5
$\begingroup$

I am examining a proof that any affine connected algebraic group is a closed subgroup of some $GL_n$, and I am stuck on a fine point.

Let $G$ be an affine algebraic group in the sense that it is a group and also an affine variety over an algebraically closed field $k$, and such that the multiplication map $G\times G\rightarrow G$ and inverse map $G\rightarrow G$ are morphisms of varieties. Suppose that it is also connected.

$G$'s action on itself by (right or left) multiplication induces an action of $G$ on its coordinate ring $k[G]$. I have been convinced by the proof I am studying that there exists a finite-dimensional $k$-linear subspace $W\subset k[G]$ that is $G$-invariant and on which $G$ acts faithfully: essentially, if $f_1,\dots,f_m$ are $k$-algebra generators for $k[G]$, then all the $g(f_i)$'s live in a finite-dimensional subspace of $k[G]$ and this is $W$. This gives us an inclusion

$$\varphi:G\hookrightarrow GL(W)$$

so that $G$ is realized as a subgroup of $GL(W)\cong GL_n$ for $n=\dim W$. The argument I am studying now asserts that $\varphi$ is a morphism (with the variety structure on $GL(W)$ given by its identification with $GL_n\subset k^{n\times n}\cong \mathbb{A}_k^{n^2}$) and deduces that the image of $\varphi$ is a closed subvariety of $GL(W)$. I believe the argument except that I am having trouble explaining to myself why $\varphi$ has to be a morphism.

Why is $\varphi$ a morphism?

Here are my (very few) thoughts so far: selecting a basis $w_1,\dots,w_n$ for $W$, we can describe the $g$-action on $W$ in terms of its action on the basis:

$$ g(w_i)=\sum_j c_{ij}(g)w_j$$

Then $g\mapsto (c_{ij}(g))$ is an explicit description of $\varphi$. $c_{ij}:G\rightarrow k$ is the pullback of the $ij$th coordinate function on $GL(W)$ under $\varphi$. I need to know that this is a regular function on $G$. It must be implied by the fact that the multiplication map $G\times G\rightarrow G$ is a morphism.

This would be sufficient. Technically I also need to know that the pullback of $1/\delta$ is regular, where $\delta$ is the determinant function on $GL(W)$ (this is because the coordinate ring of $GL(W)$ is generated by the coordinate functions and $1/\delta$). However, it follows from $c_{ij}$ being regular for each $i,j$, because the pullback of $1/\delta$ is $1/\det (c_{ij}(g))$. If the $c_{ij}$ are all regular, then $\det (c_{ij})$ is regular, and nonvanishing, so $1/\det (c_{ij})$ is regular.

That's as far as I've gotten. Thanks in advance.

$\endgroup$
3
$\begingroup$

Given any finite dimensional vector space $V$ with basis $\{e_1, \ldots, e_n\}$ and an algebraic action $\phi\colon G \times V \to V$ the induced map $\phi\colon G \to \operatorname{GL}_n(k)$ is a morphism of varieties.

As you said you need only show that the composition $c_{ij} = x_{ij}\phi$ of $\phi$ with the coordinate function $x_{ij}$ giving the $(i, j)^\text{th}$ entry of the matrix is regular. But this is true because $c_{ij}$ is a composition of regular functions: $$\begin{matrix}G & \xrightarrow{g \mapsto (g, e_j)} & G \times V & \xrightarrow{\quad\phi\quad} & V & \xrightarrow{\quad x_i\quad} & k\end{matrix}$$ where $x_i$ is dual to $e_i$.

| cite | improve this answer | |
$\endgroup$

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.