Closed Form Multiplicity Polynomials Attached to Hook-type Springer Fibers for SL(n, C). Matthew Housley housley@math.utah.edu University of Utah January 6th, 2011 Joint Mathematics Meetings - Special Session on Analytic and Geometric Methods in Representation Theory 1 / 43 Slides are available at www.math.utah.edu/~housley → talks. 2 / 43 Setting GR = SU(p, q) gR = su(p, q) g = sl(n, C), n = p + q h = diagonal elements in g. U(g) = universal enveloping algebra of g Maximal compact: KR = S(U(p) × U(q)), K = S(GL(p) × GL(q)) 3 / 43 Objects of study: irreducible (U(g), K )-modules. (Assume regular integral infinitesimal character.) 4 / 43 Commutative Algebra I Construct gr U(g) by using degree filtration. 5 / 43 Commutative Algebra I I Construct gr U(g) by using degree filtration. PBW: gr U(g) ∼ = S(g). (Algebra of polynomials in elements of g.) 6 / 43 Commutative Algebra I I Construct gr U(g) by using degree filtration. PBW: gr U(g) ∼ = S(g). (Algebra of polynomials in elements of g.) Let X be an irreducible (U(g), K )-module at regular integral infinitesimal character. We would like to define an (S(g), K )-module. 7 / 43 I Let X0 be a K -invariant generating subspace of X . 8 / 43 I Let X0 be a K -invariant generating subspace of X . I Xn = Un (g) · X0 . (Degree filtration.) 9 / 43 I Let X0 be a K -invariant generating subspace of X . I Xn = Un (g) · X0 . (Degree filtration.) I Xn is K -invariant. 10 / 43 I Let X0 be a K -invariant generating subspace of X . I Xn = Un (g) · X0 . (Degree filtration.) I Xn is K -invariant. I gr X is an (S(g/k), K )-module. 11 / 43 I Let X0 be a K -invariant generating subspace of X . I Xn = Un (g) · X0 . (Degree filtration.) I Xn is K -invariant. I gr X is an (S(g/k), K )-module. S(g/k) identifies with polynomial functions on (g/k)∗ . 12 / 43 The structure of gr X depends on the choice of X0 . 13 / 43 The structure of gr X depends on the choice of X0 . Definition (Vogan) Associated Variety: AV(X ) = {λ ∈ (g/k)∗ |p(λ) = 0 for all p ∈ Ann(gr X )} where Ann(X ) is the annihilator in S(g/k) of gr X . 14 / 43 Associated Variety Theorem (Vogan) AV(X ) does not depend on the choice of X0 . 15 / 43 Associated Variety Theorem (Vogan) AV(X ) does not depend on the choice of X0 . The components of AV(X ) consist of closures of nilpotent K -orbits on (g/k)∗ : [ AV(X ) = Oi i 16 / 43 Associated Cycle AV(X ) = [ Oi i 17 / 43 Associated Cycle AV(X ) = [ Oi i Oi ←→ Pi where Pi is a minimal prime ideal in S(g/k) containing Ann(gr X ). 18 / 43 Associated Cycle AV(X ) = [ Oi i Oi ←→ Pi where Pi is a minimal prime ideal in S(g/k) containing Ann(gr X ). I Associated cycle: components of AV(X ) with multiplicity. X AC(X ) = mi Oi i 19 / 43 Associated Cycle AV(X ) = [ Oi i Oi ←→ Pi where Pi is a minimal prime ideal in S(g/k) containing Ann(gr X ). I Associated cycle: components of AV(X ) with multiplicity. X AC(X ) = mi Oi i I Roughly speaking, mi measures number of copies of S(g/k)/Pi in gr X . 20 / 43 Like AV(X ), AC(X ) is independent of our initial choices. We would like to the calculate multiplicities mi . 21 / 43 Let X be an irreducible (U(g), K )-module for SU(p, q) with trivial infinitesimal character in the block of a finite dimensional representation. (p 6= q : there is only one block.) 22 / 43 Coherent Continuation 23 / 43 Coherent Continuation I Coherent family: At each dominant regular integral weight, we get a module X (λ), with X (ρ) = X . X (λ) has infinitesimal character λ. I At other weights: virtual modules. (Formal sums of irreducible modules.) 24 / 43 Multiplicity Polynomials I In our setting, AV(X (λ)) is irreducible, so we only need to calculate one integer to get AC(X (λ)). 25 / 43 Multiplicity Polynomials I In our setting, AV(X (λ)) is irreducible, so we only need to calculate one integer to get AC(X (λ)). I Let pX (λ) equal the multiplicity for AC(X (λ)) when λ is dominant and regular. 26 / 43 Multiplicity Polynomials I In our setting, AV(X (λ)) is irreducible, so we only need to calculate one integer to get AC(X (λ)). I Let pX (λ) equal the multiplicity for AC(X (λ)) when λ is dominant and regular. Theorem (Vogan) pX (λ) extends to a polynomial on h∗ . New goal: calculate pX (λ). 27 / 43 Cells I X is contained in an equivalence class of modules with the same infinitesimal character called a Harish-Chandra cell. Let X be in the cell C. 28 / 43 Cells I X is contained in an equivalence class of modules with the same infinitesimal character called a Harish-Chandra cell. Let X be in the cell C. I Each Y ∈ C is in a coherent family and has an attached multiplicity polynomial. 29 / 43 Cells I X is contained in an equivalence class of modules with the same infinitesimal character called a Harish-Chandra cell. Let X be in the cell C. I Each Y ∈ C is in a coherent family and has an attached multiplicity polynomial. Definition M(C) = spanC {pY (λ)|Y ∈ C} 30 / 43 Cells I X is contained in an equivalence class of modules with the same infinitesimal character called a Harish-Chandra cell. Let X be in the cell C. I Each Y ∈ C is in a coherent family and has an attached multiplicity polynomial. Definition M(C) = spanC {pY (λ)|Y ∈ C} Theorem (Joseph) M(C) is an irreducible W -representation. 31 / 43 Strategy I Find one X ∈ C such that pX (λ) is easy to calculate. 32 / 43 Strategy I Find one X ∈ C such that pX (λ) is easy to calculate. I Find pY (λ) for other modules in the cell by using the W action on M(C). 33 / 43 Question: how do we understand M(C) as a W -representation? 34 / 43 Question: how do we understand M(C) as a W -representation? Answer: we have the following isomorphisms: Sp(C) ←− Coh(C) −→ M(C) 35 / 43 Question: how do we understand M(C) as a W -representation? Answer: we have the following isomorphisms: Sp(C) ←− Coh(C) −→ M(C) I Sp(C) is a Springer representation: basis consists of components of a certain Springer fiber. 36 / 43 Question: how do we understand M(C) as a W -representation? Answer: we have the following isomorphisms: Sp(C) ←− Coh(C) −→ M(C) I Sp(C) is a Springer representation: basis consists of components of a certain Springer fiber. I Coh(C) is a cell representation: basis consists of modules in C. 37 / 43 Question: how do we understand M(C) as a W -representation? Answer: we have the following isomorphisms: Sp(C) ←− Coh(C) −→ M(C) I Sp(C) is a Springer representation: basis consists of components of a certain Springer fiber. I Coh(C) is a cell representation: basis consists of modules in C. I These isomorphisms respect basis elements: each Y ∈ C maps right to pY (λ) and left to a component Sp(Y ) of the Springer fiber. 38 / 43 Hook-type Case I Sp(C) is an irreducible representation of Sn . Such representations are parameterized by Young diagrams with n boxes. 39 / 43 Hook-type Case I Sp(C) is an irreducible representation of Sn . Such representations are parameterized by Young diagrams with n boxes. I Let X ∈ C. We say that X is of hook type if Sp(C) corresponds to a hook-type Young diagram. In this case, Sp(C) is particularly easy to understand. 40 / 43 Result We get a formula for pX (λ) where X is a Harish-Chandra module of hook-type: 41 / 43 Result We get a formula for pX (λ) where X is a Harish-Chandra module of hook-type: ! Y X X 1 sgn(σ 0 )σ 0 · zτm−i+1 pX (λ) = Am Q sgn(σ)σ· (i) |τk |! 0 σ∈Sτ σ ∈Sm i=1...m (zj are standard basis elements of the Cartan h.) The relevant combinatorial parameters are obtained by looking at the Springer fiber component Sp(X ). 42 / 43 Details Details are available at www.math.utah.edu/~housley → research. housley@math.utah.edu 43 / 43