Bar-Natan Homology Overview

• Bar-Natan homology is a deformation of Khovanov homology defined via a rank-2 Frobenius system and dotted cobordism relations to study knot invariants.
• It employs spectral sequences and filtered specializations, connecting the theory to Rasmussen invariants, instanton homology, and concordance obstructions.
• The approach extends to equivariant, virtual, and four-dimensional settings, offering computational tools for analyzing torsion, unknotting, and ribbon distance.

Bar-Natan homology is a deformation of Khovanov homology that is defined from the same cube of resolutions but with a deformed rank-$2$ Frobenius system, and it is equally expressible in a dotted-cobordism formalism. In the characteristic-$2$ formulations emphasized across the literature, the deformation parameter is written either $h$ or $H$, ordinary Khovanov homology is recovered by specialization, and filtered specializations at $h=1$ or $H=1$ connect the theory to Rasmussen-type invariants, instanton spectral sequences, and symmetry-sensitive refinements (Alishahi, 2017, Kronheimer et al., 2019, Kim, 10 Apr 2026).

1. Algebraic definition and Frobenius-system realizations

A standard characteristic-$2$ presentation defines Bar-Natan homology from the formal Bar-Natan complex built from all $0/1$-resolutions of a link diagram and a $2$-dimensional TQFT associated to the Frobenius algebra

$$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$

Applying this TQFT produces the chain complex $2$0 of $2$1-modules, and its homology is $2$2 (Kim, 10 Apr 2026).

Closely related papers package the same deformation in Frobenius-system language. One formulation uses

$2$3

with

$2$4

while another starts from the universal characteristic-$2$5 Frobenius system

$2$6

and identifies Bar-Natan homology as the specialization $2$7, yielding $2$8 and the graded homology $2$9 (Gujral, 2020, Kronheimer et al., 2019).

The grading conventions retain the Khovanov-style bigrading, with the formal parameter of bidegree

$h$0

and a filtered specialization arises by setting $h$1. In the periodic-link literature this is expressed as the base change $h$2, producing a filtered Bar-Natan differential $h$3 with deformation term of bidegree $h$4 (Alishahi, 2017, Borodzik et al., 2017).

2. Cobordism categories, reduced theories, and functorial variants

The geometric realization of Bar-Natan homology begins with a formal complex assembled from a cube of resolutions in a quotient cobordism category. In one formulation this category is generated by dotted cobordisms and quotiented by relations $h$5, $h$6, $h$7, $h$8-trading, and the derived $h$9-Tu relation; for a strongly invertible knot $H$0, Sano’s construction gives an involution $H$1 on the formal Bar-Natan complex, and the involutive Bar-Natan complex is defined by

$H$2

After applying the same TQFT one obtains $H$3, whose homology $H$4 is the involutive Bar-Natan homology; in that paper the tilde notation is reserved for the equivariant or involutive theory, not for a reduced theory (Kim, 10 Apr 2026).

At the bicategorical level, the Bar-Natan bicategory $H$5 has objects given by collections of boundary points, $H$6-morphisms given by flat tangles, and $H$7-morphisms given by dotted cobordisms modulo the usual local relations. A central structural theorem identifies Blanchet’s web-and-foam bicategory with a weighted Bar-Natan bicategory $H$8, and the resulting equivalence transports Khovanov-type theories factoring through $H$9 to a foam setting in which functoriality is strict rather than merely projective (Beliakova et al., 2019).

Reduced Bar-Natan theories are obtained by choosing a basepoint and separating generators according to whether the basepoint circle is labeled $h=1$0 or $h=1$1. Writing $h=1$2 for the subcomplex with basepoint label $h=1$3 and $h=1$4 for the quotient identified with generators having basepoint label $h=1$5, one has a short exact sequence

$h=1$6

Over $h=1$7, the Bar-Natan homology splits as

$h=1$8

and the two reduced theories are in fact isomorphic at chain level: $h=1$9 as graded chain complexes over $H=1$0 (Wigderson, 2015).

3. Spectral sequences, immersed curves, and stable homotopy refinements

Bar-Natan homology appears naturally as the $H=1$1-page of gauge-theoretic spectral sequences. In characteristic $H=1$2, one obtains

$H=1$3

and a reduced version

$H=1$4

where the instanton target is defined with a specialized local system. A related paper further states, at the abstract level, that a spectral sequence from Bar-Natan’s variant of Khovanov homology to a deformation of instanton homology yields concordance invariants, including a $H=1$5-parameter family of homomorphisms $H=1$6 from the concordance group to the reals with the potential to provide independent bounds on the genus and number of double points for immersed surfaces with boundary a given knot (Kronheimer et al., 2019, Kronheimer et al., 2019).

For $H=1$7-ended tangles, Bar-Natan’s universal invariant admits a geometric reinterpretation in terms of multicurves in the $H=1$8-punctured sphere. The associated invariant $H=1$9 satisfies a gluing theorem recovering reduced Bar-Natan homology from wrapped Lagrangian Floer homology, and this framework is used to prove that Conway mutation preserves reduced Bar-Natan homology over $2$0 and Rasmussen’s $2$1-invariant over any field (Kotelskiy et al., 2019).

Bar-Natan homology also has a spatial refinement. For any link diagram $2$2, there is a CW-spectrum $2$3 whose reduced cellular cochain complex is the Bar-Natan complex, and the stable homotopy type is a link invariant. The resulting Bar-Natan homotopy type is stably equivalent to a wedge sum of canonical cells indexed by orientations, and the same work conjectures a lift of the quantum filtration to the spatial level, leading to a proposed cohomotopical refinement $2$4 of the Rasmussen invariant (Sano, 2021).

4. Torsion, unknotting, ribbon distance, and concordance-type invariants

A major theme in Bar-Natan homology is the extraction of numerical invariants from $2$5- or $2$6-torsion. For a torsion class $2$7 with $2$8, its order defines the invariant

$2$9

and explicit crossing-change maps $0/1$0 satisfy

$0/1$1

From this one obtains the inequality

$0/1$2

The same paper relates this torsion order to the Bar-Natan spectral sequence at $0/1$3, notes that if the spectral sequence collapses on the $0/1$4-th page then $0/1$5, and gives examples $0/1$6 for which $0/1$7 but $0/1$8 (Alishahi, 2017).

The same torsion philosophy extends to ribbon distance. Defining

$0/1$9

one has, if $2$0 is the ribbon distance between knots $2$1 and $2$2,

$2$3

and

$2$4

In particular,

$2$5

so the Bar-Natan lower bound on ribbon distance agrees with Alishahi’s lower bound on unknotting number when one of the knots is the unknot (Gujral, 2020).

For strongly invertible knots, the involutive theory provides an equivariant analogue. Writing

$2$6

where $2$7 count the three types of $2$8-symmetric crossing changes, the main theorem is

$2$9

The paper computes $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$0 for all prime strongly invertible knots of crossing number at most $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$1 in Sakuma’s table and identifies five knots,

$$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$2

for which $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$3 but $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$4; more precisely, $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$5 for $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$6 and $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$7, while $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$8 for $$V = F[H]\langle 1,x\rangle,\qquad \Delta(1)=1\otimes x + x\otimes 1 + H\,1\otimes 1,\qquad \Delta(x)=x\otimes x.$$9 (Kim, 10 Apr 2026).

5. Extensions to periodic, virtual, surface, and projective settings

Bar-Natan homology admits equivariant forms for periodic links. If $2$00 is $2$01-periodic and $2$02, then the Khovanov-type chain complex becomes a complex of $2$03-modules, and applying the characteristic-$2$04 Bar-Natan deformation gives equivariant Bar-Natan homology $2$05. The resulting spectral sequence has

$2$06

and converges to $2$07; for semisimple coefficients, the equivariant groups decompose according to the orbit structure of orientations under the cyclic symmetry (Borodzik et al., 2017).

The cobordism formalism also extends beyond classical links in $2$08. For virtual links and tangles, one replaces the classical cobordism category by a category $2$09 of possibly non-orientable cobordisms with boundary decorations, introducing a non-orientable generator $2$10 to represent saddles that preserve the number of circles. This construction recovers the classical theory on classical diagrams and, in characteristic $2$11, extends Bar-Natan’s $2$12-theory in two non-equivalent ways, distinguished by whether $2$13 or $2$14. For framed links in thickened surfaces $2$15, a different extension replaces diagrammatic generators by embedded foams, produces an infinite family of homology theories, and yields a simpler theory whose graded Euler characteristic is exactly the Kauffman bracket of the link in the surface (Tubbenhauer, 2011, 0810.5566).

For links in $2$16, two distinct Bar-Natan-type extensions appear in the supplied literature. One defines $2$17 for null homologous links over $2$18, with the essential new input of twisted orientations and a $2$19-to-$2$20 bifurcation map forced to be $2$21; if all components are null homologous then

$2$22

with canonical basis indexed by twisted orientations, and the induced $2$23-invariant gives genus bounds for twisted orientable slice surfaces (Chen, 2023). Another constructs a doubled Bar-Natan theory $2$24 for links in $2$25 by combining pair-of-pants maps with once-punctured Möbius-band maps; $2$26 is the target of a spectral sequence from doubled Khovanov homology, has rank

$2$27

depending on degeneracy, and is stated to be distinct from Chen’s theory even though the two have equivalent rank and homological-support information in the nullhomologous case (Rushworth, 11 Mar 2026).

6. Computational uses and four-dimensional extensions

Bar-Natan homology has become a central computational tool in low-dimensional topology. One recent proof that the Khovanov homology of the closure of a $2$28-strand braid has only $2$29-torsion relies on Bar-Natan’s version of Khovanov homology for tangles, Bar-Natan’s delooping and Gaussian-elimination techniques, and the reduced integral Bar-Natan–Lee–Turner spectral sequence. A parallel line of work introduces discrete Morse theory in Bar-Natan’s dotted cobordism category, producing recursive descriptions of the complexes of $2$30- and $2$31-torus braids and new torsion results for closures of $2$32-braids in the families $2$33 (Schuetz, 20 Jan 2025, Kelomäki, 2023).

The Bar-Natan package also extends into $2$34-manifold topology through skein lasagna modules. In this setting the input Frobenius algebra is

$2$35

and the resulting module $2$36 is naturally an $2$37-module in which multiplication by $2$38 is interpreted geometrically as attaching a $2$39-handle to a skein surface. This identification leads to a notion of lasagna $2$40-torsion order and is used to construct exotic pairs of surfaces with boundary in $2$41-manifolds for which one internal stabilization is not enough; the flagship example glues Hayden’s exotic pair to the exceptional sphere in $2$42 (Sullivan, 4 Apr 2025).

Bar-Natan homology therefore occupies several roles simultaneously: a deformation of Khovanov homology, a cobordism-theoretic and functorial package for tangles, a source of torsion-based unknotting and concordance obstructions, a bridge to instanton and Floer-theoretic constructions, and a framework that continues to admit new generalizations in equivariant, virtual, projective, and four-dimensional settings.