak{g}$. I showed that: $$ \frac{d}{dt}(\exp(tX)\exp(tY))= d_e (l_{\exp(tX)\exp(tY)}) (Ad(\exp(-tY))(X)) + d_e l_{\exp(tX)\exp(tY)}(Y) $$ or at least I hope I have, if my calculations are correct. If this formula is correct, then obviously if $G$ is abelian then $Ad$ is trivial and $\exp(tX)\exp(tY)$ is an integral curve of $X+Y$ as wanted.

But is it true for any Lie group that $\exp(X+Y)=\exp(X)\exp(Y)$ if and only if $[X,Y] = 0$ (this question suggests it should be true also for Lie groups)? The formula above seems to get pretty close to it, but I don't know how to make progress from there.

I know also that in general on a manifold $[X,Y] = 0$ if and only if the flows commute, which in this case just means $\exp(tX)\exp(sY) = \exp(sY)\exp(tX)$, but I don't know if this helps.

Essentially, is it true that $Ad(\exp(-tY)(X)) = X$ for any $t$ if and only if $[X,Y]=0$?

rosecabbage
  • 1,645
  • Your question is not equivalent to your final reformulation. It could be that $\exp(X+Y)$ ends up being "coincidentally" equal to $\exp(X)\exp(Y)$ for global reasons even though $\exp(tX)\exp(tY)$ is not equal to $\exp(t(X+Y))$ for most values of $t$. – Eric Wofsey Jun 22 '22 at 14:20
  • @EricWofsey That seems only fair. But is $[X,Y] = 0$ strong enough to imply $\exp(X)\exp(Y) = \exp(X+Y)$ at least? – rosecabbage Jun 22 '22 at 14:24

1 Answers1

4

The condition $[X,Y]=0$ does indeed imply that $\exp(X) \exp(Y) = \exp(X+Y)$.

This can be proved by applying the Baker-Cambell-Hausdorff formula which gives a Lie-algebra formula for a vector $Z$, expressed as a formal infinite series in terms of iterated commutators, that satisfies the equation $$\exp(X) \exp(Y) = \exp(Z) $$ (one may need to assume that $X,Y$ are "sufficiently small" in order to guarantee convergence of the formal series).

It's quite a complicated formula. However, in the special case that the 1st order commutator $[X,Y]$ satisfies $[X,Y]=0$, all the higher iterated commutators are also $0$ and the entire formula collapses to $Z=X+Y$.

Lee Mosher
  • 120,280
  • 1
    On the other hand, as Eric points out, it's possibly that $e^{X+Y}$ just happens to be equal to $ e^X e^Y$ with $[X,Y]\neq 0$. And, indeed, this does occur. For example, for $G = Sp(1)$, the unit length quaternions, this occurs when $X = 3\pi i$, $Y = 4\pi j$. – Jason DeVito - on hiatus Jun 22 '22 at 16:02
  • Right, hence my answer is only about the implication asked for in the OP's comment. – Lee Mosher Jun 22 '22 at 16:16
  • Yes, I agree that your answer successfully addresses the OP's comment (and I upvoted it). I was just trying to bump up Eric's comment from "it's possible" to "it definitely happens". – Jason DeVito - on hiatus Jun 22 '22 at 16:30