Let $G$ be a semi-simple group over an algebraically closed field of characteristic zero. In which cases there is a unique wonderful compactification of $G$ (modulo isomorphism)?

For instance, is the wonderful compactification of $SL(n,\mathbb{C})$ unique?

  • 1
    $\begingroup$ Wonderful compactification is for an algebraic group with an action on a variety. When you just refer to $G$, what is the group and what is the variety? $G$ acting on itself by left translation? $G$ acting on itself by conjugation? $G\times G$ acting on $G$ by left-right translation? $\endgroup$ – YCor Feb 4 '18 at 0:04
  • $\begingroup$ For the definition: en.wikipedia.org/wiki/Wonderful_compactification $\endgroup$ – YCor Feb 4 '18 at 16:37

I assume you mean the variety $G$ considered as a $G \times G$ variety via the action $(g,h) \cdot x = gxh^{-1}$, which is the standard interpretation in the literature. The variety $G$ is spherical as a $G \times G$ variety, meaning that it contains a dense $B \times B$-orbit (where $B$ is a Borel subgroup of $G$). This statement is the content of the well-known Bruhat decomposition of $G$.

There is a more general statement: If $X$ is a spherical $G$-variety, where $G$ is reductive algebraic group over an algebraically closed field of any characteristic, then if $X$ admits a wonderful compactification, it is unique up to $G$-isomorphism.

The variety $G$ does have a wonderful compactification, which is therefore unique up to $G \times G$-isomorphism. The existence of the wonderful compactification of $G$ goes back to the original work on wonderful compactifications, Complete symmetric varieties by De Concini and Procesi (MSN). I don't know where the first proof of uniqueness appears, but a good introduction to the subject that includes a full proof is Pezzini's survey article, Lectures on spherical and wonderful varieties.

Here's a sketch of the main ideas involved. Any wonderful compactification of $G$, which I will denote $Y$, contains a big cell $Y_0$ with the following properties:

  1. $Y_0$ is dense in $Y$
  2. $Y_0$ is $B \times B$-invariant and isomorphic to affine space
  3. $G \cdot Y_0 = Y$

So suppose $Y$ and $Y'$ were wonderful compactifications of $G$ with big cells $Y_0$ and $Y'_0$. Let $g, g' \in G$ and consider the unique birational $G \times G$-equivariant map $f: Y \dashrightarrow Y'$, i.e. $f(g_1 g g_2^{-1}) = g_1 g' g_2^{-1}$. This map induces an isomorphism $Y_0 \cong Y'_0$ because they have isomorphic coordinate rings (an argument is needed why the map can be extended to the big cell). Then $f$ extends to an isomorphism $Y \cong Y'$ since $(G \times G) \cdot Y_0 = Y$.

  • $\begingroup$ Thanks a lot for the answer. Indeed I was considering $G = SL(n)$ and the action of $G\times G$ on $G$ given by $((A,B),M)\mapsto AMB^{t}$ (where $B^{t}$ is the transpose of $B$). Is it still fine if we consider this action instead of $AMB^{-1}$? $\endgroup$ – J_Cole Feb 4 '18 at 12:00
  • 2
    $\begingroup$ @J_Cole Yes, because it's just another system of coordinates for the same action. More formally, let $H_1$ acts on $X,Y$ and $f:X\to Y$ is $H_1$-equivariant. If $H_2$ is another algebraic group and $u:G\to H$ an isomorphism, then $H_2$ acts on $X,Y$ through $u$ and $f$ is $H_2$-equivariant, and vice-versa. So $X\to Y$ is a wonderful compactification for the $H_1$-action iff it's a wonderful compactification for the $H_2$-action. Now apply this to $G=SL(n)$, $H_1=H_2=G\times G$ and $u$ the automorphism given by $(A,B)\mapsto (A,(B^t)^{-1})$. $\endgroup$ – YCor Feb 4 '18 at 16:35
  • 2
    $\begingroup$ Isn't de Concini-Procesi only for $G$ adjoint (so not $SL(n)$)? $\endgroup$ – Allen Knutson Feb 25 '18 at 15:43
  • 3
    $\begingroup$ Allen is right: $SL(n)$ has no wonderful compactification unless $n=2$. This is because wonderfulness entails smoothness. Nevertheless, any semisimple group has has a certain canonical compactification which is wonderful if it happens to be smooth. That's the case for all adjoint groups, for $SL(2)$ and maybe some others. $\endgroup$ – Friedrich Knop Mar 18 at 19:22

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.