Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Two-parameter Noncommutative Central Limit Theorem Natasha Blitvić Vanderbilt University January 11, 2013 N. Blit. 11/1/2013 1 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models (Classical) Central Limit Theorem — CLT (Classical) Probability space: (Ω, B, P). Theorem (Classical CLT) Let X1 , X2 , . . . be a sequence of independent and identically distributed random variables with E(Xi ) = 0 and E(Xi2 ) = 1 for all i ∈ N. Let PN i=1 Xi SN = √ . N Then, as N → ∞, SN ⇒ dµ1 (x) = N. Blit. 11/1/2013 √1 exp(−x 2 /2) 2π 2 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Free Central Limit Theorem — CLT ∗-probability space: (A, ϕ) Theorem (Free CLT) Let a1 , a2 , . . . be a sequence of freely independent and identically distributed self-adjoint elements of A with ϕ(ai ) = 0 and ϕ(ai2 ) = 1 for all i ∈ N. Let PN i=1 ai . SN = √ N Then, as N → ∞, SN ⇒ dµ0 (x) = N. Blit. 11/1/2013 1 2π √ 4 − x 2, x ∈ [−2, 2] 3 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Speicher’s Non-commutative CLT Condition (Speicher 1992) Consider a ∗-algebra A, state ϕ : A → C and {ai }i∈N ∈ A satisfying: 1. (vanishing means) for all i ∈ N, ϕ(ai ) = ϕ(ai∗ ) = 0. 2. (normalized second moments) for all for all i, j ∈ N with i < j and , 0 ∈ {1, ∗}, ϕ(ai ai∗ ) = 1 0 and ϕ(ai ai ) = 0 for (, 0 ) 6= (1, ∗). 3. (uniform moment bounds) for all n ∈ N and all n Y (i) j(1), . . . , j(n) ∈ N, (1), . . . , (n) ∈ {1, ∗}, |ϕ( aj(i) )| ≤ αn . i=1 N. Blit. 11/1/2013 4 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models 4. (“independence”) If Ai = ∗-alg(ai ) (i = 1, 2, . . .), then ϕ(x1 x2 . . . xn ) = ϕ(x1 ) . . . ϕ(xn ) whenever x1 ∈ Ai1 , . . . , xn ∈ Ain for i1 < . . . < in . Assume additionally that for all i 6= j and all , 0 ∈ {1, ∗}, ai and 0 aj satisfy the commutation relation 0 0 ai aj = s(j, i) aj ai , N. Blit. 11/1/2013 s(j, i) ∈ {−1, 1}. 5 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Theorem (Non-commutative CLT, Speicher 1992) A non-commutative probability space (A, ϕ) and {ai }i∈N in A satisfying Condition 1. Fix q ∈ [−1, 1], and let {s(i, j)}1≤i<j be independent, identically distributed random variables taking values in {−1, 1} with E(s(i, j)) = q. N 1 X (ai + ai∗ ). Let SN = √ N i=1 Then, for a.e. sign sequence {s(i, j)}1≤i<j , as N → ∞, SN ⇒ q-Gaussian. N. Blit. 11/1/2013 6 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 7 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 8 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 9 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 10 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 11 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 12 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 13 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 14 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 15 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 16 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 17 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 18 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models q-Gaussian N. Blit. 11/1/2013 19 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Combinatorial View Definition Let P2 (2n) = {all pair-partitions of {1, . . . , 2n}}. For V = {(w1 , z1 ), . . . , (wn , zn )} ∈ P2 (2n), pairs (wi , zi ) and (wj , zj ) are said to cross if wi < wj < zi < zj . Then, lim ϕ(SN2n−1 ) = 0, lim ϕ(SN2n ) = N→∞ 1 N. Blit. 11/1/2013 2 3 4 N→∞ 5 6 7 X q cross(V ) , V ∈P2 (2n) 8 9 10 11 12 20 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models More general commutation structure? Why s(i, j) ∈ {−1, 1}? j µ(i,j) i For µ(i, j) ∈ R, • Limit exists? • q-Gaussian? N. Blit. 11/1/2013 21 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models More general combinatorics? Definition Let P2 (2n) = {all pair-partitions of {1, . . . , 2n}}. For V = {(w1 , z1 ), . . . , (wn , zn )} ∈ P2 (2n), pairs (wi , zi ) and (wj , zj ) are said to nest if wi < wj < zj < zi . 1 ei ... ej ... zi ... zj 2n 1 ei ... ej ... zj ... zi 2n The number of nestings of V is: nest(V ) = #{(wi , zi ), (wj , zj ) ∈ V | i < j, wi < wj < zj < zi }. N. Blit. 11/1/2013 22 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Generalized Non-commutative CLT Condition Consider a ∗-algebra A, state ϕ : A → C and {ai }i∈N ∈ A satisfying: 1. (vanishing means) for all i ∈ N, ϕ(ai ) = ϕ(ai∗ ) = 0. 2. (normalized second moments) for all for all i, j ∈ N with i < j and , 0 ∈ {1, ∗}, ϕ(ai ai∗ ) = 1 0 and ϕ(ai ai ) = 0 for (, 0 ) 6= (1, ∗). 3. (uniform moment bounds) for all n ∈ N and all n Y (i) j(1), . . . , j(n) ∈ N, (1), . . . , (n) ∈ {1, ∗}, |ϕ( aj(i) )| ≤ αn . i=1 N. Blit. 11/1/2013 23 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models 4. (“independence”) If Ai = ∗-alg(ai ) (i = 1, 2, . . .), then ϕ(x1 x2 . . . xn ) = ϕ(x1 ) . . . ϕ(xn ) whenever x1 ∈ Ai1 , . . . , xn ∈ Ain for i1 < . . . < in . Assume additionally that for all i 6= j and all , 0 ∈ {1, ∗}, ai and 0 aj satisfy the commutation relations for ai aj = µ1,1 (j, i) aj ai , ai∗ aj∗ = µ∗,∗ (j, i) aj∗ ai∗ ai aj∗ = µ∗,1 (j, i) aj∗ ai , ai∗ aj = µ1,∗ (j, i) aj ai∗ µ1,1 (j, i), µ∗,∗ (j, i), µ∗,1 (j, i), µ1,∗ (j, i) ∈ R. N. Blit. 11/1/2013 24 / 39 Classical, Free, NC CLT Moments and Combinatorics 0 0 ai aj = µ0 , (j, i) aj ai , Generalized NC CLT Random Matrix Models µ0 , (j, i) ∈ R Lemma Coefficients {µ0 , (i, j)} satisfy the consistency relations: 1 µ,0 (i, j) 1 µ1,1 (i, j) = , µ∗,∗ (i, j) µ0 , (j, i) = for all i 6= j, , 0 ∈ {1, ∗}, (1) 1 µ∗,1 (i, j) (2) µ1,∗ (i, j) = In addition, when ϕ is assumed to be positive, µ∗,∗ (i, j) µ∗,1 (i, j) = . | µ∗,1 (i, j) | | µ∗,∗ (i, j) | N. Blit. 11/1/2013 (3) 25 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Theorem (Generalized non-commutative CLT, B. 2012) Consider a ∗-probability space (A, ϕ) and {ai }i∈N in A satisfying Condition 2. In addition, suppose that {µ∗,∗ (i, j)}1≤i<j satisfy µ∗,1 (i, j) = t µ∗,∗ (i, j) (3’) for some t > 0. N 1 X Let SN = √ (ai + ai∗ ). N i=1 N. Blit. 11/1/2013 26 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models For fixed q ∈ R, let {µ∗,∗ (i, j)}1≤i<j be independent, identically distributed random variables with E(µ∗,∗ (i, j)) = q/t and E(µ∗,∗ (i, j)2 ) = 1, and populate the remaining µ,0 (i, j), for , 0 ∈ {1, ∗} and i 6= j (i, j ∈ N), by consistency conditions (1)-(3’). N. Blit. 11/1/2013 27 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models For fixed q ∈ R, let {µ∗,∗ (i, j)}1≤i<j be independent, identically distributed random variables with E(µ∗,∗ (i, j)) = q/t and E(µ∗,∗ (i, j)2 ) = 1, and populate the remaining µ,0 (i, j), for , 0 ∈ {1, ∗} and i 6= j (i, j ∈ N), by consistency conditions (1)-(3’). Then, for a.e. coefficient sequence {µ∗,∗ (i, j)}1≤i<j , Sn ⇒ (q, t)-Gaussian. lim ϕ(SN2n−1 ) = 0, N→∞ N. Blit. 11/1/2013 lim ϕ(SN2n ) = N→∞ X q cross(V ) t nest(V ) V ∈P2 (2n) 28 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Proof sketch Show that ZN = a1 + . . . + aN √ converges in mixed moments. N Fix k ∈ N and a ∗-pattern (1), . . . , (k) ∈ {1, ∗}, and look at the mixed moment X 1 (1) (k) (1) (k) ϕ ai(1) . . . ai(k) . ϕ(ZN . . . ZN ) = k/2 N i(1),...,i(k)∈[N] (1) (k) Repetition patterns of indices in ϕ ai(1) . . . ai(k) induce set partitions. N. Blit. 11/1/2013 29 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Note: • By standard arguments, only pair-partitions can contribute to the limit. Set k = 2n. (1) (n) • If limN→∞ E(ϕ(ZN . . . ZN )) exists, Markov inequality + more careful estimates yield convergence a.e. Remains to factor (1) (2n) ϕ ai(1) . . . ai(2n) If i(1) = i(2) < i(3) = i(4) < . . . < i(2n − 1) = i(2n), then (2n−1) (2n) (1) (2) (1) (2n) ϕ ai(1) . . . ai(2n) = ϕ ai(1) ai(2) . . . ϕ ai(2n−1) ai(2n) e.g. N. Blit. 11/1/2013 ϕ 0 0 ai j ai j aji aj i 0 0 j j = ϕ ai ai ϕ aji aj i i <j 30 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Otherwise, use the commutation relation. For i < j, need to factor 0 j j i 0i ϕ aj aj ai ai 0 0 0 j i j 0i i j i j ϕ ai aj ai aj and ϕ aj ai aj ai 0 0 0 j i 0i j i j j i ϕ ai aj aj ai and ϕ aj ai ai aj N. Blit. 11/1/2013 31 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Coefficients incurred: µi ,0j (j, i)µi ,j (j, i)µ0i ,0j (j, i)µ0i ,j (j, i) N. Blit. 11/1/2013 µ0i ,j (i, j) and µ0j ,i (j, i) µ0i ,j (i, j)µ0i ,0j (i, j) and µj ,i (j, i)µj ,0i (j, i) 32 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models 0 If ϕ(ai ai∗ ) = 1 and ϕ(ai ai ) = 0 for (, 0 ) 6= (1, ∗), µ1,∗ (j, i)µ1,1 (j, i)µ∗,∗ (j, i)µ∗,1 (j, i) N. Blit. 11/1/2013 µ∗,1 (i, j) and µ∗,1 (j, i) µ∗,1 (i, j)µ∗,∗ (i, j) and µ1,1 (i, j)µ∗,1 (i, j) 33 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Recall: 0 0 ai aj = µ0 , (j, i) aj ai , µ0 , (j, i) ∈ R And coefficients {µ0 , (i, j)} satisfy the consistency relations: 1 µ,0 (i, j) 1 µ1,1 (i, j) = , µ∗,∗ (i, j) µ0 , (j, i) = for all i 6= j, , 0 ∈ {1, ∗}, (1) 1 µ∗,1 (i, j) (2) µ1,∗ (i, j) = µ∗,1 (i, j) = t µ∗,∗ (i, j) N. Blit. 11/1/2013 (3’) 34 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Applying consistency relations (1)-(3’) and taking E 1 N. Blit. 11/1/2013 q and q t and t 35 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Theorem (Extended Jordan-Wigner Transform, Biane 1997) Non-commutative probability space (A, ϕ) with A = M2 (C)⊗n and ϕ(a) = hae0⊗n , e0⊗n i. Fix q ∈ [−1, 1] and s(i, j) ∈ {−1, 1} for all 1 ≤ i < j ≤ n. Let 1 0 1 0 0 1 σ1 = , σ−1 = , γ= 0 1 0 −1 0 0 and, for i = 1, . . . , n, let the element ai ∈ M2 (C)⊗n be given by ai = σs(1,i) ⊗ σs(2,i) ⊗ . . . ⊗ σs(i−1,i) ⊗ γ ⊗ σ1 ⊗ . . . ⊗ σ1 . | {z } ⊗(n−i) =σ1 Then, for every n ∈ N, a1 , . . . , an satisfy Condition 1. N. Blit. 11/1/2013 36 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Theorem (Two-parameter Jordan-Wigner Transform, B. 2012) Non-commutative probability space (A, ϕ) with A = M2 (C)⊗n and ϕ(a) = hae0⊗n , e0⊗n i. Fix |q| ≤ t and µ,0 (i, j) ∈ R for all 1 ≤ i < j ≤ n, , 0 ∈ {1, ∗}. Let 1 √0 0 1 , γ= σx = 0 0 tx 0 and, for i = 1, . . . , n, let the element ai ∈ M2 (C)⊗n be given by ai = σµ(1,i) ⊗ σµ(2,i) ⊗ . . . ⊗ σµ(i−1,i) ⊗ γ ⊗ σ1 ⊗ . . . ⊗ σ1 . | {z } ⊗(n−i) =σ1 Then, for every n ∈ N, a1 , . . . , an satisfy Condition 2. N. Blit. 11/1/2013 37 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Looking ahead • Connections. q = -1 free q=0 anti-symmetric q=1 symmetric q=0<t<1 • More general commutation structure. • Beyond i.i.d. • Practical applications? N. Blit. 11/1/2013 38 / 39 Classical, Free, NC CLT Moments and Combinatorics Generalized NC CLT Random Matrix Models Taknh yuo N. Blit. 11/1/2013 39 / 39