Computations of Floer Lasagna Modules for Traces of Negative $L$-Space Knots

A skein lasagna module is an invariant of a pair $(X,L)$, where $X$ is a 4-manifold and $L\subset\partial X$ is a link, that was introduced by Morrison, Walker, and Wedrich in [19]. This invariant was recently used by Ren and Willis to give the first analysis-free proof of the existence of exotic compact orientable 4-manifolds in [24]. This result has since been reproved without the use of lasagna modules by Nahm in [20]. The invariant is constructed using Khovanov-Rozansky homology [11], which is a generalisation of Khovanov’s categorification of the Jones polynomial [12]. Further, the module is defined by considering the vector space over a field $\mathbb{F}$ generated by what are called Lasagna fillings of the pair $(X,L)$, modulo some equivalence relation. This definition is unwieldy and near impossible to compute except in very simple cases. However, for 2-handlebodies, that is 4-manifolds obtained by attaching 2-handles to $D^{4}$ along a framed link $L$, it was shown by Manolescu and Neithalath in [15] that the skein lasagna module can be computed in terms of the direct sum of the Khovanov-Rozansky homologies of cables of $L$ quotiented by relations coming from cobordism maps. This work was then extended by Manolescu, Walker, and Wedrich in [17] to any 4-manifold given a handle decomposition. However, even with these breakthroughs it remains difficult to compute the skein lasagna module as there are, in general, no formulas for the Khovanov-Rozansky homology of cables of a knot, let alone a link.

In another direction Ozsvath and Szabo introduced Heegaard Floer homology in [21], the hat flavour is denoted $\widehat{HF}$, which assigns an isomorphism class of Abelian groups to a closed oriented 3-manifold. They later extended this in [23] to define an isomorphism class of bigraded $\mathbb{F}_{2}$-vector spaces for pointed null-homologous links in three manifolds called link Floer homology and denoted $\widehat{HFL}$. It was then proven by Juhasz in [10] that given a decorated link cobordism, see Definition 2.3, $\Sigma$ from $L$ to $L^{\prime}$, there is a well defined linear map $F_{\Sigma}\colon\widehat{HFL}(L)\to\widehat{HFL}(L^{\prime})$. Using different techniques Zemke defined linear maps for all of the flavours of link Floer homology in [28]. These two techniques were shown to agree for the hat flavour in [8].

Using link Floer homology and these cobordism maps, Chen generalised the definition of a skein lasagna module to define a Floer lasagna module in [3], which is denoted $\mathcal{FL}(X,L)$. He also proved that similar to the skein lasagna module case there is a 2-handlebody formula for Floer lasagna modules in terms of the cables of the link defining the attaching spheres. Further, for an $L$-space knot $K$ and framing $n\geq 2g(K)$, Gorsky and Hom in [6] give a formula for the link Floer homology of the $(r,rn)$-cable of $K$ for all $r\geq 1$. Hence, using this it should, in theory, be possible to compute the Floer lasagna module of a knot trace $X_{n}(K)$, that is the 2-handlebody obtained by attaching a single 2-handle to $D^{4}$ along $K$ with framing $n$. However, in the formula for the Floer lasagna module of the 2-handle maps infinitely many cobordism maps must be computed. This presents a challenge as these maps are defined in terms of holomorphic triangle counts.

In this paper we present a method for reducing the complexity of computing this set of infinitely many cobordism maps to a single computation of a cobordism map $F_{K_{-n}}$ in the case of negative $L$-space knots. Recall that a negative $L$-space knot is a knot that admits a negative surgery where the Heegaard Floer homology of the resulting 3-manifold is as simple as possible, see [22]. The cobordism $F_{K_{-n}}$ is defined in Definition 6.7 and is between cables of $K$, where the cable depends on $n$ and the Alexander polynomial $\Delta_{K}(t)$. Formally we prove the following.

We compute the map $F_{U_{-n}}$ for $U$ the unknot and all $n>0$ in Lemma 6.12. Using this lemma and the above theorem we prove the following.

One of the main technical ingredients in the proof of the above corollary is the holomorphic triangle counts carried out in Proposition 6.11. This uses a particularly nice Heegaard triple and so does not easily generalise to other negative $L$-space knots. However, it is possible there is a more general method for doing these computations, which leads to the following question.

Before explaining the organisation of the paper we mention a detail about the proof of the main theorem that may be of independent interest. The main theorem relies on a technical result about how the maps on link Floer homology behave under spectral sequences. In particular we consider Ozsvath and Szabo’s spectral sequence [23] for forgetting link components, that is the spectral sequence from $\widehat{HFL}(L)$ to $\widehat{HFL}(L\backslash L_{0})\otimes\mathbb{F}^{2}$, where $L_{0}\subset L$ is a single component. We prove that a band map $F_{\Sigma}\colon\widehat{HFL}(L)\to\widehat{HFL}(L^{\prime})$ induces a map of spectral sequences and the map on the $E^{\infty}$-page is $F_{\Sigma\backslash L_{0}\times I}\otimes\text{Id}_{\mathbb{F}^{2}}$, see Section 4 for the relevant definitions and Proposition 4.16 for the exact statement. We also prove this for the quasi-stabilisation map $T^{+}$ and as a consequence pair of pants cobordisms. We believe that there should be a general version of these results for a link cobordism $\Sigma$ relating the $E^{\infty}$-page to the map induced by the cobordism obtained by forgetting all the components intersecting $L_{0}$. We plan to investigate this in future work.

We assume the reader has some familiarity with Heegaard Floer homology and in particular link Floer homology. All of our links will be null-homologous in some closed but not necessarily connected 3-manifold $Y$, and we will denote by $\mathbb{L}=(L,\mathbf{w},\mathbf{z})$ a pointed link where we assume there is at least one $w$ and one $z$ basepoint on each component. Further, if $\mathcal{H}$ is a multipointed Heegaard diagram for the pair $(Y,\mathbb{L})$ then we denote by $\widehat{CFL}(\mathcal{H})$ the hat flavour chain complex associated to this Heegaard diagram over the field $\mathbb{F}:=\mathbb{F}_{2}$. We will always implicitly assume that the Heegaard diagrams mentioned satisfy the necessary admissibility requirements. In the case that the link is in $S^{3}$ there is a $\mathbb{Z}$-grading called the Maslov grading and a $\mathbb{Z}^{|L|}$-grading called the Alexander multigrading. We will denote by $\widehat{HFL}_{i}(\mathcal{H},\mathbf{k})$ the homology in Maslov grading $i$ and Alexander grading $\mathbf{k}$ and we denote by $\widehat{HFK}_{i}(\mathcal{H},k)$ the vector space obtained by collapsing the Alexander multigrading to a single grading. That is, the vector space,

where $|\mathbf{k}|=\mathbf{k}_{1}+\ldots+\mathbf{k}_{|L|}$. As we will be considering maps associated to decorated link cobordisms we will need the notion of a transitive system from [7], but we follow the notation of [28].

An important thing to note from this is that if $F$ is a morphism of transitive systems, then it is completely determined by one of the maps $F_{i\to j}$. The two categories we are particularly interested in are vector spaces and chain complexes over $\mathbb{F}$. In the case of chain complexes we ask for the distinguished morphisms to be chain homotopy equivalences and we will only ask that the diagram (2.1) commutes up to chain homotopy equivalence. Further, given a transitive system of vector spaces $\{V_{i}\}_{i\in I}$ a canonical vector space $V_{I}$ can be defined in such a way that if $F$ is a morphism of transitive systems of vector spaces from $\{V_{i}\}_{i\in I}$ to $\{W_{j}\}_{j\in J}$, then there is a canonically defined linear map $F\colon V_{I}\to W_{J}$, see [7] for details. What is proven in [7] is that the collection $\{\widehat{CFL}(\mathcal{H})\}_{\mathcal{H}}$ over all Heegaard diagrams for $(Y,\mathbb{L})$ forms a transitive system of chain complexes. Formally, we have the following.

We will denote this transitive system by $\widehat{CFL}(Y,\mathbb{L})$ and by $\widehat{CFL}(\mathbb{L})$ when $Y\cong S^{3}$. Further, note that by taking homology of each of these we get a transitive system of vector spaces $\{\widehat{HFL}(\mathcal{H})\}_{\mathcal{H}}$ that preserves the gradings when they exist. Hence, there is a well defined vector space $\widehat{HFL}(Y,\mathbb{L})$ associated to a multipointed link in $Y$ and it has a $\mathbb{Z}\oplus\mathbb{Z}^{|L|}$ grading in $S^{3}$. To define morphisms between these it is not enough to just consider link cobordisms; instead, the notion of decorated link cobordisms is needed. These were introduced by Juhasz in [10], but we will use the reformulated definition from [28].

The colourings above are technically “type-partitioned, indexed colourings” in the language of [28], however we do not need this level of generality. Further, the main colouring that we consider is $\mathbb{J}=\mathbb{J}_{w}\cup\mathbb{J}_{z}$ where each of $\mathbb{J}_{w},\mathbb{J}_{z}$ is the set of connected components of $\Sigma$, and $J$ maps a connected component of $\Sigma_{\boldsymbol{\delta}}$ to the connected component of $\Sigma$ it lies on in $\mathbb{J}_{\boldsymbol{\delta}}$ for $\boldsymbol{\delta}\in\{\mathbf{w},\mathbf{z}\}$. For the Alexander grading this amounts to collapsing the Alexander gradings of all of the components that lie on the same connected component of $\Sigma$. Also we will often abuse notation and write $\Sigma\colon\mathbb{L}_{0}\to\mathbb{L}_{1}$ when the decorated link cobordism is in $S^{3}\times I$ and the set $\mathcal{A}$ is clear.

To a decorated link cobordism both [10] and [28] associate morphisms of transitive systems of chain complexes. The following theorem is a combination of theorems from the previous papers and the grading change formulas from [27].

Note from this that if we compute such a cobordism map for a single Heegaard diagram then this determines the cobordism maps for all Heegaard diagrams. There are two link cobordism maps that we will use throughout the paper so we recall their definitions now for ease of reading. The first is a quasi-stabilisation and we first recall the following topological preliminary.

Note that $\mathcal{H}^{+}$ is a Heegaard diagram for the same underlying link $L$ except the component $L_{i}$ now has two additional basepoints. We will denote this new pointed link by $\mathbb{L}^{+}$. In [26] it is shown that this construction can be used to define what is called a quasi-stabilisation map and that the resulting map is a morphism of transitive systems. We recall the definition now.

Note that in [28] there are three other quasi-stabilisation maps but we will not need these. The second cobordism we will need is a band map. We recall the definition of an $\alpha$-band.

There is clearly a saddle cobordism corresponding to such a band as pictured in Figure 2. To define the corresponding map on link Floer complexes we need the following definition of a Heegaard triple subordinate to this band.

Let $\mathcal{T}=(\Sigma,\boldsymbol{\alpha}^{\prime},\boldsymbol{\alpha},\boldsymbol{\beta},\mathbf{w},\mathbf{z})$ be a Heegaard triple subordinate to some $\alpha$-band $B$, then we denote by

Then from this definition we have that $\mathcal{T_{\alpha,\beta}}$ and $\mathcal{T}_{\alpha^{\prime},\beta}$ are Heegaard diagrams for $\mathbb{L}$ and $\mathbb{L}(B)$ respectively. It was also shown in [28] that $\mathcal{T}_{\alpha^{\prime},\alpha}$ is a Heegaard diagram for a $(|\mathbf{w}|-1)$-component unlink in $(S^{1}\times S^{2})^{\#g(\Sigma)}$, with two basepoints on each component except one, which has four. Due to this, $\widehat{HFL}(\mathcal{T}_{\alpha^{\prime},\alpha})$ has a unique element of highest Maslov grading. Finally, before defining the maps associated to these band maps we recall the definition of the holomorphic triangle count map associated to a Heegaard triple.

Combining all of these we get the map associated to the cobordism associated to an $\alpha$-band $B$ with the decorations as in Figure 2.

Finally, in this section we recall the definition of cylindrical boundary degenerations of curves. We will need this for some neck stretching arguments. Recall that in the Lipshitz cylindrical setup we consider certain $J$-holomorphic embeddings,

where $S$ is a Riemann surface with $2n$ punctures $\{p_{1},\ldots,p_{n},q_{1}\ldots,q_{n}\}\subseteq\partial S$ and $J$ is an almost complex structure satisfying certain properties, see [14] for details. Now recall the definition of cylindrical boundary degenerations, we use the formulation from [29].

We recap here the definition of Floer lasagna modules and the construction of $c\widehat{HFK}(K_{n})$ from [3]. For some properties and more exposition for Floer lasagna modules see [3] and for the definition of skein lasagna modules that this is based on see [19]. Further, see [24] for a more categorical approach to the definition. We start by recalling the definition of a lasagna filling.

We can then consider the infinite dimensional vector space $\mathbb{F}\{\text{lasagna fillings of }(W,\mathbb{L})\}$. To turn this into a more interesting invariant we define an equivalence relation on this vector space.

Using this equivalence relation we can define the Floer lasagna module of the pair $(W,\mathbb{L})$.

An observation made in [3] is that if any two lasagna fillings are equivalent then the surfaces must represent the same element in $H_{2}(W,L)$. Hence, we can write $\mathcal{FL}(W,\mathbb{L};\alpha)$ for the lasagna fillings in the class $\alpha$ and

In fact Chen defines a further bigrading on $\mathcal{FL}(W,\mathbb{L})$ coming from the Maslov and Alexander gradings, see [3] for the proof.

With the definition in hand, we now define $c\widehat{HFK}_{n}(K)$, which by a result from [3] is isomorphic as a graded vector space to $\mathcal{FL}(X_{n}(K))$. This vector space was defined by Chen in [3] and was based on a similar formula from [15]. We will only recall the definition in the case of a knot and no link in the boundary as this is all we will need for the paper.

The motivation for the definition comes from the observation that for a knot trace $X_{n}(K)$, if we consider the subspace

that is a neighbourhood of the cocore of the 2-handle and the boundary, then $B:=X_{n}(K)\backslash Y$ is a 4-ball. Further, any surface $\Sigma\subseteq int(X_{n}(K))$ intersects $Y$ as a collection of parallel copies of $(D^{2}\times\{0\})\cap Y$. Therefore, every lasagna filling is equivalent to a lasagna filling,

where $\tilde{K}$ is an $r$-component cable of $K$ with some orientations and $v\in\widehat{HFK}(\tilde{K})$. These fillings are then used to give the isomorphism with $c\widehat{HFK}(K_{n})$. We start by defining the cables of $K$ that are needed.

The definition takes a direct sum over all of the link Floer homology of these and then quotients out by certain images of some cobordisms. We recall the definition of these cobordisms now starting with the basepoint moving map.

The only thing we need to note about this map is that $F_{m_{j}}$ is clearly an isomorphism, as the inverse is given by the decoration going the other way. Further, by applying the equations in Theorem 2.4 we can see it also preserves the gradings. The next cobordisms are those associated to the braid action.

Note that again this is an isomorphism, with inverse given by $F_{\tau^{-1}}$, and preserves grading. To define the final cobordism map we need the following fact about link Floer homology.

Finally, using this we can define the pair of pants cobordism.

Note that we will often abuse notation and use $F_{P}$ to refer to both the map on homology and chain complexes, when the meaning is clear. Further, by using Theorem 2.4 we can see that $F_{P}$ preserves the Alexander grading and shifts the Maslov grading by $-1$. Finally, note that $F_{P}$ is a composition of a quasi-stabilisation map and a band map. With these cobordism maps defined we can now state the definition of $c\widehat{HFK}(K)$ and the theorem that it is isomorphic to $\mathcal{FL}(X_{n}(K))$ from [3].

As noted in [3] the above construction inherits a bigrading from the Maslov and Alexander gradings of the $K_{n}(a_{+},a_{-})$, as the shift cancels out the shift from the cobordism maps. It also admits a $H_{2}(X_{n}(K))\cong\mathbb{Z}$ grading given by summing only over $a_{+}$ and $a_{-}$ such that $a_{+}-a_{-}=\alpha$. We denote the part in Maslov grading $i$, Alexander grading $j$ and homological grading $\alpha$ by $c\widehat{HFK}_{i}(K_{n};\alpha,j)$. Hence, we can state the following theorem. Note in [3] it is proved for links but we will not need this.

The main aim of this section is to examine how the maps associated to bands, quasi-stabilisations and pair of pants cobordisms on link Floer homology interact with the forgetting components spectral sequence. We split the section into four parts. In the first we recap some definitions and results about spectral sequences coming from filtered chain complexes and set up notation. The second subsection is then dedicated to showing that the identification of the $E^{\infty}$-page is canonical. Then the next subsection is dedicated to the statement and proof of how the maps interact with the spectral sequence, and the final subsection to giving an example.