Ad-nilpotent ideals of Borel subalgebras: combinatorics and representation theory

Transcription

1 Ad-nilpotent ideals of Borel subalgebras: combinatorics and representation theory Paolo Papi Sapienza Università di Roma Preliminary version Winter School: Geometry, Algebra and Combinatorics of Moduli Spaces and Configurations; Dobbiaco, February 19-25, 2017 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 1 / 116

2 General Plan 1 Lecture 1 Two combinatorial Problems Outline of the course: statement of results 2 Lecture 2 Background 3 Lecture 3 Affine root systems and affine Weyl groups ad-nilpotent ideals of Borel subalgebras 4 Lecture 4 Panyushev s theory of rootlets Abelian ideals and g 5 Lecture 5 Kostant-Macdonald formulas Applications to number theory Application to affine Lie algebras Developments and open problems Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 2 / 116

3 Lecture 1 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 3 / 116

4 Lecture 1 Two combinatorial Problems Introduction: two combinatorial problems Consider the staircase shape T n = (n, n 1,..., 1) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 4 / 116

5 Lecture 1 Two combinatorial Problems Introduction: two combinatorial problems Label the boxes as matrix entries with row (resp. column) indices increasing from left to right (resp. from top to bottom). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 4 / 116

6 Lecture 1 Two combinatorial Problems Two combinatorial problems A subdiagram D of T n, is a shape like let h D denote the hook length of box (1, 1), i.e. the Upper-Left corner box. In the example at hand, h D = 7. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 5 / 116

7 Lecture 1 Two combinatorial Problems Two combinatorial problems Problems We want to count 1 the number N adnilp of subdiagrams of T n ; 2 the number N ab of subdiagrams D of T n such that h D n. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 6 / 116

8 Lecture 1 Two combinatorial Problems Two combinatorial problems Problems We want to count 1 the number N adnilp of subdiagrams of T n ; 2 the number N ab of subdiagrams D of T n such that h D n. Answers If C n = 1 n+1( 2n n ) denotes the Catalan number, then N adnilp = C n+1. Moreover, N ab = 2 n. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 6 / 116

9 Lecture 1 Two combinatorial Problems Reformulation of Problem 1 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 7 / 116

10 Lecture 1 Two combinatorial Problems Reformulation of Problem 1 It is clearly equivalent to count the following objects: Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 7 / 116

11 Lecture 1 Two combinatorial Problems Reformulation of Problem 1 It is clearly equivalent to count the following objects: 1 subdiagrams of T n ; 2 lattice paths from (0, n + 1) to (n + 1, 0) lying over the diagonal with horizontal rightwards and vertical upwards steps; 3 Dyck paths of semilength n + 1; 4 (weak) Dyck words of length 2(n + 1) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 7 / 116

12 Lecture 1 Two combinatorial Problems Reformulation of Problem 1 It is clearly equivalent to count the following objects: 1 subdiagrams of T n ; 2 lattice paths from (0, n + 1) to (n + 1, 0) lying over the diagonal with horizontal rightwards and vertical upwards steps; 3 Dyck paths of semilength n + 1; 4 (weak) Dyck words of length 2(n + 1) The notions of Dyck path/word are best illustrated by an example: Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 7 / 116

13 Lecture 1 Two combinatorial Problems Reformulation of Problem 1 It is clearly equivalent to count the following objects: 1 subdiagrams of T n ; 2 lattice paths from (0, n + 1) to (n + 1, 0) lying over the diagonal with horizontal rightwards and vertical upwards steps; 3 Dyck paths of semilength n + 1; 4 (weak) Dyck words of length 2(n + 1) The notions of Dyck path/word are best illustrated by an example: Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 7 / 116

14 Lecture 1 Two combinatorial Problems Reformulation of Problem 1 It is clearly equivalent to count the following objects: 1 subdiagrams of T n ; 2 lattice paths from (0, n + 1) to (n + 1, 0) lying over the diagonal with horizontal rightwards and vertical upwards steps; 3 Dyck paths of semilength n + 1; 4 (weak) Dyck words of length 2(n + 1) The notions of Dyck path/word are best illustrated by an example: abaabbaabababbaabb Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 7 / 116

15 Lecture 1 Two combinatorial Problems Counting Dyck words Notation W a,b : set of words w in the alphabet {A, B} with N A (w) = a, N B (w) = b. W a,b (A): set of words in W a,b starting with A [w] k : k-th prefix of w A,B (w) = N A (w) N B (w). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 8 / 116

16 Lecture 1 Two combinatorial Problems Counting Dyck words Notation W a,b : set of words w in the alphabet {A, B} with N A (w) = a, N B (w) = b. W a,b (A): set of words in W a,b starting with A [w] k : k-th prefix of w A,B (w) = N A (w) N B (w). Example W 2,2 = {aabb, abab, abba, bbaa, baba, baab} W 2,2 (A) = {aabb, abab, abba} w = abbba, [w] 1 = a, [w] 2 = ab, [w] 3 = abb,... A,B (w) = 1, A,B ([w] 4 ) = 2 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 8 / 116

17 Lecture 1 Two combinatorial Problems Counting Dyck words Notation W a,b : set of words w in the alphabet {A, B} with N A (w) = a, N B (w) = b. W a,b (A): set of words in W a,b starting with A [w] k : k-th prefix of w A,B (w) = N A (w) N B (w). Definition A weak (strong) Dyck word is a word w W a,b such that A,B ([w] k ) 0 (resp. A,B ([w] k ) > 0) for all 1 k a + b. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 8 / 116

18 Lecture 1 Two combinatorial Problems Counting Dyck words We want first to to enumerate the set T of strong Dyck words in W a,b. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 9 / 116

19 Lecture 1 Two combinatorial Problems Counting Dyck words We want first to to enumerate the set T of strong Dyck words in W a,b. Note that such a word belongs to W a,b (A), and that ( ) a + b 1 W a,b (A) =. b Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 9 / 116

20 Lecture 1 Two combinatorial Problems Counting Dyck words Now remark that T c = {w W a,b (A) A,B ([w] k ) = 0 for some k}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 9 / 116

21 Lecture 1 Two combinatorial Problems Counting Dyck words Now remark that T c = {w W a,b (A) A,B ([w] k ) = 0 for some k}. Let w T c and k = 2m be the rightmost position such that A,B ([w] k ) = 0. Then w has the form w = [w] k α, N A (α) = a m, N B (α) = b m. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 9 / 116

22 Lecture 1 Two combinatorial Problems Counting Dyck words Now remark that T c = {w W a,b (A) A,B ([w] k ) = 0 for some k}. Let w T c and k = 2m be the rightmost position such that A,B ([w] k ) = 0. Then w has the form w = [w] k α, N A (α) = a m, N B (α) = b m. Replace now α by its complement α (i.e., make in α the switch a b) and consider the word w = [w] k α Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 9 / 116

23 Lecture 1 Two combinatorial Problems Counting Dyck words Now remark that T c = {w W a,b (A) A,B ([w] k ) = 0 for some k}. Let w T c and k = 2m be the rightmost position such that A,B ([w] k ) = 0. Then w has the form w = [w] k α, N A (α) = a m, N B (α) = b m. Replace now α by its complement α (i.e., make in α the switch a b) and consider the word w = [w] k α We have N A (w ) = m + (b m) = b, N B (w ) = m + (a m) = a, so that w W b,a, indeed w W b,a (A). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 9 / 116

24 Lecture 1 Two combinatorial Problems Counting Dyck words Claim The map Φ : T c W b,a (A), Φ(w) = w is a bijection. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 10 / 116

25 Lecture 1 Two combinatorial Problems Counting Dyck words Claim The map Φ : T c W b,a (A), Φ(w) = w is a bijection. Example W 2,3 (A) w w abba a abbab abab a ababb ab aab abbba aabb a aabbb aabab aaabb Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 10 / 116

26 Lecture 1 Two combinatorial Problems Counting Dyck words Claim The map Φ : T c W b,a (A), Φ(w) = w is a bijection. Corollary ( a + b 1 T = W a,b (A) T c = b = a b ( ) a + b. a + b b ) ( ) a + b 1 a Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 10 / 116

27 Lecture 1 Two combinatorial Problems Counting Dyck words Now remark that Hence w W a,b is weak Dyck iff Aw W a+1,b is strong Dyck. Proposition The number of weak Dyck words in W a,b is ( ) a b + 1 a + b. a + 1 a In particular, for a = b = n + 1, N adnilp = 1 ( ) 2n + 2 = C n+1. n + 2 n + 1 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 11 / 116

28 Lecture 1 Two combinatorial Problems Elementary combinatorial calculation of N ab We claim that the number N k of subdiagrams of T n with hook lenght k is 2 k 1, k 1. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 12 / 116

29 Lecture 1 Two combinatorial Problems Elementary combinatorial calculation of N ab We claim that the number N k of subdiagrams of T n with hook lenght k is 2 k 1, k 1. B A Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 12 / 116

30 Lecture 1 Two combinatorial Problems Elementary combinatorial calculation of N ab We claim that the number N k of subdiagrams of T n with hook lenght k is 2 k 1, k 1. B A Fix a hook, and let A, B denote the end boxes in the leg and arm of cell (1, 1). We have to count all lattice paths from A to B inside the small rectangle. If the arm and leg length of the hook are h + 1, k h, they are ( k 1 h ). So k 1 ( ) k 1 N k = = 2 k 1. h h=0 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 12 / 116

31 Lecture 1 Two combinatorial Problems Elementary combinatorial calculation of N ab We claim that the number N k of subdiagrams of T n with hook lenght k is 2 k 1, k 1. Fix a hook, and let A, B denote the end boxes in the leg and arm of cell (1, 1). We have to count all lattice paths from A to B inside the small rectangle. If the arm and leg length of the hook are h + 1, k h, they are ( k 1 h ). So Now simply remark that N ab = 1 + k 1 ( ) k 1 N k = = 2 k 1. h h=0 n N k = 1 + k=1 n 2 k 1 = 2 n k=1 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 12 / 116

32 Lecture 1 Two combinatorial Problems Relationships between combinatorial and algebraic objects Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 13 / 116

33 Lecture 1 Two combinatorial Problems Relationships between combinatorial and algebraic objects The diagrams we have seen are instances (more precisely, combinatorial encodings) of objects which make sense for any simple Lie algebra g, the ad-nilpotent and abelian ideals of a Borel subalgebra. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 13 / 116

34 Lecture 1 Two combinatorial Problems Relationships between combinatorial and algebraic objects The diagrams we have seen are instances (more precisely, combinatorial encodings) of objects which make sense for any simple Lie algebra g, the ad-nilpotent and abelian ideals of a Borel subalgebra. We will now explain (in a rough way) several occurences of these algebraic objects in theorems involving the following keywords Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 13 / 116

35 Lecture 1 Two combinatorial Problems Relationships between combinatorial and algebraic objects The diagrams we have seen are instances (more precisely, combinatorial encodings) of objects which make sense for any simple Lie algebra g, the ad-nilpotent and abelian ideals of a Borel subalgebra. We will now explain (in a rough way) several occurences of these algebraic objects in theorems involving the following keywords the Euler product n=1 (1 x n ); Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 13 / 116

36 Lecture 1 Two combinatorial Problems Relationships between combinatorial and algebraic objects The diagrams we have seen are instances (more precisely, combinatorial encodings) of objects which make sense for any simple Lie algebra g, the ad-nilpotent and abelian ideals of a Borel subalgebra. We will now explain (in a rough way) several occurences of these algebraic objects in theorems involving the following keywords the Euler product n=1 (1 x n ); the structure of g as a g-module; Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 13 / 116

37 Lecture 1 Two combinatorial Problems Relationships between combinatorial and algebraic objects The diagrams we have seen are instances (more precisely, combinatorial encodings) of objects which make sense for any simple Lie algebra g, the ad-nilpotent and abelian ideals of a Borel subalgebra. We will now explain (in a rough way) several occurences of these algebraic objects in theorems involving the following keywords the Euler product n=1 (1 x n ); the structure of g as a g-module; u-cohomology; Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 13 / 116

38 Lecture 1 Outline of the course: statement of results Plan In the following slides I shall state a number of theorems, basically involving a simple Lie algebra and related algebraic objects. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 14 / 116

39 Lecture 1 Outline of the course: statement of results Plan In the following slides I shall state a number of theorems, basically involving a simple Lie algebra and related algebraic objects. The goal of my lectures is to gradually clarify all these statements and the relationships among them, providing the necessary background setting. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 14 / 116

40 Lecture 1 Outline of the course: statement of results Plan In the following slides I shall state a number of theorems, basically involving a simple Lie algebra and related algebraic objects. The goal of my lectures is to gradually clarify all these statements and the relationships among them, providing the necessary background setting. The role of ad-nilpotent and abelian ideal of a Borel subalgebras, which we have seem embodied in one special case, will naturally emerge. In particular, the two combinatorial results we have proved will be part of a more general and intrinsic theory. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 14 / 116

41 Lecture 1 Outline of the course: statement of results Macdonald-Kostant Theorem Notation g complex simple finite dimensional Lie algebra; b Borel subalgebra, with Cartan component h and nilradical n root system of (g, h), + set of positive roots with basis Π and fundamental chamber C ρ = 1/2 α + α Weyl vector, P+ dominant integral weights W Weyl group of g, (, ) Killing form of g V λ irreducible g-module of highest weight λ, χ λ character of V λ Cas(λ) = (λ, λ + 2ρ), eigenvalue of the Casimir operator Ω g on V λ Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 15 / 116

42 Lecture 1 Outline of the course: statement of results Standard Lie algebra notation Euler product Let φ(x) = (1 x n ) C[[x]] n=1 denote the Euler product. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 16 / 116

43 Lecture 1 Outline of the course: statement of results Standard Lie algebra notation Euler product Let φ(x) = (1 x n ) C[[x]] n=1 denote the Euler product.recall that 1 φ(x) = p(n)x n n 0 where p(n) is the classical partition function. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 16 / 116

44 Lecture 1 Outline of the course: statement of results Standard Lie algebra notation Theorem Let φ(x) = n=1 (1 x n ) C[[x]] denote the Euler product. Then φ(x) dim g = χ λ (e 2π 1 2ρ ) dim V λ x Cas(λ). λ P + Moreover χ λ (e 2π 1 2ρ ) { 1, 0, 1} for λ P +. Problem 1 Single out the subset of P + consisting of weights giving nonzero contribution to the sum. Find the coefficients b k in φ(x) dim g = b k x k. k=0 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 16 / 116

45 Lecture 1 Outline of the course: statement of results First answer to problem 1 Q coroot lattice, Ŵ = W Q Aff (h R ) A 1 fundamental alcove, A w = wa 1, w Ŵ. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 17 / 116

46 Lecture 1 Outline of the course: statement of results First answer to problem 1 Q coroot lattice, Ŵ = W Q Aff (h R ) A 1 fundamental alcove, A w = wa 1, w Ŵ. Ŵ + = {w Ŵ A w C}, λ w = w(2ρ)/2 ρ, D + = {λ w } w Ŵ +. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 17 / 116

47 Lecture 1 Outline of the course: statement of results First answer to problem 1 Ŵ + = {w Ŵ A w C}, λ w = w(2ρ)/2 ρ, D + = {λ w } w Ŵ +. Theorem (Kostant, 2004) χ λ (e 2π 1 2ρ ) = { ( 1) l(w) λ = λ w, w Ŵ +, 0 otherwise. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 17 / 116

48 Lecture 1 Outline of the course: statement of results First answer to problem 1 Ŵ + = {w Ŵ A w C}, λ w = w(2ρ)/2 ρ, D + = {λ w } w Ŵ +. Theorem (Kostant, 2004) χ λ (e 2π 1 2ρ ) = b k = { ( 1) l(w) λ = λ w, w Ŵ +, 0 otherwise. ( 1) l(w) dim V λ w. w Ŵ +,Cas(λ w )=k Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 17 / 116

49 Lecture 1 Outline of the course: statement of results Nilradical homology for affine algebras Notation ĝ = C[t, t 1 ] g CK Cd affine Kac-Moody algebra attached to g u = tg[t], u = t 1 g[t 1 ] opposite niradicals in ĝ. Bigrading on u u = (n,k) Z 0 Z 0 ( n u ) where the subscript k denotes the subspace of t-degree k. This grading descends to homology. k. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 18 / 116

50 Lecture 1 Outline of the course: statement of results Nilradical homology for affine algebras Theorem As a g-module Moreover H (u) = H (u ) = V λw. w Ŵ + H n (u ) k = V λ w. w Ŵ +,l(w)=n,cas(λ w )=k Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 19 / 116

51 Lecture 1 Outline of the course: statement of results Interlude: ad-nilpotent and abelian ideals of Borel subalgebras Let i be an ideal of b contained in n. It consists of ad-nilpotent elements, so we ll call it an ad-nilpotent ideal and we ll denote by I the set of ad-nilpotent ideals. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 20 / 116

52 Lecture 1 Outline of the course: statement of results Interlude: ad-nilpotent and abelian ideals of Borel subalgebras Let i be an ideal of b contained in n. It consists of ad-nilpotent elements, so we ll call it an ad-nilpotent ideal and we ll denote by I the set of ad-nilpotent ideals.it turns out that i = α Φ i g α where Φ i + is dual order ideal of the root poset. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 20 / 116

53 Lecture 1 Outline of the course: statement of results Interlude: ad-nilpotent and abelian ideals of Borel subalgebras Let i be an ideal of b contained in n. It consists of ad-nilpotent elements, so we ll call it an ad-nilpotent ideal and we ll denote by I the set of ad-nilpotent ideals.it turns out that i = α Φ i g α where Φ i + is dual order ideal of the root poset. I contains the remarkable subset of abelian ideals of b: I ab = {i I [x, y] = 0 x, y i}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 20 / 116

54 Lecture 1 Outline of the course: statement of results ad-nilpotent and abelian ideals of Borel subalgebras Theorem 1 I ab = 2 rk (g). 2 If h denotes the Coxeter number and m i are the exponents of g then I = rk g i=1 (h + m i + 1). W Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 21 / 116

55 Lecture 1 Outline of the course: statement of results ad-nilpotent and abelian ideals of Borel subalgebras Theorem 1 I ab = 2 rk (g). 2 If h denotes the Coxeter number and m i are the exponents of g then I = rk g i=1 (h + m i + 1). W Example If g = sl(n + 1), then rk g = n, W = S n+1 = (n + 1)!, h = n + 1, m i = i, so that N ab = 2 n, N adnilp = C n+1. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 21 / 116

56 Lecture 1 Outline of the course: statement of results Relationships between I, I ab, Ŵ The above theorem has a more significant formulation. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 22 / 116

57 Lecture 1 Outline of the course: statement of results Relationships between I, I ab, Ŵ Let θ be the highest root of and set, for i I i = α Φ i α, Ŵ + 2 = {w Ŵ A w 2A 1 }. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 22 / 116

58 Lecture 1 Outline of the course: statement of results Relationships between I, I ab, Ŵ Let θ be the highest root of and set, for i I i = α Φ i α, Ŵ + 2 = {w Ŵ A w 2A 1 }. Theorem There are natural bijections η : I Q /(h + 1)Q, ζ : I ab Ŵ + 2. Moreover, for i I ab i = λ ζ(i). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 22 / 116

59 Lecture 1 Outline of the course: statement of results Relationships between I, I ab, Ŵ Let θ be the highest root of and set, for i I i = α Φ i α, Ŵ + 2 = {w Ŵ A w 2A 1 }. Theorem There are natural bijections η : I Q /(h + 1)Q, ζ : I ab Ŵ + 2. Moreover, for i I ab i = λ ζ(i). Next we ll see some representation theoretic applications. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 22 / 116

60 Lecture 1 Outline of the course: statement of results The structure of g as a g-module Notation If a = k i=1 Cv i is an abelian subalgebra of g, set v a = v 1... v k k g. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 23 / 116

61 Lecture 1 Outline of the course: statement of results The structure of g as a g-module Notation If a = k i=1 Cv i is an abelian subalgebra of g, set v a = v 1... v k k g. m k is the maximum eigenvalue of Ω g on k g M k eigenspace of Ω g on k g of eigenvalue k C k = Span(v a a abelian, dim(a) = k) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 23 / 116

62 Lecture 1 Outline of the course: statement of results Kostant Theorems m k is the maximum eigenvalue of Ω g on k g M k eigenspace of Ω g on k g of eigenvalue k C k = Span(v a a abelian, dim(a) = k), C = k C k Theorem 1 m k k, and m k = k iff C k. In such a case M k = C k. 2 C is a multiplicity-free g-module. Moreover C k = V i = i I ab, dim i=k w Ŵ + 2, l(w)=k V λ ζ(i). 3 If d is the Chevalley-Eilenberg differential affording Lie algebra cohomology, then g = C dg, where dg denotes the ideal generated by dg under wedge multiplication. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 24 / 116

63 Lecture 1 Outline of the course: statement of results Final results Theorem The following numbers are equal: Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 25 / 116

64 Lecture 1 Outline of the course: statement of results Final results Theorem The following numbers are equal: 1 dim C k 2 dim M k 3 dim H k (u ) k Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 25 / 116

65 Lecture 1 Outline of the course: statement of results Final results Theorem The following numbers are equal: 1 dim C k 2 dim M k 3 dim H k (u ) k If moreover k h, the dual Coxeter number of g, they are also equal to ( 1) k b k Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 25 / 116

66 Lecture 1 Outline of the course: statement of results References P. Cellini and P. Papi. ad-nilpotent ideals of a Borel subalgebra I, II, J. Algebra, 225, (2000), and 58, (2002), B. Kostant, Eigenvalues of a Laplacian and commutative Lie subalgebras Topology, 3, suppl. 2 (1965), B. Kostant, On Macdonald s η-function formula, the Laplacian and Generalized exponents, Adv. Math. 20 (1976), B. Kostant, The set of abelian ideals of a Borel subalgebra, Cartan decompositions, and discrete series representations, Int. Math. Res. Notices (1998), no. 5, B. Kostant, Powers of the Euler product and commutative subalgebras of a complex simple Lie algebra, Invent. Math., 158 (2004), D. Panyushev. Abelian ideals of a Borel subalgebra and long positive roots, Intern. Math. Res. Notices 5 (2003), no. 35, Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 26 / 116

67 Lecture 2 Background A crash course in semisimple Lie algebras 1 Examples: the classical Lie algebras sl(n, C) = {A M n (C) tr(a) = 0} A B v so(2n + 1, C) = { C A t u B = B t, C = C t, u, v C n } v t u t 0 ( ) A B sp(2n, C) = { C A t B = B t, C = C t } ( ) A B so(2n, C) = { C A t B = B t, C = C t } Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 27 / 116

68 Lecture 2 Background A crash course in semisimple Lie algebras 2 General Definitions A complex finite-dimensional Lie algebra g is said to be 1 simple if it is not abelian and has no nontrivial ideals 2 semisimple if has no solvable ideals Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 28 / 116

69 Lecture 2 Background A crash course in semisimple Lie algebras 2 General Definitions A complex finite-dimensional Lie algebra g is said to be 1 simple if it is not abelian and has no nontrivial ideals 2 semisimple if has no solvable ideals Characterization of semisimple Lie algebras g is semisimple if and only if one of the following conditions is verified 1 g is a direct sum of simple ideals. 2 The Killing form of g, defined as (x, y) = tr(ad(x) ad(y)), is nondegenerate. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 28 / 116

70 Lecture 2 Background Interlude Why are semisimple Lie algebras important? Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 29 / 116

71 Lecture 2 Background Interlude Why are semisimple Lie algebras important? Let G a Lie group (for instance a closed subgroup of GL(n)). Then g = {c (0) c : R G, C curve with c(0) = I }. has a natural Lie algebra structure, which makes g a first-order approximation of G. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 29 / 116

72 Lecture 2 Background Interlude Why are semisimple Lie algebras important? Let G a Lie group (for instance a closed subgroup of GL(n)). Then g = {c (0) c : R G, C curve with c(0) = I }. has a natural Lie algebra structure, which makes g a first-order approximation of G. Theorem If G is compact then g is reductive, i.e. is a direct sum as Lie algebras of a semisimple Lie algebra and an abelian one. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 29 / 116

73 Lecture 2 Background A crash course in semisimple Lie algebras 3 Structure Theory Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 30 / 116

74 Lecture 2 Background A crash course in semisimple Lie algebras 3 Structure Theory Recall that and element x is said to be semisimple if ad(x) is diagonalizable as an endomorphism of g. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 30 / 116

75 Lecture 2 Background A crash course in semisimple Lie algebras 3 Structure Theory Recall that and element x is said to be semisimple if ad(x) is diagonalizable as an endomorphism of g. By Engel s theorem, if g is semisimple then there exist semisimple elements, so we can consider subalgebras formed by semisimple elements, and in turn subalgebras maximal w.r.t. this property, which are called Cartan subalgebras. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 30 / 116

76 Lecture 2 Background A crash course in semisimple Lie algebras 3 Structure Theory Recall that and element x is said to be semisimple if ad(x) is diagonalizable as an endomorphism of g. By Engel s theorem, if g is semisimple then there exist semisimple elements, so we can consider subalgebras formed by semisimple elements, and in turn subalgebras maximal w.r.t. this property, which are called Cartan subalgebras. A Cartan subalgebra h turns out to be abelian, hence is a set of commuting diagonalizable operators on g. We can therefore consider the corresponding eigenspace decomposition. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 30 / 116

77 Lecture 2 Background A crash course in semisimple Lie algebras 4 Root Space decomposition Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 31 / 116

78 Lecture 2 Background A crash course in semisimple Lie algebras 4 Root Space decomposition g = α h g α, g α = {x g [h, x] = α(h)x h h}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 31 / 116

79 Lecture 2 Background A crash course in semisimple Lie algebras 4 Root Space decomposition g = α h g α, g α = {x g [h, x] = α(h)x h h}. Since h is self-centralizing, we can rewrite the previous decomposition as g = h α g α where h \ {0} is a certain finite set, called the root system of g w.r.t. h. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 31 / 116

80 Lecture 2 Background A crash course in semisimple Lie algebras 5 Basic Theorems Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 32 / 116

81 Lecture 2 Background A crash course in semisimple Lie algebras 5 Basic Theorems 1 Root systems, as we shall see, can be studied and classified combinatorially. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 32 / 116

82 Lecture 2 Background A crash course in semisimple Lie algebras 5 Basic Theorems 1 Root systems, as we shall see, can be studied and classified combinatorially. 2 One proves that the classification of roots systems induces the classification of semisimple Lie algebras, meaning that there is no dependence, up to isomorphism, on the choice of the Cartan subalgebra and other choices which should be done in classifying root systems. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 32 / 116

83 Lecture 2 Background A crash course in semisimple Lie algebras 5 Basic Theorems 1 Root systems, as we shall see, can be studied and classified combinatorially. 2 One proves that the classification of roots systems induces the classification of semisimple Lie algebras, meaning that there is no dependence, up to isomorphism, on the choice of the Cartan subalgebra and other choices which should be done in classifying root systems. 3 The final outcome is that there are the four infinite series we have seen in a previous slide (named A n, B n, C n, D n ) plus five exceptional Lie algebras (named E 6, E 7, E 8, G 2, F 4 ). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 32 / 116

84 Lecture 2 Background Finite root systems Reflections Let E be an Euclidean space. If 0 α E, the reflection in α is the orthogonal transformation defined by s α (v) = v 2(v, α) (α, α) α. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 33 / 116

85 Lecture 2 Background Finite root systems Reflections Let E be an Euclidean space. If 0 α E, the reflection in α is the orthogonal transformation defined by s α (v) = v 2(v, α) (α, α) α. Definition A finite set E of nonzero vectors is a root system in E if 1 E = Span R ; 2 if α then cα c = ±1; 3 s α ( ) α ; 4 2 (β,α) (α,α) Z α, β Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 33 / 116

86 Lecture 2 Background Finite root systems: examples in rank 2 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 34 / 116

87 Lecture 2 Background Finite root systems: structure Notice that to a root system we can associate the central hyperplane arrangement in E given by the equations (α, x) = 0, α ; a reflection group, i.e. the subgroup W of O(E) generated by s α, α. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 35 / 116

88 Lecture 2 Background Finite root systems: structure Notice that to a root system we can associate the central hyperplane arrangement in E given by the equations (α, x) = 0, α ; a reflection group, i.e. the subgroup W of O(E) generated by s α, α. The complement E = E \ α α is a union of convex cones acted on by W. We say that a vector v E is regular. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 35 / 116

89 Lecture 2 Background Example: central arrangement Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 36 / 116

90 Lecture 2 Background Example: affine arrangement Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 37 / 116

91 Lecture 2 Background Finite root systems: structure Fix a regular vector γ and set + = {α (γ, α) > 0}, so that = + +. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 38 / 116

92 Lecture 2 Background Finite root systems: structure Fix a regular vector γ and set + = {α (γ, α) > 0}, so that = + +. Proposition Let Π = {α 1,..., α r } be the set of roots + which are not sum of two roots from +. Then 1 Π is a linear basis of E; 2 + = { r i=1 a iα i a i 0}; 3 W acts simply transitively on chambers and bases. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 38 / 116

93 Lecture 2 Background Finite root systems: structure Fix a regular vector γ and set + = {α (γ, α) > 0}, so that = + +. Proposition Let Π = {α 1,..., α r } be the set of roots + which are not sum of two roots from +. Then 1 Π is a linear basis of E; 2 + = { r i=1 a iα i a i 0}; 3 W acts simply transitively on chambers and bases. Corollary One associates to Π a graph with some combinatorial data, the Dynkin diagram. The classification of the possible Dynkin diagrams, affords the classification of root systems. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 38 / 116

94 Lecture 2 Background Recollections from the theory of semisimple Lie algebras Triangular decomposition g semisimple, h Cartan, roots, + positive roots, Π simple roots. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 39 / 116

95 Lecture 2 Background Recollections from the theory of semisimple Lie algebras Triangular decomposition g semisimple, h Cartan, roots, + positive roots, Π simple roots. g = h n + n, n ± = α ± + g α Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 39 / 116

96 Lecture 2 Background Recollections from the theory of semisimple Lie algebras Triangular decomposition g semisimple, h Cartan, roots, + positive roots, Π simple roots. g = h n + n, n ± = α ± + g α g α = {x g [h, x] = α(h)x h h} is one-dimensional. g α g α [g α, g α ] is a copy of sl(2, C). b = h n +, a Borel subalgebra, is a maximal solvable subalgebra of g. n + is the nilradical of b. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 39 / 116

97 Lecture 2 Background Weyl group action Notation Let V = h R. A vector in E = V \ α α is said to be regular The connected components of E are called chambers. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 40 / 116

98 Lecture 2 Background Weyl group action Notation Let V = h R. A vector in E = V \ α α is said to be regular The connected components of E are called chambers. Proposition W acts simply transitively on chambers and bases. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 40 / 116

99 Lecture 2 Background Weyl group action Notation Let V = h R. A vector in E = V \ α α is said to be regular The connected components of E are called chambers. Proposition W acts simply transitively on chambers and bases. Corollary Fixing a chamber C and labelling it by 1 W, the map w C w := w C 1 is a bijection between W and the set of chambers. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 40 / 116

100 Lecture 2 Background Weyl groups as Coxeter groups Recall that W is generated by the reflections s α, α. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 41 / 116

101 Lecture 2 Background Weyl groups as Coxeter groups Recall that W is generated by the reflections s α, α. It turns out that it is generated just by the set S = {s α, α Π}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 41 / 116

102 Lecture 2 Background Weyl groups as Coxeter groups Recall that W is generated by the reflections s α, α. It turns out that it is generated just by the set S = {s α, α Π}.Moreover, the corresponding relations have a particular nice form. Coxeter relations (s α s β ) m α,β = 1 where α, β Π m α,β N { }, m α,α = 1. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 41 / 116

103 Lecture 2 Background Weyl groups as Coxeter groups Recall that W is generated by the reflections s α, α. It turns out that it is generated just by the set S = {s α, α Π}.Moreover, the corresponding relations have a particular nice form. Coxeter relations (s α s β ) m α,β = 1 where α, β Π m α,β N { }, m α,α = 1. Remark One has a natural length function l on W. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 41 / 116

104 Lecture 2 Background Example: sl(n) Take g = sl(n). Then one can choose h = diagonal traceless matrices. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 42 / 116

105 Lecture 2 Background Example: sl(n) Take g = sl(n). Then one can choose h = diagonal traceless matrices. Denote by e ij the standard matrix unit. The basic computation is as follows; if h = diag(h 1,..., h n ), and ɛ i is the i-th coodinate function of h, then [h, e ij ] = (h i h j )e ij = (ɛ i ɛ j )(h)e ij. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 42 / 116

106 Lecture 2 Background Example: sl(n) Take g = sl(n). Then one can choose h = diagonal traceless matrices. Denote by e ij the standard matrix unit. The basic computation is as follows; if h = diag(h 1,..., h n ), and ɛ i is the i-th coodinate function of h, then [h, e ij ] = (h i h j )e ij = (ɛ i ɛ j )(h)e ij. Hence = {ɛ i ɛ j i j}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 42 / 116

107 Lecture 2 Background Example: sl(n) One can choose + = {ɛ i ɛ j i < j}. so that Π = {ɛ i ɛ i+1 1 i n 1}. since ɛ i ɛ j = (ɛ i ɛ i+1 ) + (ɛ i+1 ɛ i+2 ) +... ((ɛ j 1 ɛ j )). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 43 / 116

108 Lecture 2 Background Example: sl(n) One can choose + = {ɛ i ɛ j i < j}. so that Π = {ɛ i ɛ i+1 1 i n 1}. since ɛ i ɛ j = (ɛ i ɛ i+1 ) + (ɛ i+1 ɛ i+2 ) +... ((ɛ j 1 ɛ j )).In turn W = S n via s ɛi ɛ i+1 (i, i + 1). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 43 / 116

109 Lecture 2 Background Example: sl(n) One can choose + = {ɛ i ɛ j i < j}. so that Π = {ɛ i ɛ i+1 1 i n 1}. since ɛ i ɛ j = (ɛ i ɛ i+1 ) + (ɛ i+1 ɛ i+2 ) +... ((ɛ j 1 ɛ j )).In turn W = S n via s ɛi ɛ i+1 (i, i + 1). Note that the Coxeter relations read (i, i + 1) 2 = 1, (i, i + 1)(h, h + 1) = (h, h + 1)(i, i + 1) if i + 1 < h, (i + 1, i + 2)(i, i + 1)(i + 1, i + 2) = (i, i + 1)(i + 1, i + 2)(i, i + 1) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 43 / 116

110 Lecture 2 Background Example Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 44 / 116

111 Lecture 2 Background Interpretation of the length function Definition w W N(w) = {α + w 1 (α) + }. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 45 / 116

112 Lecture 2 Background Interpretation of the length function Definition w W N(w) = {α + w 1 (α) + }. Proposition 1 l(w) = N(w) = # hyperplanes separating C 1, C w. More precisely, α N(w) iff (α, x) = 0 separates C 1, C w. 2 If w = s i1 s ik is a reduced expression, then N(w) = {α i1, s i1 (α i2 ),..., s i1 s ik 1 (α ik )}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 45 / 116

113 Lecture 2 Background Interpretation of the length function Proposition 1 l(w) = N(w) = # hyperplanes separating C 1, C w. More precisely, α N(w) iff (α, x) = 0 separates C 1, C w. 2 If w = s i1 s ik is a reduced expression, then N(w) = {α i1, s i1 (α i2 ),..., s i1 s ik 1 (α ik )}. Example If σ S n, the N(σ) is the set of its inversions: ( ) σ = = s s 2 s 5 s 4 s 3 s 2 N(σ) = {ɛ 1 ɛ 2, ɛ 1 ɛ 3, ɛ 5 ɛ 6, ɛ 4 ɛ 6, ɛ 1 ɛ 6, ɛ 3 ɛ 6 } Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 45 / 116

114 Lecture 2 Background An important technical Lemma Definition Say that L + is root-closed if α, β L, α + β = α + β L Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 46 / 116

115 Lecture 2 Background An important technical Lemma Definition Say that L + is root-closed if Proposition α, β L, α + β = α + β L Given L +, there exists a (unique) w W such that L = N(w) if and only if both L and + \ L are root-closed. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 46 / 116

116 Lecture 2 Background An important technical Lemma Definition Say that L + is root-closed if Proposition α, β L, α + β = α + β L Given L +, there exists a (unique) w W such that L = N(w) if and only if both L and + \ L are root-closed. Remark The fact that + \ L is root-closed means that α L, α = β + γ, β, γ + = β L or γ L. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 46 / 116

117 Lecture 2 Background An important technical Lemma Example-Algorithm Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 47 / 116

118 Lecture 2 Background An important technical Lemma Example-Algorithm Given L biclosed, choose a simple root α L and iterate starting from s α (L \ {α}). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 47 / 116

119 Lecture 2 Background An important technical Lemma Example-Algorithm Given L biclosed, choose a simple root α L and iterate starting from s α (L \ {α}).for instance, in type A 5 L = {α 1, α 1 + α 2, α 1 + α 2 + α 3, α 5, α 1 + α 2 + α 3 + α 4 + α 5 }. s 1 {α 2, α 2 + α 3, α 5, α 2 + α 3 + α 4 + α 5 } s 1 s 2 {α 3, α 5, α 3 + α 4 + α 5 } s 1 s 2 s 3 {α 3, α 3 + α 4 } s 1 s 2 s 3 s 5 {α 4 } s 1 s 2 s 3 s 5 s 4 = w Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 47 / 116

120 Lecture 2 Background Another technical Lemma Definition Weyl vector ρ = 1/2 α + α. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 48 / 116

121 Lecture 2 Background Another technical Lemma Definition Weyl vector Proposition For w W ρ = 1/2 α + α. ρ w(ρ) = α. α N(w) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 48 / 116

122 Lecture 2 Background Another technical Lemma Proposition For w W ρ w(ρ) = α. α N(w) Proof. By induction on l(w). If w = s α, α Π, it is known that s α ( + \ {α}) + \ {α}, hence ρ s α (ρ) = ρ (ρ α) = α = β N(w) β. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 48 / 116

123 Lecture 2 Background Another technical Lemma Proposition For w W ρ w(ρ) = α. α N(w) Proof. Now assume w = s α w, α Π, l(w ) = l(w) 1 ρ s α w (ρ) = s α (ρ) + α s α w (ρ) = s α (ρ w (ρ)) + α = s α ( β) + α = β β N(w ) β N(w) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 48 / 116

124 Lecture 2 Background Representations Recall that a representation V of g is a Lie algebra homomorphism g gl(v ). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 49 / 116

125 Lecture 2 Background Representations Basic theorems For semisimple Lie algebras: 1 finite dimensional representations are completely reducible. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 49 / 116

126 Lecture 2 Background Representations Basic theorems For semisimple Lie algebras: 1 finite dimensional representations are completely reducible. 2 finite dimensional representations are in bijection with the set of dominant weights { } P + = λ h 2(λ, α) R (α, α) Z 0 for any simple root α. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 49 / 116

127 Lecture 2 Background Representations Abstract construction For λ P +, the attached irreducible representation V λ is the unique irreducible quotient of M λ = U(g) U(b) C λ, where C λ is the b-module with basis v λ and action x.v λ = 0, x b, h.v λ = λ(h)v λ and U(g) is the universal enveloping algebra of g. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 50 / 116

128 Lecture 2 Background Representations Abstract construction For λ P +, the attached irreducible representation V λ is the unique irreducible quotient of M λ = U(g) U(b) C λ, where C λ is the b-module with basis v λ and action x.v λ = 0, x b, h.v λ = λ(h)v λ and U(g) is the universal enveloping algebra of g. Any quotient of M λ, in particular V λ is a highest weight module, i.e. it is generated under U(g) by a vector v such that n +.v = 0, h.v = λ(h)v h h. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 50 / 116

129 Lecture 2 Background Representations Theorem (Weyl dimension formula) If ν P +, then dim V ν = β β + (ν + ρ, β) + (ρ, β) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 51 / 116

130 Lecture 2 Background Cohomology of Lie algebras Definition Let g be (any) Lie algebra and V be a representation of g. The Lie algebra cohomology H (g, V ) is the cohomology of the complex 0 C 0 d 0 C 1 d 1 C 2... C p dp C p+1... where C p = Hom( p g, V ) and (d p ω)(x 1... x p+1 ) = ( 1) i+j ω([x i, x j ] x 1... ˆx i... ˆx j... x p+1 ) i<j + i ( 1) i+1 x i.ω(x 1... ˆx i... x p+1 ) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 52 / 116

131 Lecture 2 Background Cohomology of Lie algebras General Facts H 0 (g, V ) = V g H 1 (g, V ) = Der(g, V )/InnDer(g, V ) H 2 (g, V ) = iso-classes of abelian extension of g by V Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 53 / 116

132 Lecture 2 Background Cohomology of Lie algebras General Facts H 0 (g, V ) = V g H 1 (g, V ) = Der(g, V )/InnDer(g, V ) H 2 (g, V ) = iso-classes of abelian extension of g by V Proposition Write H (g) for cohomology with trivial coefficients. If g is semisimple then H 1 (g, V ) = 0 (implies complete reducibility of reps). If g is semisimple then H 2 (g, V ) = 0 (implies Levi decomposiiton). If G is compact H DR (G) = H (g) = ( g) g Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 53 / 116

133 Lecture 2 Background Dual version, homology Complex where Λ p p p 1 Λ p 1... Λ 1 1 Λ0 0 Λ p = Λ p (g, V ) = p g V and p (x 1... x p v) = i<j [x i, x j ] x 1... ˆx i... ˆx j... x p v + i ( 1) i x 1... ˆx i... x p x i.v Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 54 / 116

134 Lecture 2 Background Computing homology Kostant s approach One can put on C = g V a Hilbert space structure, and then one defines a positive semidefinite operator L V on C by putting L V = dd + d d where d is the Hermitian adjoint of d. One then has a natural isomorphism KerL V = H (g, V ) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 55 / 116

135 Lecture 2 Background Computing homology Kostant s approach One can put on C = g V a Hilbert space structure, and then one defines a positive semidefinite operator L V on C by putting L V = dd + d d where d is the Hermitian adjoint of d. One then has a natural isomorphism KerL V = H (g, V ) Kostant has a nice spectral resolution for L V for a class of subalgebras which includes the parabolic subalgebras of semisimple Lie algebras (i.e., the subalgebras containing a Borel subalgebra). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 55 / 116

136 Lecture 2 Background Kostant theorem on u-cohomology Theorem Let p be a parabolic subalgebra of a semisimple Lie algebra g with Levi decomposition p = l u. Then, as l-modules, H p (u, V λ ) = V (w(λ + ρ) ρ) w W, l(w)=p where ρ is the Weyl vector and W is the set of minimal length right coset representatives for W l \W. Moreover, a representative for the highest weight vector is given by the decomposable vector x β1... x βp v wλ, where N(w) = {β 1,..., β p } and v w(λ) is a nonzero weight vector of weight w(λ). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 56 / 116

137 Lecture 3 Affine root systems and affine Weyl groups Affine root systems Affine root system Let F be the space of affine-linear functions on V = h R = R Z Q, where Q = α Π Zα is the coroot lattice. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 57 / 116

138 Lecture 3 Affine root systems and affine Weyl groups Affine root systems Affine root system Let F be the space of affine-linear functions on V = h R = R Z Q, where Q = α Π Zα is the coroot lattice. Endow F = V R with a symmetric bilinear form induced by (, ) on the linear part and extended by zero on the affine part. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 57 / 116

139 Lecture 3 Affine root systems and affine Weyl groups Affine root systems Affine root system Let F be the space of affine-linear functions on V = h R = R Z Q, where Q = α Π Zα is the coroot lattice. Endow F = V R with a symmetric bilinear form induced by (, ) on the linear part and extended by zero on the affine part. For α, j Z define a α,j (v) = α(v) + j and set = {a α,j α, j Z} Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 57 / 116

140 Lecture 3 Affine root systems and affine Weyl groups Affine root systems Affine root system Let F be the space of affine-linear functions on V = h R = R Z Q, where Q = α Π Zα is the coroot lattice. Endow F = V R with a symmetric bilinear form induced by (, ) on the linear part and extended by zero on the affine part. For α, j Z define a α,j (v) = α(v) + j and set = {a α,j α, j Z} Affine Weyl group For α, j Z let s α,j be the affine reflection around α(x) = j: s α,j (v) = v a α, j (v)α. Let Ŵ be the subgroup of Isom(V ) generated by {s α,j a α,j } Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 57 / 116

141 Lecture 3 Affine root systems and affine Weyl groups Affine Weyl groups Proposition Let t v be the translation by v. 1 Ŵ = W Q where Q is viewed inside Ŵ via α t α 2 Ŵ is a Coxeter group with generating set s 0 = s θ,1 = t θ s θ,0, s i = s αi,0, i = 1,..., n. Here θ = n i=1 c iα i is the highest root of. 3 A fundamental domain for the action of Ŵ on V is given by {v V α(v) 0 α +, θ(v) 1}. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 58 / 116

142 Lecture 3 Affine root systems and affine Weyl groups Alcoves Identifying V and V by means of (, ), we can also define an action of Ŵ on V ; Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 59 / 116

143 Lecture 3 Affine root systems and affine Weyl groups Alcoves Identifying V and V by means of (, ), we can also define an action of Ŵ on V ; then Ā 1 = {λ V (α, λ) 0 α +, (θ, λ) 1} is a fundamental domain for this action, called the fundamental alcove. We will refer to the alcoves as the Ŵ -translates of A 1 (i.e., A w = wa 1 ). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 59 / 116

144 Lecture 3 Affine root systems and affine Weyl groups Example Disclaimer Although I kept the Killing form since the beginning (and this will be important in the sequel), in many of the following pictures I use the more usual normalization (θ, θ) = 2). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 60 / 116

145 Lecture 3 Affine root systems and affine Weyl groups Positive systems The set + = {a α,j α, j > 0} {a α,0 α + } can be shown to be a set of positive roots in and the corresponding set of simple roots is Π = {α 0,..., α n }, where α 0 = a θ,1 and we identify α i with a αi,0, i = 1,..., n. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 61 / 116

146 Lecture 3 Affine root systems and affine Weyl groups Positive systems The set + = {a α,j α, j > 0} {a α,0 α + } can be shown to be a set of positive roots in and the corresponding set of simple roots is Π = {α 0,..., α n }, where α 0 = a θ,1 and we identify α i with a αi,0, i = 1,..., n. Ŵ acts on F (as functions on V ) and this action preserves and fixes δ, the constant function 1. If we set c 0 = 1 we have δ = n i=0 c iα i, so that we might write + = {α + nδ α, n > 0} + Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 61 / 116

147 Lecture 3 Affine root systems and affine Weyl groups Algebraic intepretation The elements of can be regarded as (part of the) roots of an infinite dimensional Lie algebra. Here is a sketch of its construction. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 62 / 116

148 Lecture 3 Affine root systems and affine Weyl groups Algebraic intepretation The elements of can be regarded as (part of the) roots of an infinite dimensional Lie algebra. Here is a sketch of its construction. Start with a fd-simple Lie algebra g and form the loop algebra g = g C[t, t 1 ], [x p(t), y q(t)] = [x, y] p(t)q(t). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 62 / 116

149 Lecture 3 Affine root systems and affine Weyl groups Algebraic intepretation The elements of can be regarded as (part of the) roots of an infinite dimensional Lie algebra. Here is a sketch of its construction. Start with a fd-simple Lie algebra g and form the loop algebra g = g C[t, t 1 ], [x p(t), y q(t)] = [x, y] p(t)q(t). One shows that H 2 ( g) = Cψ, ψ(x p(t), y q(t)) = (x, y)res t ( d p(t) dt One can therefore form an infinite dimensional Lie algebra ĝ = g C[t, t 1 ] CK Cd q(t)). where the (canonical) central extension is determined by ψ, K is central and d acts as the Euler operator t d dt. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 62 / 116

150 Lecture 3 Affine root systems and affine Weyl groups Algebraic intepretation Facts If h g is a Cartan subalgebra, the subalgebra of ĝ is ĥ = h CK Cd ĝ is maximal commutative adĝ-diagonalizable. ĝ has an invariant nondegenerate bilinear form If δ ĥ is defined by δ(h) = δ(d) = 0, δ(k) = 1, then δ generates the kernel of the restriction of the bilinear form to [ĝ, ĝ]. One has a root space decomposition w.r.t. ĥ; the root system is = re ±Nδ, re = {α + nδ α }. Note that re is our previous. the simple systems Π give rise to the extended Dynkin diagrams of g, i.e. ordinary Dynkin diagrams to which θ is added as an independent simple root. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 63 / 116

151 Lecture 3 ad-nilpotent ideals of Borel subalgebras ad-nilpotent of Borel subalgebras Let g be a simple Lie algebra and b be a Borel subalgebra. Let h be the Cartan component and + the positive system. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 64 / 116

152 Lecture 3 ad-nilpotent ideals of Borel subalgebras ad-nilpotent of Borel subalgebras Let g be a simple Lie algebra and b be a Borel subalgebra. Let h be the Cartan component and + the positive system. Definition Let i be an ideal of b contained in n. It consists of ad-nilpotent elements, so we ll call it an ad-nilpotent ideal and we denote by I the set of ad-nilpotent ideals. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 64 / 116

153 Lecture 3 ad-nilpotent ideals of Borel subalgebras ad-nilpotent of Borel subalgebras If i I, then i is h-stable, hence it admits a decomposition i = α Φ i g α where as usual g α = {x g [h, x] = α(h)x h h} and Φ i + is dual order ideal of the root poset. More precisely, recall the partial order on Q defined by α β = β α γ + Z 0 γ Then it is clear that i I α Φ i, β +, α + β + = α + β Φ i. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 64 / 116

154 Lecture 3 ad-nilpotent ideals of Borel subalgebras Encoding ad-nilpotent ideals Definition For i I set Φ 1 i = Φ i, Φ i i = (Φi 1 i + Φ i ) + and L i = k 1( Φ k i + kδ) +. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 65 / 116

155 Lecture 3 ad-nilpotent ideals of Borel subalgebras Encoding ad-nilpotent ideals Definition For i I set Φ 1 i = Φ i, Φ i i = (Φi 1 i + Φ i ) + and L i = k 1( Φ k i + kδ) +. Theorem 1 L i is biclosed in +, hence there exists a unique w i Ŵ such that L i = N(w i ). 2 Given w Ŵ, there exists i I such that w = w i if and only if w 1 (α) > 0, α Π (i.e., w i Ŵ + ); If w(α) < 0 for α Π, then there exists β + such that w(α) = β δ. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 65 / 116

156 Lecture 3 ad-nilpotent ideals of Borel subalgebras Abelian ideals Definition We denote by I ab the set of abelian ideals of b. Clearly I ab I. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 66 / 116

157 Lecture 3 ad-nilpotent ideals of Borel subalgebras Abelian ideals Definition We denote by I ab the set of abelian ideals of b. Clearly I ab I. Theorem The following statements are equivalent 1 i I ab 2 L i = Φ i + δ is biclosed, hence there exists a unique w i Ŵ such that L i = N(w i ). 3 w i (A 1 ) 2A 1 (i.e. w i Ŵ + 2 ). In particular, I ab = 2 rk g (Peterson s abelian ideal Theorem) Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 66 / 116

158 Lecture 3 ad-nilpotent ideals of Borel subalgebras Example Paolo Papi (Sapienza Universita di Roma) Ad-nilpotent ideals of Borel subalgebras 67 / 116

159 Lecture 3 ad-nilpotent ideals of Borel subalgebras Abelian ideals Proof (1) (2). If i is abelian, it is clear that if α, β Φ i, then (α + β) + 2δ / + ; now assume α + δ = ξ + η, α Φ i, ξ, η +. Then ξ = ξ 0 + δ, η +, so that α = ξ 0 + η; in particular ξ 0 and since Φ i is a dual order ideal, ξ 0 = α + η Φ i, as required. The converse is easy. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 68 / 116

160 Lecture 3 ad-nilpotent ideals of Borel subalgebras Abelian ideals Proof (1) (2). If i is abelian, it is clear that if α, β Φ i, then (α + β) + 2δ / + ; now assume α + δ = ξ + η, α Φ i, ξ, η +. Then ξ = ξ 0 + δ, η +, so that α = ξ 0 + η; in particular ξ 0 and since Φ i is a dual order ideal, ξ 0 = α + η Φ i, as required. The converse is easy. (2) (3). It is obvious that A w 2A 1, otherwise the hyperplane θ = 2 separates A 1 and A w, and θ + 2δ N(w), against the assumption. Conversely, if A w 2A 1, then each hyperplane which separates A 1 and A w intersects 2A 1. Now α + kδ N(w) iff α = k separates A 1, A w But for each x 2A 1 and for each α + we have 0 < (x, α) < (x, θ) < 2. Therefore if a bounding hyperplane α = k intersects 2A 1, we have 0 < k < 2. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 68 / 116

161 Lecture 3 ad-nilpotent ideals of Borel subalgebras Another encoding of ad-nilpotent ideals Recall that Ŵ = Q W. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 69 / 116

162 Lecture 3 ad-nilpotent ideals of Borel subalgebras Another encoding of ad-nilpotent ideals Recall that Ŵ = Q W. Proposition 1 The map φ : i v 1 i (τ i ), where w i = t τi v i is a bijection I D = {τ Q (τ, α) 1 α Π, (τ, θ) 2}. 2 The map φ restricts to a bijection I ab D ab = {τ Q (τ, α) {1, 0, 1, 2} α + }. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 69 / 116

163 Lecture 3 ad-nilpotent ideals of Borel subalgebras Another encoding of ad-nilpotent ideals Recall that Ŵ = Q W. Proposition 1 The map φ : i v 1 i (τ i ), where w i = t τi v i is a bijection I D = {τ Q (τ, α) 1 α Π, (τ, θ) 2}. 2 The map φ restricts to a bijection I ab D ab = {τ Q (τ, α) {1, 0, 1, 2} α + }. Corollary If h denotes the Coxeter number, and m i the exponents, then I = rk g i=1 (m i h). W Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 69 / 116

164 Lecture 3 ad-nilpotent ideals of Borel subalgebras Proof of the proposition Let t τi v i = w i, t τj v j = w j for some i and j in I. Assume v 1 i (τ i ) = v 1 j (τ j ). Since τ i, τ j C 1, which is a fundamental domain for W, we have τ i = τ j and v i v 1 j (τ i ) = τ i. Hence t τi v i (A 1 ) = t τi v i v 1 j v j (A 1 ) = v i v 1 j (t τi v j (A 1 )) = v i v 1 j (t τj v j (A 1 )) v i v 1 j (C 1 ). But t τi v i (A 1 ) C 1, hence v i v 1 j = 1. Thus F is injective. Next let σ D. We first see that there exists v W such that t v(σ) v(a 1 ) C 1 : simply take the unique v W such that v(σ + A 1 ) C 1. Now it is immediate that, since σ D, t v(σ) v also satisfies the second condition of part 2 in our characterization Theorem, hence t v(σ) v = w i for some i in I. It is obvious that F maps t v(σ) v to σ, thus F is surjective. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 70 / 116

165 Lecture 3 ad-nilpotent ideals of Borel subalgebras Proof of the Corollary Let X = {x V (x, α i ) 1 for each i {1,..., n} and (x, θ) 2} = t ρ w 0 (A h+1 ). where ρ = ω ω n. One can show that there exists w Ŵ such that X = w(a h+1 ). Such a w gives a bijection from A h+1 Q D = X Q. If i I and w i = t τi v i, with τ i Q and v i W, then we obtain that w 1 v 1 i (τ i ) belongs to A h+1 Q and i w 1 v 1 i (τ i ), I A h+1 Q is a bijection. Since elements in A h+1 Q are a natural set of representatives of the W -orbits of Q /(h + 1)Q, we are done. The combinatorial enumeration is due to Haiman. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 71 / 116

166 Lecture 3 ad-nilpotent ideals of Borel subalgebras ρ-points Definitions Take as invariant form on h the Killing form 1 The ρ-points are the Ŵ -orbit of 2ρ. 2 The weight of i I ab is i = g α i α. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 72 / 116

167 Lecture 3 ad-nilpotent ideals of Borel subalgebras ρ-points Definitions Take as invariant form on h the Killing form 1 The ρ-points are the Ŵ -orbit of 2ρ. 2 The weight of i I ab is i = g α i α. Then 2ρ A 1, and 2(λ w + ρ) = w(2ρ). Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 72 / 116

168 Lecture 3 ad-nilpotent ideals of Borel subalgebras ρ-points Definitions Take as invariant form on h the Killing form 1 The ρ-points are the Ŵ -orbit of 2ρ. 2 The weight of i I ab is i = g α i α. Then 2ρ A 1, and 2(λ w + ρ) = w(2ρ). Introducing a linear version of Ŵ as a subgroup of O(ĥ ), one can prove that λ w i = i. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 72 / 116

169 Lecture 3 ad-nilpotent ideals of Borel subalgebras ρ-points Recall that an antichain A in a poset P is a set of mutually non-comparable elements Proposition The following sets are in bijection with I ab : 1 the set of abelian dual order ideals in + ; 2 the set Ŵ the set of alcoves in 2A 1 ; 4 the set of ρ-points in 2A 1 ; in Ŵ (the minuscule elements); 5 the set of weights of abelian ideals. 6 the set D ab = {η Q η(α) { 2 1, 0, 1} α + }; 7 the set of antichains A + such that for any α, β A we have α + β θ. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 72 / 116

170 Lecture 3 ad-nilpotent ideals of Borel subalgebras Afterwords Lemma (Kostant) Let i 1, i 2 I be such that i 1 = i 2. Then i 1 = i 2. Proof. Set Φ i = Φ ii, Φ := Φ 1 Φ 2. Assume by contradiction that Φ 1 Φ 2. Then since Φ 1 = Φ 2 both Φ 1 Φ and Φ 2 Φ are nonempty. Pick ϕ i Φ i Φ (i = 1, 2). We must have (ϕ 1 ϕ 2 ) 0. Otherwise ϕ 1 ϕ 2 would be a root which can be assumed positive by possibly interchanging the indices 1 and 2. By the ideal property Φ i + + Φ i we then have ϕ 1 = ϕ 2 + (ϕ 1 ϕ 2 ) Φ 2, a contradiction. Thus (ϕ 1 ϕ 2 ) 0. Hence since Φ 1 Φ = Φ 2 Φ we obtain and so Φ = Φ 1 = Φ 2. 0 Φ i Φ 2 = ( Φ 1 Φ Φ 2 Φ ) 0 Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 73 / 116

171 Lecture 4 Panyushev s theory of rootlets Panyushev s theory of rootlets For α + l define I ab (α) = {i I w 1 i ( θ + 2δ) = α}, Ŵ α = s β β Π, β α Ŵ, W α = s β β Π, β α Ŵ. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 74 / 116

172 Lecture 4 Panyushev s theory of rootlets Panyushev s theory of rootlets For α + l define I ab (α) = {i I w 1 i ( θ + 2δ) = α}, Ŵ α = s β β Π, β α Ŵ, W α = s β β Π, β α Ŵ. Theorem 1 The set I ab of nonzero abelian ideals of b decomposes as I ab = I ab (α) α + l 2 There an explicit poset isomorphism M : I ab (α) Ŵα/W α. In particular I ab (α) has minimum and maximum. 3 M gives rise to a natural bijection maximal abelian ideals of b and Π l. Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 74 / 116

173 Lecture 4 Panyushev s theory of rootlets Paolo Papi (Sapienza Università di Roma) Ad-nilpotent ideals of Borel subalgebras 75 / 116