ScalingStacks

8.4 Invariants

A plane diagram DD of an oriented (1,1)(1,1)-tangle LL is a generic projection of LL onto ℝ×[0,1].\mathbb{R}\times[0,1].

If DD is a plane diagram of an oriented (1,1)(1,1)-tangle, define x⁡(D)x(D) and y⁡(D)y(D) in the same way as for plane diagrams of oriented links (see Section 2.4).

Fix a graded AA-module M.M. Let nn be the number of double points of DD, so that n=x⁡(D)+y⁡(D)n=x(D)+y(D) and ℐ\mathcal{I} the set of double points of D.D. To MM and DD associate a commutative ℐ\mathcal{I}-cube VDMV_{D}^{M} over the category A​-mod0A\mbox{-mod}_{0} of graded AA-modules as follows.

For ℒ⊂ℐ\mathcal{L}\subset\mathcal{I} the ℒ\mathcal{L}-resolution D⁡(ℒ)D(\mathcal{L}) of DD consists of a disjoint union of circles and an interval. The functor FMF_{M} (see Section 8.2) assigns a graded AA-module to D⁡(ℒ).D(\mathcal{L}). Define

VDM​(ℒ)=FM​(D⁡(ℒ))​{−|ℒ|}.V_{D}^{M}(\mathcal{L})=F_{M}(D(\mathcal{L}))\{-|\mathcal{L}|\}. (203)

Maps between VDM​(ℒ)V_{D}^{M}(\mathcal{L}) for various subsets ℒ\mathcal{L} are defined by the procedure completely analogous to the one described in Section 4.2. Due to shifts {−|ℒ|},\{-|\mathcal{L}|\}, these maps of graded AA-modules are grading-preserving, rather than just graded maps, so that VDMV_{D}^{M} is a commutative cube over A​-mod0.A\mbox{-mod}_{0}.

Example: For a diagram D,D, depicted below,

[Uncaptioned image]

resolutions of DD are

[Uncaptioned image]

so that

VDM​(∅)\displaystyle V_{D}^{M}(\emptyset) =\displaystyle= A⊗M\displaystyle A\otimes M
VDM​(ℐ)\displaystyle V_{D}^{M}(\mathcal{I}) =\displaystyle= M​{−1}\displaystyle M\{-1\}

and the structure map VDM​(∅)⟶VDM​(ℐ)V_{D}^{M}(\emptyset)\longrightarrow V_{D}^{M}(\mathcal{I}) is the multiplication map mM:A⊗M→M⁡{−1}.m_{M}:A\otimes M\to M\{-1\}.

Next we transform the commutative ℐ\mathcal{I}-cube VDMV_{D}^{M} into a skew commutative ℐ\mathcal{I}-cube by putting minus signs in front of some structure maps of VDM,V_{D}^{M}, or, equivalently, by tensoring it with Eℐ.E_{\mathcal{I}}. Denote by C¯M​(D)\overline{C}_{M}(D) the complex C¯​(VDM⊗Eℐ)\overline{C}(V_{D}^{M}\otimes E_{\mathcal{I}}) of graded AA-modules. Define

CM​(D)=C¯M​(D)​[x⁡(D)]​{y⁡(D)−2​x​(D)}C_{M}(D)=\overline{C}_{M}(D)[x(D)]\{y(D)-2x(D)\} (204)

Denote the ii-th cohomology group of the complex CM​(D)C_{M}(D) by Hi​(D,M).H^{i}(D,M). These cohomology groups are graded AA-modules. Denote the jj-th graded component of Hi​(D,M)H^{i}(D,M) by Hi,j​(D,M)H^{i,j}(D,M)

0PMP

Theorem 3 For a graded AA-module M,M, an oriented (1,1)(1,1)-tangle LL and a diagram DD of L,L, isomorphism classes of graded AA-modules Hi​(D,M)H^{i}(D,M) do not depend on the choice of DD and are invariants of L.L.

Our proof of Theorem 1 immediately generalizes without essential modifications to a proof of Theorem 3. Proposition 38 is used to establish direct sum decompositions of CM​(D)C_{M}(D), analogous to decompositions of C⁡(D)C(D), for suitable D,D, given by Propositions 11, 14, 18.1, 21.1.

□\square

Cohomology groups Hi​(D),H^{i}(D), defined in Section 4.2, is a special case of groups Hi​(D,M),H^{i}(D,M), as the next proposition explains.

0PMQ

Proposition 39 Let DD be a diagram of an oriented (1,1)(1,1)-tangle LL and denote by cl⁡(D){\mathrm{cl}}(D) the associated diagram of the marked oriented link cl⁡(L).{\mathrm{cl}}(L). Considering AA as a graded AA-module, we have a canonical isomorphism of cohomology groups (as graded RR-modules)

Hi​(D,A)≅Hi​(cl⁡(D)),i∈ℤ.H^{i}(D,A)\cong H^{i}({\mathrm{cl}}(D)),\hskip 36.135pti\in\mathbb{Z}. (205)

□\square

Given a finitely-generated graded AA-module M,M, define the graded Euler characteristic χ^​(M)\widehat{\chi}(M) by

χ^​(M)=∑j∈ℤdimℚ​(Mj⊗ℤℚ)\widehat{\chi}(M)=\sum_{j\in\mathbb{Z}}{\mathrm{dim}}_{\mathbb{Q}}(M_{j}\otimes_{\mathbb{Z}}\mathbb{Q}) (206)
0PMR

Proposition 40 Let MM be a finitely-generated graded AA-module, LL an oriented (1,1)(1,1)-tangle and DD a diagram of LL. Then

K⁡(cl⁡(L))​χ^​(M)q+q−1=∑i,j∈ℤ(−1)i​qj​dimℚ​(Hi,j​(D,M)⊗ℤℚ).\frac{K({\mathrm{cl}}(L))\widehat{\chi}(M)}{q+q^{-1}}=\sum_{i,j\in\mathbb{Z}}(-1)^{i}q^{j}{\mathrm{dim}}_{\mathbb{Q}}(H^{i,j}(D,M)\otimes_{\mathbb{Z}}\mathbb{Q}). (207)

that is, the Kauffman bracket of cl⁡(L){\mathrm{cl}}(L) is proportional to the Euler characteristic of groups Hi,j​(D,M).H^{i,j}(D,M).

□\square

Given two graded AA-modules M,NM,N and a grading-preserving homomorphism f:M→Nf:M\to N, it induces a map of commutative cubes VDM→VDN,V_{D}^{M}\to V_{D}^{N}, which, in turn, induces a map of complexes CM​(D)→CN​(D)C_{M}(D)\to C_{N}(D) and a map of cohomology groups Hi​(D,M)→Hi​(D,N).H^{i}(D,M)\to H^{i}(D,N). So, in fact, each diagram DD of an oriented long link defines functors HDiH^{i}_{D} from the category A​-mod0A\mbox{-mod}_{0} of graded AA-modules to itself, HDi​(M)=Hi​(D,M).H^{i}_{D}(M)=H^{i}(D,M). If two diagrams D1,D2D_{1},D_{2} are related by a Reidemeister move, constructions of Section 5 extend to functor isomorphism HD1i⟶≅HD2i.H^{i}_{D_{1}}\stackrel{{\scriptstyle\cong}}{{\longrightarrow}}H^{i}_{D_{2}}. Let us frame this observation into a proposition.

0PMS

Proposition 41 For an oriented (1,1)-tangle LL and a diagram DD of LL isomorphism classes of functors

HDi:A​-mod0⟶A​-mod0H^{i}_{D}:A\mbox{\rm-mod}_{0}\longrightarrow A\mbox{\rm-mod}_{0} (208)

do not depend on the choice of DD and are invariants of L.L.

Oriented long links with one component correspond one-to-one to oriented knots in ℝ3.\mathbb{R}^{3}. Thus, Proposition 41 gives invariants of oriented knots in ℝ3.\mathbb{R}^{3}. Moreover, if a diagram DD represents an oriented knot and D′D^{\prime} is the diagram obtained from DD by reversing the orientation of the underlying curve, there is a natural in MM isomorphism Hi​(D,M)=Hi​(D′,M).H^{i}(D,M)=H^{i}(D^{\prime},M). Consequently, for knots, isomorphism classes of functors HDiH^{i}_{D} do not depend on the orientation, and HDiH^{i}_{D} provide “functor-valued” invariants of non-oriented knots. Of course, these invariants depend on how the ambient 3-space is oriented.

Let DD be the 3-crossing diagram of the left-hand trefoil (knot T2,3T_{2,3} in the notations of Section 6.2). The functors HDiH^{i}_{D} are written below

HD−3​(M)\displaystyle H^{-3}_{D}(M) =\displaystyle= ker​2​X​(M)​{8},\displaystyle{\mathrm{ker}}\hskip 3.61371pt2X(M)\{8\},
HD−2​(M)\displaystyle H^{-2}_{D}(M) =\displaystyle= (M/(2​X​M))​{6},\displaystyle(M/(2XM))\{6\},
HD0​(M)\displaystyle H^{0}_{D}(M) =\displaystyle= M​{2},\displaystyle M\{2\},
HDi​(M)\displaystyle H^{i}_{D}(M) =\displaystyle= 0​ for all other values of ​i,\displaystyle 0\mbox{ for all other values of }i,

where ker​2​X​(M)={t∈M|2​X​t=0}.{\mathrm{ker}}\hskip 3.61371pt2X(M)=\{t\in M|2Xt=0\}.

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

Mikhail Khovanov

Original source: arXiv:math/9908171v2