10
$\begingroup$

I don't know if this is already answered somewhere in MO. The dynkin involution of $SL_{2n}$ that is $\alpha_i \mapsto \alpha_{2n-i}$ gives an outer automorphism of $SL_{2n}$ and then the maximal parabolic $P_n$ corresponding to the simple root $\alpha_n$ is invariant under this automorphism. So it induces an automorphism of $Gr(n,2n)=SL_{2n}/P_n$. Do you get the Lagrangian Grassmannian from here say as a fixed points of this automorphism or you need to twist it by an inner automorphism of $SL_{2n}$ to get the Lagrangian Grassmannian $LG(n,2n)$ ? How is this automorphism related to the involution $V \mapsto V^{\perp}$ on $Gr(n,2n)$ ? Of course here one needs to fix a symplectic form on $\mathbb C^{2n}$ in order to define the map.

$\endgroup$
3
$\begingroup$

I think the confusion comes from the term "gives an outer automorphism" which has to be made precise. The automorphism, say $\phi$, is only well-defined up to multipliciation by conjugation with an element of the torus. To make it unique one usually requires that a "pinning" $e_\alpha$ is fixed. For the $SL(n)$ this means that, on the Lie algebra level, the elementary matrix $E_{i,i+1}$ is mapped to $E_{2n-i,2n-i+1}$. It is easily verfied that $\phi(A)=J^{-1}(A^t)^{-1}J$ does the job where $J=\text{antidiag}(1,-1,1,\ldots,-1)$.

Observe that $J$ is swewsymmetric so the fixed point set of $\phi$ is a symplectic group. To get the symplectic form $\omega$ one wants $A\mapsto\phi(A)^{-1}$ to be the adjoint $A^*$ with respect to $\omega$: $$ \omega(Ax,y)=\omega(x,\phi(A)^{-1}y)\text{ or, equivalently, }\omega(Ax,\phi(A)y)=\omega(x,y). $$ Another easy calculation shows that $\omega(x,y)=x^tJy$ fits. Any other solution is a non-zero multiple.

At this stage, the perpendicular space $U^\perp$ with respect to $\omega$ is defined. It is intertwined with $\phi$ via the formual $$ (AU)^\perp=\phi(A)U^\perp. $$ Thus if you let $\phi$ act on $Gr(n,2n)$ as $\phi(U)=U^\perp$ then one gets $\phi(AU)=\phi(A)\phi(U)$ and then $LG(n,2n)$ is precisely the fixed point set of $\phi$.

$\endgroup$
  • 1
    $\begingroup$ Thank you for your beautiful answer. Indeed, that was the confusion. Combinatorially, how does the fixed points of $\phi$ (in the 1st paragraph of your answer) on $SL_{2n}/P_n$ look like ? How is it embedded in $Gr(n,2n)$ ? $\endgroup$ – Mark May 18 '17 at 18:43

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.