Let $G$ be a lie group and $\mathfrak{g}$ its lie algebra. Consider the category $Rep(G)$ of finite dimensional representations of $G$ and the category $Rep(\mathfrak{g})$ of finite dimensional representations of $\mathfrak{g}$. Morphisms in those categories are just $G$- resp. $\mathfrak{g}$ equivariant linear maps. There is an obvious functor $$d\colon Rep(G)\rightarrow Rep(\mathfrak{g}),\pi\mapsto d_e\pi$$ which maps a representation of $G$ to a representation of $\mathfrak{g}$ by taking the derivative at the neutral element. On morphisms, $d$ is just the identity because $G$ equivariant maps are also $\mathfrak{g}$ equivariant, what can be seen by putting 1-parameter subgroups of $G$ in the defining definition of beeing $G$ equivariant and taking the derivative.

If $G$ is simply connected, $d$ is bijective on objects, because Lie group homomoprhism are in bijection to the morphisms of their lie algebras, if the domain is simply connected.

Is $d$ bijective on morphisms too? So $d$ is an isomorphism of categories?

The question reduces to the question, whether every $\mathfrak{g}$ equivariant linear map between vector spaces is $G$ equivariant.

  • 1
    $\begingroup$ Welcome to Math.SE Tina $\endgroup$
    – magma
    Jan 17, 2014 at 10:00
  • $\begingroup$ If you consider real Lie groups, $Rep(G)$ are unitary representations, then not every representation of $\mathfrak g$ is equal to $d\pi,\ \pi\in Rep(G).$ See e.g. the book by Barut,Raczka. $\endgroup$
    – yurius
    Jan 17, 2014 at 11:05
  • 1
    $\begingroup$ If $G$ is a simply connected lie group and $H$ and arbitrary Lie group, then for each lie algebra homomoprhism $\alpha\colon\mathfrak{g}\rightarrow\mathfrak{h}$ there is a unique Lie homomorphism $f\colon G\rightarrow H$, such that $\alpha=d_ef$. See the introduction of tinyurl.com/m6qhl6a for example. Representations of $G$ are just Lie-homomorphisms $G\rightarrow GL(V)$, where $V$ is a real or complex vector space, so it is indeed correct, that representations of $G$ are in 1:1 correspondance to the one of $\mathfrak{g}$ in the simply connected case. $\endgroup$
    – Tina
    Jan 17, 2014 at 11:18
  • $\begingroup$ Okay maybe I should add that I restrict me to finite dimensional representations, because I want GL(V) to be a (finite dimensional) Lie group. $\endgroup$
    – Tina
    Jan 17, 2014 at 11:42

1 Answer 1


Let $V$ and $W$ be two representations of $G$ and $f : V \rightarrow W$ be invariant by $\frak{g}$, i.e. for all $X \in \frak{g}$ and $v \in V$: $$f(X.u) = X.f(u).$$ Let $X \in \frak{g}$ and $g=\exp(X)$. Define $$ \phi : \mathbb{R} \longrightarrow \mathrm{End}(V,W), \quad t \mapsto f \circ \exp(tX).$$ and $$ \psi : \mathbb{R} \longrightarrow \mathrm{End}(V,W), \quad t \mapsto \exp(tX) \circ f.$$ Then $$\frac{d}{dt} \phi(t) = f \circ X \circ \exp(tX) = X \circ f \circ \exp(tX) = X.\phi(t).$$ and $$\frac{d}{dt} \psi(t) = X \circ \exp(tX) \circ f = X.\psi(t).$$ Hence $\phi$ and $\psi$ both satisfy the differential equation. $$\begin{cases} \frac{d}{dt} y(t) &= X.y(t) \\ y(0) & = f \end{cases} $$ So $\phi=\psi$ by Cauchy-Lipschitz, and $f.g = g.f$.

  • 3
    $\begingroup$ Note that this argument makes no use of the hypothesis that $G$ is simply connected; it suffices that $G$ be connected. $\endgroup$ Aug 25, 2015 at 9:11
  • $\begingroup$ What is $t \mapsto f \circ \exp(tX)$ a one-parameter subgroup of? I thought one-parameter subgroups were Lie group homomorphisms from $\mathbb{R}$ to another Lie group $G$, but obviously this cannot be a homomorphism $\mathbb{R} \to GL(W)$. $\endgroup$
    – ಠ_ಠ
    Nov 9, 2016 at 0:53
  • 1
    $\begingroup$ @ಠ_ಠ You are perfectly right. I have fixed the proof. $\endgroup$
    – user10676
    Nov 30, 2016 at 17:00

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.