Planar cube edges merge or split circles #
Along an edge of the cube of resolutions, from a state s to the state t differing from it
only at the crossing c, two circles of s through c merge into one circle of t, one circle
of s through c splits into two circles of t, or one circle through c is replaced by one
circle (TauCeti.PDCode.card_circlesAt_add_card_circlesAt_le_three). The last case occurs for
the nonplanar code of two circles meeting in a single crossing. This file proves that it never
occurs for a planar code: then exactly three circles pass through c in the two smoothings
together (TauCeti.PDCode.IsPlanar.card_circlesAt_add_card_circlesAt_eq_three), so every edge of
the cube of resolutions is a merge or a split, and the number of circles changes by exactly one.
These are the two kinds of edges carrying the merge and split maps of the Khovanov complex.
Main results #
TauCeti.PDCode.IsPlanar.card_circlesAt_add_card_circlesAt_eq_three: for a planar code, three circles pass through the crossing at which two states differ, in the two smoothings together.TauCeti.PDCode.IsPlanar.card_circlesAt_eq_two_or_card_circlesAt_eq_two: every edge of the cube of resolutions of a planar code is a merge or a split.TauCeti.PDCode.IsPlanar.stateLoopCount_eq_add_one_or_stateLoopCount_eq_add_one: along every edge the number of circles changes by exactly one.
Implementation notes #
The proof is a finite Euler-characteristic count, read off from the transposition bound
TauCeti.card_add_orbitCount_le_length_add_two_mul_card_orbits. Smoothing the crossings
according to s gives the perfect matching σ of the half-edges, and the rotation ρ to the
next slot at each crossing is σ * (σ * ρ), where σ * ρ transposes two opposite slots at each
crossing. So the face permutation α * ρ of the underlying graph, with α the arc matching, is
conjugate to (σ * ρ) * (α * σ), and α * σ is the inverse of the traversal
TauCeti.PDCode.statePerm, whose cycles are the circles of s in both directions of travel.
Writing this product with n transpositions for σ * ρ, and 2 * n each for α and σ, the
bound is Euler's bound TauCeti.PDCode.faceCount_le, attained exactly by planar codes.
Now suppose one circle of s and one circle of t pass through c. Then the two slots x and
y transposed by σ * ρ at c lie on one circle of s, but they do not lie on one directed
traversal orbit: otherwise, the reversal of directed orbits along the smoothing would put the
other two slots at c on one other directed orbit, and resmoothing c would split both, leaving
t one circle more than s. A transposition joining two directed orbits of one circle merges two
faces without joining two components, so the bound is not attained
(TauCeti.card_add_orbitCount_add_two_le_length_add_two_mul_card_orbits), and the code is not
planar.
References #
- M. Khovanov, A categorification of the Jones polynomial, Duke Math. J. 101 (2000), 359-426, §2 and §4.2 (the merge and split maps on the edges of the cube).
- D. Bar-Natan, On Khovanov's categorification of the Jones polynomial, Algebr. Geom. Topol. 2 (2002), 337-370, §2.
- S. K. Lando, A. K. Zvonkin, Graphs on Surfaces and Their Applications, Encyclopaedia of Mathematical Sciences 141, Springer 2004, §1.3 (faces of rotation systems).
The crossing transpositions of a smoothing #
Joining two directed orbits of a circle is not planar #
One circle through a crossing in both smoothings #
Planar cube edges #
Every edge of the cube of resolutions of a planar code is a merge or a split. If two
states differ exactly at the crossing c of a planar code, then three circles pass through c in
the two smoothings together: two circles of one smoothing and one circle of the other.
Every edge of the cube of resolutions of a planar code is a merge or a split. If two
states differ exactly at the crossing c of a planar code, then two circles pass through c in
one of the two smoothings: two circles of s merge into one circle of t, or one circle of s
splits into two circles of t.
Along every edge of the cube of resolutions of a planar code, the number of circles changes by exactly one.