Classification of Modular Categories Eric Rowell XX CLA, Lima, Perú December 8-12 2014 Supported by USA NSF grant DMS1108725 joint work with Richard Ng, Paul Bruillard, Zhenghan Wang, César Galindo1 , Julia Plavnik 1 Organizing a school on Quantum Computation at U. de los Andes, 5/2015 Definition A modular category C (over C) is monoidal: (⊗, 1), L semisimple: X ∼ = i mi Xi , linear: Hom(X , Y ) ∈ VecC , rigid: X ∗ ⊗ X 7→ 1 7→ X ⊗ X ∗ , finite rank: Irr(C) = {1 = X0 , . . . , Xr −1 }, spherical: uX θX : X ∼ = X ∗∗ , dim(X ) ∈ R, braided: cX ,Y : X ⊗ Y ∼ = Y ⊗ X, modular: det(TrC (cXi ,Xj∗ cXj∗ ,Xi )) 6= 0. Remark I uX is Drinfeld isomorphism, θ is ribbon structure, satisfying, e.g.: θX ⊗Y = θX ⊗ θY (cY ,X cX ,Y ) I TrC : End(X ) → C is the pivotal trace. Heuristic Definitions I Definition (Physics) The algebraic models of elementary excitations of topological phases of matter are modular categories. I Definition (Algebra) Modular categories axiomatize Rep(U) where U is a modular Hopf algebra. E.g. Drinfeld doubles: D ω G . I Definition (Topology) Modular categories classify [some] 2 + 1-topological quantum field theories. Key Data I I L fusion rules: Xi ⊗ Xj ∼ = k Nijk Xk S-matrix: Sij := TrC (cXi ,Xj∗ cXj∗ ,Xi ) I T -matrix: Tij = δij θi I Level: N = ord(T ) I Dimension: dim(C) := I Ambient fields: Q(S) := Q(Sij : 0 ≤ i, j ≤ r − 1) and Q(θi : 0 ≤ i ≤ r − 1) = Q(T ). Pr −1 i=0 2 Si0 I Definition (In Coordinates) For S, T ∈ C(r ,r ) define dj := S0j , θj := Tjj , D 2 := P p± := j dj2 θj±1 . (S, T ) admissible if: P j dj2 , t 1. S = S t , SS = D 2 Id, T diagonal, ord(T ) = N < ∞ N 2. (ST )3 = p+ S 2 , p+ p− = D 2 , pp−+ =1 P S S S 3. Nijk := a iaD 2jada ka ∈ N P k 4. θi θj Sij = a Ni ∗ j dk θk where Nii0∗ uniquely defines i ∗ . n P 5. νn (k) := D12 i,j Nijk di dj θθji ∈ Z[ζN ] satisfies: ν2 (k) ∈ {0, ±1} 6. Q(S) ⊂ Q(T ), AutQ Q(S) ⊂ Sr , AutQ(S) Q(T ) ∼ = (Z2 )k . 2 7. Prime (ideal) divisors of hD i and hNi coincide in Z[ζN ]. Conjecture: Any admissible (S, T ) determines a modular category. Note: 1&2 imply (S, T ) yields ρ : SL(2, Z) → PU(r ). Sources I Pointed: C(A, q), A finite abelian group, q non-degenerate quad. form on A. I Rep(D ω G ), ω a 3-cocycle on G a finite group. Quantum groups/Kac-Moody algebras: subquotients of Rep(Uq g) at q = e πi/` or level k integrable ĝ-modules. e.g. I I I I I I SU(N)k = C(slN , N + k), SO(N)k , Sp(N)k , for gcd(N, k) = 1, PSU(N)k ⊂ SU(N)k “even half” Drinfeld center: Z(D) for spherical fusion category D. Example SU(2)`−2 from Rep(Uq sl2 ) at q = e πi/` I simple objects {X0 = 1, X1 , . . . , X`−2 } I Sij = I X1 ⊗ Xk ∼ = Xk−1 ⊕ Xk+1 for 1 ≤ k ≤ ` − 3 I θj = e I sin( (i+1)(j+1)π ) ` sin( π` ) πi(j 2 +2j) 2` so ord(T ) = 4` P 2 D 2 = dim SU(2)`−2 = `−1 n=1 [n] where [n] = Remark q n −q −n . q−q −1 √ Special case: ` = 5: dim SU(2)3 = 5 + 5, ord(T ) = 20. PSU(2)`−2 for ` odd: subcategory with simple objects X2i , 0 ≤ i. For the rest of this talk, we assume dim(X ) > 0 for all X . Problem Classify “small” modular categories. E.g.: I Bounded dimension: dim(C) < M I Bounded rank: r < M, I few primes: |{p ∈ Spec Z[ζN ] : p | dim(C)}| < M I Bounded level: ord(T ) < M Rank-Finiteness Theorem (Bruillard,Ng,R,Wang 2013) There are finitely many modular categories of a given rank r . History: I (2003) Conjectured by Wang, verified for rank = 2 (Ostrik). I (2005) Verified for: fusion categories with dim(C) ∈ N (Etingof, Nikshych and Ostrik), rank= 3 (Ostrik). I (2009-10) Verified for: rank = 4 (R,Stong,Wang), rank = 5 non-self-dual (R,Hong). Warm Up Theorem (E. Landau 1903) For any r ∈ N, there are finitely many groups G with |Irr (G )| = r . Proof. Use class equation: |G | = r X |gi |, i=1 gi distinct conjugacy classes. Set xi = [G : C (gi )] (index of centralizers) to get r X 1 1= . xi i=1 xi ≤ a(r ) where a(1) = 2, a(2) = 3, a(n) = a(n − 1)a(n − 2) + 1 is Sylvester’s sequence. Therefore |G | = maxi xi is bounded. Recall dim(C) := P j dim(Xi )2 . Lemma For any M > 0, there are finitely many rank = r fusion categories C with dim(C) < M. Proof. k ≤ dim(C) so follows from Ocneanu Rigidity† . Ni,j † There are finitely many fusion (and modular) categories with given fusion rules Nijk . A little number theory... For N = ord(T ) and ζN a primitive Nth root of 1, set: SC = {p ∈ Spec Z[ζN ] : p|hdim(C)i}. Example (C = SU(2)3 ) √ I N = ord(T ) = 20 and dim(C) = 5 + 5 I Prime factorization: in Z[ζ20 ], hdim(C)i = h2ih1 − (ζ20 )4 i2 I Thus: SC = {h2i, h1 − (ζ20 )4 i} For each r define: [ Sr := SC rank(C)=r Proposition (BNRW) |Sr | < ∞ implies rank-finiteness. Proof requires (analytic) number theory... Let K be a number field and S ⊂ Spec OK be finite. The S-units are: Y × OK,S = {x ∈ K : hxi = pαp } p∈S where αp ∈ Z. Theorem (Evertse 1984) There are finitely many solutions to 0 = 1 + x0 + · · · + xr −1 with × xi ∈ OK,S such that no sub-sum of 1 + x0 + · · · + xr −1 vanishes. Proof of Proposition. |Sr | < ∞ & Evertse’s Theorem imply: |{(− dim(C), (d1 )2 , . . . , (dr −1 )2 )}| < ∞. In particular, dim(C) < M(r ). This gives rank-finiteness. For C with N = ord(T ) set MC = {p ∈ Spec Z[ζN ] : p|hNi}, Mr = [ MC rank(C)=r Example (SU(2)3 ) 4 i4 , so M 4 hNi = h20i = h2i2 h1 − ζ20 SU(2)3 = {h2i, h1 − ζ20 i}. Proposition |Mr | < ∞ implies rank-finiteness. Follows from: Cauchy Theorem for Modular Categories Proposition (Bruillard,Ng,R.,Wang) Sr = Mr , i.e. hdim(C)i and hNi have the same prime divisors in Z[ζN ]. Lemma |Mr | < ∞. Proof. Enough to bound N < M(r ). Ng-Schauenburg Congruence Subgroup Theorem implies modular representation ρ : SL(2, Z) factors over SL(2, ZN ), minimal dimension of faithful SL(2, ZN ) representation bounds r < J(N), hence N < M(r ). Rank-Finiteness for modular categories follows. Low Rank Modular Categories Theorem (R,Hong,Stong,Bruillard,Ng,Wang,Ostrik) If 2 ≥ rank(C) ≤ 5 then C has the same fusion rules as one of: I PSU(2)3 (Fibonacci), SU(2)1 (pointed) I PSU(2)5 , SU(2)2 (Ising), SU(3)1 (pointed) I PSU(2)7 , SU(2)3 , SU(4)1 , products. I SU(2)4 , PSU(2)9 , SU(5)1 , PSU(3)4 . Theorem (R,Bruillard) I Suppose C has ord(T ) ∈ {2, 3, 4}. Then C ⊂ Rep(D ω G ) for some 2- or 3-group G . I Suppose dim(C) is odd and rank(C) ≤ 11. Then C is pointed. Remark I ord(T ) = 6 implies integral, but not C ⊂ Rep(D ω G ). I Open: dim(C) odd, 13 ≤ rank(C) ≤ 23 implies pointed? Tambara Yamagami Categories C is weakly integral if dim(C) ∈ N. Example Let A be a finite abelian group, χ a non-degenerate bicharacter on A and ν a sign. Tambara and Yamagami defined a spherical fusion category TY (A, χ, ν) with simple objects A ∪ {m} with: P I m⊗m = a∈A a I m⊗a=m I a ⊗ b = ab dim(TY (A, χ, ν)) = 2|A|. ZTY (A, χ, ν) is modular. Low Dimension Sample results Theorem (Natale) dim(C) < 36 and integral implies C is group-theoretical. Theorem (Bruillard,Galindo,Hong,Kashina,Naidu,Natale,Plavnik,R) dim(C) = 36 and integral non-group theoretical is: I twist-equivalent to SU(3)3 I TY (Z3 × Z3 , χ, τ )Z2 , χ elliptic. Theorem (R,Ng,Bruillard,Wang,Galindo,Plavnik) If C is: I weakly integral with rank(C) ≤ 7 or I dim(C) = 4m with m odd and square-free then C ≡ D F where D, F ⊂ ZTY (A, χ, ν). More explicitly: D, F are pointed, Ising or TY (Z2k+1 , χ, ν)Z2 . Few Primes Some classifications are known for C integral with: I dim(C) ∈ {p n , pq 2 , pq 3 , pq 4 , pq 5 , (pq)2 }. (various authors) I dim(X ) ∈ {1, p} (Natale,Plavnik). I dim(C) = pq n group-theoretical iff nilpotent (Dong,Tucker) Thank you!