1
$\begingroup$

I'm reading Humphreys' Linear Algebraic Groups, trying to understand the following argument found on the top of pg. 76.

Let $G$ be an algebraic group over some field $k$, with $x\in G$. Let $\gamma_x:G\to G$ be the commutator map sending $y$ to $yxy^{-1}x^{-1}$. Note that the fiber over $e\in G$ is exactly the centralizer of $x$, $C_G(x)$. Earlier it is shown that the differential $(d\gamma_x)_e:\mathfrak{g}\to\mathfrak{g}$ is $1-\mathrm{Ad}\;x$, where $\mathfrak{g}$ is the Zariski tangent space of $G$ at the identity. Notice then that the kernel of $(d\gamma_x)_e$ is $\mathfrak{c}_{\mathfrak{g}}(x)=\{a\in\mathfrak{g}\mid a=\mathrm{Ad}\;x(a)\}$.

I understand all of this just fine. What I don't understand is the next claim. Humphreys states that from the above setup, we can conclude that $\mathscr{L}(C_G(x))\subset\mathfrak{c}_{\mathfrak{g}}(x)$. How can we conclude this? I must not have a firm grasp of how to think about the Lie algebra associated to an algebraic group. The two definitions I'm familiar with are the space of left invariant derivations of $k[G]$, or the Zariski tangent space of $G$ at $e$. Neither one helps me see why the above containment should be true.

$\endgroup$
1
$\begingroup$

Let $f:X\to Y$ be a morphism of affine varieties. Let $y\in Y$ and denote by $X_y=f^{-1}(y)$ the fiber. It is a closed subvariety of $X$. Let $x\in X_y$ be any point and $I=I(X_y)$. Then we have $I\subseteq \mathfrak m_x$. Denote by $T_x:=(\mathfrak{m}_x/\mathfrak{m}_x^2)^\ast$ the Zariski tangent space of $x$ and similarly, define $T_y$. Let $T'_x$ be the tangent space of $X_y$ at $x$. We have $T'_x\subseteq T_x$ in a way which I will explain in the following. The goal is to show that $T'_x$, as a subspace of $T_x$, lies in the kernel of the differential map $T_x\to T_y$.

The morphism $f$ induces a ring homomorphism $\phi:\mathcal{O}_{Y,y}\to\mathcal{O}_{X,x}$ with the property that $\mathcal O_{X,x}\cdot \phi(\mathfrak m_y)=I\subseteq \mathfrak m_x$. The dual map $\phi^\ast:\mathfrak m_x^\ast \to \mathfrak m_y^\ast$ has the kernel $K:=\{ \ell\in\mathfrak m_x^\ast \mid \ell(I)=0 \}$, because $0=\phi^\ast(\ell)=\ell\circ\phi$ if and only if $I=\operatorname{im}(\phi)\subseteq\ker(\ell)$. The tangent space $T'_x$ of $X_y$ at $x$ is the dual of $(\mathfrak m_x/I)/(\mathfrak m_x/I)^2$ and the dual map of $\rho:\mathfrak m_x \to \mathfrak m_x/I$ will satisfy $\operatorname{im}(\rho)\subseteq K$. Therefore, when we pass to quotients, the inclusion $T'_x\hookrightarrow T_x$ will map into the kernel of $T_x\to T_y$.

This is precisely what we wanted.

$\endgroup$

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.