I am reading the book: Lie Groups, Lie Algebras, and Representations: An Elementary Introduction by Brian C. Hall. I am stuck at the following exercise: exercise 11, chapter 2 . Can you help me?

Suppose $G$ is a matrix lie group in $GL(n, \mathbb{C})$ with the lie algebra $\mathfrak{g}$. suppose that $A$ is in G and that $||A-I||<1$,so the power series for $\log A$ is convergent. Is it necessarily the case that $\log A$ is in $\mathfrak{g}$? prove or give a counterexample.

In this book, $\log A$ is defined as a power series: $$ log A = \sum _{n=1}^{\infty} (-1)^{n+1}(A-I)^n/n$$ whenever the series converges.

I know that if $A$ is small enough, then we can prove that it is true.

  • $\begingroup$ Simon, did you ever figure this one out? I got stuck in the same place. I think you could prove it on the assumption that $G \cap B_1(I)$ is path connected, but it's not clear to me that this must always be the case. $\endgroup$ – Nathaniel Mayer May 5 '18 at 6:35

I reposted a closely related question here and got a good answer.

To summarize: No, $\log(A)$ may not be in $\mathfrak{g}$. Counterexamples can be found in the group of $n$-th roots of unity for large $n$, or in the line on the torus $$G_a = \left\{\begin{pmatrix} e^{2\pi it} & 0 \\ 0 & e^{2\pi ita} \end{pmatrix} : t \in \mathbb{R} \right\}$$ for $a \in \mathbb{Q}$ with large denominator.


In the cases of ordinary numbers, under the definition $$ \log x = \sum _{n=1}^{\infty} (-1)^{n+1}(x-1)^n/n, $$ we can expand $$ e^{\log x} = C_0 + C_1 (x-1)+ C_2 (x-1)^2 + \cdots .$$ Because $ e^{\log x} =x $, we should get $$ C_0 =1, C_1 =1 ,C_i=0 \, (i\geq 2) . $$ In the cases of matricies, under the definition $$ \log A = \sum _{n=1}^{\infty} (-1)^{n+1}(A-I)^n/n, $$ we can also expand $ e^{\log x} . $ Because $ \left[ (A-I)^k,(A-I)^l \right]=0, $ the coefficients of this series are the same as in the cases of ordinary numbers. So $$ e^{\log A} = I + (A-I) = A $$ Thus it is necessarily the case that $\log A$ is in $\mathfrak{g}$.

  • $\begingroup$ It is not enough to show that $e^{\log A}\in G$, we need to show that $e^{t \log A}\in G$ for all $t$. $\endgroup$ – Simon Zhu Oct 7 '16 at 4:26
  • $\begingroup$ "A Lie algebra" is a vector space, not the basis of the vector space. we don't need to show that $e^{t \log A}\in G$ for all $t$. $\endgroup$ – GotchaP Oct 9 '16 at 3:20
  • $\begingroup$ No, you do need $e^{t\log A} \in G$ for all $t$. For example, $e^{2\pi i I} = I \in G$ for any matrix Lie group $G$, but $2\pi i I$ is not in every Lie subalgebra of $\mathfrak{gl}(n)$ $\endgroup$ – Nathaniel Mayer May 5 '18 at 6:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.