A unified approach on Springer fibers in the hook, two-row and two-column cases (Q707906)

Consider a nilpotent endomorphism \(u : {\mathbb C}^n \to {\mathbb C}^n\) and the associated Springer fiber \({\mathcal B}_u\) of complete flags of \({\mathbb C}^n\) that are preserved by \(u\). The nilpotent endomorphism \(u\) is determined by the sizes of its Jordan blocks and hence can be associated with a partition \(Y(u)\) of \(n\). The irreducible components of \({\mathcal B}_u\) are in one-to-one correspondence with standard Young tableaux of shape \(Y(u)\). For a standard tableaux \(T\), let \({\mathcal K}^T\) denote the corresponding component of \({\mathcal B}_u\). Given a row-standard Young tableaux \(\tau\) of shape \(Y(u)\), the author constructs a flag \(F_{\tau} \in {\mathcal B}_{u}\) which is fixed by the action of the standard torus associated to the Jordan basis of \(u\). The main question of interest in this paper is determining when \(F_{\tau}\) lies in \({\mathcal K}^T\) for a given \(\tau\) and \(T\) as above. Note that while the definition of \(F_{\tau}\) depends on a specific choice of basis for \(u\), the question under consideration depends only on the Jordan form of \(u\). Further, the question of whether \(F_{\tau} \in {\mathcal K}^T\) is equivalent to whether or not \({\mathcal Z}_{\tau} \subset {\mathcal K}^T\) where \({\mathcal Z}_{\tau}\) is the \(Z(u)\)-orbit of \(F_{\tau}\) in \({\mathcal B}_u\) for \(Z(u) = \{g \in GL_n({\mathbb C}) : gu = ug\}\). This paper focuses on diagrams having one of three shapes: hook type (the diagram contains at most one row of length greater than one), two-row type (the diagram contains at most two rows), or two-column type (the diagram contains at most two columns). For these three cases, the author gives two characterizations of pairs \((\tau,T)\) for which \(F_{\tau} \in {\mathcal K}^T\). The first characterization involves the dominance relation on a collection of Young diagrams associated to \(\tau\) and \(T\). It is shown that this characterization fails for arbitrary shapes. The second characterization involves a tableaux construction algorithm determined by a pair \((\tau, T)\). Starting with the empty tableaux, the algorithm gives a procedure for adding successively a box containing 1, then a box containing 2, and so on through a box containing \(n\). It is shown that \(F_{\tau} \in {\mathcal K}^T\) if and only if the construction (based on the pair (\(\tau,T)\)) results in \(\tau\). The author then uses his techniques to study the question of when the intersection of two components \({\mathcal K}^S\) and \({\mathcal K}^T\) has codimension 1. In the three special cases as above, it is shown that if the intersection has codimension 1, then either \({\mathcal F}_S \in {\mathcal K}^T\) or \({\mathcal F}_T \in {\mathcal K}^S\). In the two-row case, by adding a condition on the entries in the first rows of \(S\) and \(T\), the author gives a precise characterization of the pairs \((S,T)\) for which \({\mathcal K}^S\cap{\mathcal K}^T\) has codimension 1. The paper concludes with some further applications of these results.