Invariance and the knot Floer cube of resolutions (Q259628)

Knot theory involves two major approaches. The first is analytic: one thinks of a knot as a geometric object and hopes to extract topological information from the three-dimensional geometry. The work of constructing knot invariants then includes showing that an alleged invariant indeed is independent of the geometric choices. Alexander's original definition of the Alexander polynomial is of this type. The second is algebraic: one thinks of a knot as a planar diagram and hopes to extract topological information from the combinatorics. The work then includes showing that an alleged invariant is preserved by the Reidemeister moves. The Jones and HOMFLYPT polynomials are famous examples of this second type of invariant. It is, of course, vital to be able to move between these two approaches. Conway's skein-theoretic description of the Alexander polynomial is the paragon example, and allows for algebraic verification of invariance. This excellent paper provides such a link between algebraic and analytic approaches for the knot Floer homology of Ozsváth and Szabó. Knot Floer homology is a categorification of the Alexander polynomial, and like its shadow is traditionally defined analytically, although there are by now many descriptions. It stands in contrast to the algebraically-defined categorifications of Jones and HOMFLYPT polynomials due to Khovanov and Rozansky. These latter categorifications are almost uniformly studied in terms of ``cubes of resolutions'': one takes a knot diagram and builds a cube of modules whose vertices correspond to ways to ``resolve'' each crossing; the invariant is then the homology of the total complex of the cube. Ozsváth and Szabó have provided a ``cube of resolutions'' description of knot Floer homology; this paper studies a slight improvement of their cube. The cube depends not just on a diagram for the knot but a presentation of the knot as a braid closure (for comparison, Jones' original description of the Jones polynomial also was in terms of braid closures), and indeed on a choice of braid word. The paper then shows, using purely algebraic means, that the total complex of the cube is invariant under braidlike Reidemeister moves II and III and under conjugation; therefore the homology is an invariant of the conjugation class of the element of the braid group. To show that it is a knot invariant requires checking invariance also under stabilization, a particular type of Reidemeister move I. The behavior under this move is investigated, but the result is not as satisfying as the main theorems. The paper is long and technical and absolutely worth reading for active participants in the subject: the author leads the reader through detailed calculations, providing ample opportunities to hone one's techniques.