Matrices with Integer Eigenvalues 4 1 1 4 1 0 0 1 −1 −1 1 0 −1 0 1 −1 3 0 0 0 3 0 0 0 3 Ron Adin Bar-Ilan University radin@math.biu.ac.il 1 A Recent Monthly Paper Almost All Integer Matrices Have No Integer Eigenvalues, G. Martin and E. B. Wong, Amer. Math. Monthly 116 (2009), 588-597. We want to deal with (specific) exceptions. 2 Highlights Certain matrices with integer entries have, conjecturally, integer eigenvalues only. Partial results are known. The multiplicity of the zero eigenvalue has an algebraic interpretation. 3 A Signed Graph: Vertices [0, r ] := {0,1,… , r} For k ≥ 0 and 0 ≤ a, b ≤ k + 1 let Ω k ( a, b) := set of all pairs (U , V ) s.t. U , V ⊆ [0, k ], U = a, V = b. Weight: wt (U , V ) := ∑ u + ∑ v u∈U v∈V Ω k ( a, b, w) := set of all (U , V ) ∈ Ω k ( a, b) with weight w. 4 A Signed Graph: Vertices (cont.) Example: a = b = 2, w = 4. k = 0,1: Ω k (2, 2, 4) = ∅ k = 2 : Ω k (2, 2, 4) = {(12, 01), (02, 02), (01,12)} k = 3 : Ω k (2, 2, 4) = {(03, 01), (12, 01), (02, 02), (01,12), (01, 03)} Note: wmin a b a b = + , wmax = k (a + b) − − 2 2 2 2 5 A Signed Graph: Edges For (U , V ), (U , V ) ∈ Ω k (a, b, w) and u ∈ U , v ∈ V , z ∈ », write (U , V ) ∼ (U , V ) if ( u ,v , z ) 1. U = (U \ {u}) ∪ {u + z} 2. V = (V \ {v}) ∪ {v − z} 3. u + v ≤ k 6 A Signed Graph: Edges For (U , V ), (U , V ) ∈ Ω k (a, b, w) and u ∈ U , v ∈ V , z ∈ », write (U , V ) ∼ (U , V ) if ( u ,v , z ) 1. U = (U \ {u}) ∪ {u + z} 2. V = (V \ {v}) ∪ {v − z} 3. u + v ≤ k U u v V u+v U u+z v − z V 7 A Signed Graph Gk ( a, b, w) = graph with vertex set Ω k ( a, b, w) and edges corresponding to the various (U , V ) ∼ (U , V ). ( u ,v , z ) Loops and multiple edges may occur. Attach signs to edges: 8 Signs For u = (u1 ,… , ua ), u = (u1 ,… , ua ) define: ε (u , u ) = 0 if u or u has repeated elements; ε (u , u ) = sign(σ ) sign(τ ) if uσ (1) < … < uσ ( a ) , uτ (1) < … < uτ ( a ) . For U = {u1 , … , ua }, U = {u1 ,… , ui + z ,… , ua } where u1 < … < ua , define ε (U , U ) := ε ((u1 ,… , ua ), (u1 ,… , ui + z ,… , ua )). ε ((U , V ), (U , V )) := ε (U , U ) ⋅ ε (V , V ) 9 Signed Adjacency Matrix Tk ( a, b, w) = the (signed) adjacency matrix of the graph Gk (a, b, w). Note: Diagonal elements are nonnegative integers. Off-diagonal elements are 0, 1 or −1. 10 Signed Adjacency Matrix Example: k = 3, a = 2, b = 3, w = 8 4 1 1 4 Tk ( a, b, w) = 1 0 0 1 −1 −1 1 0 −1 0 1 −1 3 0 0 0 3 0 0 0 3 11 Conjectures Conjecture: [Hanlon, ’92] All the eigenvalues of Tk (a, b, w) are nonnegative integers. 12 Conjectures (cont.) Let M k ( x, y, λ ) := ∑ m ( a , b, r ) x k a yλ b r a ,b , r where mk (a, b, r ) = multiplicity of r as an e.v. of Tk ( a, b) = ⊕ Tk ( a, b, w). w Conjecture: [Hanlon, ’92] M k ( x, y , λ ) = ∏ (1 + x + y + λ i +1 xy ) k Still open (in general)! i =0 13 Background Macdonald’s root system conjecture 14 Background Macdonald’s root system conjecture Hanlon’s property M conjecture: H * ( L ⊗ »[t ] /(t k +1 )) ≅ H * ( L ) ⊗ ( k +1) where the Lie algebra L is either * semisimple, or * upper triangular (nilpotent), or * Heisenberg (nilpotent) 15 Background Macdonald’s root system conjecture Hanlon’s property M conjecture: H * ( L ⊗ »[t ] /(t )) ≅ H * ( L ) where the Lie algebra L is either k +1 ⊗ ( k +1) * semisimple, or * upper triangular (nilpotent), or * Heisenberg (nilpotent) 16 Background Macdonald’s root system conjecture TRUE [Cherednik, ’95] Hanlon’s property M conjecture: H * ( L ⊗ »[t ] /(t )) ≅ H * ( L ) where the Lie algebra L is either k +1 ⊗ ( k +1) * semisimple, or * upper triangular (nilpotent), or * Heisenberg (nilpotent) 17 Background Macdonald’s root system conjecture TRUE [Cherednik, ’95] Hanlon’s property M conjecture: H * ( L ⊗ »[t ] /(t )) ≅ H * ( L ) where the Lie algebra L is either k +1 * semisimple, or * upper triangular (nilpotent), or * Heisenberg (nilpotent) ⊗ ( k +1) TRUE [FGT, ’08] 18 Background Macdonald’s root system conjecture TRUE [Cherednik, ’95] Hanlon’s property M conjecture: H * ( L ⊗ »[t ] /(t )) ≅ H * ( L ) where the Lie algebra L is either k +1 * semisimple, or * upper triangular (nilpotent), or * Heisenberg (nilpotent) ⊗ ( k +1) TRUE [FGT, ’08] FALSE [Kumar, ’99] 19 Background Macdonald’s root system conjecture TRUE [Cherednik, ’95] Hanlon’s property M conjecture: H * ( L ⊗ »[t ] /(t )) ≅ H * ( L ) where the Lie algebra L is either k +1 * semisimple, or * upper triangular (nilpotent), or * Heisenberg (nilpotent) ⊗ ( k +1) TRUE [FGT, ’08] FALSE [Kumar, ’99] ? 20 Background (cont.) 0 * * For the 3-dim Heisenberg L3 = 0 0 * , with basis{e, f , x}, 0 0 0 the Laplacian ∂∂* + ∂ *∂ for L3 ⊗ »[t ] /(t k +1 ) is ⊕ Tk ( a, b, w) (in a suitable basis). a ,b , w The property M conjecture is then equivalent to (an extension of) M k ( x, y, 0) = (1 + x + y ) k +1 21 Partial Results [Hanlon, ’92] Explicit eigenvalues in the stable case: * a≤b a b * w ≤ 2 + 2 a b * k ≥ (a − 1) + (b − 1) + w − − 2 2 22 Partial Results (cont.) Theorem: [Hanlon, ’92] 1:1 In the stable case, Ω k (a, b) ↔ pairs (λ , µ ) of partitions with λ + µ = w − wmin . The eigenvalues of Tk (a, b) are λ − µ + ab. The proof uses Schur functions in two sets of variables. 23 Partial Results (cont.) [A.-Athanasiadis, ’96] 1. a = 1, b = 2, ∀k : The nonzero eigenvalues of ⊕ w Tk (a, b, w) are λ = 1,… , k + 1, each with multiplicity k (and explicit distribution over w). 2. a = 1, ∀b, k : The multiplicity of the zero e.v.: k +1 k mk (1, b, 0) = = ( k + 1) 1 b k − b b 24 Partial Results (cont.) [Hanlon-Wachs, ’02] Extend result 2 above to ∀a, b, k : k +1 mk ( a, b, 0) = a b k +1− a − b 25 Partial Results (cont.) [Kuflik, ’06] Distribution over w (∀a, b, k ) : mk ( a, b, w, 0) = coefficient of q w− wmin in where k +1 a b k + 1 − a − b q [ m]!q := [1]q ⋅ [2]q [ m]q , [ m]q := 1 + q + … + q m −1. 26 תודה על ההקשבה! 27