Dror Bar-Natan: Talks: Oberwolfach-1405: ωεβ≔http://www.math.toronto.edu/˜drorbn/Talks/Oberwolfach-1405/ Some very good formulas for the Alexander polynomial, 1 Abstract. I will describe some very good formulas for a (matrix plus Meta-Associativity Runs. scalar)-valued extension of the Alexander polynnomial to tangles, then say that everything extends to virtual tangles, then roughly to simply knotted balloons and hoops in 4D, then the target space extends to (free Lie algebras plus cyclic words), and theresult is a universalfinite type of the knotted objects in its domain. Taking a cue from the BF topological R3 ... divide and conquer! quantum field theory, everything should extend (with some modifica- 2 tions) to arbitrary codimension-2 knots in arbitrary dimension and in 6 5 particular, to arbitrary 2-knots in 4D. But what is really going on is still 4 1 3 a mystery. 2 1 Tangles. 4 6 ⊔ r 5 TS, S T 3
a b c b ab g mc T T +6 ++10 + 13 14 15 16 Why Tangles? 9 5 ab ac bc ab 12 8 • (meta-associativity: m m = m m ) 7 11 Finitely presented. a a b a 1 32 4 • Divide and conquer proofs and computations. − − − − • “Algebraic Knot Theory”: If K is ribbon, 817 cl cl ab Z(K) ∈{cl2(Z): cl1(Z) = 1}. 1 2 Weaknesses, • m is non-linear. T c (Genus and crossing number a- T T • The product ωA is always Laurent, but proving this takes indu- re also definable properties). ction with exponentially many conditions. trivial ribbon example Theorem 1. ∃! an invariant γ: {pure framed S -component K bh(H; T). balloons / tails tangles} → R × M (R), where R = R = Z((T ) ) is the ring S ×S S a a∈S T of rational functions in S variables, intertwining “Simply- u v ω1ω2 S 1 S 2 knotted ω1 S 1 ω2 S 2 ⊔ simple R4 1. , −−−−−→ S 1 A1 0 , balloons ∞ S 1 A1 S 2 A2 embeddings 4 S 2 0 A2 and hoops” S ω a b S H ab ω c S a α β θ mc x y z / 2. −−−−−→ c γ + αδ/ ǫ + δθ/ , hoops heads b γ δ ǫ ≔1−β S φ + αψ/ Ξ + ψθ/ Examples. x S φ ψ Ξ Ta,Tb→Tc δ 1 a b ǫx: x γ
1 a ±1 t " and satisfying |a; ! , −→ ; a 1 1 − T . a b b a a 1 a ǫu: u b 0 T ±1 a y x In Addition, • This is really “just” a stitching + − ρux: u ρux: u formula for Burau/Gassner [LD, KLW, CT]. x • L → ω is Alexander, mod units. “the generators” Shin Satoh • L → (ω, A) → ω det′(A − I)/(1 − T ′) is the Disturbing R2 R3 MVA, mod units. Conjecture = = • The “fastest” Alexander algorithm. VR3 • ? VR1 VR2 There are also formulas for strand deletion, K bh = u v = = = reversal, and doubling. M OC • Every step along the computation is the invariant of something. = = • Extends to and more naturally defined on v/w-tangles. x CP CP • Fits in one column, including propaganda & implementation. UC Dictionary. = = Implementation key idea: ωεβ/Demo ↔ = (ω, A = (αab)) ↔ = “v-xing” ↔ (ω, λ = αabtahb) = ↔ ↔ blue is never “over” =: wK bh no! yet not OC: as UC: Dror Bar-Natan: Talks: Oberwolfach-1405: ωεβ≔http://www.math.toronto.edu/˜drorbn/Talks/Oberwolfach-1405/ Some very good formulas for the Alexander polynomial, 2 Operations Connected ⊔ Theorem 3 [BND, BN]. ∃! a homomorphic expansion, aka a ho- Punctures & Cuts Sums. momorphic universal finite type invariant Z of w-knotted balloons H uv and hoops. ζ ≔ log Z takes values in FL(T) × CW(T). If X is a space, π1(X) K: K tmw : ζ is computable! ζ of the Borromean tangle, to degree 5: is a group, π2(X) is an u v w Abelian group, and π + cyclic colour 1 permutations, acts on π2. for trees Proposition. The gene- x y x y rators generate.
xy ux K hmz : K tha : u v u v
z x y
Definition. lxu is the linking number of hoop x with balloon u. ≔ lxu Z Proposition [BN]. Modulo all re- For x ∈ H, σx u∈T Tu ∈ R = RT = ((Ta)a∈T ), the ring of rational functions in T variables. lations that universally hold for = − Theorem 2 [BNS]. ∃! an invariant β: wK bh(H; T) → R × the 2D non-Abelian Lie alge- bra and after some changes-of- MT×H(R), intertwining variable, ζ reduces to β and the [u, v] = cuv− cvu ω1ω2 H1 H2 ω1 H1 ω2 H2 ⊔ KBH operations on ζ reduce to the formulas in Theorem 2. 1. , −−−−−→ T1 A1 0 , T1 A1 T2 A2 T2 0 A2 A Big Question. Does it all extend to arbitrary 2-knots (not neces- ω H sarily “simple”)? To arbitrary codimension-2 knots? uv ω H 1 4 2 ∗ u α tmw BF Following [CR]. A ∈ Ω (M = R , g), B ∈ Ω (M, g ), 2. −−−−−→ w α + β , v β ≔ T Ξ S (A, B) B, FA . T Ξ Tu,Tv→Tw M xy 2 0 1 ∗ Rossi ω x y H hmz ω z H With κ: (S = R ) → M, β ∈ Ω (S, g), α ∈ Ω (S, g ), set 3. −−−−−→ , Ξ + Ξ T α β T α σxβ i ∗ O ≔ D D ∗ + ω x H νω x H (A, B,κ) β α exp β, dκ Aα κ B . thaux S 4. u α θ −−−−−→ u σxα/ν σxθ/ν , ν≔1+α The BF Feynman Rules. For an edge e, let Φe be its T φ Ξ T φ/ν Ξ − φθ/ν 3 1 direction, in S or S . Let ω3 and ω1 be volume forms β ± 1 x 1 1 x 3 Cattaneo and satisfying ǫx; ǫu; ρ −→ ; ; . on S and S 1. Then ux u u T ±1 − 1 u [D] ∗ ∗ Proposition. If T is a u-tangle and β(δT) = (ω, A), then ZBF = Φeω3 Φeω1 |Aut(D)| R2 R2 R4 R4 ab diagrams red black γ(T) = (ω, σ − A), where σ = diag(σa)a∈S . Under this, mc ↔ D e∈D e∈D ab ab ab S -vertices M-vertices tha tmc hmc . (modulo some STU- and IHX -like relations). References. [BN] D. Bar-Natan, Balloons and Hoops and their Universal Finite Type I- M 1 2 1 nvariant, BF Theory, and an Ultimate Alexander Invariant, ωεβ/KBH, 2 air 2 piece S S arXiv:1308.1721. 1 1 rattle S 2 1 1 2 [BND] D. Bar-Natan and Z. Dancso, Finite Type Invariants of W-Knotted Ob- 2 1 0 S S S M 2 S jects I-II, ωεβ/WKO1, ωεβ/WKO2, arXiv:1405.1956, arXiv:1405.1955. 1 0 1 [BNS] D. Bar-Natan and S. Selmani, Meta-Monoids, Meta-Bicrossed Products, 1 1 S 1 0 1 2 1 0 2 1 0 and the Alexander Polynomial, J. of Knot Theory and its Ramifications 22-10 S S S 1 0 1 S 2 1 S (2013), arXiv:1302.5689. 2 0 S ground 0 piece S [CR] A. S. Cattaneo and C. A. Rossi, Wilson Surfaces and Higher Dimen- S 0 1 2 1 1 2 1 sional Knot Invariants, Commun. in Math. Phys. 256-3 (2005) 513–537, S S S degree = #(rattles) arXiv:math-ph/0210037. Issues. • Signs don’t quite work out, and BF seems to reproduce [CT] D. Cimasoni and V. Turaev, A Lagrangian Representation of Tangles, To- only “half” of the wheels invariant. pology 44 (2005) 747–767, arXiv:math.GT/0406269. • There are many more configuration space integrals than BF [KLW] P. Kirk, C. Livingston, and Z. Wang, The Gassner Representation for String Links, Comm. Cont. Math. 3 (2001) 87–136, arXiv:math/9806035. Feynman diagrams and than just trees and wheels. [LD] J. Y. Le Dimet, Enlacements d’Intervalles et Repr´esentation de Gassner, • I don’t know how to define “finite type” for arbitrary 2-knots. Comment. Math. Helv. 67 (1992) 306–315. “God created the knots, all else in topology is the work of mortals.” Leopold Kronecker (modified) www.katlas.org Videos of all talks are available at their respective web pages Meta−Groups, Meta−Bicrossed−Products, and the Alexander Polynomial, 1 Dror Bar−Natan in Montreal, June 2013. http://www.math.toronto.edu/~ drorbn/Talks/Montreal-1306/ Abstract. I will define “meta-groups” and explain how one specificAlexander Issues. meta-group, which in itself is a “meta-bicrossed-product”, gives rise• Quick to compute, but computation departs from topology. to an “ultimate Alexander invariant” of tangles, that contains the• Extends to tangles, but at an exponential cost. Alexander polynomial (multivariable, if you wish), has extremely• Hard to categorify. good composition properties, is evaluated in a topologically mean- Idea. Given a group G and two “YB” ingful way, and is least-wasteful in a computational sense. If you ± ± pairs R± = (g±, g±) ∈ G2, map them Z go believe in categorification, that’s a wonderful playground. o u ± This work is closely related to work by Le Dimet (Com-to xings and “multiply along”, so that gu ment. Math. Helv. 67 (1992) 306–315), Kirk, Livingston and Wang (arXiv:math/9806035) and Cimasoni and Turaev + + + − − + + + + + + Z go gu go gu go gu go gu (arXiv:math.GT/0406269). − − gu go See also Dror Bar-Natan and Sam Selmani, Meta-Monoids, − − Meta-Bicrossed Products, and the Alexander Polynomial, arXiv:1302.5689. Sam Selmani ± ∓ ± ∓ This Fails! R2 implies that go go = e = gu gu and then R3 u v + + implies that go and gu commute, so the result is a simple 4D u v counting invariant. K A Group Computer. Given G, can store group elements and x y perform operations on them:
z xy x : g1 m z u : g2 xy ∗ u : g2 ...so that m u v : g3 xy v : g3 uz yz xu K hm mv = mu mv z y : g4 uz xy z : g1g4 (or mv ◦ mu = xu ◦ yz Riddle. People often w u v G{x,u,v,y} mv mu , in old- G{u,v,z} 1 study π1(X) = [S ,X] speak). 2 and π2(X) = [S ,X]. Also has Sx for inversion, ex for unit insertion, dx for register dele- z x Why not πT (X) := tion, ∆ for element cloning, ρ for renamings, and (D1,D2) 7→ x y x y xy y [T,X]? D1 ∪ D2 for merging, and many obvious composition axioms relat- b b ing those. 1 2 b P = {x : g ,y : g } ⇒ P = {dyP }∪{dxP } b uv th S1 S2 T K tmw K swux A Meta-Group. Is a similar “computer”, only its internal 817 structure is unknown to us. Namely it is a collection of sets 6 4 {Gγ } indexed by all finite sets γ, and a collection of opera- xy z x +6 8 ++10 + 9=1 tions mz , Sx, ex, dx, ∆ (sometimes), ρ , and ∪, satisfying 13 14 15 16 xy y 9 5 + + + the exact same linear properties. 112 2 8 γ 7 11 Example 0. The non-meta example, Gγ := G . 12 3 4 3 − − − − − − Example 1. Gγ := Mγ×γ(Z), with simultaneous row and 5 column operations, and “block diagonal” merges. Here if 7 x : a b “divide and conquer” P = then dyP = (x : a) and dxP = (y : d) so y : c d R1 R2 R3 x : a 0 = = = {dyP } ∪ {dxP } = 6= P . So this G is truly meta. y : 0 d ± A Standard Alexander Formula. Label the arcs 1 throughClaim. From a meta-group G and YB elements R ∈ G2 we (n + 1) = 1, make an n × n matrix as below, delete one rowcan construct a knot/tangle invariant. and one column, and compute the determinant: Bicrossed Products. If G = HT is a group presented as a b c a b c product of two of its subgroups, with H ∩ T = {e}, then also G = TH and G is determined by H, T , and the “swap” map + a c −1 1 − X X swth : (t, h) 7→ (h′,t′) defined by th = h′t′. The map sw b a b c c satisfies (1) and (2) below; conversely, if sw : T ×H → H ×T a− c −X X − 1 1 satisfies (1) and (2) (+ lesser conditions), then (3) defines a group structure on H × T , the “bicrossed product”.
t1 t2 h4 t1 t2 h4 11 2 2 H (1) (2) = = (3) 3 3 12 12 12 12 12 T tm1 sw14 =sw24 sw14 tm1 gm3 :=sw12 tm3 hm3 ′ ′ ′ ′ h1t1h2t2 = h1(h2t1)t2 = (h1h2)(t1t2)= h3t3 Videos of all talks are available at their respective web pages Meta−Groups, Meta−Bicrossed−Products, and the Alexander Polynomial, 2 A Meta-Bicrossed-Product is a collection of sets β(η,τ) andI mean business! uv xy th operations tmw , hmz and swux (and lesser ones), such that tm and hm are “associative” and (1) and (2) hold (+ lesser conditions). A meta-bicrossed-product defines a meta-group with Gγ := β(γ, γ) and gm as in (3). Example. Take β(η,τ) = Mτ×η(Z) with row operations for the tails, column operations for the heads, and a trivial swap. β Calculus. Let β(η,τ) be
ω h1 h2 · · · hj ∈ η, ti ∈ τ, and ω and t1 α11 α12 · the αij are rational func- , t2 α21 α22 · tions in a variable X with . . ω(1) = 1 and αij(1) = 0 (1) Some . · · · = testing ω · · · ω1 η1 ω2 η2 ω · · · ∪ τ1 α1 τ2 α2 uv tu α 2 tm : 7→ tw α + β , ω ω η η , 6 w tv β 1 2 1 2 . 5 . . γ = τ1 α1 0 4 1 . γ 3 τ2 0 α2 2 ω hx hy · · · ω hz · · · 4 1 hmxy : 7→ , 6 z . . 5 . α β γ . α + β + hαiβ γ 3 ... divide and conquer!
ω hx · · · ωǫ hx · · · 817 swth : tu α β 7→ tu α(1 + hγi/ǫ) β(1 + hγi/ǫ) , ux . . . γ δ . γ/ǫ δ − γβ/ǫ where ǫ := 1+ α and hci := i ci, and let P 1 ha hb 1 ha hb p m −1 Rab := ta 0 X − 1 Rab := ta 0 X − 1 . tb 0 0 tb 0 0 Theorem. Zβ is a tangle invariant (and more). Restricted to knots, the ω part is the Alexander polynomial. On braids, it 817, cont. is equivalent to the Burau representation. A variant for links contains the multivariable Alexander polynomial. Why Happy? • Applications to w-knots. • Everything that I know about the Alexander polynomial can be expressed cleanly in this language (even if without proof), except HF, but including genus, ribbonness, cabling, v-knots, knotted graphs, etc., and there’s potential for vast James generalizations. Waddell • The least wasteful “Alexander for tangles” Alexander I’m aware of. A Partial To Do List. 1. Where does it more • Every step along the computation is the in- simply come from? T variant of something. 2. Remove all the denominators. • Fits on one sheet, including implementation 3. How do determinants arise in this context? trivial & propaganda. 4. Understand links (“meta-conjugacy classes”). Further meta-monoids. Π (and variants), A (and quotients),5. Find the “reality condition”. T vT , ... 6. Do some “Algebraic Knot Theory”. −→ Further meta-bicrossed-products. Π (and variants), A (and7. Categorify. ribbon bh rbh quotients), M0, M, K , K , ... 8. Do the same in other natural quotients of the Meta-Lie-algebras. A (and quotients), S, ... v/w-story. −→ Meta-Lie-bialgebras. A (and quotients), . . . "God created the knots, all else in I don’t understand the relationship between gr and H, as it topology is the work of mortals." Leopold Kronecker (modified) example appears, for example, in braid theory. www.katlas.org Videos of all talks are available at their respective web pages
Trees and Wheels and Balloons and Hoops Kbh(T ; H). 15 Minutes on Topology Dror Bar-Natan, Zurich, September 2013 balloons / tails “Ribbon- ωεβ:=http://www.math.toronto.edu/~drorbn/Talks/Zurich-130919 T knotted u v 15 Minutes on Algebra balloons 4 Let T be a finite set of “tail labels” and H a finite set of and hoops” ribbon R ∞ embeddings “head labels”. Set S4 H M1/2(T ; H) := FL(T ) , H “H-labeled lists of elements of the degree-completed free Lie x y z hoops / heads algebra generated by T ”. Examples. x δ 1 anti-symmetry ǫx: x FL(T )= 2t2 − [t1, [t1,t2]] + . . . 2 Jacobi t ǫ : u . . . with the obvious bracket. u y u v u v v x Shin Satoh u + − x → y → − 22 ρux: u ρux: u M1/2(u, v; x,y)= λ = , 7 . . . x x y y “the generators” Operations M1/2 → M1/2. newspeak! uv More on δ δ Tail Multiply tmw is λ → λ (u, v → w), satisfies “meta- = uv uw vw uv δ associativity”, tmu tmu = tmv tmu . xy “v” Head Multiply hmz is λ → (λ\{x,y}) ∪ (z → bch(λx, λy)), satisfies R123, VR123, D, and no! where yet not [α,β] [α,[α,β]]+[[α,β],β] OC: as bch(α, β) := log(eαeβ)= α + β + + + . . . UC: 2 12 bh satisfies bch(bch(α, β), γ) = log(eαeβeγ ) = bch(α, bch(β, γ))• δ injects u-knots into K (likely u-tangles too). xy xz yz xy • δ maps v-tangles to Kbh; the kernel contains the above and and hence meta-associativity, hmx hmx = hmy hmx . conjecturally (Satoh), that’s all. Tail by Head Action thaux is λ → λ RCλx , where u • Allowing punctures and cuts, δ is onto. C−γ : FL → FL is the substitution u → e−γ ueγ, or more u Operations precisely, Connected ∗ Punctures & Cuts Sums. −γ − ad γ 1 Cu : u → e (u)= u − [γ, u]+ [γ, [γ, u]] − ..., uv 2 If X is a space, π1(X) K: K tmw : −β γ −γ −1 bch(α,β) α RCu β is a group, π2(X) u v w and RCu = (Cu ) . Then Cu = Cu Cu hence bch(α,β) β RCα is an Abelian group, RC = RCα RC u hence “meta uxy = (ux)y”, u u u and π acts on π . xy uz ux uy xy 1 2 hmz tha = tha tha hmz , uv −γ Riddle. People often x y x y uv γ tmw γ RCv γ uv and tmw Cw = Cu Cv tmw and hence “metastudy π (X) = [S1,X] x x x uv wx ux vx uv 1 newspeak! (uv) = u v ”, tm tha = tha tha tm . 2 w w and π2(X) = [S ,X]. xy ux K hmz : K tha : Wheels. Let M(T ; H) := M (T ; H) × CW(T ), whereWhy not πT (X) := 1/2 [T,X]? u v u v CW(T ) is the (completed graded) vector space of cyclic words b b b on T , or equaly well, on FL(T ): b 1 2 u v u v uv u v S S T z x y = − “Meta-Group-Action” v v ωεβ/antiq-ave Properties. uv xy • Associativities: mab mac = mbc mab, for m = tm,hm. Operations. On M(T ; H), define tmw and hmz as before, a a b a ux x x x uv wx ux vx uv and tha by adding some J-spice: • “(uv) = u v ”: tmw tha = tha tha tmw , (xy) x y xy uz ux uy xy λx • “u = (u ) ”: hmz tha = tha tha hmz . (λ; ω) → (λ, ω + Ju(λx)) RCu , 1 ab ab Tangle concatenations → π1 ⋉ π2. With dmc := tha where J (γ):= ds div (γ RCsγ) C−sγ, and ab ab u u u u tmc hmc , 0 u v u v u v Alekseev ab a bmc c u u u δ δ divu a bab c + dmc ab c γ Torossian Finite type invariants make − − Theorem Blue. All blue identities still hold. sense in the usual way, and Merge Operation. (λ1; ω1)∗(λ2; ω2) := (λ1 ∪ λ2; ω1 + ω2). “algebra” is (the primitive part of) “gr” of “topology”. Videos of all talks are available at their respective web pages Trees and Wheels and Balloons and Hoops: Why I Care Moral. To construct an M-valued invariant ζ of (v-)tangles,The β quotient is M divi- and nearly an invariant on Kbh, it is enough to declare ζ onded by all relations that uni- the generators, and verify the relations that δ satisfies. versally hold when when g is = − The Invariant ζ. Set ζ(ǫx) = (x → 0; 0), ζ(ǫu) = ((); 0), and the 2D non-Abelian Lie alge- x Q [u, v] = c v− c u u u bra. Let R = J{cu}u∈T K and u v ζ:u ; 0 u − ; 0 Lβ := R ⊗ T with central R and with [u, v] = cuv − cvu for x x x u, v ∈ T . Then FL → Lβ and CW → R. Under this, Theorem. ζ is (log of) the unique homomor- phic universal finite type invariant on Kbh. → ((λx); ω) with λx = λuxux, λux,ω ∈ R, (. . . and is the tip of an iceberg) u ∈T c + c ecu − 1 ecv − 1 Paper in progress with Dancso, ωεβ/wko u v cu bch(u, v) → c +c u + e v , e u v − 1 cu cv if γ = γvv then with cγ := γvcv,