I'm trying to understand how to go from the general definition of a tangent space to the explicit definition of the Lie algebra of a matrix group. Let $G$ be a Lie group, the Lie algebra is the tangent space at the identity

$$\mathfrak{g}=T_e G$$

We can see define the tangent space as follows: let $\gamma: (-a,a)\to G$ be a smooth curve such that $\gamma(0)=e$, we define an equivalence relation: $\gamma \sim \tau$ if for any chart $\phi:U\to\mathbb{R}^n$ such that $e\in U$ $(\phi\circ\gamma)'(0)=(\phi\circ\tau)'(0)$, we note by $[\gamma]$ the equivalence class of $\gamma$. For a curve $\gamma$ we can define a derivation

$$\begin{aligned}\gamma'(0):& C^\infty(G)\to\mathbb{R} \\ &f\mapsto(f\circ \gamma)'(0)\end{aligned}$$

then $$T_e G =\{[\gamma]'(0): \gamma:(-a,a)\to G \textrm{ smooth, }\gamma(0)=e\}$$ where by the derivative of the equivalence class I mean the one of a representative.

First question: do I have this definition correct?

Now suppose that $G$ is a matrix group, i.e. $G\subset M_n(\mathbb{R})$, the Lie algebra is usually defined as the space of the matrices

$$\Omega =\frac{d}{dt}R(t)|_{t=0} $$

where $R(t):(-a,a)\to G$ is a smooth curve. The derivative is well defined as $R(t)$ is a matrix for any $t$. To reconcile the two definitions my idea is to consider the functions

$$\begin{aligned}x^{ij}:&G\to \mathbb{R}\\&A\mapsto A_{ij}=\langle e_i, A e_j\rangle \end{aligned}$$

for some basis $\{e_i\}$, then $R'(0)(x^{ij})=R'(0)_{ij}$ and we automatically have a matrix associated to any derivation in $T_eG$. My question is, can we reconstruct the image of the derivation applied to any smooth function on $G$ via this matrix, i.e., is it true that

$$R'(0)(f)=\sum_{ij}\alpha_{ij} R'(0)(x^{ij})$$.

For some coefficient $a_{ij}$?. This seems to rely on the fact that in some way $f(A)=\sum \alpha_{ij} x^{ij}$, hence that $f$ is linear, so this doesn't seem like the right way to do it. How can I reconcile these two definitions?


Your definition of the tangent space at the identity is correct.

You seem to be trying to identify $M_n(R)$ with $\mathbb{R}^{n\times n}$, which is a valid approach. The $x^{ij}$ are the coordinate projections.

We can write the curve $$R(t)= \pmatrix{R_{11}(t) & R_{12}(t) &\cdots & R_{1n}(t) \\ \vdots & \ddots &\\ R_{n1}(t) & R_{n2}(t) &\cdots & R_{nn}(t) \\} .$$

The only step you're missing is that $f: G \to \mathbb{R}$ now has to be identified as a function from a subset of $\mathbb{R}^{n\times n}$. Thus, we have $$f\circ R: I\subset \mathbb{R} \to \mathbb{R}.$$ Applying the chain rule we have, $$\frac{d}{dt} (f\circ R(t))\lvert_0 \ = \ df\vert_{R(0) }\circ \frac{d}{dt} R\vert_0.$$

Here $df$ is the derivative (linearisation/jacobian) of $f$ at $R(0)$. It is a linear map from the space of $n\times n$ real matrices to $\mathbb{R}$ (or from $\mathbb{R}^{n\times n}$ to $\mathbb{R}$).

Careful consideration of indices will show that your $\alpha_{ij}$ are in fact the directional derivatives of $f$ at $R(0)$.


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.