Directed polymers RSK correspondence Geometric RSK Conclusions Exact solvability in directed random polymer models PhD student: Elia Bisi Supervisor: Nikos Zygouras Department of Statistics - University of Warwick 25 June 2015 1 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Summary 1 Directed polymers 2 RSK correspondence 3 Geometric RSK 2 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Polymers Polymer: large molecule consisting of smaller molecules called monomers and tied together by chemical bonds. Linear polymer: linear structure without multiple cross connections. Examples of linear polymers: DNA, RNA, proteins. Figure: A DNA helix. 3 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Directed random polymers in a random potential Model: set of allowed paths Xn = x = (i, xi )ni=0 ∈ (N×Zd )n+1 : x0 = 0, xi−1 ∼ xi ∀i ; random potential: i.i.d. field on (Ω, F, P, E) {V (i, x) : i ∈ N, x ∈ Zd } ; Hamiltonian function Hnβ (x) = −β n X V (i, xi ) , x ∈ Xn . i=1 4 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Directed random polymers in a random potential Model: set of allowed paths Xn = x = (i, xi )ni=0 ∈ (N×Zd )n+1 : x0 = 0, xi−1 ∼ xi ∀i ; random potential: i.i.d. field on (Ω, F, P, E) {V (i, x) : i ∈ N, x ∈ Zd } ; Hamiltonian function Hnβ (x) = −β n X V (i, xi ) , x ∈ Xn . i=1 4 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Directed random polymers in a random potential Model: set of allowed paths Xn = x = (i, xi )ni=0 ∈ (N×Zd )n+1 : x0 = 0, xi−1 ∼ xi ∀i ; random potential: i.i.d. field on (Ω, F, P, E) {V (i, x) : i ∈ N, x ∈ Zd } ; Hamiltonian function Hnβ (x) = −β n X V (i, xi ) , x ∈ Xn . i=1 4 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Polymer measure in 1 + d dimensions Polymer measure on Xn : Pnβ (x) = 1 Znβ β e −Hn (x) Pn (x) , Pn : law of the n-step simple random walk on Zd ; Znβ : partition sum n X XY β β e βV (i,xi ) Znβ = En e −Hn = e −Hn (x) Pn (x) = . 2d x∈Xn x∈Xn i=1 5 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Infinite and zero temperature limits β = 0 (or constant potential): simple random walk on Zd . β → ∞: Znβ X n 1 X = exp β V (i, xi ) (2d)n i=1 x∈Xn n X 1 exp β max V (i, xi ) . x∈Xn (2d)n i=1 connection to directed last passage percolation: n X 1 log Znβ = max V (i, xi ) =: Tn β→∞ β x∈Xn lim i=1 6 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Infinite and zero temperature limits β = 0 (or constant potential): simple random walk on Zd . β → ∞: Znβ X n 1 X = exp β V (i, xi ) (2d)n i=1 x∈Xn n X 1 exp β max V (i, xi ) . x∈Xn (2d)n i=1 connection to directed last passage percolation: n X 1 log Znβ = max V (i, xi ) =: Tn β→∞ β x∈Xn lim i=1 6 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions (Generalized) RSK correspondence RSK RSK Define the map : (R≥0 )n×n −−→ (R≥0 )n×n , W −−→ T : li,j replaces the submatrix wi−1,j−1 wi−1,j wi,j−1 wi,j with its image under the map a b min(b, c) − a b → ; c d c max(b, c) + d ( l1,j−i+1 ◦ · · · ◦ li−1,j−1 ◦ li,j i ≤ j i %j := li−j+1,1 ◦ · · · ◦ li−1,j−1 ◦ li,j i ≥ j ; Dk : composition of all %ij such that i + j = k + 1; RSK: D2n−1 ◦ · · · ◦ D1 . 7 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Properties of RSK By T = (Z , Z 0 ) we mean: t1,1 t1,2 t2,1 tn,1 t1,n tn−1,n tn,n−1 tn,n = 0 zn−1,n−1 λn 0 z1,1 z n−1,n−1 0 zn−1,1 z1,1 zn−1,1 λ1 . Properties: j X zj,i − i=1 i X j=1 j−1 X zj−1,i = i=1 0 zi,j − i−1 X j=1 X wi,j ∀j ; wi,j ∀i . i 0 zi−1,j = X j 8 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Properties of RSK By T = (Z , Z 0 ) we mean: t1,1 t1,2 t2,1 tn,1 t1,n tn−1,n tn,n−1 tn,n = 0 zn−1,n−1 λn 0 z1,1 z n−1,n−1 0 zn−1,1 z1,1 zn−1,1 λ1 . Properties: j X zj,i − i=1 i X j=1 j−1 X zj−1,i = i=1 0 zi,j − i−1 X j=1 X wi,j ∀j ; wi,j ∀i . i 0 zi−1,j = X j 8 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Last passage percolation RSK W −−→ T = (Z , Z 0 ); λ = (λ1 , λ2 , . . . ): shape of Z , (n, n) Z 0; Πn : directed paths (1, 1) → (n, n); X tn,n = λ1 = max wi,j . π∈Πn (1, 1) (i,j)∈π tn,n is a point-to-point version of the maximal passage time T2n−2 . (0, 0) (2n − 2, 0) 9 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Last passage percolation RSK W −−→ T = (Z , Z 0 ); λ = (λ1 , λ2 , . . . ): shape of Z , (n, n) Z 0; Πn : directed paths (1, 1) → (n, n); X tn,n = λ1 = max wi,j . π∈Πn (1, 1) (i,j)∈π tn,n is a point-to-point version of the maximal passage time T2n−2 . (0, 0) (2n − 2, 0) 9 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Last passage percolation with geometric weights If wi,j are independent and Geometric(pj qi ): P(shape = λ) = c · sλ (p1 , . . . , pn ) · sλ (q1 , . . . , qn ) ; X P(tn,n ≤ t) = c sλ (p1 , . . . , pn ) · sλ (q1 , . . . , qn ) . λ: λ1 ≤t Here: c= n Y (1 − pj qi ) ; i,j=1 sλ is a Schur polynomial. 10 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Geometric RSK Formally change operations: a+b → a·b, a−b → a , b max(a, b) → a + b . Now li,j replaces the submatrix wi−1,j−1 wi−1,j wi,j−1 wi,j with its image under the map a b bc/[a(b + c)] b → . c d c (b + c)d gRSK The new map : (R>0 )n×n −−−→ (R>0 )n×n is called geometric RSK. 11 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Geometric RSK Formally change operations: a+b → a·b, a−b → a , b max(a, b) → a + b . Now li,j replaces the submatrix wi−1,j−1 wi−1,j wi,j−1 wi,j with its image under the map a b bc/[a(b + c)] b → . c d c (b + c)d gRSK The new map : (R>0 )n×n −−−→ (R>0 )n×n is called geometric RSK. 11 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Geometric RSK and directed polymers gRSK W −−−→ T = (Z , Z 0 ); (n, n) λ = (λ1 , λ2 , . . . ): shape of Z , Z 0 ; Πn : directed paths (1, 1) → (n, n); X Y tn,n = λ1 = wi,j . (1, 1) π∈Πn (i,j)∈π β tn,n is a point-to-point version of the polymer partition sum Z2n−2 . (0, 0) (2n − 2, 0) 12 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Geometric RSK and directed polymers gRSK W −−−→ T = (Z , Z 0 ); (n, n) λ = (λ1 , λ2 , . . . ): shape of Z , Z 0 ; Πn : directed paths (1, 1) → (n, n); X Y tn,n = λ1 = wi,j . (1, 1) π∈Πn (i,j)∈π β tn,n is a point-to-point version of the polymer partition sum Z2n−2 . (0, 0) (2n − 2, 0) 12 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Properties of geometric RSK n Y Qj n Y Qj i=1 zj,i wi,j = Qj−1 i=1 zj−1,i i=1 j=1 0 i=1 zj,i 0 i=1 zj−1,i wi,j = Qj−1 X 1 1 = + wi,j t1,1 i,j X (i,j)6=(1,1) ∀j ; ∀i ; ti−1,j + ti,j−1 ti,j log(wi,j )1≤i,j≤n → log(ti,j )1≤i,j≤n has Jacobian ±1. 13 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Inverse gamma polymer Theorem (Corwin, O’Connell, Seppäläinen, Zygouras 2014) −1 If wi,j are independent and wi,j ∼ Γ(αj + α̂i , 1), the Laplace transform of tn,n is Y E e −θtn,n = i,j 1 Γ(αj + α̂i ) Z −1 e −θλ1 −λn ψα (λ)ψα̂ (λ) (R>0 )n n Y dλi i=1 λi . Here, ψα and ψα̂ are Whittaker functions: −αj Z Y n Qj i=1 zj,i exp ψα (λ) = Qj−1 i=1 zj−1,i j=1 X zi−1,j + zi+1,j+1 − zi,j (i,j) ! Y 1≤j≤i<n dzi,j , zi,j where the integral is over all triangular patterns (zi,j ) of height n and fixed shape λ = (zn,1 , . . . , zn,n ). 14 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Variations and open problems Variations on the polymer model with inverse gamma weights (O’Connell, Seppäläinen and Zygouras 2014): symmetrized polymers; polymers forced to stay below a hard wall. Open problems: antisymmetrized polymers; polymers not allowed to go outside a diagonal strip. 15 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Variations and open problems Variations on the polymer model with inverse gamma weights (O’Connell, Seppäläinen and Zygouras 2014): symmetrized polymers; polymers forced to stay below a hard wall. Open problems: antisymmetrized polymers; polymers not allowed to go outside a diagonal strip. 15 / 16 Directed polymers RSK correspondence Geometric RSK Conclusions Recap We have seen: (1 + d)-dim. directed random polymers in random environment; RSK correspondence (1 + 1)-dim. last passage percolation and the exactly solvable case of geometric weights; geometric RSK correspondence (1 + 1)-dim. polymer partition sum and the exactly solvable case of inverse gamma weights. 16 / 16