1: \documentclass[a4paper,12pt]{amsart}
2: \usepackage{amscd,amsmath,amssymb,amsxtra,boldsect}
3: \usepackage{hyperref}
4: \usepackage[xdvi]{graphicx}
5:
6: \textheight=8.5in
7: \textwidth=5.5in
8:
9: \newcommand{\Q}{\mathbb Q}
10: \newcommand{\R}{\mathbb R}
11: \newcommand{\Z}{\mathbb Z}
12: \newcommand{\Ga}{\alpha}
13: \newcommand{\Gb}{\beta}
14: \newcommand{\Gs}{\sigma}
15: \newcommand{\p}{\partial}
16: \newcommand{\rnk}{\operatorname{rk}}
17: \newcommand{\cal}{\mathcal}
18: \newcommand{\bfit}{\bfseries\itshape}
19:
20:
21: \title{Remarks on definition of Khovanov homology}
22:
23: \author{Oleg Viro}
24: \dedicatory{Uppsala University, Uppsala, Sweden\break
25: POMI, St.\ Petersburg, Russia}
26: \address{Department of Mathematics, Uppsala University,
27: Box 480, S-751 06 Uppsala, Sweden
28: }
29: \email{oleg@math.uu.se}
30: \begin{document}
31: \begin{abstract}
32:
33: Mikhail Khovanov defined, for a diagram of an oriented classical
34: link, a collection of groups numerated by pairs of
35: integers. These groups were constructed as homology groups of
36: certain chain complexes. The Euler characteristics of these complexes
37: are coefficients of the Jones polynomial of the link. The goal of this
38: note is to rewrite this construction in terms more friendly to
39: topologists. A version of Khovanov homology for framed links is
40: introduced. For framed links whose Kauffman brackets are involved in a
41: skein relation, these homology groups are related by an exact sequence.
42: \end{abstract}
43: \maketitle
44:
45: \section{Introduction}\label{s0}
46:
47: For a diagram $D$ of an oriented link $L$,
48: Mikhail Khovanov \cite{Kh} constructed a collection of groups
49: $\cal H^{i,j}(D)$ such that
50: $$K(L)(q)=\sum_{i,j}q^{j}(-1)^i\dim_{\Q}(\cal H^{i,j}\otimes\Q),$$
51: where $K(L)$ is a version of the Jones polynomial of $L$.
52: These groups are constructed as homology groups of chain complexes.
53: The primary goal of this note is to rewrite this construction in terms more
54: pleasant for topologists.
55:
56: To some extent this has been done recently by Dror Bar-Natan in his
57: preprint \cite{BN}. This came together with progress in
58: calculation and understanding of Khovanov homology, see \cite{Kh2} and
59: \cite{Lee}. The outbreak of activity that happened two years after
60: the release of initial paper has proved that stripping Khovanov's
61: construction of its fancy formal decorations was useful.
62:
63: I hope that a further chewing\footnote{From Bar-Natan's paper: ``Not being
64: able to really digest it (construction of Khovanov's invariant -- O.V.)
65: we decided to just chew some, and then provide our output
66: as a note containing a description of his construction, complete and
67: consistent and accompanied by computer code and examples but stripped
68: of all philosophy and of all the linguistic gymnastics that is necessary
69: for the philosophy but isn't necessary for the mere purpose of having
70: a working construction. Such a note may be more accessible than the
71: original papers. It may lead more people to read Khovanov at the source,
72: and maybe somebody reading such a note will figure out what the Khovanov
73: invariants really are. Congratulations! You are reading this note right
74: now.''} of Khovanov's construction that is presented below occur to
75: be useful. I am grateful to Khovanov,
76: who made these my considerations possible by not only the very
77: initiating of the subject, but also by keeping (at least
78: during the Fall of 1998) his text secret, while giving talks about his
79: work at various seminars. The rumors reached me, and I tried to figure
80: out how homology with the Euler characteristic equal to the Kauffman
81: bracket could be defined. When the preprint became available, I found
82: that Khovanov's construction can be reformulated in the way that I had
83: guessed. I could upload then most of the text presented below, but
84: hesitated, since I had no real new results based on my chewing. I still
85: have no them, but feel that a text showing how a transition from Kauffman
86: bracket to Khovanov homology could be motivated for a topologist may be
87: appreciated. Some irresponsible speculations about possible
88: generalizations are included at the end of Section \ref{s1}.
89:
90: The Khovanov homology is closer to Kauffman bracket, which is an invariant
91: of non-oriented, but framed links, rather than to the Jones polynomial, which
92: is an invariant of oriented, but non-framed links. The corresponding
93: modification of Khovanov homology is presented in Section
94: \ref{s2}. This allows us to write down a categorification of the Kauffman
95: skein relation for Kauffman bracket. The skein relation gives rise to a
96: homology sequence.
97:
98:
99: \section{From Kauffman bracket to Khovanov homology}\label{s1}
100:
101: Since the Euler characteristic does not change under passing from a
102: chain complex to its homology, it is natural to expect that $\cal
103: H^{i,j}$ appear as the homology groups of a chain complex $\cal
104: C^{i,j}$ such that its polynomial
105: Euler characteristic $\sum_{i,j}q^j(-1)^i\rnk \cal C^{i,j}$ is the
106: Jones polynomial $K(L)(q)$.
107:
108: Khovanov constructs such a complex starting with the Kauffman state sum
109: presentation of the Jones polynomial. However, the construction
110: proceeds as a chain of algebraic constructions of auxiliary objects.
111: This hides a simple geometric meaning. A natural generators of
112: Khovanov's complex can be represented as enhancements of the states
113: from the Kauffman state sum presentation of the Jones polynomial.
114:
115: \subsection{Kauffman's states of a link diagram}\label{s1.1}Recall that
116: for a state sum
117: representation of the Jones polynomial, Kauffman
118: \cite{Ka} introduced the following states of a link
119: diagram. Each state is a collection of markers. At each crossing of the
120: diagram there should be a marker specifying a pair of vertical angles.
121: See Figure \ref{f-markers}.
122: \begin{figure}[htb]
123: \centerline{\includegraphics{fig/markers.eps}}
124: \caption{Markers comprising a Kauffman state.}
125: \label{f-markers}
126: \end{figure}
127:
128: \subsection{Numerical characteristics of states and
129: diagrams}\label{s1.2} Each marker of a Kauffman state has a {\em sign},
130: which is $+$ (or, rather, $+1$) if the direction of rotation of the
131: upper string
132: towards the lower one through the specified pair of vertical angles is
133: counter-clock-wise, and $-$ (or $-1$) otherwise. For a state $s$ of a
134: diagram $D$ denote by $\Gs(s)$ the difference between the numbers of
135: positive and negative markers. A state of a diagram defines a {\em
136: smoothing\/} of the diagram: at each of its double points the marked
137: angles are united in a connected area, see Figure \ref{f-smoothings}.
138:
139: \begin{figure}[htb]
140: \centerline{\includegraphics{fig/smoothings.eps}}
141: \caption{Smoothing of a diagram according to markers.}
142: \label{f-smoothings}
143: \end{figure}
144:
145: The result of the smoothing is a collection of circles embedded into the
146: plane. Denote the union of these circles by $D_s$. The number of the
147: circles is denoted by $|s|$.
148:
149: With a crossing point of a diagram of an oriented
150: link we associate {\em local writhe number} equal to\begin{itemize}
151: \item $+1$, if at the
152: point the diagram looks like
153: $\vcenter{\hbox{\includegraphics{fig/pcross.eps}}}$, and
154: \item $-1$, if it looks like
155: $\vcenter{\hbox{\includegraphics{fig/ncross.eps}}}$.\end{itemize}
156: The sum of local writhe numbers over all crossing points of a link
157: diagram $D$ is called the {\it writhe number\/} of $D$ and denoted by $w(D)$.
158:
159:
160: To a Kauffman state $s$ of an oriented link diagram $D$ we assign a
161: polynomial
162: $$V_s(A)=A^{\Gs(s)}(-A)^{-3w(D)}(-A^2-A^{-2})^{|s|},$$
163: where $w(D)$ is the writhe number of $D$.
164:
165: \subsection{Jones Polynomial}\label{s1.3} The sum
166: of $V_s(A)$ over all states $s$ of the diagram is a version the Jones
167: polynomial of $L$. It is denoted by $V_L(A)$.
168:
169: Khovanov \cite{Kh} uses another variable $q$ instead of $A$. It is
170: related to $A$ by $q=-A^{-2}$. Following Khovanov, we denote the
171: corresponding version of the Jones polynomial by $K(L)$. It is defined by
172: $K(L)(-A^{-2})=V_L(A)$.
173:
174: Certainly, $K(L)$ can be presented as the sum of the corresponding
175: versions of $V_s(A)$ over all Kauffman states $s$. The summand of
176: $K(L)$ corresponding to $s$ can be defined via $K(s)(-A^2)=V_s(A)$ or,
177: more directly, by
178: $$K(s)(q)=(-1)^{\frac{w(D)-\Gs(s)}2}q^{\frac{3w(D)-\Gs(s)}2}(q+q^{-1})^{|s|}.$$
179:
180: \subsection{Enhanced Kauffman states}\label{s1.4}
181: By an {\em enhanced Kauffman state\/} $S$ of an oriented link diagram
182: $D$ we shall mean a collection of markers comprising a usual Kauffman
183: state $s$ of $D$ enhanced by an assignment of a plus or minus sign to
184: each of the circles obtained by the smoothing of $D$ according to the
185: markers.
186:
187: Denote by $\tau(S)$ the difference between the numbers of pluses and
188: minuses assigned to the circles of $D_s$. Put
189: $$j(S)=-\frac{\Gs(s)+2\tau(S)-3w(D)}2.$$
190: Observe that both $\Gs(s)$ and $w(D)$ are congruent modulo 2 to the
191: number of crossing points. Therefore $j(S)$ is an integer.
192:
193: \subsection{Monomial state sum representation of Jones
194: polynomial}\label{s1.5} Recall that
195: $V_s(A)=A^{\Gs(s)}(-A)^{-3w(D)}(-A^2-A^{-2})^{|s|}$. We can associate
196: to each component of $D_s$ one of $|s|$ factors $(-A^2-A^{-2})$ of
197: the right hand side of this formula.
198:
199: Passing to the enhanced Kauffman states, we associate the summand
200: $-A^2$ of the sum $-A^2-A^{-2}$ to a component of $D_{|s|}$ equipped
201: with $+$. To the same component, but equipped with $-$, we associate
202: the summand $-A^{-2}$.
203:
204: Thus an enhancement of a Kauffman state gives rise to a choice of a
205: summand from each of the binomial factors in the formula defining
206: $V_s(A)$. Now by opening the brackets we obtain the following
207: presentation for $V_s(A)$:
208: \begin{multline}\label{SpltVs(A)}
209: V_s(A)= \\
210: \sum_{S\in s}A^{\Gs(s)}(-A^2)^{\tau(S)}(-A)^{-3w(D)}= \\
211: \sum_{S\in s}(-1)^{\tau(S)+w(D)}A^{-2j(S)}.
212: \end{multline}
213: Denote by $V_S(A)$ the summand $(-1)^{\tau(S)+w(D)}A^{-2j(S)}$
214: which corresponds to $S$.
215: Hence $V_s(A)=\sum_{S\in s}V_S(A)$ and
216: \begin{multline}\label{V_L(A)}
217: V_L(A)=\sum_{\substack{\text{Kauffman}\\
218: \text{ states }s \text{ of }D}} V_s(A)= \\
219: \sum_{\substack{\text{enhanced Kauffman}\\
220: \text{ states }S \text{ of }D}} V_S(A)=
221: \sum_{\substack{\text{enhanced Kauffman}\\
222: \text{ states }S \text{ of }D}}
223: (-1)^{\tau(S)+w(D)}A^{-2j(S)}.\end{multline}
224:
225: \subsection{Change of variable}\label{s1.6}
226: Let us switch to variable $q$ used by Khovanov.
227: \begin{multline} V_S(A)=
228: (-1)^{\tau(S)+w(D)}A^{-2j(S)}=\\
229: (-1)^{\tau(S)+w(D)}(A^{-2})^{j(S)}=\\
230: (-1)^{\tau(S)+w(D)-j(S)}(-A^{-2})^{j(S)}=\\
231: (-1)^{\tau(S)+w(D)+\frac12\Gs(s)+\tau(S)-\frac32w(D)}q^{j(S)}=\\
232: (-1)^{\frac{\Gs(s)-w(D)}2}q^{j(S)}=\\
233: (-1)^{\frac{w(D)-\Gs(s)}2}q^{j(S)}.
234: \end{multline}
235: Denote $(-1)^{\frac{w(D)-\Gs(s)}2}q^{j(S)}$ by $K(S)(q)$.
236: Clearly,
237: \begin{equation}\label{V_L(q)}
238: K(L)(q)=
239: \sum_{\substack{\text{enhanced Kauffman}\\
240: \text{ states }S \text{ of }D}}
241: (-1)^{\frac{w(D)-\Gs(s)}2}q^{j(S)}
242: =\sum_{\substack{\text{enhanced Kauffman}\\
243: \text{ states }S \text{ of }D}} K(S)(q).\end{equation}
244:
245: \subsection{Khovanov chain groups}\label{s1.7}
246: Denote the free abelian group generated by enhanced Kauffman states of
247: a link diagram $D$ by $\cal C(D)$. Denote by $\cal C^j(D)$ the subgroup of $\cal C(D)$
248: generated by enhanced Kauffman states $S$ of $D$ with $j(S)=j$. Thus
249: $\cal C(D)$ is a $\Z$-graded free abelian group:
250: $$\cal C(D)=\bigoplus_{j\in\Z}\cal C^j(D).$$
251:
252: For an enhanced Kauffman state $S$ belonging to a Kauffman state $s$ of
253: a link diagram $D$, put
254: $$i(S)=\frac{w(D)-\Gs(s)}2.$$
255: Denote by $\cal C^{i,j}(D)$ the subgroup of $\cal C^j(D)$ generated by enhanced
256: Kauffman states $S$ with $i(S)=i$.
257:
258: Notice that as follows from \eqref{V_L(q)},
259: \begin{equation}\label{chiHij}
260: K(L)(q)=\sum_{j=-\infty}^{\infty}q^j\sum_{i=-\infty}^{\infty}
261: (-1)^i\rnk \cal C^{i,j}(D).
262: \end{equation}
263:
264: {\bfit Remark.\/}
265: The Kauffman choice of variable ($A$ instead of $q$) in the Jones
266: polynomial discussed above suggests another choice of the second
267: grading of $\cal C(D)$. In
268: fact, the choices incorporate already two different $\Z_2$-grading:
269: $\tau(S)+w(D)\pmod 2$ and $\frac{w(D)-\Gs(s)}2\pmod 2$. However
270: these $\Z_2$-grading differ just by the reduction modulo 2 of the first
271: one, $$\tau(S)+w(D)-\frac{w(D)-\Gs(s)}2=j(S)\mod2.$$
272: One can prove that the Kauffman choice of variable gives rise to the
273: same homology groups.
274:
275:
276: \subsection{Differential}\label{s1.8}
277: Let $D$ be a diagram of an oriented link $L$. In \cite{Kh} Khovanov
278: defined in $\cal C^{i,j}(D)$ a differential of bidegree $(1,0)$ and proved
279: that its homology groups $\cal H^{i,j}(D)$ do not depend on $D$. The
280: construction of the differential depends on ordering of crossing points
281: of $D$, although the homology groups do not. Suppose that the crossing
282: points of $D$ are numerated by natural numbers $1,\dots,n$.
283:
284: Khovanov's description of the differential is somewhat complicated by
285: several auxiliary algebraic constructions. I give here its simplified
286: version. I just describe the matrix elements. In the context of chain
287: complex the matrix elements are traditionally called {\em incidence
288: numbers}. For enhanced Kauffman states $S_1$ and $S_2$, denote their
289: incidence number by $(S_1:S_2)$.
290:
291:
292: Recall that $\cal C^{i,j}(D)$ is generated by enhanced Kauffman states.
293: Thus an incidence number is a function of two enhanced Kauffman
294: states, say $S_1$ and $S_2$ (which are generators of $\cal C^{i,j}(D)$ and
295: $\cal C^{i+1,j}(D)$, respectively). Enhanced Kauffman states with a nonzero
296: incidence number are said to be {\it incident} to each other. Pairs of
297: incident states satisfy natural restrictions. Surprisingly, these
298: restrictions give exact description of the set of incident
299: states: each pair of enhanced Kauffman states which is not thrown away
300: by the restrictions consists of incident states.
301:
302: The first restriction emerges from our desire to have a differential
303: of bidegree $(1,0)$. Since the differential preserves $j$ and
304: increases $i$ by one, $j(S_1)=j(S_2)$ and $i(S_2)=i(S_1)+1$. Recall
305: that $i(S)=\frac{w(D)-\Gs(s)}2$. Therefore $i(S_2)=i(S_1)+1$ implies
306: that $\Gs(s_2)=\Gs(s_1)-2$, in other words the number of negative
307: markers of $S_2$ is greater by one than the number of negative markers
308: of $S_1$.
309:
310: It is natural to enforce this numerological restriction in the
311: following way:
312:
313: {\it The incidence number is zero, unless only at one crossing point
314: of $D$ the markers of $S_1$ and $S_2$ differ and at this crossing the
315: marker of $S_1$ is positive, while the marker of $S_2$ is negative.}
316:
317: In our description of the incidence number of $S_1$ and $S_2$ we
318: will assume that this is the case. The crossing point where the
319: markers differ is called the {\em difference point} of $S_1$ and
320: $S_2$. Let it have number $k$.
321:
322: Since exactly at one crossing the markers differ, $D_{S_2}$
323: is obtained from $D_{S_1}$ by a single oriented Morse
324: modification of index 1. Hence $|S_1|-|S_2|=\pm1$. In other words
325: either $D_{S_2}$ is obtained from $D_{S_1}$ by joining two circles or
326: by splitting a circle of $D_{S_1}$ into two circles.
327:
328: Here is the next natural restriction on incident states:
329:
330: {\em The incidence number of $S_1$ and $S_2$ vanishes, unless the
331: common circles of $D_{S_1}$ and $D_{S_2}$ have the same signs.}
332:
333: This, together with equality $j(S_2)=j(S_1)$, gives a strong
334: restriction
335: also on the signs of the circles of $D_{S_l}$ adjacent to the $k$th
336: crossing point. Indeed,
337: \begin{multline*}
338: j(S_2)=\frac{3w(D)-\Gs(s_2)-2\tau(S_2)}2\\=
339: j(S_1)=\frac{3w(D)-\Gs(s_1)-2\tau(S_1)}2=\frac{3w(D)-\Gs(s_2)-2-2\tau(S_1
340: )}2,\end{multline*}
341: thus $\tau(S_2)=\tau(S_1)+1$.
342:
343: Now we can easily list all the
344: situations which satisfy these restrictions (see Figure \ref{f-Xinc}):
345: \begin{enumerate}
346: \item If $|S_2|=|S_1|-1$ and
347: both joining circles of $D_{S_1}$ are negative then the resulting
348: circle of $S_2$ should be negative.
349: \item If $|S_2|=|S_1|-1$ and
350: the joining circles of $D_{S_1}$ have different signs then the resulting
351: circle of $S_2$ should be positive.
352: \item If the joining circles of $D_{S_1}$
353: are positive, none $S_2$ can be incident.
354: \item If $|S_2|=|S_1|+1$ and the
355: splitting circle of $D_{S_1}$ is positive then both of the circles of
356: $D_{S_2}$ obtained from it should be positive.
357: \item If $|S_2|=|S_1|+1$ and the
358: splitting circle of $D_{S_1}$ is negative then the circles of
359: $D_{S_2}$ obtained from it should be of different signs.
360: \end{enumerate}
361:
362: \begin{figure}[thbp]
363: \centerline{\includegraphics{fig/x-inc.eps}}
364: \caption{Pairs of incident enhanced Kauffman states.
365: Dotted arcs show how the fragments of $D_s$ at a crossing point are
366: connected in the whole $D_s$.}
367: \label{f-Xinc}
368: \end{figure}
369:
370:
371: In each of the cases listed above the incidence number $(S_1:S_2)$ is
372: $\pm1$. The sign depends on the ordering of the crossing points: it is
373: equal to $(-1)^t$ where $t$ is the number of negative markers in $S_1$
374: numerated with numbers greater than $k$.
375:
376: This sign is needed to make the differential satisfying identity
377: $d^2=0$. However, the proof of this identity requires some routine check.
378: At first glance, Khovanov \cite{Kh} escaped it. In fact, it is hidden in
379: the checking (which was left to the reader) that $F$ is a functor, see
380: \cite{Kh}, Section 2.2.
381:
382: \subsection{Enhanced Kauffman states with polynomial
383: coefficients}\label{s1.9} Khovanov constructed not only groups
384: $\mathcal H^{i,j}$, but also graded modules $H^i$ over the ring
385: $\Z[c]$ of polynomials with integer coefficients in variable $c$ of
386: degree 2. The grading is a representation of $H^i$ as a direct sum of
387: abelian subgroups $H^{i,j}$ such that multiplication by $c$ in $H^i$
388: gives rise to a homomorphism $H^{i,j}\to H^{i,j+2}$.
389:
390: To construct homology groups $H^{i,j}$, let us define the corresponding
391: complex of graded $\Z[c]$-modules $C^i$. The module $C^i$ is the sum of
392: its subgroups $C^{i,j}$. The group $C^{i,j}$ is generated by formal
393: products $c^kS$, where $k\ge0$ and $S$ is an enhanced Kauffman states
394: $S$ with $i(S)=i$ and $j(S)=j-2k$.
395:
396: The differential is defined almost in the same way. The states which
397: were adjacent above are adjacent here, as well as their products by the
398: same power of $c$. Products of enhanced Kauffman states by different
399: powers of $c$ are not adjacent, besides the following situations:
400: $c^{k+1}S_2$ is adjacent to $c^kS_1$, if the markers of $S_1$ and $S_2$
401: differ exactly at one point, where the marker of $S_1$ is positive and
402: the marker of $S_2$ negative, the signs of $S_1$ and $S_2$ on the common
403: circles of $D_{S_1}$ and $D_{S_2}$ are the same, $|S_2|=|S_1|+1$,
404: the splitting circle of $D_{S_1}$ is negative, and the circles of
405: $D_{S_2}$ obtained from it are positive, see Figure \ref{c-inc}, where
406: this situation is shown symbolically in the style of Figure \ref{f-Xinc}.
407: \begin{figure}[htb]
408: \centerline{\includegraphics{fig/c-inc.eps}}
409: \caption{}
410: \label{c-inc}
411: \end{figure}
412:
413: Each group $C^{i,j}$ is finitely generated, but there are infinitely
414: many non-trivial groups. Groups $H^{i,j}$ with fixed $i$ and
415: sufficiently large $j$ are isomorphic to each other.
416:
417: \subsection{Speculations}\label{s1.10}
418: How can one understand construction of Khovanov homology? By
419: understanding I mean a formulation which would give, true or false,
420: indication to possible generalizations. The general program of
421: categorification pictured by Khovanov in \cite{Kh} is an attractive
422: approach to the problem of constructing combinatorial counterparts for
423: Donaldson and Seiberg-Witten invariants, although it does not promise a
424: fast and technically easy track.
425:
426: In the construction above the specific of Kauffman bracket is used so
427: much that it is difficult to imagine a geralization to other polynomial
428: link invariants. A state sum over smoothings along Kauffman markers does not
429: appear in formulas representing any other quantum invariant. Of course,
430: there are many state sum formulas. One can try to decompose any of them
431: to a sum of monomials and consider decompositions as chains. Attach to them
432: appropriate dimensions and define differentials such that the homology
433: groups would be invariant under Reidemeister moves. Success of the
434: whole project depends on quality of several guesses. The most difficult
435: of the guesses seems to be the choice of differentials. It may involve and
436: rely on a delicate analysis of the effects of Reidemeister moves. However
437: a wrong choice at the previous step can make any efforts at this last step
438: unfruitful.
439:
440: An additional hint comes from Khovanov's mention of
441: Lusztig's canonical bases. Indeed, signs which are attached to circles
442: in a smoothened link diagram can be replaced with orientations and an
443: orientation was interpreted by Frenkel and Khovanov \cite{FK} as an
444: element of Lusztig's canonical basis. Smoothings at crossing points
445: correspond then to monomials of the entries of the $R$-matrix. The
446: decomposition of the entries of $R$-matrix to monomials not as clear as
447: it seams: in the only known case considered above a zero entry of
448: $R$-matrix is decomposed to two non-zero monomials canceling each
449: other. However, the main guess is still to be done after the entries of
450: the $R$-matrix in crystal bases are decomposed to monomials: one should
451: find a differential that would give homology invariant under
452: Reidemeister moves.
453:
454: If the guesses made above about the r\^oles of crystal bases and
455: monomials of $R$-matrices are correct, $1+1$-dimensional TQFTs would not
456: appear in the future more involved categorifications. Instead of cobordisms
457: of 1-manifolds, there appear probably cobordisms of 4-valent
458: graphs with some additional structure.
459:
460: How can one estimate chances for further categorification of quantum
461: topology? Still, they do not seem to be clear, despite of high optimism
462: of Khovanov.As for chain groups, the observations made above
463: are encouraging. However, we are not even able to discuss the choice
464: of differentials on a comparable level of generality.
465:
466: It would be very interesting, especially for prospective 4-dimensional
467: applications, to find a categorification of face state models.
468:
469:
470:
471: \section{Frame version}\label{s2}
472:
473: \subsection{Kauffman bracket versus Jones polynomial}\label{s2.1}
474: Involvement of an orientation of link in the definition of the Jones
475: polynomial in Section \ref{s1.3} is easy to localize. It is limited to
476: to the writhe number $w(D)$ of $D$. If we remove $w(D)$, we get the
477: {\it Kauffman bracket\/} of diagram $D$. To eliminate fractional powers,
478: we stay with Kauffman's variable $A$. Thus Kauffman bracket of a link
479: diagram $D$ is
480: $$\langle D\rangle=
481: \sum_{\substack{\text{ Kauffman }\\ \text{ states }
482: s\text{ of }D}}A^{\Gs(s)}(-A^2-A^{-2})^{|s|}.$$
483: It is not invariant under first Reidemeister moves. Hence it is not
484: an ambient isotopy invariant of link. However it is invariant with respect to
485: second and third Reidemeister moves.
486:
487: A link diagram defines the blackboard framing on the link. The
488: isotopies which accompany second and third Reidemeister moves can be
489: extended to the isotopy of the corresponding framings. Moreover,
490: embedding of the diagrams to the 2-sphere makes this relation
491: two-sided: link diagrams on $S^2$ can be converted to each other by
492: a sequence of second and third Reidemeister moves iff the corresponding
493: links with blackboard framings are isotopic in the class of framed
494: links. Since the Kauffman bracket can be defined for a link diagram on
495: 2-sphere, this makes the Kauffman bracket invariant for non-oriented
496: framed links.
497:
498: The Kauffman bracket is categorified below.
499:
500: \subsection{Framed Khovanov homology}\label{s2.2}
501: For an enhanced Kauffman state of a link diagram $D$, put
502: $$J(S)=\Gs(s)+2\tau(S)\quad\text{ and }\quad I(S)=\Gs(s). $$
503: If this is an oriented link, and hence $w(D)$ is defined, then
504: $$J(S)=3w(D)-2j(S)\quad\text{ and }\quad I(S)=w(D)-2i(S).$$
505: In terms of $J(S)$ and $I(S)$, the Kauffman bracket is expressed as
506: follows:
507: $$\langle D\rangle=\sum_{\substack{\text{enhanced Kauffamn}\\
508: \text{states } S \text{ of } D}}(-1)^{\frac{I(S)}2}A^{J(S)}.$$
509:
510: Denote the free abelian group generated by enhanced Kauffman states $S$ of
511: $D$ with $I(S)=i$ and $J(S)=j$ by $\cal C_{i,j}(D)$. If one orients the
512: link, the Khovanov chain groups $\cal C^{i,j}(D)$ appear, and
513: $$\cal C_{i,j}(D)=\cal C^{\frac{w(D)-i}2,\frac{3w(D)-j}2}(D).$$
514: Under this identification, the differentials of the Khovanov complex
515: turn into differentials
516: $$\p:\cal C_{i,j}(D)\to \cal C_{i-2,j}(D)$$
517: (the construction of the differentials does not involve orientation of
518: the link, hence it does not matter what orientation is used). Denote the
519: homology group of the complex obtained by $H_{i,j}(D)$.
520:
521: \subsection{Skein homology sequence}\label{s2.3} Let $D$ be a link
522: diagram, $c$ its crossing and $D_+$, $D_-$ be link diagrams obtained
523: from $D$ by smoothing at $c$ along positive and negative markers,
524: respectively. As is well-known, the Kauffman brackets of $D$, $D_+$ and
525: $D_-$ are related as follows
526: $$\langle D\rangle=A\langle D_+\rangle+A^{-1}\langle D_-\rangle.$$
527: This equality is called the {\it Kauffman skein relation.\/} It allows one
528: to define the Kauffman bracket up to a normalization (which can be
529: done by fixing the Kauffman bracket of the unknot).
530:
531: Let us categorify this skein relation.
532: Consider map
533: $$\Ga:\cal C_{i,j}(D_-)\to \cal C_{i-1,j-1}(D)$$
534: which sends an enhanced Kauffman state $S$ of $D_-$ to the enhanced Kauffman
535: state of $D$ smoothing along which coincides with the smoothing of
536: $D_-$ along $S$ and signs of the ovals are the same, too.
537: The collection of these maps is a homomorphism of complex, that is they
538: commute with $\p$, provided $c$ is the last crossing of $D$ in the
539: ordering of crossings which is used in the construction of $\p$.
540: Indeed, the incidence coefficients are the same for an enhanced
541: Kauffman states of $D_-$ and its images in $D$, and the latter cannot
542: contain in the boundary a state with positive marker at $c$.
543:
544: Now consider map
545: $$\Gb:\cal C_{i,j}(D)\to \cal C_{i-1,j-1}(D_+)$$
546: which sends each enhanced Kauffman state with negative marker at $c$ to
547: 0 and each enhanced Kauffman state with positive marker at $c$ to the
548: enhanced Kauffman state of $D_+$ with the same smoothing and signs of
549: the ovals. This is again a complex homomorphism.
550:
551: Homomorphisms $\Ga$ and $\Gb$ form a short exact sequence of complexes:
552: $$\begin{CD}
553: 0@>>>\cal C_{*,*}(D_-)@>{\Ga}>> \cal C_{*-1,*-1}(D)@>{\Gb}>>
554: \cal C_{*-2,*-2}(D_+) @>>> 0
555: \end{CD}
556: $$
557: It induces collection of long homology sequences:
558: $$
559: \begin{CD}
560: @>{\p}>>H_{i,j}(D_-)@>{\Ga_*}>> H_{i-1,j-1}(D)@>{\Gb_*}>>
561: H_{i-2,j-2}(D_+)\\ @>{\p}>>H_{i-2,j}(D_-)
562: @>{\Ga_*}>> H_{i-3,j-1}(D)@>{\Gb_*}>>H_{i-4,j-2}(D_+) @>{\p}>>
563: \end{CD}
564: $$
565: A special case of this sequence, which relates the groups of connected
566: sum and disjoint sum of knots, can be found in Section 7.4 of
567: Khovanov's paper \cite{Kh}. This special case is the only one which can
568: be formulated for the original version of homology depending on
569: orientations of links.
570:
571: I was not able to categorify the other skein relation, the one which involves
572: the Jones polynomial of {\it oriented\/} links. It follows from a couple of
573: skein relations of the type considered above.
574:
575:
576:
577: \begin{thebibliography}{9999999}
578:
579: \bibitem{BN} Dror Bar-Natan, Khovanov's categorification of the Jones
580: polynomial, arXiv: math.QA /0201043.
581:
582: \bibitem{FK} I.~B.~Frenkel, M.~Khovanov, Canonical bases in tensor
583: products and graphical calculus for $U_q(\mathfrak{sl}_2)$, {\it Duke
584: Math. Journal,\/} v. 87, (1997), 409--480.
585: \bibitem{Ka} L.H.Kauffman, State models and the Jones polynomial,
586: {\it Topology,\/} {\bf 26:3} (1987), 395--407.
587:
588: \bibitem{Kh} Mikhail Khovanov, A categorification of the Jones
589: polynomial, {\it Duke Math. J.,\/} 101 (3), 359--426, 1999,
590: arXiv: math.QA /9908171.
591:
592: \bibitem{Kh1} Mikhail Khovanov, A functor-valued invariant of tangles,
593: arXiv: math.QA /0103190
594:
595: \bibitem{Kh2} Mikhail Khovanov, Patterns in knot cohomology I,
596: arXiv: math.QA /0201306.
597:
598: \bibitem{Lee} E.~S.~Lee, The support of the Khovanov's invariants for
599: alternating knots, arXiv: math.GT /0201105.
600:
601: \end{thebibliography}
602:
603:
604: \end{document}
605:
606: