--- title: Functoriality of Khovanov Homology url: https://www.emergentmind.com/topics/functoriality-of-khovanov-homology type: topic --- # Functoriality of Khovanov Homology Searching arXiv for recent and foundational papers on functoriality of Khovanov homology. Functoriality of Khovanov homology is the problem of assigning to an oriented link cobordism \(S\colon L_0\to L_1\) a morphism on Khovanov-type complexes or homology groups in a way compatible with isotopy and composition. In the classical Bar–Natan formalism, a movie of Reidemeister and Morse moves produces a chain map \(BN(S)\), but the map is only well defined up to an overall sign; equivalently, the induced map on homology is defined only up to \(\pm\). A central theme of the subject is therefore the elimination, control, or reinterpretation of this ambiguity. The resulting literature contains several distinct outcomes: strict functoriality on the chain level, functoriality only up to sign, functoriality up to global invertible scalars, and functoriality only after passage to homotopy or spectra [2008.02131][1405.7246][2410.23455][2104.12907]. ## 1. Classical formulation and the source of the sign ambiguity For an oriented link diagram \(D\) with \(n\) crossings, the ordinary Khovanov complex is built from the cube of resolutions. Each vertex \(I\in\{0,1\}^n\) determines a smoothing with \(k(I)\) circles, and one sets \[ C^i(D)=\bigoplus_{|I|=i+n_-} A^{\otimes k(I)}, \] where \(A=R[X]/(X^2-hX-t)\) is the chosen Frobenius algebra. The differential is the sum over cube edges \(I\to I'\) of multiplication \(m\) when two circles merge and comultiplication \(\Delta\) when one circle splits. To force \(d^2=0\), one chooses an ordering of the crossings, which produces the usual edge-signs \(( -1)^{\sigma(I,e)}\). These signs are harmless for the homology of a fixed diagram, but they are the source of the familiar “up to sign” behavior for cobordism maps [2008.02131]. Bar–Natan’s formalism assigns to each elementary cobordism \(S\) — a Reidemeister move or a Morse critical point — a chain map \(BN(S)\colon C(D)\to C(D')\). By the movie-move argument, these satisfy \[ BN(S_2\circ S_1)\simeq \pm BN(S_2)\circ BN(S_1), \] so the induced map on homology is well defined only up to an overall sign. This point is often obscured by the phrase “functoriality of Khovanov homology”: in the original \(sl_2\) theory, the issue is not invariance of the homology groups, but the nonuniqueness of the cobordism maps. The anomaly appears already in movie moves involving Reidemeister moves and persists until additional structure or explicit sign corrections are introduced [2008.02131][0806.0601]. ## 2. Movie moves, elementary cobordisms, and what strict functoriality means A link cobordism in \(\mathbb R^3\times I\) is represented by a movie consisting of planar isotopies, Reidemeister moves, births, deaths, and saddles. Carter–Saito movie moves generate isotopies of movies, so functoriality reduces to checking that two movies related by these local moves induce the same chain map. In the classical setting, the check typically yields equality only up to sign. A strict theory is one in which every movie-move relation becomes an exact equality of chain maps, not merely equality up to \(\pm\) or up to homotopy with a residual sign [2008.02131]. This distinction is visible across the literature. Blanchet’s oriented model produces a chain map \(F(\Sigma)\colon K(D_0)\to K(D_1)\) of bidegree \((\chi(\Sigma),0)\) and proves \[ F(\mathrm{id}_D)=\mathrm{id}_{K(D)},\qquad F(\Sigma_2\circ\Sigma_1)=F(\Sigma_2)\circ F(\Sigma_1) \] in \(K^b(\mathbb Z\text{-Mod})\), with no sign ambiguity. By contrast, the generalized bracket of Migdail–Wehrli yields a pseudo-functor in which \[ \Phi_{G\circ F}\simeq c(F,G)\cdot(\Phi_G\circ\Phi_F), \] where \(c(F,G)\in\{\pm1,\pm\pi\}\), and strict functoriality appears only after quotienting by these units. At the stable-homotopy level, Khovanov spectra are functorial only up to pointed homotopy and sign. The phrase “functoriality” therefore refers to a family of progressively stronger statements rather than a single theorem [1405.7246][2410.23455][2104.12907]. ## 3. Explicit sign correction in the classical Bar–Natan setup A particularly direct resolution of the sign problem stays inside the ordinary Khovanov complex and modifies only the signs of the elementary cobordism maps. For a Reidemeister move \(D\to D'\), write \(w(D)\) for the writhe, \(r(D)\) for the number of Seifert circles, set \[ \delta w=w(D')-w(D),\qquad \delta r=r(D')-r(D),\qquad j=\frac{\delta w-\delta r}{2}, \] and let \(c=\sqrt{h^2+4t}\). On canonical Lee-style generators, the standard Bar–Natan map sends \[ [(D)]\mapsto \pm c^j[(D')],\qquad [\overline{(D)}]\mapsto \pm(-c)^j[\overline{(D')}], \] with the two signs differing by \((-1)^j\). The corrected Reidemeister map is defined by \[ f'_R(S)=( -1)^jBN(S), \] equivalently \(BN'(S)=\varepsilon_R(S)\cdot BN(S)\), with \(\varepsilon_{RII}=+1\) and the \(RI\) and \(RIII\) signs given by \(( -1)^{(\delta w-\delta r)/2}\). After correction, one has \[ BN'(S)[(D)]=c^j[(D')],\qquad BN'(S)[\overline{(D)}]=(-c)^j[\overline{(D')}], \] with no leftover \(\pm\) [2008.02131]. For an elementary Morse cobordism \(S\), let \(\delta r=r(D')-r(D)\) and \(\chi(S)\) be the Euler characteristic of the surface piece. Then \[ j=\frac{-\delta r-\chi(S)}{2},\qquad \varepsilon_M(S)=( -1)^j, \] and one defines \(BN'(S)=\varepsilon_M(S)\cdot BN(S)\). In particular, the correction is specified separately for births, deaths, and saddles. The claim is then checked case by case, or by Sano’s movie-move analysis: with these \(\varepsilon\)-factors inserted, every Carter–Saito movie move becomes an exact equality of chain maps. Consequently, \[ D\mapsto C(D),\qquad S\mapsto BN'(S) \] defines a genuine functor \[ Cob^4(\emptyset)\to Kom(Mat(Cob^3)), \] and after applying the TQFT \(\mathfrak F_{h,t}\) and taking homology, one obtains for every variant \(H_{h,t}(-;R)\) of Khovanov homology a strictly functorial cobordism map without sign ambiguity [2008.02131]. The significance of this construction is methodological. Earlier fixes typically changed the algebraic or foam-theoretic framework. Here the classical cube-of-resolutions model is retained, and strict functoriality is recovered by an explicit normalization of the Reidemeister and Morse maps. This suggests that, in the ordinary rank-two setting, the obstruction can be packaged as a finite combinatorial sign bookkeeping problem rather than as a defect of the entire theory. ## 4. Oriented and rank-two models with strict functoriality A different line of resolution replaces the unoriented surface formalism by an oriented one. Blanchet gives an oriented \(sl(2)\) state model over \(\mathbb Z\), with chain groups \[ K^{i,j}(D)=\bigoplus_{s(D)=i} V(D_s)\{-(w(D)+s(D))\}\otimes \Lambda^{d_s}\Delta_s, \] and differential \[ \partial=\sum_{c,\;s\to s'} \epsilon(c,s)\,[V(\Sigma_c)\otimes \mu_c], \] where \(\epsilon(c,s)\) is the usual Koszul sign and \(\mu_c\) is exterior multiplication or contraction depending on the crossing type. The boundary operator is defined as a twisted TQFT action on trivalent graphs and oriented trivalent cobordisms, and the resulting theory is strictly functorial over integers [1405.7246]. In this oriented framework, local relations for bubbles, bands, and tubes encode the signs intrinsically, rather than correcting them afterward. A later exposition formulates the theory in a foam category \(F(B)\) of blue-red oriented trivalent graphs and decorated foams, with a tautological functor \(G\), an oriented differential \[ d=\sum_{s\to s'}(-1)^{\sum_{k