1: \section{Computation of $\widehat{HFK}(D_+(K,t),1)$ and $\tau(D_+(K,t))$ for all $t$}
2: In this section we use the skein exact sequence for knot Floer homology to interpolate between the case when $t\gg0$ and the case when $t\ll0$. This will enable us to determine \HFD \ for all values of $t$. We will also determine $\tau(\WDnm)$. The analysis of the skein sequence will be similar to the technique used in \cite{STau} in the special case of Whitehead doubles of the $(2,2n+1)$ torus knots. The main result of this section will be:
3: \begin{prop}
4: Let $K\subset S^3$ be a knot with Seifert genus $g(K)=g$. Then for $t\ge 2\tau(K)$ we have:
5: $$\HFKa_*(\WDnm,1)\cong \Z_{(1)}^{t-2g-2} \bigoplus_{i=-g}^{g}[H_{*-1}(\Filt(K,i))]^2, $$
6: and $\tau(\WDnm)=0$. For $t< 2\tau(K)$ the following holds:
7: $$\HFKa_*(\WDnm,1)\cong \Z_{(1)}^{2\tau(K)-2g-2} \oplus\Z_{(0)}^{2\tau(K)-t} \bigoplus_{i=-g}^{g}[H_{*-1}(\Filt(K,i))]^2, $$
8: and $\tau(\WDnm)=1.$
9: \end{prop}
10: \noindent {\bf Remark:} This takes care of the top and bottom group for the Whitehead double, by the symmetry of knot Floer homology about $\Filt=0$. It also proves Theorem \ref{thm:tau} stated in the introduction.
11:
12: \begin{proof} It will be helpful to first rephrase Theorem \ref{thm:largen}:
13: \begin{theorem}
14: \label{thm:largen2}
15: Let $\companion\subset S^3$ be a knot, and suppose $g(\companion)=g$ denotes the Seifert genus of $\companion$. Then for all $t>T>0$ there are isomorphisms of absolutely $\Z$-graded abelian groups:
16: $$\HFDgnm \cong \Z_{(1)}^{t-2g-2} \bigoplus_{i=-g} ^{g}[ H_{*-1}(\Filt(\companion,i))]^2$$
17: $$\HFDgnmneg \cong \Z_{(0)}^{|t|-2g}\bigoplus_{i=-g} ^{g}[ H_{*}(\frac{\CFa(S^3)}{\Filt(\companion,i)})]^2$$
18: \end{theorem}
19: \begin{proof} This follows from the adjunction inequality for knot Floer homology, which implies that $H_*(\Filt(\companion,i))\cong0$ for $i<g$ and $H_*(\Filt(\companion,i))\cong Z_{(0)}$ for $i\ge g$.
20: \end{proof}
21:
22: As in \cite{STau}, we note that it is possible to change \WD \ to \WDneg \ by a sequence of $2t$ crossing changes, each of which change a negative crossing in the twisting region to a positive crossing. Theorem $8.2$ of \cite{Knots} shows that corresponding to each crossing change, there is a long exact sequence relating the knot Floer homology groups of \WD, \WDm, and the two component link obtained from the oriented resolution of the crossing which we change. In each case, this link is the positive Hopf link, which we denote by $H$. Summarizing the discussion of \cite{STau}, the skein exact sequence for the top filtration level takes the following form: $$
23: \begin{CD}
24: ... @>>> \HFDnm
25: @>{f_1}>> \F_{(\OneHalf)} @>{f_2}>> \HFDnmm @>{f_3}>>...
26: \end{CD}
27: $$
28:
29: \noindent Here the maps $f_1$ and $f_2$ lower homological degree by one-half and $f_3$ is non-increasing in the homological degree. We wish to understand the maps in this sequence. To aid this cause, we determine how the ranks of the groups in each homological dimension differ between the cases when $t>T>0$ and $t<-T<0$.
30: \begin{lemma}
31: \label{lem:rank}
32: Let $\companion\subset S^3$ be a knot with genus $g$, and let $t>T>0$ be an integer so that Theorem \ref{thm:largen2} holds. Then
33: $$\left. \ \begin{array}{ll}
34: rk \HFKa_*(\WDnm,1) = rk \HFKa_*(\WDnmneg,1) & {\text if} *\ne0,1 \\
35: rk \HFKa_1(\WDnm,1) = rk \HFKa_1(\WDnmneg,1) + t -2\tau(K) & \\
36: rk \HFKa_0(\WDnm,1) = rk \HFKa_0(\WDnmneg,1) -t - 2\tau(K) & \\
37: \end{array} \right. $$
38:
39: \end{lemma}
40:
41: \begin{proof} The lemma follows from Theorem \ref{thm:largen2}, the definition of $\tau(K)$, and the long exact in homology coming from the short exact sequence of chain complexes,
42: $$
43: \begin{CD}
44: 0 @>>> \Filt(\companion,j)
45: @>{i}>> \CFa(S^3) @>{p}>> \frac{\CFa(S^3)}{\Filt(\companion,j)} @>>> 0 \end{CD}
46: .$$
47:
48: \noindent Since $\HFa(S^3)\cong \F_{(0)}$, the long exact sequence shows that $H_{*-1}(\Filt(\companion,j))\cong H_{*}(\frac{\CFa(S^3)}{\Filt(\companion,j)})$ if $*\ne0,-1$, from which the first part of the lemma follows (taking into account the grading shift in the first part of Theorem \ref{thm:largen2}). For the second two parts, recall that $\tau(K)$ is defined as:
49: $$\tau(K)=\mathrm{min}\{j\in\Z|i_*:H_*(\Filt(K,j))\longrightarrow \HFa(S^3)\ {\text{~is non-trivial}}\}.$$
50:
51: \noindent In the long exact sequence we have:
52: $$
53: \begin{CD}
54: 0 @>>> H_{1}(\frac{\CFa(S^3)}{\Filt(\companion,j)}) @>{\delta}>> H_0(\Filt(\companion,j))
55: @>{i_*}>> \Z_{(0)} @>{p_*}>>
56: \\ H_{0}(\frac{\CFa(S^3)}{\Filt(\companion,j)}) @>{\delta}>> H_{-1}(\Filt(\companion,j)) @>>> 0 \end{CD}
57: ,$$
58: \noindent and the map $i_*$ is trivial precisely when $j < \tau(K)$ and non-trivial otherwise. Taking the sum over each $j$ from $-g,\ldots ,g$, and examining ranks yields the second two parts of the lemma.
59:
60: \end{proof}
61:
62: Next, we observe that the map $f_3$ in the Skein exact sequence, which {\em a priori} is non-increasing in the absolute degree, in fact preserves degree.
63:
64: \begin{prop}
65: In the exact sequence relating \HFD,\newline \HFDm, $\HFKa(H,1)$, the map $f_3$ preserves degree.
66: \end{prop}
67: \begin{proof} This follows from the fact that the maps $f_1$ and $f_2$ lower degree by one-half and $f_3$ is non-increasing in the degree, together with the preceding lemma and the fact that the Floer homology of $H$ is supported in degree one-half.
68: \end{proof}
69:
70: \noindent{\bf Claim:} In the $2t$ applications of the skein sequence connecting \WD \ and \WDneg, $f_2$ is trivial exactly $t-2\tau(K)$ times.
71:
72: \begin{proof} The preceding proposition tells us that $f_3$ preserves degree. Since $\HFKa(H,1)$ has rank one, supported in degree one-half, \HFDm \ is determined by \HFD \ and whether or not $f_2$ is trivial. This holds for all $t$, and Lemma \ref{lem:rank} tells us the difference in the ranks of the groups for $t>T>0$ and $-t<-T<0$, thus determining the number of times $f_2$ is trivial.
73: \end{proof}
74:
75: Next, recall Proposition $2.4$ of \cite{STau}:
76:
77: \begin{prop}
78: \label{prop:skein}
79: In the exact sequence above, the map $f_2$ is non-trivial if and only if $\tau(\WDnmm)=1$. Furthermore, if $\tau(\WDnmm)\ne1$, then it is equal to $0$.
80: \end{prop}
81:
82: It is proved in \cite{Livingston1} and \cite{FourBall} that $\tau(K)$ satisfies the following inequality under the operation of changing a crossing in a projection of $K$: $$\tau(K_+)-1\le \tau(K_-) \le \tau(K_+),$$
83:
84: \noindent where $K_+$ (resp. $K_-$) denote the diagram with the positive (resp. negative) crossing.
85: Since each application of the skein sequence arose from changing a single negative crossing to a positive crossing, the above inequality becomes (for $k>0$):
86:
87: $$\tau(D_+(\companion,t'-k))-k\le \tau(D_+(\companion,t')) \le \tau(D_+(\companion,t'-k)).$$
88:
89: If $f_2$ were non-trivial for some $t'$ and trivial for $t'-k$, then Proposition \ref{prop:skein} would imply $\tau(D_+(\companion,t'-1))=1$ and $\tau(D_+(\companion,t'-k-1))= 0$, violating the inequality. Thus $f_2$ is trivial for the first $t-2\tau(K)$ applications of the skein sequence and non-trivial thereafter. Now \HFDm \ is determined by \HFD \ and knowledge of $f_2$, while $\tau(\WDnm)$ is determined by $f_2$, so this completes our proof of the proposition.\end{proof}
90:
91:
92: