Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK and random polymers in antisymmetric environment PhD student: Elia Bisi Supervisor: Nikos Zygouras Department of Statistics – University of Warwick 8 December 2015 1 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Summary 1 Geometric RSK 2 Inverse gamma polymers 3 Antisymmetric gRSK 4 Inverse gamma antisymmetric polymers 2 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK For 1 ≤ i ≤ m, 1 ≤ j ≤ n, define the maps acting on a matrix m×n X ∈ R>0 : ai,j replaces xi,j with −1 1 1 1 + ; xi,j xi+1,j xi,j+1 bi,j replaces xi,j with −1 1 1 1 (xi−1,j + xi,j−1 ) + ; xi,j xi+1,j xi,j+1 σi,j := b(i−j)∨0+1,(j−i)∨0+1 ◦ · · · ◦ bi−1,j−1 ◦ bi,j ; %i,j := σi,j ◦ ai,j ; ri := %i,n ◦ · · · ◦ %i,2 ◦ %i,1 . The map g := rm ◦ · · · ◦ r1 is the geometric RSK (gRSK). 3 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK For 1 ≤ i ≤ m, 1 ≤ j ≤ n, define the maps acting on a matrix m×n X ∈ R>0 : ai,j replaces xi,j with −1 1 1 1 + ; xi,j xi+1,j xi,j+1 bi,j replaces xi,j with −1 1 1 1 (xi−1,j + xi,j−1 ) + ; xi,j xi+1,j xi,j+1 σi,j := b(i−j)∨0+1,(j−i)∨0+1 ◦ · · · ◦ bi−1,j−1 ◦ bi,j ; %i,j := σi,j ◦ ai,j ; ri := %i,n ◦ · · · ◦ %i,2 ◦ %i,1 . The map g := rm ◦ · · · ◦ r1 is the geometric RSK (gRSK). 3 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK For 1 ≤ i ≤ m, 1 ≤ j ≤ n, define the maps acting on a matrix m×n X ∈ R>0 : ai,j replaces xi,j with −1 1 1 1 + ; xi,j xi+1,j xi,j+1 bi,j replaces xi,j with −1 1 1 1 (xi−1,j + xi,j−1 ) + ; xi,j xi+1,j xi,j+1 σi,j := b(i−j)∨0+1,(j−i)∨0+1 ◦ · · · ◦ bi−1,j−1 ◦ bi,j ; %i,j := σi,j ◦ ai,j ; ri := %i,n ◦ · · · ◦ %i,2 ◦ %i,1 . The map g := rm ◦ · · · ◦ r1 is the geometric RSK (gRSK). 3 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK For 1 ≤ i ≤ m, 1 ≤ j ≤ n, define the maps acting on a matrix m×n X ∈ R>0 : ai,j replaces xi,j with −1 1 1 1 + ; xi,j xi+1,j xi,j+1 bi,j replaces xi,j with −1 1 1 1 (xi−1,j + xi,j−1 ) + ; xi,j xi+1,j xi,j+1 σi,j := b(i−j)∨0+1,(j−i)∨0+1 ◦ · · · ◦ bi−1,j−1 ◦ bi,j ; %i,j := σi,j ◦ ai,j ; ri := %i,n ◦ · · · ◦ %i,2 ◦ %i,1 . The map g := rm ◦ · · · ◦ r1 is the geometric RSK (gRSK). 3 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK For 1 ≤ i ≤ m, 1 ≤ j ≤ n, define the maps acting on a matrix m×n X ∈ R>0 : ai,j replaces xi,j with −1 1 1 1 + ; xi,j xi+1,j xi,j+1 bi,j replaces xi,j with −1 1 1 1 (xi−1,j + xi,j−1 ) + ; xi,j xi+1,j xi,j+1 σi,j := b(i−j)∨0+1,(j−i)∨0+1 ◦ · · · ◦ bi−1,j−1 ◦ bi,j ; %i,j := σi,j ◦ ai,j ; ri := %i,n ◦ · · · ◦ %i,2 ◦ %i,1 . The map g := rm ◦ · · · ◦ r1 is the geometric RSK (gRSK). 3 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric RSK For 1 ≤ i ≤ m, 1 ≤ j ≤ n, define the maps acting on a matrix m×n X ∈ R>0 : ai,j replaces xi,j with −1 1 1 1 + ; xi,j xi+1,j xi,j+1 bi,j replaces xi,j with −1 1 1 1 (xi−1,j + xi,j−1 ) + ; xi,j xi+1,j xi,j+1 σi,j := b(i−j)∨0+1,(j−i)∨0+1 ◦ · · · ◦ bi−1,j−1 ◦ bi,j ; %i,j := σi,j ◦ ai,j ; ri := %i,n ◦ · · · ◦ %i,2 ◦ %i,1 . The map g := rm ◦ · · · ◦ r1 is the geometric RSK (gRSK). 3 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Example of gRSK For m = n = 2, the image of a matrix w w1,2 W = 1,1 w2,1 w2,2 under gRSK is " T = g (W ) = 1 w1,2 + 1 −1 w2,1 w1,1 w2,1 # w1,1 w1,2 . w1,1 w2,2 (w1,2 + w2,1 ) 4 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Equivalent notation and symmetry property Match a matrix X = (xi,j , 1 ≤ i ≤ m, 1 ≤ j ≤ n) to the pair of triangles/trapezoids (U, V ): U := (xi,j , 1 ≤ i ≤ m, 1 ≤ j ≤ i + n − m) , V := (xi,j , 1 ≤ i ≤ m, i + n − m ≤ j ≤ n) . E.g., for m = 3, n = 4, v2,2 v1,1 u3,3 u4,3 = v3,3 . u3,2 u4,2 = v3,2 v2,1 X = (U, V ) = u2,2 u1,1 u2,1 u3,1 u4,1 = v3,1 The highlighted diagonal is the common shape of U and V . Symmetry property of the gRSK If g (W ) = (U, V ) then g (W > ) = (V , U). In particular, the image of a symmetric matrix under gRSK is symmetric. 5 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Equivalent notation and symmetry property Match a matrix X = (xi,j , 1 ≤ i ≤ m, 1 ≤ j ≤ n) to the pair of triangles/trapezoids (U, V ): U := (xi,j , 1 ≤ i ≤ m, 1 ≤ j ≤ i + n − m) , V := (xi,j , 1 ≤ i ≤ m, i + n − m ≤ j ≤ n) . E.g., for m = 3, n = 4, v2,2 v1,1 u3,3 u4,3 = v3,3 . u3,2 u4,2 = v3,2 v2,1 X = (U, V ) = u2,2 u1,1 u2,1 u3,1 u4,1 = v3,1 The highlighted diagonal is the common shape of U and V . Symmetry property of the gRSK If g (W ) = (U, V ) then g (W > ) = (V , U). In particular, the image of a symmetric matrix under gRSK is symmetric. 5 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers gRSK and random polymers (m, n) gRSK W −−−→ T ; Πm,n : directed paths (1, 1) → (m, n); X Y tm,n = wi,j . π∈Πm,n (i,j)∈π (1, 1) We may think of tm,n as the partition function of a polymer model in a random environment given by weights wi,j ’s. 6 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers gRSK and random polymers (m, n) gRSK W −−−→ T ; Πm,n : directed paths (1, 1) → (m, n); X Y tm,n = wi,j . π∈Πm,n (i,j)∈π (1, 1) We may think of tm,n as the partition function of a polymer model in a random environment given by weights wi,j ’s. 6 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma weights Assume: m = n and α = (α1 , . . . , αn ), α0 = (α10 , . . . , αn0 ) ∈ Rn>0 ; wi,j ’s are independent; −1 wi,j ∼ Γ(αj + αi0 , 1). n×n Then the joint law of W on R>0 is Y X Y dwi,j 1 1 −αj −α0i να,α0 (dw ) = wi,j exp − , Zα,α0 wi,j wi,j i,j where Zα,α0 = Q i,j i,j i,j Γ(αj + αi0 ). 7 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma weights Assume: m = n and α = (α1 , . . . , αn ), α0 = (α10 , . . . , αn0 ) ∈ Rn>0 ; wi,j ’s are independent; −1 wi,j ∼ Γ(αj + αi0 , 1). n×n Then the joint law of W on R>0 is Y X Y dwi,j 1 1 −αj −α0i να,α0 (dw ) = wi,j exp − , Zα,α0 wi,j wi,j i,j where Zα,α0 = Q i,j i,j i,j Γ(αj + αi0 ). 7 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Properties of gRSK n(n+1)/2 For a triangle V ∈ R>0 and 1 ≤ i ≤ n, define Qi j=1 vi,j (type V )i := Qi−1 j=1 vi−1,j , E(V ) := X 1≤j≤i≤n vi−1,j + vi+1,j+1 . vi,j Properties of gRSK Let W ∈ Rn×n >0 and T = (U, V ) := g (W ). Then: Q Q 1 i wi,j = (type U)j for all i and j wi,j = (type V )i for all j; P −1 −1 2 i,j wi,j = t1,1 + E(U) + E(V ); 3 The gRSK is volume preserving in logarithmic variables. 8 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Properties of gRSK n(n+1)/2 For a triangle V ∈ R>0 and 1 ≤ i ≤ n, define Qi j=1 vi,j (type V )i := Qi−1 j=1 vi−1,j , E(V ) := X 1≤j≤i≤n vi−1,j + vi+1,j+1 . vi,j Properties of gRSK Let W ∈ Rn×n >0 and T = (U, V ) := g (W ). Then: Q Q 1 i wi,j = (type U)j for all i and j wi,j = (type V )i for all j; P −1 −1 2 i,j wi,j = t1,1 + E(U) + E(V ); 3 The gRSK is volume preserving in logarithmic variables. 8 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma polymer partition function Theorem (Corwin, O’Connell, Seppäläinen, Zygouras 2014) The Laplace transform of tn,n under να,α0 is E e −θtn,n = 1 Zα,α0 Z e Rn>0 −θλ1 − λ1 n i=1 Here, Ψnα and Ψnα0 are Whittaker functions: Z Y n Ψα (λ) := (type V )−α e −E(V ) n(n−1)/2 R>0 n Y dλi Ψnα (λ)Ψnα0 (λ) λi . dvi,j , vi,j 1≤j≤i≤n−1 where the integral is over all triangles V of height n with the same fixed shape λ = (λ1 , . . . , λn ) = (vn,1 , . . . , vn,n ). 9 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma polymer partition function Theorem (Corwin, O’Connell, Seppäläinen, Zygouras 2014) The Laplace transform of tn,n under να,α0 is E e −θtn,n = 1 Zα,α0 Z e Rn>0 −θλ1 − λ1 n i=1 Here, Ψnα and Ψnα0 are Whittaker functions: Z Y n Ψα (λ) := (type V )−α e −E(V ) n(n−1)/2 R>0 n Y dλi Ψnα (λ)Ψnα0 (λ) λi . dvi,j , vi,j 1≤j≤i≤n−1 where the integral is over all triangles V of height n with the same fixed shape λ = (λ1 , . . . , λn ) = (vn,1 , . . . , vn,n ). 9 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers The antisymmetric case We aim to: study the restriction of gRSK to antisymmetric matrices, i.e. n × n matrices W s.t. wi,j = wj ∗ ,i ∗ for all i, j, where i ∗ := n − i + 1; deduce analogous properties as in the “independent” case, in order to obtain an integral formula for the polymer partition function in an inverse gamma antisymmetric environment. Example: " w w 1,2 2,1 w1,1 w1,2 g w1,2 +w2,1 W = 7→ T = − w2,1 w1,1 w1,1 w2,1 # w1,1 w1,2 . 2 (w w1,1 1,2 + w2,1 ) 10 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers The antisymmetric case We aim to: study the restriction of gRSK to antisymmetric matrices, i.e. n × n matrices W s.t. wi,j = wj ∗ ,i ∗ for all i, j, where i ∗ := n − i + 1; deduce analogous properties as in the “independent” case, in order to obtain an integral formula for the polymer partition function in an inverse gamma antisymmetric environment. Example: " w w 1,2 2,1 w1,1 w1,2 g w1,2 +w2,1 W = 7→ T = − w2,1 w1,1 w1,1 w2,1 # w1,1 w1,2 . 2 (w w1,1 1,2 + w2,1 ) 10 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric Schützenberger involution (1) Define the following maps: W 7→ W c reverses the column order of W ; W 7→ W r reverses the row order of a W ; T 7→ T s is such that W cr - cr W g g ? s - ? Ts T We call T s the geometric Schützenberger involution of T . 11 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric Schützenberger involution (1) Define the following maps: W 7→ W c reverses the column order of W ; W 7→ W r reverses the row order of a W ; T 7→ T s is such that W cr - cr W g g ? s - ? Ts T We call T s the geometric Schützenberger involution of T . 11 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric Schützenberger involution (2) For 1 ≤ j ≤ n, define: sj := σm,j ◦ σm,j−1 ◦ · · · ◦ σm,1 , qj := s1 ◦ s2 ◦ · · · ◦ sj . Then > ◦ qn−1 (T ) . T s = (U s , V s ) = qm−1 (qn−1 (T )> )> =: qm−1 The map U 7→ U s is a birational involution that preserves the shape of triangles/trapezoids. 12 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Geometric Schützenberger involution (2) For 1 ≤ j ≤ n, define: sj := σm,j ◦ σm,j−1 ◦ · · · ◦ σm,1 , qj := s1 ◦ s2 ◦ · · · ◦ sj . Then > ◦ qn−1 (T ) . T s = (U s , V s ) = qm−1 (qn−1 (T )> )> =: qm−1 The map U 7→ U s is a birational involution that preserves the shape of triangles/trapezoids. 12 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers gRSK bijection in the antisymmetric case Lemma Let W be an n × n antisymmetric matrix. Then: 1 W cr = W > . 2 g (W )s = g (W )> . Namely, if g (W ) = (U, V ), then U = V s . n(n+1)/2 So the maps R>0 n(n+1)/2 → R>0 (wi,j , 1 ≤ i ≤ j ∗ ≤ n) 7→ V = (ti,j , 1 ≤ i ≤ j ≤ n) , (wi,j , 1 ≤ i ≤ j ∗ ≤ n) 7→ U = (ti,j , 1 ≤ j ≤ i ≤ n) are bijections. 13 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers gRSK bijection in the antisymmetric case Lemma Let W be an n × n antisymmetric matrix. Then: 1 W cr = W > . 2 g (W )s = g (W )> . Namely, if g (W ) = (U, V ), then U = V s . n(n+1)/2 So the maps R>0 n(n+1)/2 → R>0 (wi,j , 1 ≤ i ≤ j ∗ ≤ n) 7→ V = (ti,j , 1 ≤ i ≤ j ≤ n) , (wi,j , 1 ≤ i ≤ j ∗ ≤ n) 7→ U = (ti,j , 1 ≤ j ≤ i ≤ n) are bijections. 13 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Volume preserving property in the antisymmetric case Conjecture Let W ∈ Rn×n >0 be antisymmetric and T = (U, V ) := g (W ). Then the Jacobian determinants of the maps (log wi,j , 1 ≤ i ≤ j ∗ ≤ n) 7→ log V = (log ti,j , 1 ≤ i ≤ j ≤ n) , (log wi,j , 1 ≤ i ≤ j ∗ ≤ n) 7→ log U = (log ti,j , 1 ≤ j ≤ i ≤ n) are both ±1. 14 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inductive step 1 Lemma Let W ∈ Rn×n >0 be antisymmetric. Then g (W ) = rn where wn,1 Ss , · · · wn,n > 1,n−1 g (W2,n ) S = rn . wn,2 · · · wn,n 15 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inductive step 2 Lemma Let W ∈ Rn×n Un−1 are the lower >0 be antisymmetric. If U and 1,n−1 triangular parts of g (W ) and g W2,n respectively, then U = −1 > (sn−1 ) −1 sn−1 ◦ rn> U n−1 ◦ rn wn,1 ··· wn,2 .. . . wn,n wn,n Question Is the map (Un−1 , wn,1 , . . . , wn,n ) 7→ U volume preserving? 16 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inductive step 2 Lemma Let W ∈ Rn×n Un−1 are the lower >0 be antisymmetric. If U and 1,n−1 triangular parts of g (W ) and g W2,n respectively, then U = −1 > (sn−1 ) −1 sn−1 ◦ rn> U n−1 ◦ rn wn,1 ··· wn,2 .. . . wn,n wn,n Question Is the map (Un−1 , wn,1 , . . . , wn,n ) 7→ U volume preserving? 16 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma antisymmetric weights Assume: α = (α1 , . . . , αn ) ∈ Rn>0 ; wi,j ’s are independent for 1 ≤ i ≤ j ∗ ≤ n; −1 wi,j ∼ Γ(αi + αj ∗ , 1) for 1 ≤ i < j ∗ ≤ n; −1 wi,i ∗ ∼ Γ(αi , 1/2) for 1 ≤ i ≤ n; n(n+1)/2 Then the joint law of W on R>0 is # P " 1 Y −αi −αj ∗ Y −αi − i<j ∗ να (dw ) := w · wi,i ∗ e Zα i<j ∗ i,j i 1 wi,j − 1 i 2w ∗ i,i P · Y dwi,j , wi,j ∗ i≤j where P Zα := 2 i αi · Y i Γ(αi ) · Y Γ(αi + αj ∗ ) . i<j ∗ 17 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma antisymmetric weights Assume: α = (α1 , . . . , αn ) ∈ Rn>0 ; wi,j ’s are independent for 1 ≤ i ≤ j ∗ ≤ n; −1 wi,j ∼ Γ(αi + αj ∗ , 1) for 1 ≤ i < j ∗ ≤ n; −1 wi,i ∗ ∼ Γ(αi , 1/2) for 1 ≤ i ≤ n; n(n+1)/2 Then the joint law of W on R>0 is # P " 1 Y −αi −αj ∗ Y −αi − i<j ∗ να (dw ) := w · wi,i ∗ e Zα i<j ∗ i,j i 1 wi,j − 1 i 2w ∗ i,i P · Y dwi,j , wi,j ∗ i≤j where P Zα := 2 i αi · Y i Γ(αi ) · Y Γ(αi + αj ∗ ) . i<j ∗ 17 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma antisymmetric polymer partition function Properties of antisymmetric gRSK Let W ∈ Rn×n >0 be an antisymmetric matrix and T = (U, V ) := g (W ). Then Q −αi −αj ∗ Q −αi −α ; 1 · i wi,i ∗ = (type V ) i<j ∗ wi,j P P −1 −1 = (2v )−1 + E(V ); 2 n,n i<j ∗ wi,j + i (2wi,i ∗ ) 3 The antisymmetric gRSK is volume preserving in logarithmic variables (conjecture). Theorem The Laplace transform of tn,n under να is 1 E e −θtn,n = Zα Z e Rn>0 −θλ1 − 2λ1 n Ψnα (λ) n Y dλi i=1 λi . 18 / 18 Geometric RSK Inverse gamma polymers Antisymmetric gRSK Inverse gamma antisymmetric polymers Inverse gamma antisymmetric polymer partition function Properties of antisymmetric gRSK Let W ∈ Rn×n >0 be an antisymmetric matrix and T = (U, V ) := g (W ). Then Q −αi −αj ∗ Q −αi −α ; 1 · i wi,i ∗ = (type V ) i<j ∗ wi,j P P −1 −1 = (2v )−1 + E(V ); 2 n,n i<j ∗ wi,j + i (2wi,i ∗ ) 3 The antisymmetric gRSK is volume preserving in logarithmic variables (conjecture). Theorem The Laplace transform of tn,n under να is 1 E e −θtn,n = Zα Z e Rn>0 −θλ1 − 2λ1 n Ψnα (λ) n Y dλi i=1 λi . 18 / 18