ScalingStacks

4. Khovanov–Rozansky homology in S3S^{3}

From now on, we will only consider framed oriented links and framed oriented link cobordisms. Furthermore, all diffeomorphisms are oriented.

4.1. Link homology in abstract 3-balls

The purpose of this section is to define a functorial Khovanov–Rozansky link homology for links in abstract 33-manifolds (abstractly) diffeomorphic to ℝ3{\mathbb{R}}^{3}, which is functorial under link cobordisms in abstract 44-manifolds diffeomorphic to ℝ3×[0,1]{\mathbb{R}}^{3}\times{[0,1]}. The framework set up in this section could have been developed immediately after the initial construction of functorial link invariants, but to our knowledge it has not been developed in the literature. We hope that the careful presentation of this improvement of the invariant will be a helpful warm-up for the following section, where we employ a very similar strategy to build invariants of links in abstract 3-spheres.

{link embeddings in oriented ​B≅ℝ3link cobordisms in or.W≅ℝ3×[0,1] up to isotopy rel ∂}→KhRN{bigraded abelian groupshomogeneous homomorphisms}\begin{Bmatrix}\textrm{link embeddings in oriented }B\cong{\mathbb{R}}^{3}\\ \textrm{link cobordisms in or.}W\cong{\mathbb{R}}^{3}\times{[0,1]}\textrm{ up to isotopy rel }\partial\end{Bmatrix}\xrightarrow{\mathrm{KhR}_{N}}\begin{Bmatrix}\textrm{bigraded abelian groups}\\ \textrm{homogeneous homomorphisms}\end{Bmatrix}

We will call such an invariant a link homology for links in 33-balls.

Throughout this section, BB will denote a 3-ball: an oriented 33-manifold that is diffeomorphic to ℝ3{\mathbb{R}}^{3} via some (unspecified!) diffeomorphism. We say a link embedding LL in ℝ3{\mathbb{R}}^{3} is generic if it is in generic position with respect to the projection along the zz-axis to ℝ2{\mathbb{R}}^{2} and all crossings in the resulting link diagram have distinct yy coordinates. In this case, we consider the crossings as ordered from smallest to largest yy coordinate. We say a link embedding LL in ℝ3{\mathbb{R}}^{3} is blackboard-framed if the framing is parallel to ℝ2{\mathbb{R}}^{2}.

0NAN

Lemma 4.1. Let L⊂BL\subset B be a link embedded in a 3-ball. Let ϕ0\phi_{0} and ϕ1\phi_{1} be two diffeomorphisms from BB to ℝ3{\mathbb{R}}^{3} such that ϕ0​(L)\phi_{0}(L) and ϕ1​(L)\phi_{1}(L) are generic. Then we have the following:

  1. (1)

    There exists a continuous family of diffeomorphisms ϕt\phi_{t} for t∈[0,1]t\in{[0,1]}, such that ϕt​(L)\phi_{t}(L) is generic for all but finitely many t∈[0,1]t\in{[0,1]}, at which a Reidemeister move occurs or the crossing height order changes.

  2. (2)

    Given two such families ϕt,0\phi_{t,0} and ϕt,1\phi_{t,1}, both interpolating between ϕ0\phi_{0} and ϕ1\phi_{1}, then there exists a continuous family ϕt,s\phi_{t,s} of diffeomorphisms interpolating between the families ϕt,0\phi_{t,0} and ϕt,1\phi_{t,1}, for which the parameter space [0,1]×[0,1]{[0,1]}\times{[0,1]} is stratified such that:

    • •

      ϕt,s​(L)\phi_{t,s}(L) is generic for (t,s)(t,s) in any codimension-0 stratum,

    • •

      ϕt,s​(L)\phi_{t,s}(L) undergoes a Reidemeister move or the crossing height order changes as (t,s)(t,s) crosses through a codimension-1 stratum,

    • •

      ϕt,s​(L)\phi_{t,s}(L) has a movie move as monodromy if (t,s)(t,s) loops around a codimension-2 stratum.

0NAQ

Definition 4.2. Let BB be an oriented 33-manifold diffeomorphic to ℝ3{\mathbb{R}}^{3}. We define

M⁡(B)=def{diffeomorphisms ​ϕ:B→ℝ3}.M(B)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\{\textrm{diffeomorphisms }\phi\colon B\to{\mathbb{R}}^{3}\}.

Given an embedded link L⊂BL\subset B, we define the subspace

M(B,L)=def{ϕ∈M(B)|ϕ(L) is z-generic and blackboard-framed},M(B,L)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\{\phi\in M(B)|\phi(L)\text{ is }z\text{-generic and blackboard-framed}\},

and consider the bundle π:T⁡(B,L)→M⁡(B,L)\pi\colon T(B,L)\to M(B,L) of bigraded abelian groups, whose fiber at the point ϕ\phi is KhRN​(ϕ​(L))\mathrm{KhR}_{N}(\phi(L)).

For a path ϕt\phi_{t} in M⁡(B)M(B) between points ϕ0,ϕ1∈M⁡(B,L)⊂M⁡(B)\phi_{0},\phi_{1}\in M(B,L)\subset M(B), we define the grading-preserving isomorphism

(KhRN(ϕt):T(B,L)ϕ0→T(B,L)ϕ1)=def(KhRN(ϕt(L)):KhRN(ϕ0(L))→KhRN(ϕ1(L))),\left(\mathrm{KhR}_{N}(\phi_{t})\colon T(B,L)_{\phi_{0}}\to T(B,L)_{\phi_{1}}\right)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\left(\mathrm{KhR}_{N}(\phi_{t}(L))\colon\mathrm{KhR}_{N}(\phi_{0}(L))\to\mathrm{KhR}_{N}(\phi_{1}(L))\right),

where the latter denotes the homomorphism associated to the trace of the link isotopy ϕt​(L)\phi_{t}(L) in ℝ3×[0,1]{\mathbb{R}}^{3}\times{[0,1]}. This is well-defined by Theorem 2.4, even though for some tt the embeddings ϕt​(L)\phi_{t}(L) can be highly non-generic with respect to projection in the zz-coordinate. Also note that while Reidemeister I moves induce qq-grading shifts on the level of KhRN\mathrm{KhR}_{N}, any isotopy of framed links features such moves in pairs, leading to a grading-preserving isomorphism.

0NAR

Lemma 4.3. The parallel transport isomorphisms KhRN​(ϕt)\mathrm{KhR}_{N}(\phi_{t}) define a flat connection on T⁡(B,L)T(B,L).

0NAS

Proof. Lemma 4.1 (1) implies that we have such parallel transport maps KhRN​(ϕt)\mathrm{KhR}_{N}(\phi_{t}) between the fibers over any pair of points ϕ0\phi_{0} and ϕ1\phi_{1} in the base. Note that π1​(M⁡(B),ϕ0)≅π1​(S​O​(3))≅ℤ/2​ℤ\pi_{1}(M(B),\phi_{0})\cong\pi_{1}(SO(3))\cong{\mathbb{Z}}/2{\mathbb{Z}} is generated by the class of the loop obtained from ϕ0\phi_{0} by rotating through 360360 degrees around the zz-axis, to which KhRN\mathrm{KhR}_{N} assigns the identity map, see also Section 6.4. Lemma 4.1 (2) and Theorem 2.4 thus imply that the parallel transport maps between the fibers do not depend on the choice of the path ϕt\phi_{t}. ∎

0NAT

Definition 4.4. Let L⊂BL\subset B be a link embedded in a 3-ball. Then we define the Khovanov–Rozansky homology of LL in BB to be

KhRN​(B,L)=defΓflat​(T⁡(B,L)),\mathrm{KhR}_{N}(B,L)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\Gamma_{\textrm{flat}}(T(B,L)),

the bigraded abelian group of flat sections of the bundle T⁡(B,L)T(B,L).

Note that every diffeomorphism ϕ:B→ℝ3\phi\colon B\to{\mathbb{R}}^{3} such that ϕ⁡(L)\phi(L) is generic and blackboard-framed induces a grading-preserving isomorphism KhRN​(B,L)→KhRN​(ϕ⁡(L))\mathrm{KhR}_{N}(B,L)\to\mathrm{KhR}_{N}(\phi(L)) by evaluating sections at the point ϕ\phi.

Remark. An alternative way to describe this definition is as follows. Consider the groupoid with set of objects given by M⁡(B,L)M(B,L) and with morphisms given by paths ϕt\phi_{t} modulo isotopies ϕt,s\phi_{t,s} as in Lemma 4.1. Then KhRN\mathrm{KhR}_{N} restricts to a functor from this groupoid to bigraded abelian groups and KhRN​(B,L)\mathrm{KhR}_{N}(B,L) is defined as its (co)limit.

0NAU

Definition 4.5. Consider a link cobordism Σ⊂W\Sigma\subset W in a 4-manifold WW diffeomorphic to ℝ3×[0,1]{\mathbb{R}}^{3}\times{[0,1]}. Let Σin⊂Win\Sigma_{\mathrm{in}}\subset W_{\mathrm{in}} and Σout⊂Wout\Sigma_{\mathrm{out}}\subset W_{\mathrm{out}} denote the boundary links in the incoming and outgoing boundary 3-balls of WW. Then we define

KhRN​(W,Σ):KhRN​(Win,Σin)→KhRN​(Wout,Σout)\mathrm{KhR}_{N}(W,\Sigma)\colon\mathrm{KhR}_{N}(W_{\mathrm{in}},\Sigma_{\mathrm{in}})\to\mathrm{KhR}_{N}(W_{\mathrm{out}},\Sigma_{\mathrm{out}})

in two steps. First we pick a diffeomorphism ϕ:W→ℝ3×[0,1]\phi\colon W\to{\mathbb{R}}^{3}\times{[0,1]}, such that ϕout:=ϕ|Wout:Wout→ℝ3\phi_{\mathrm{out}}:=\phi|_{W_{\mathrm{out}}}\colon W_{\mathrm{out}}\to{\mathbb{R}}^{3} and ϕin:=ϕ|Win:Win→ℝ3\phi_{\mathrm{in}}:=\phi|_{W_{\mathrm{in}}}\colon W_{\mathrm{in}}\to{\mathbb{R}}^{3} are such that ϕin​(Σin)\phi_{\mathrm{in}}(\Sigma_{\mathrm{in}}) and ϕout​(Σout)\phi_{\mathrm{out}}(\Sigma_{\mathrm{out}}) are both generic and blackboard-framed. Then we declare KhRN​(W,Σ)​(η)\mathrm{KhR}_{N}(W,\Sigma)(\eta), for a flat section η∈KhRN​(Win,Σin)\eta\in\mathrm{KhR}_{N}(W_{\mathrm{in}},\Sigma_{\mathrm{in}}), to be the unique flat section of KhRN​(Wout,Σout)\mathrm{KhR}_{N}(W_{\mathrm{out}},\Sigma_{\mathrm{out}}) with value:

KhRN​(W,Σ)​(η)​(ϕout)=KhRN​(ϕ⁡(Σ))​(η⁡(ϕin))\mathrm{KhR}_{N}(W,\Sigma)(\eta)(\phi_{\mathrm{out}})=\mathrm{KhR}_{N}(\phi(\Sigma))(\eta(\phi_{\mathrm{in}}))
0NAV

Lemma 4.6. KhRN​(W,Σ)\mathrm{KhR}_{N}(W,\Sigma) is independent of the choices of ϕin\phi_{\mathrm{in}}, ϕout\phi_{\mathrm{out}} and ϕ\phi, and thus well-defined.

0NAW

Proof. We first show independence of ϕ\phi, given a fixed choice of ϕin\phi_{\mathrm{in}} and ϕout\phi_{\mathrm{out}}. Suppose that ϕ′:W→ℝ3×[0,1]\phi^{\prime}\colon W\to{\mathbb{R}}^{3}\times{[0,1]} is another diffeomorphism restricting to ϕin\phi_{\mathrm{in}} and ϕout\phi_{\mathrm{out}} on WinW_{\mathrm{in}} and WoutW_{\mathrm{out}} respectively.

Lemma 4.7, proved below, implies that the link cobordisms ϕ⁡(Σ)\phi(\Sigma) and ϕ′​(Σ)\phi^{\prime}(\Sigma) are isotopic rel boundary in ℝ3×[0,1]{\mathbb{R}}^{3}\times{[0,1]} and we have KhRN​(ϕ⁡(Σ))=KhRN​(ϕ′​(Σ))\mathrm{KhR}_{N}(\phi(\Sigma))=\mathrm{KhR}_{N}(\phi^{\prime}(\Sigma)) by Theorem 2.4.

Next we show independence of ϕin\phi_{\mathrm{in}}, given a fixed choice of ϕout\phi_{\mathrm{out}}. Let ϕin′:Win→ℝ3\phi^{\prime}_{\mathrm{in}}\colon W_{\mathrm{in}}\to{\mathbb{R}}^{3} be another diffeomorphism such that ϕin′​(Σin)\phi^{\prime}_{\mathrm{in}}(\Sigma_{\mathrm{in}}) is generic and blackboard-framed, and ϕ′\phi^{\prime} another diffeomorphism W→ℝ3×[0,1]W\to{\mathbb{R}}^{3}\times{[0,1]} restricting to ϕin′\phi^{\prime}_{\mathrm{in}} on WinW_{\mathrm{in}} but still to ϕout\phi_{\mathrm{out}} on WoutW_{\mathrm{out}}. Then, by Lemma 4.1 (1), we can find a family ϕin,t\phi_{\mathrm{in},t} connecting ϕin\phi_{\mathrm{in}} to ϕin′\phi^{\prime}_{\mathrm{in}}. By definition of parallel transport, we have:

η⁡(ϕin′)=KhRN​(ϕin,t​(Σin))​(η⁡(ϕin))\eta(\phi^{\prime}_{\mathrm{in}})=\mathrm{KhR}_{N}(\phi_{\mathrm{in},t}(\Sigma_{\mathrm{in}}))(\eta(\phi_{\mathrm{in}}))

Now we obtain a new diffeomorphism ϕ′∘ϕin,t:W→ℝ3×[0,1]\phi^{\prime}\circ\phi_{\mathrm{in},t}\colon W\to{\mathbb{R}}^{3}\times{[0,1]} and by the previous independence result and Theorem 2.4, we have:

KhRN​(ϕ′​(Σ))​(η⁡(ϕin′))=KhRN​(ϕ′​(Σ)∘ϕin,t​(Σin))​(η⁡(ϕin))=KhRN​(ϕ⁡(Σ))​(η⁡(ϕin))\mathrm{KhR}_{N}(\phi^{\prime}(\Sigma))(\eta(\phi^{\prime}_{\mathrm{in}}))=\mathrm{KhR}_{N}(\phi^{\prime}(\Sigma)\circ\phi_{\mathrm{in},t}(\Sigma_{\mathrm{in}}))(\eta(\phi_{\mathrm{in}}))=\mathrm{KhR}_{N}(\phi(\Sigma))(\eta(\phi_{\mathrm{in}}))

Thus, the definition was independent of the choice of ϕin\phi_{\mathrm{in}}. An analogous argument also establishes independence of the choice of ϕout\phi_{\mathrm{out}}. ∎

It remains to prove the lemma referenced above.

0NAX

Lemma 4.7. Let Σ⊂ℝ3×[0,1]\Sigma\subset{\mathbb{R}}^{3}\times{[0,1]} be a link cobordism and let f:ℝ3×[0,1]→ℝ3×[0,1]f:{\mathbb{R}}^{3}\times{[0,1]}\to{\mathbb{R}}^{3}\times{[0,1]} be a diffeomorphism which restricts to the identity in a neighborhood of the boundary ℝ3×{0,1}{\mathbb{R}}^{3}\times\{0,1\}. Then Σ\Sigma is isotopic rel boundary to f⁡(Σ)f(\Sigma).

0NAY

Proof. The proof would be easy if we knew that ff were isotopic to the identity, but π0​(Diff+​(ℝ3×[0,1],ℝ3×{0,1}))\pi_{0}(\mathrm{Diff}^{+}({\mathbb{R}}^{3}\times{[0,1]},{\mathbb{R}}^{3}\times\{0,1\})) is unknown. We can, however, replace ff with a diffeomorphism f′f^{\prime} which is isotopic (rel boundary) to ff, or replace ff with a diffeomorphism f′f^{\prime} which coincides with ff in a neighborhood of Σ\Sigma. In both cases, proving that f′​(Σ)f^{\prime}(\Sigma) is isotopic to Σ\Sigma easily implies that f⁡(Σ)f(\Sigma) is isotopic to Σ\Sigma.

Choose a point p∈ℝ3p\in{\mathbb{R}}^{3} such that p×[0,1]p\times{[0,1]} is disjoint from Σ\Sigma. There is no obstruction to modifying (post-composing) ff by an isotopy which takes f⁡(p×[0,1])f(p\times{[0,1]}) to p×[0,1]p\times{[0,1]}, so we may assume that ff restricts to the identity on p×[0,1]p\times{[0,1]}. (Note that this modification changes f⁡(Σ)f(\Sigma) as well as f⁡(p×[0,1])f(p\times{[0,1]}), so there is no issue of the image of p×[0,1]p\times{[0,1]} getting “caught" on the image of Σ\Sigma.)

Next consider the tangent map of ff along p×[0,1]p\times{[0,1]}. We would like to deform the tangent map to the identity, but there is an obstruction living in π1​(S​O​(3))≅ℤ/2\pi_{1}(SO(3))\cong{\mathbb{Z}}/2. We can modify (precompose) ff in a neighborhood of p×[0,1]p\times{[0,1]} (and away from Σ\Sigma) so that this obstruction vanishes. (Specifically, let f:[0,1]→[0,1]f:[0,1]\to[0,1] be a smooth function such that f⁡(t)=0f(t)=0 for tt near 0 and f⁡(t)=1f(t)=1 for tt near 1. Let γ:[0,1]→S​O​(3)\gamma:[0,1]\to SO(3) be a representative of the nontrivial element of π1​(S​O​(3))\pi_{1}(SO(3)), with γ⁡(0)=γ⁡(1)=𝟏\gamma(0)=\gamma(1)=\mathbf{1}. Let B3B^{3} be the unit ball in ℝ3{\mathbb{R}}^{3}, and for p∈B3p\in B^{3}, let |p||p| denote the distance from pp to the origin. Define a diffeomorphism of [0,1]×B3[0,1]\times B^{3} by

(s,p)↦(s,γ⁡(f⁡(s⋅(1−|p|)))​(p)).(s,\>p)\mapsto(s,\>\gamma(f(s\cdot(1-|p|)))(p)).

This diffeomorphism is the identity near [0,1]×∂B3[0,1]\times\partial B^{3} and it effects a full twist on the tangent space along [0,1]×{0}[0,1]\times\{0\}.)

Once the above obstruction vanishes we can isotope ff to a map which is the identity on a neighborhood NN of p×[0,1]∪ℝ3×{0,1}p\times{[0,1]}\cup{\mathbb{R}}^{3}\times\{0,1\}.

Choose a family of diffeomorphisms gt:ℝ3×[0,1]→ℝ3×[0,1]g_{t}:{\mathbb{R}}^{3}\times{[0,1]}\to{\mathbb{R}}^{3}\times{[0,1]}, with t∈[0,1]t\in{[0,1]}, such that g0g_{0} is the identity, gtg_{t} restricted to ℝ3×{0,1}{\mathbb{R}}^{3}\times\{0,1\} is the identity for all tt, and g1​(Σ)⊂Ng_{1}(\Sigma)\subset N. The family of surfaces f​(gt​(Σ))f(g_{t}(\Sigma)) provides an isotopy from f⁡(Σ)=f⁡(g0​(Σ))f(\Sigma)=f(g_{0}(\Sigma)) to f​(g1​(Σ))f(g_{1}(\Sigma)). But ff is the identity on NN and g1​(Σ)⊂Ng_{1}(\Sigma)\subset N, so f⁡(g1​(Σ))=g1​(Σ)f(g_{1}(\Sigma))=g_{1}(\Sigma). The family of surfaces gt​(Σ)g_{t}(\Sigma) provides an isotopy from g1​(Σ)g_{1}(\Sigma) to g0​(Σ)=Σg_{0}(\Sigma)=\Sigma. Composing these two isotopies provides the desired isotopy from f⁡(Σ)f(\Sigma) to Σ\Sigma. ∎

4.2. Link homology in abstract 3-spheres

0NAZ

Definition 4.8. A link homology for links in 33-spheres is a functor

{link embeddings in oriented ​S≅S3link cobordisms in oriented Y≅S3×[0,1] up to isotopy rel ∂}→{bigraded abelian groupshomogeneous homomorphisms}\begin{Bmatrix}\textrm{link embeddings in oriented }S\cong S^{3}\\ \textrm{link cobordisms in oriented }Y\cong S^{3}\times{[0,1]}\textrm{ up to isotopy rel }\partial\end{Bmatrix}\xrightarrow{}\begin{Bmatrix}\textrm{bigraded abelian groups}\\ \textrm{homogeneous homomorphisms}\end{Bmatrix}
0NB0

Theorem 4.9. KhRN\mathrm{KhR}_{N} extends to a link homology theory for links in 33-spheres.

The proof occupies the remainder of this subsection.

Remark. In the proof of Theorem 4.9, we will show that the sweep-around property from Theorem 1.1 is sufficient to extend a link homology for links in 33-balls to 33-spheres, without using any special properties of KhRN\mathrm{KhR}_{N}.

0NB1

Definition 4.10. Let SS be an oriented 33-manifold diffeomorphic to S3S^{3}. For any point p∈S∖Lp\in S\setminus L, we consider LL as a link in the 33-ball S∖{p}S\setminus\{p\} and denote by π:T⁡(S,L)→S∖L\pi\colon T(S,L)\to S\setminus L the bundle of bigraded abelian groups, whose fiber at the point p∈S∖Lp\in S\setminus L is KhRN​(S∖{p},L)\mathrm{KhR}_{N}(S\setminus\{p\},L) as defined in Definition 4.4.

For any path ptp_{t} in S∖LS\setminus L, we have that L×[0,1]⊂S×[0,1]∖{(pt,t)}t∈[0,1]≅ℝ3×[0,1]L\times{[0,1]}\subset S\times{[0,1]}\setminus\{(p_{t},t)\}_{t\in{[0,1]}}\cong{\mathbb{R}}^{3}\times{[0,1]}. By the results of §4.1, this cobordism induces a parallel transport isomorphism

(KhRN(pt):T(S,L)p0→T(S,L)p1)=defKhRN(S×[0,1]∖{(pt,t)}t∈[0,1],L×[0,1])(\mathrm{KhR}_{N}(p_{t})\colon T(S,L)_{p_{0}}\to T(S,L)_{p_{1}})\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\mathrm{KhR}_{N}(S\times{[0,1]}\setminus\{(p_{t},t)\}_{t\in{[0,1]}},L\times{[0,1]})
0NB2

Lemma 4.11. The parallel transport isomorphisms KhRN​(pt)\mathrm{KhR}_{N}(p_{t}) define a flat connection on T⁡(S,L)T(S,L).

0NB3

Proof. We have to show that the parallel transport isomorphisms associated to closed loops ptp_{t} in S∖LS\setminus L are identity maps. Suppose first that ptp_{t} is a contractible loop. Then the pair (S×[0,1]∖{(pt,t)}t∈[0,1],L×[0,1])(S\times{[0,1]}\setminus\{(p_{t},t)\}_{t\in{[0,1]}},L\times{[0,1]}) is diffeomorphic to a pair (ℝ3×[0,1],Σ)({\mathbb{R}}^{3}\times{[0,1]},\Sigma) where Σ\Sigma is isotopic to an identity link cobordism, which implies that the parallel transport isomorphism KhRN​(pt)\mathrm{KhR}_{N}(p_{t}) is the identity. This also implies that the parallel transport isomorphisms associated to isotopic paths between two points p0p_{0} and p1p_{1} in S∖LS\setminus L are equal. Now suppose that the loop ptp_{t} is a small meridian around a component of LL. Then the pair (S×[0,1]∖{(pt,t)}t∈[0,1],L×[0,1])(S\times{[0,1]}\setminus\{(p_{t},t)\}_{t\in{[0,1]}},L\times{[0,1]}) is diffeomorphic to a pair (ℝ3×[0,1],Σ)({\mathbb{R}}^{3}\times{[0,1]},\Sigma) where Σ\Sigma is a sweep-around cobordism as in (1.1). By Theorem 1.1, it follows that the parallel transport isomorphism KhRN​(pt)\mathrm{KhR}_{N}(p_{t}) is the identity. Since π1​(S∖L)\pi_{1}(S\setminus L) is generated by such small meridian loops, it follows that the parallel transport isomorphism for every loop ptp_{t} is the identity. ∎

0NB4

Definition 4.12. Let L⊂SL\subset S be a link embedded in a 3-sphere. Then we define the Khovanov–Rozansky homology of LL in SS to be

KhRN​(S,L)=defΓflat​(T⁡(S,L)),\mathrm{KhR}_{N}(S,L)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\Gamma_{\textrm{flat}}(T(S,L)),

the bigraded abelian group of flat sections of the bundle T⁡(S,L)T(S,L).

Note that every point p∈S∖Lp\in S\setminus L induces a grading-preserving isomorphism

KhRN​(S,L)→KhRN​(S∖{p},L)\mathrm{KhR}_{N}(S,L)\to\mathrm{KhR}_{N}(S\setminus\{p\},L)

of evaluating sections at the point pp.

0NB5

Definition 4.13. Consider a link cobordism Σ⊂W\Sigma\subset W in a 4-manifold WW diffeomorphic to S3×[0,1]S^{3}\times{[0,1]}. Let Σin⊂Win\Sigma_{\mathrm{in}}\subset W_{\mathrm{in}} and Σout⊂Wout\Sigma_{\mathrm{out}}\subset W_{\mathrm{out}} denote the boundary links in the incoming and outgoing boundary 3-spheres of WW. Now we define

KhRN​(W,Σ):KhRN​(Win,Σin)→KhRN​(Wout,Σout)\mathrm{KhR}_{N}(W,\Sigma)\colon\mathrm{KhR}_{N}(W_{\mathrm{in}},\Sigma_{\mathrm{in}})\to\mathrm{KhR}_{N}(W_{\mathrm{out}},\Sigma_{\mathrm{out}})

by first choosing a path pt⊂W∖Σp_{t}\subset W\setminus\Sigma from p0∈Win∖Σinp_{0}\in W_{\mathrm{in}}\setminus\Sigma_{\mathrm{in}} to p1∈Wout∖Σoutp_{1}\in W_{\mathrm{out}}\setminus\Sigma_{\mathrm{out}}. Then we have W∖{(pt,t)}t∈[0,1]≅ℝ3×[0,1]W\setminus\{(p_{t},t)\}_{t\in{[0,1]}}\cong{\mathbb{R}}^{3}\times{[0,1]} and we declare KhRN​(W,Σ)​(η)\mathrm{KhR}_{N}(W,\Sigma)(\eta), for a flat section η∈KhRN​(Win,Σin)\eta\in\mathrm{KhR}_{N}(W_{\mathrm{in}},\Sigma_{\mathrm{in}}), to be the unique flat section of KhRN​(Wout,Σout)\mathrm{KhR}_{N}(W_{\mathrm{out}},\Sigma_{\mathrm{out}}) with value

KhRN​(W,Σ)​(η)​(pout)=KhRN​(W∖{(pt,t)}t∈[0,1],Σ)​(η⁡(pin))\mathrm{KhR}_{N}(W,\Sigma)(\eta)(p_{\mathrm{out}})=\mathrm{KhR}_{N}(W\setminus\{(p_{t},t)\}_{t\in{[0,1]}},\Sigma)(\eta(p_{\mathrm{in}}))
0NB6

Lemma 4.14. KhRN​(W,Σ)\mathrm{KhR}_{N}(W,\Sigma) is independent of the choice of the path ptp_{t}.

0NB7

Proof. Let us first fix a choice of endpoints p0∈Win∖Σinp_{0}\in W_{\mathrm{in}}\setminus\Sigma_{\mathrm{in}} and p1∈Wout∖Σoutp_{1}\in W_{\mathrm{out}}\setminus\Sigma_{\mathrm{out}}. Then any two choices of paths pt∈p_{t}\in and pt′p^{\prime}_{t} from p0p_{0} to p1p_{1} can be related by isotopy in W∖ΣW\setminus\Sigma or splicing in a little loop linking a component of Σ\Sigma. As before, isotopic paths give rise to isotopic surfaces in ℝ3×[0,1]{\mathbb{R}}^{3}\times{[0,1]}, which induce equal maps. Similarly, in the case of a linking loop, we can choose a standard local model and then notice that the sweep-around property from Theorem 1.1 implies that the two paths induce the same map. Finally, the independence from the choice of endpoints p0∈Win∖Σinp_{0}\in W_{\mathrm{in}}\setminus\Sigma_{\mathrm{in}} and p1∈Wout∖Σoutp_{1}\in W_{\mathrm{out}}\setminus\Sigma_{\mathrm{out}} follows as in the proof of Lemma 4.6. ∎

This completes the proof of Theorem 4.9.

4.3. Monoidality

Links in 33-balls and their cobordisms form a symmetric monoidal category under boundary connect sum, which is respected by KhRN\mathrm{KhR}_{N} as we will now see.

0NB8

Proposition 4.15. The Khovanov–Rozansky homologies KhRN\mathrm{KhR}_{N} are lax symmetric monoidal functors.

0NB9

Proof. Let L1∈B1L_{1}\in B_{1} and L2∈B2L_{2}\in B_{2} and write LL for the resulting split disjoint union in B=defB1​#∂​B2B\stackrel{{\scriptstyle\mathrm{def}}}{{=}}B_{1}\#_{\partial}B_{2}. We can find a diffeomorphism ϕ:B→ℝ3\phi\colon B\to{\mathbb{R}}^{3} such that not only is LL generic and blackboard-framed, but also the zz-projections of the L1L_{1} and L2L_{2} components of LL are contained in disjoint disks in ℝ2{\mathbb{R}}^{2}. Then, monoidality on the chain level is manifest in the definition of KhRN\mathrm{KhR}_{N}, and we get

KhRN​(B1,L1)⊗KhRN​(B2,L2)\displaystyle\mathrm{KhR}_{N}(B_{1},L_{1})\otimes\mathrm{KhR}_{N}(B_{2},L_{2}) ≅KhRN​(ϕ⁡(L1))⊗KhRN​(ϕ⁡(L2))\displaystyle\cong\mathrm{KhR}_{N}(\phi(L_{1}))\otimes\mathrm{KhR}_{N}(\phi(L_{2}))
→KhRN​(ϕ⁡(L1⊔L2))\displaystyle\to\mathrm{KhR}_{N}(\phi(L_{1}\sqcup L_{2}))
≅KhRN​(B1​#∂​B2,L1⊔L2)\displaystyle\cong\mathrm{KhR}_{N}(B_{1}\#_{\partial}B_{2},L_{1}\sqcup L_{2})

where the map in the second line comes from the Künneth theorem (this is guaranteed to be an isomorphism when working with field coefficients). The compatibility on the level of morphisms is verified similarly. ∎

Given a finite collection of links in 33-balls Li⊂BiL_{i}\subset B_{i}, we can also define

KhRN(⊔iBi,⊔iLi)=def⨂iKhRN(Bi,Li).\mathrm{KhR}_{N}(\sqcup_{i}B_{i},\sqcup_{i}L_{i})\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\bigotimes_{i}\mathrm{KhR}_{N}(B_{i},L_{i}).

Then the proof of the proposition implies that the boundary connect sum of 33-balls induces natural homomorphisms (and even isomorphisms when working with field coefficients)

KhRN(⊔iBi,⊔iLi)→KhRN(#∂Bi,⊔iLi).\mathrm{KhR}_{N}(\sqcup_{i}B_{i},\sqcup_{i}L_{i})\to\mathrm{KhR}_{N}(\#_{\partial}B_{i},\sqcup_{i}L_{i}).

Remark. This monoidality property can be interpreted as saying that KhRN\mathrm{KhR}_{N} categorifies the 𝔤​𝔩N\mathfrak{gl}_{N} skein algebra of ℝ2{\mathbb{R}}^{2}. For more on skein algebra categorification we refer to [QW21].

Original mathematics by the credited authors. Source collection and HTML conversion remain in progress.

Scott Morrison, Kevin Walker, Paul Wedrich

Original source: arXiv:1907.12194v5