This paper computes various versions of link Floer homology for two-component L-space links, using Alexander polynomials to explicitly determine topological invariants like the Thurston polytope and norm.
Contribution
It introduces a method to compute link Floer homology for two-component L-space links based on Alexander polynomials, providing explicit topological invariants.
Findings
01
Explicit formulas for link Floer homology of 2-component L-space links.
02
Determination of Thurston polytope and norm from Alexander polynomials.
03
Method applicable to all 2-component L-space links.
Abstract
We compute different versions of link Floer homology HFL− and HFL for any L-space link with two components. The main approach is to compute the h-function of the filtered chain complex which is determined by the Alexander polynomials of every sublink of the L-space link. As an application, Thurston polytope and Thurston norm of any 2-component L-space link are explicitly determined by Alexander polynomials of the link and the link components.
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Full text
Heegaard Floer Homology Of L-space Links With Two Components
Beibei Liu
Department of Mathematics, UC Davis, One shields Avenue, Davis CA 95616, USA
We compute different versions of link Floer homology HFL− and HFL for any L-space link with two components. The main approach is to compute the h-function of the filtered chain complex which is determined by the Alexander polynomials of every sublink of the L-space link. As an application, Thurston polytope and Thurston norm of any 2-component L-space link are explicitly determined by Alexander polynomials of the link and the link components.
1. Introduction
In this article, we present a general method to compute the Heegaard Floer link homology for any L-space link with two components. Usually, it is very hard to compute the Heegaard Floer link homology HFL− and HFL. For L-space links, the computation is easier. Yajing Liu computed the link Floer homology HFL− for any L-space link with two components [5]. We revisit his computation and compute link Floer homology HFL of any L-space link with two components. As a consequence, we compute the Thurston polytope and Thurston norm of the link complement. Recall that for an L-space link with r components with a given generic admissible multipointed Heegaard diagram, we can associate the generalized Floer complexesA−(s) filtered by Alexander gradings. Here we work over F=F_2 and s∈H where H is some r-dimensional lattice (see Definition 2.3 and [6]). If the link L is an L-space link, there is an important result for the generalized Floer complexes:
Proposition 1.1**.**
[5, Proposition 1.11]**
For any L-space link, H_∗(A−(s))=F[[U]] with s∈H.
Here U has homological grading −2. Define −2h(s) as the homological grading of the unique generator in H_∗(A−(s)). By the work of E. Gorsky, A. Némethi and Yajing Liu, h(s) is determined by the Alexander polynomials Δ_L(t_1,t_2),Δ_L_1(t) and Δ_L_2(t) for any L-space link with two components L=L_1∪L_2 and s∈H. For any 2-component L-space link, there is a spectral sequence which converges to HFL−(L,s) [3]. In the spectral sequence, the E1-page is combinatorially determined by h(s) and the spectral sequence collapses at E2-page [3, Theorem 2.2.10] [5].
The computation of HFL(L,s) is more complicated. We introduce a bigraded “iterated cone” complex (C(s_1,s_2),d+d_1) in Section 3. There exists a spectral sequence associated to this bigraded complex where E1-page is defined by HFL− and E3=HFL(L,s_1,s_2). Theorem 3.2 shows that the E1-page of this spectral sequence is HFL−(s_1+1,s_2+1)⊕HFL−(s_1,s_2+1)⊕HFL−(s_1+1,s_2)⊕HFL−(s_1,s_2) and the differential d_1 is induced by the actions of U_1 and U_2. Lemma 3.3 in Section 3 indicates how U_i act on the Heegaard Floer link homology HFL−(L,s) for any s∈H. So we can compute E2-page of the spectral sequence. If d_2=0, the spectral sequence collapses at E2-page. If d_2 is nonzero, we need to use another strategy to compute HFL(L,s). We first find all possible cases where d_2 may be nontrivial. In order to compute HFL(L,s), we use the symmetry of Heegaard Floer link homology: HFL(L,s)≅HFL(L,−s) up to some grading shift [10, Equation 5]. In Section 3, we find that in all cases where d_2 may be nontrivial, the differential d_2 must vanish in the spectral sequence corresponding to HFL(L,−s). Then we can compute HFL(L,−s) and therefore HFL(L,s). Thus we compute HFL for all L-space links with two components and obtain the main theorem of this paper.
Theorem 1.2**.**
For any L-space link L=L_1∪L_2 with two components, HFL(s_1,s_2) is determined by the h-function and hence determined by the symmetrized Alexander polynomials Δ_L(t_1,t_2),Δ_L_1(t) and Δ_L_2(t) and the linking number lk of components L_1 and L_2.
Remark 1.3**.**
Heegaard Floer link homology depends on the orientation of the link. For any L-space link L=L_1∪L_2, we need to give this link an orientation which will determine the linking number of the two link components L_1 and L_2.
Yajing Liu [5] showed that rank_F(HFL−(L,s))≤2. We show that 4 is a bound for the rank of link Floer homology HFL for any L-space link with two components and give examples for all possible ranks from [math] to 4 in Section 3.
Corollary 1.4**.**
For any L-space link L=L_1∪L_2 with two components and any s∈H, rank_F(HFL(L,s))≤4. In particular, ∣χ(HFL(L,s))∣≤4.
In Section 4, we present an application of Theorem 1.2. It is known from the work of P.Ozsváth and Z.Szabó [12] that HFL(L) detects the Thurston norm of the link complement. Recall that for any compact, oriented surface with boundary (maybe disconnected) F=⋃_i=1nF_i, define its complexity as
[TABLE]
For any link L⊆S3, and any homology class h∈H_2(S3,L), there exists a compact oriented surface F with boundary embedded in S3−nd(L) which represents this homology class (i.e [F]=h). So for any homology class h∈H_2(S3,L;Z), we can assign a function:
[TABLE]
This function can be naturally extended to a semi-norm, the Thurston semi-norm, denoted by x:H_2(S3,L;R)→R [12]. The unit ball for the norm x is called Thurston polytope. By computing the convex hull of s∈H where HFL(L,s)=0, which is also called link Floer homology polytope, we can compute the dual Thurston polytope and thus Thurston norm by the work of P.Ozsváth and Z.Szabó [12]. So the Thurston polytope and Thurston norm for any 2-component L-space link L=L_1∪L_2 are determined by Alexander polynomials of every sublink, but in a very nontrivial way.
Theorem 1.5**.**
If L=L_1∪L_2 is an L-space link with two components in S3, then the Thurston norm of the link complement is determined by the Alexander polynomials Δ_L(t_1,t_2),Δ_L_1(t),Δ_L_2(t) and the linking number of the two components L_1 and L_2.
P.Ozsváth and Z.Szabó point out that for any alternating link, its Thurston polytope is dual to the Newton polytope of the multi-variable Alexander polynomial [12] and McMullen showed that the Newton polytope of the multi-variable Alexander polynomial of any link is contained in its dual Thurston polytope [7] . We compute the dual Thurston polytopes of two non-alternating L-space links with two components in Examples 4.3 and 4.4. They both agree with the Newton polytopes of Alexander polynomials. A natural question arises:
Question 1.6**.**
For any L-space link with two components which is not a split union of two L-space knots, is the Thurston polytope dual to the Newton polytope of its multi-variable Alexander polynomial?
Remark 1.7**.**
In Example 4.3, we present a 2-component L-space link where supp(HFL)={(s_1,s_2)∈H∣HFL(s_1,s_2)=0} is larger than supp(χ(HFL))={(s_1,s_2)∈H∣χ(HFL(s_1,s_2))=0}. But the convex hull of supp(HFL) is the same with the convex hull of supp(χ(HFL)) since the lattice points (s_1,s_2)∈H for which χ(HFL(s_1,s_2))=0 and HFL(s_1,s_2)=0 are inside of the convex hull of supp(χ(HFL)).
For any split L-space link, the answer to Question 1.6 is negative since its Alexander polynomial vanishes, but its dual Thurston polytope is nonempty. Example 5.5 gives the link Floer homology polytope of the split union of two right-handed trefoils. Recall that the split union of two L-space knots is an L-space link [5], and the h-function of the link satisfies that h(s_1,s_2)=h_1(s_1)+h_2(s_2) where (s_1,s_2)∈H and h_1,h_2 are h-functions of the link components L_1 and L_2 respectively. We compute HFL for any split union of two L-space knots. In general, we compute HFL for all 2-component L-space links with Alexander polynomial Δ(t_1,t_2)=0.
Theorem 1.8**.**
For any L-space link L=L_1∪L_2 with two components, if Δ_L(t_1,t_2)=0, HFL(L,s_1,s_2)≅HFL(L_1⊔L_2,s_1,s_2)≅HFL(L_1,s_1)⊗HFL(L_2,s_2)⊗(F⊕F_−1) where L_1⊔L_2 denotes the split union of L_1 and L_2 and (s_1,s_2)∈H.
ACKNOWLEDGEMENTS
I deeply appreciate Eugene Gorsky for introducing this interesting topic to me and his patiently teaching on Heegaard Floer homology, and also for his constant guidance and discussions during the project. I am also grateful to Allison Moore, Yi Ni and Jacob Rasmussen for the useful discussions on L-space links. The paper is inspired by the work of Yajing Liu, and the project is partially supported by NSF-1559338.
2. Heegaard Floer Link Homology
2.1. L-space Links
In [9], P.Ozsváth and Z.Szabó introduced the concept of L-space.
Definition 2.1**.**
A 3-manifold Y is an L-space if it is a rational homology sphere and its Heegaard Floer homology has minimal possible rank: For any spinc-structure s, HF(Y,s)=F has rank 1 and HF−(Y,s) is a free F[U]-module of rank 1.
In [4], E. Gorsky and A. Némethi define L-space links in terms of large surgeries.
Definition 2.2**.**
An l-component link L⊆S3 is an L-space link if there exist integers p_1,p_2,⋯,p_l such that for all integers n_i≥p_i,1≤i≤l, the (n_1,n_2,⋯,n_l)-surgery S_3n_1,n_2,⋯,n_l is an L-space.
The computation of Heegaard Floer link homology is not easy. However, L-space links have some nice properties and these make the computation of Heegaard Floer link homology easier. In particular, we only consider L-space links L=L_1∪L_2 with two components in this article.
For a 2-component L-space link L=L_1∪L_2 on S3, consider a generic admissible multi-pointed Heegaard diagram with each component L_i having only two basepoints w_i,z_i. Recall that in [6, Section 4], we can associate a generalized Floer complex A−(s_1,s_2) with (s_1,s_2)∈H(L) which is introduced by Manolescu and Ozsváth (see Definition 2.3). This complex is a free F[U_1,U_2]-module. The operations U_1 and U_2 are homotopic to each other on each A−(s_1,s_2), cf.[11] and both have homological degree −2.
Definition 2.3**.**
For an oriented link L=L_1∪L_2 with two components, define H(L) to be an affine lattice over Z2,
[TABLE]
where lk(L_1,L_2) denotes the linking number of L_1 and L_2.
By Proposition 1.1, for any L-space link L with two components, H_∗(A−(s_1,s_2))=F[[U]] where (s_1,s_2)∈H. Let −2h(s_1,s_2) denote the homological grading of the unique generator in H_∗(A−(s_1,s_2)). The function h(s_1,s_2) is the HFL-weight function of an L-space link defined in [3]. In this article, we will call it h-function. On each A−(s_1,s_2), the operations U_1 and U_2 are homotopic, and we denote them by U.
Let e_1=(1,0) and e_2=(0,1). For any s=(s_1,s_2)∈H, there exist inclusions j:A−(s_1,s_2)↪A−(s+e_i) for i=1 or i=2 which induce injections on homology as follows.*
[TABLE]
where j_∗=U_iδ(i) and δ(i)=0 or 1.
Remark 2.5**.**
The actions U_i induce maps U_i:A−(s+e_i)→A−(s) for i=1 and i=2. The actions also induce maps on homology. By Proposition 1.1, H_∗(A−(s))≅F[[U]] for any s∈H. Assume that a,b are unique generators of H_∗(A−(s)) and H_∗(A−(s+e_i)). Then j_∗(a)=Uδ(i)b and U_i(b)=U1−δ(i)a
Corollary 2.6**.**
For any L-space link with two components and s∈H, h(s)=h(s+e_i) or h(s)=h(s+e_i)+1 where i=1 or 2 and e_1=(1,0) and e_2=(0,1).
Proof.
By the lemma, we have −2h(s)=−2h(s+e_i)−2δ(i) where δ(i)=0 or 1. So h(s)=h(s+e_i) or h(s)=h(s+e_i)+1.
∎
Next, we will revisit Yajing Liu’s work about how to use the h-function to compute HFL−(L) for any L-space link L=L_1∪L_2 with two components [5].
Lemma 2.7**.**
[3, Lemma 2.2.9]**
For any (s_1,s_2)∈H, the chain complex CFL−(s_1,s_2) of the L-space link L=L_1∪L_2 is quasi-isomorphic to the following “iterated cone” complex:
[TABLE]
where i_1 and i_2 are the inclusion maps in Lemma 2.4.
Let d denote the differential in the generalized Floer complex A−(s_1,s_2) and i=i_2−i_1. The above “iterated cone” complex has two differentials d and i. The differential d acts in vertices of the cube and i acts in the edges. Define the cube grading ∣K∣ of the upper-right corner of cube to be [math]. The differential d decreases the homological grading by 1 and preserves the cube grading. The differential i preserves the homological grading and decreases the cube grading by 1. The total grading is defined as the sum of homological grading and the cube grading. Let D=d+i and K(s_1,s_2) denote the above “iterated cone” complex. There exists a spectral sequence whose E∞ page is the homology of the K(s_1,s_2) under the differential D=d+i.
Theorem 2.8**.**
[3, Theorem 2.2.10]**
Let L=L_1∪L_2 be an L-space link with two components. For any (s_1,s_2)∈H, there exists a spectral sequence which converges to HFL−(s_1,s_2) and collapses at its E2-page. Its E2-page is isomorphic to H_∗(H_∗(A−(s_1,s_2),d),i).
Thus HFL−(s_1,s_2) is isomorphic to H_∗(H_∗(A−(s_1,s_2),d),i) in the spectral sequence of the complex K(s_1,s_2). By Proposition 1.1, for any (s_1,s_2)∈H, H_∗(A−(s_1,s_2),d)≅F[[U]][−2h(s_1,s_2)] where −2h(s_1,s_2) is the homological grading of the unique generator in H_∗(A−(s_1,s_2),d) and U_1,U_2 act as U, homotopic to each other on A−(s_1,s_2), cf. [11]. To compute HFL−(s_1,s_2), we just need to compute the homology of the mapping cone of the inclusion map i:
[TABLE]
where a,b,c,d denote the unique generators in F[[U]][−2h(s_1,s_2)],F[[U]][−2h(s_1−1,s_2)],F[[U]][−2h(s_1−1,s_2−1)] and F[[U]][−2h(s_1,s_2−1)] respectively. Let h=h(s_1,s_2). By Corollary 2.6, there are 6 cases for the h-function corresponding to above mapping cone.
According to the h-function in Figure 1, we can compute the corresponding HFL−(s_1,s_2) in each case.
Case 1:i(b)=a,i(c)=b−d,i(d)=a and i(a)=0, so HFL−(s_1,s_2)=0.
Case 2:i(b)=a,i(c)=Ub−d,i(d)=Ua and i(a)=0, so HFL−(s_1,s_2)=0.
Case 3:i(b)=Ua,i(c)=b−Ud,i(d)=a and i(a)=0, so HFL−(s_1,s_2)=0.
Case 4:i(b)=a,i(c)=Ub−Ud,i(d)=a and i(a)=0, so HFL−(s_1,s_2)=⟨b−d⟩. Both b and d have homological grading −2h and cube grading 1. The total grading of b−d is −2h+1. Thus HFL−(s_1,s_2)=F[−2h+1].
Case 5:i(b)=Ua,i(c)=b−d,i(d)=Ua and i(a)=0, so HFL−=⟨a⟩ with total grading −2h . Thus HFL−(s_1,s_2)=F[−2h].
Case 6:i(b)=Ua,i(c)=Ub−Ud,i(d)=Ua and i(a)=0, so HFL−(s_1,s_2)=⟨a,b−d⟩. Here a has total grading −2h and b−d has total grading −2(h+1)+1=−2h−1. Thus HFL−(s_1,s_2)=F[−2h]⊕F[−2h−1].
Moreover, we can also determine the Euler characteristics χ(HFL−(s_1,s_2)) in these six cases. In Case 1, Case 2, Case 3 and Case 6, χ(HFL−(s_1,s_2))=0. In Case 4, χ(HFL−(s_1,s_2))=−1 and in Case 5, χ(HFL−(s_1,s_2))=1. Thus for any L-space link L=L_1∪L_2 with two components, once the h-function is determined, we can compute HFL−(s_1,s_2) for any (s_1,s_2)∈H.
Corollary 2.9**.**
For any L-space link with two components and (s_1,s_2)∈H, HFL−(s_1,s_2) is spanned by a or b−d or both. a has even grading and b−d has odd grading.
2.2. Alexander Polynomials of L-Space Links
In this section, we mainly introduce Yajing Liu’s work [5] about how to determine the h-function of any L-space link L=L_1∪L_2 with two components L_1,L_2 by the multi-variable Alexander polynomial Δ_L(t_1,t_2), the Alexander polynomials Δ_L_1(t) and Δ_L_2(t). Recall that for any L-space link L=L_1∪L_2 with two components, we have :
[TABLE]
[TABLE]
where f≐g means that f and g differ by multiplication by units. For any L-space link L=L_1∪L_2 with two components, Yajing Liu also defined normalization of its Alexander polynomial [5].
Definition 2.10**.**
([5, Definition 5.12])
Let the symmetrized Alexander polynomial of L be Δ_L(x_1,x_2) in the form of
[TABLE]
where t_i corresponds to the link component L_i for i=1 and i=2. Let the symmetrized Alexander polynomials of L_1 and L_2 be Δ_L_1(t),Δ_L_2(t) in the form of
[TABLE]
Let (i_0,j_0) be such that
[TABLE]
Then these Alexander polynomials are called normalized, if
(1) the leading coefficient of Δ_L_i(t) is 1 for both i=1,2.
(2) if a_j_0−lk/2+1/2L_2=1, then a_i_0,j_0L=1; while if a_L_2j_0−lk/2+1/2=0, then a_i_0,j_0L=−1 where lk is the linking number of L_1 and L_2.
For the normalized Alexander polynomials of the 2-component L-space link L=L_1∪L_2, χ(HFL−)(s_1,s_2)=a_Ls_1−1/2,s_2−1/2 and χ(HFK−(L_i,s))=a_L_is for i=1,2 [5]. Moreover, Yajing Liu gives the following formulas to determine the h-functions in [5, equation (5.8)]:
[TABLE]
Similary
[TABLE]
When s_1→+∞ or s_2→+∞
[TABLE]
[TABLE]
where h_1(s_1−lk/2),h_2(s_2−lk/2) are the corresponding h-functions for link components L_1 and L_2 respectively and s∈Z. For s sufficiently large, h_1(s)=h_2(s)=0. By using the above formulas, we can compute HFL−(s_1,s_2) for any L-space link L=L_1∪L_2 with two components and (s_1,s_2)∈H.
Remark 2.11**.**
The link components L_1 and L_2 of the 2-component L-space link are both L-space knots [5, Lemma 1.10].
Corollary 2.12**.**
[2, 3, 5]**
For any L-space link L=L_1∪L_2 with two components, HFL−(L) is determined by the Alexander polynomials Δ_L(t_1,t_2),Δ_L_1(t) and Δ_L_2(t).
3. Computation of HFL for L-space links with two components
3.1. The spectral sequence corresponding to HFL
In Section 2, we proved that for any L-space link L=L_1∪L_2 with (s_1,s_2)∈H, HFL−(s_1,s_2) is determined by the h-function. Now we are going to prove Theorem 1.2 that the Heegaard Floer link homology HFL(s_1,s_2) is also determined by the h-function.
Let C(s_1,s_2)=CFL−(s_1+1,s_2+1)⊕CFL−(s_1+1,s_2)⊕CFL−(s_1,s_2+1)⊕CFL−(s_1,s_2). For any (s_1,s_2)∈H, we have two operators U_1:CFL−(s_1,s_2)→CFL−(s_1−1,s_2) and U_2:CFL−(s_1,s_2)→CFL−(s_1,s_2−1). The action of U_1 (or U_2) is defined by h-function (See Lemma 3.3). Let D=d+d_1 where d is the differential in chain complex CFL−(s_1,s_2) with any (s_1,s_2)∈H and d_1=U_1−U_2. Then we get the “iterated cone” complex (C(s_1,s_2),d+d_1) in the following form:
[TABLE]
Lemma 3.1**.**
Suppose that L=L_1∪L_2 is an L-space link with two components L_1 and L_2. Let CFL(s_1,s_2) denote the chain complex of hat-version of Heegaard Floer link homology of L with (s_1,s_2)∈H. Then CFL(s_1,s_2) is quasi-isomorphic to the “iterated cone” complex (C(s_1,s_2),d+d_1).
Proof.
For this L-space link L=L_1∪L_2 with two components, we can write CFL(s_1,s_2) as:
[TABLE]
The quotient CFL−(s_1,s_2)/U_1(CFL−(s_1+1,s_2) can be realized as the cone of the map U_1:CFL−(s_1+1,s_2)→CFL−(s_1,s_2) and similarly the quotient CFL−(s_1,s_2+1)/U_1(CFL−(s_1+1,s_2+1)) can be realized as the cone of the map U_1:CFL−(s_1+1,s_2+1)→CFL−(s_1,s_2+1). Thus CFL(s_1,s_2) can be realized as a cone of the natural map induced by U_2 between these two cones.
∎
Theorem 3.2**.**
Let L=L_1∪L_2 be an L-space link with two components. For any (s_1,s_2)∈H, there exists a spectral sequence with the following properties:
(a) Its E2-page is isomorphic (as graded F-module) to H_∗(H_∗(C(s_1,s_2),d),d_1).
(b) Its E∞-page is isomorphic (as graded F-module) to HFL(s_1,s_2).
(c) The spectral sequence collapses at E3.
Proof.
For the “iterated cone” complex C(s_1,s_2) we mentioned above, it is doubly graded. One is the homological grading ν in the chain complex CFL−(s_1,s_2) with (s_1,s_2)∈H. We define cube grading∣C∣ in the cube of the ‘iterated cone” complex C(s_1,s_2). Fix (s_1,s_2)∈H. The cube grading is defined as (s_1+s_2)−(v_1+v_2) where (v_1,v_2)∈H. It is equivalent to saying the the cube grading of the lower-left corner of the cube is [math] and U_1 (or U_2) increases the cube grading by 1.
For this doubly-graded complex C(s_1,s_2) with two (anti)commuting differentials d and d_1, there exists a spectral sequence whose E1-page is H_∗(C(s_1,s_2),d) and converges to H_∗(C(s_1,s_2),d+d_1). By Lemma 3.1, its E∞-page is isomorphic to HFL(s_1,s_2). Its E1-page is H_∗(C(s_1,s_2),d) which can be written as HFL−(s_1+1,s_2+1)⊕HFL−(s_1+1,s_2)⊕HFL−(s_1,s_2+1)⊕HFL−(s_1,s_2). Its E2-page is H_∗(H_∗(C(s_1,s_2),d),d_1). The differential d_0=d in the spectral sequence preserves the cube degree ∣C∣ and decreases the homological degree ν by 1. The differential d_1 in E1-page increases the cube degree by 1 and decreases the homological degree ν by 2. For any nonnegative integer k, the differential d_k increases the cube degree by k, decreases the homological degree ν by k+1. The total homological degree is ν+∣C∣. By grading reason, the cube grading is less than or equal to 2. Thus for integer k>2, d_k=0 and this spectral sequence collapses at E3.
∎
By Theorem 3.2, HFL(s_1,s_2)≅E3. Then we can compute HFL(s_1,s_2) by computing E3-page of the above spectral sequence. The following lemma describes the action of U_1 (or U_2) on the E1-page of the spectral sequence in Theorem 3.2. This gives the action of d_1 on E1-page.
Lemma 3.3**.**
Consider the map U_1:HFL−(s_1+1,s_2+1)→HFL−(s_1,s_2+1). Let α be a generator of HFL−(s_1+1,s_2+1) with homological grading x. If there exists a generator β in HFL−(s_1,s_2+1) with homological grading x−2, then U_1(α)=β.
Proof.
Let a_1,b_1,c_1,d_1 denote the unique generators of H_∗(A−(s_1,s_2+1)), H_∗(A−(s_1−1,s_2+1)), H_∗(A−(s_1−1,s_2)) and H_∗(A−(s_1,s_2)) respectively in Figure 2. Let a,b,c,d denote the unique generators of H_∗(A−(s_1+1,s_2+1)), H_∗(A−(s_1,s_2+1)), H_∗(A−(s_1,s_2)) and H_∗(A−(s_1+1,s_2)) respectively in Figure 2. Here a_1 and b have different cube gradings as the generators of H_∗(A−(s_1,s_2+1)) and d_1,c have different cube gradings as the generators of H_∗(A−(s_1,s_2)). By the computation of HFL− in Section 2.1, h(s_1,s_2+1)=h(s_1+1,s_2) once HFL−(s_1+1,s_2+1) is nonempty. By the same reason, h(s_1−1,s_2+1)=h(s_1,s_2) since HFL−(s_1,s_2) is also nonempty. Assume the generator α=b−d with total homological grading −2h(s_1,s_2+1)+1. The generator a_1 has total homological grading −2h(s_1,s_2+1) and b_1−d_1 has total homological grading −2h(s_1−1,s_2+1)+1. By the assumption of this lemma, the total homological grading of β is −2h(s_1,s_2+1)−1, so β can only be b_1−d_1 and h(s_1−1,s_2+1)=h(s_1,s_2+1)+1. Now consider the map U_1:H_∗(A−(s_1,s_2+1))→H_∗(A−(s_1−1,s_2+1)) where H_∗(A−(s_1,s_2+1))=⟨b⟩ and H_∗(A−(s_1−1,s_2+1))=⟨b_1⟩. Since U_1 has homological degree −2, U_1(d)=d_1 by Lemma 2.4 and Remark 2.5. Similarly, U_1(c)=c_1. Then U_1(α)=U_1(b−d)=b_1−d_1=β. If α=a, then β=a_1 and we use the similar argument above to prove U_1(α)=β.
∎
Remark 3.4**.**
The map U_2:HFL−(s_1+1,s_2+1)→HFL−(s_1+1,s_2) can be described similarly to Lemma 3.3.
For the action of d_2 on E2-page, if it is nontrivial, we use the symmetry of Heegaard Floer link homology.
For an oriented L-space link L=L_1∪L_2 with two components and s=(s_1,s_2)∈H, there exists a relatively graded isomorphism*
[TABLE]
Remark 3.6**.**
In particular, the h-functions satisfy that h(−s)=h(s)+∣s∣ [5, Lemma 5.5] where ∣s∣=s_1+s_2.
3.2. Proof of the main Theorem
In this subsection, we are going to give the proof of Theorem 1.2 and show that 4 is an upper bound for the rank of link Floer homology HFL(s_1,s_2) for any 2-component L-space link and (s_1,s_2)∈H. Example 3.8 gives a 2-component L-space link where the rank of HFL(s_1,s_2) ranges from [math] to 4.
Proof of Theorem 1.2:
Assume that h=h(s_1+1,s_2+1). If d_2=0, then the spectral sequence in Theorem 3.2 collapses at E2-page, then we can use the computation of HFL− in Section 2.1 and Lemma 3.3 to compute HFL(s_1,s_2). For example, suppose that the h-function corresponding to HFL(s_1,s_2) is the following:
We obtain that the E2-page of the spectral sequence is:
[TABLE]
Recall that U_1 and U_2 both have homological grading −2. So U_1=U_2=0. By Theorem 3.2, we know that d_2 increases the cube grading by 2 and decreases the homological grading ν by 3, so d_2=0 by grading reason. Thus HFL(s_1,s_2)≅F[−2h−1]⊕F[−2h−1]⊕F[−2h−1]⊕F[−2h−1]. Here the cube grading for the generator in F[−2h−1] is [math]. We can use this method to compute HFL in all the cases where d_2=0. Now it suffices to consider the cases where d_2 may be nontrivial.
By grading reason, in order to have nontrivial d_2, HFL−(s_1+1,s_2+1) and HFL−(s_1,s_2) are both nonzero and contain a generator in each group such that their homological degree difference is 3. For nonzero HFL−(s_1+1,s_2+1), we have the following three possibilities for the corresponding h-function :
In Case 1, HFL−(s_1+1,s_2+1)=F[−2h+1]. In order to have nontrivial d_2, HFL−(s_1,s_2) must contain one generator with homological grading −2h−2 by grading reason. So the h-function corresponding to HFL−(s_1,s_2) can only have the pattern as in Case 2 or Case 3. Once the h-function is determined for HFL−(s_1,s_2), the h-functions for HFL−(s_1,s_2+1) and HFL−(s_1+1,s_2) are also determined by Corollary 2.6. Thus there are two possibilities for the h-function corresponding to HFL(s_1,s_2) where d_2 may be nontrivial:
In both cases, HFL−(s_1+1,s_2+1)=F[−2h+1],HFL−(s_1,s_2+1)=F[−2h]⊕F[−2h−1] and HFL−(s_1+1,s_2)=F[−2h]⊕F[−2h−1]. By Lemma 3.3, U_1a=b and U_2a=c where a is the unique generator in HFL−(s_1+1,s_2+1), and b,c are generators with homological grading −2h−1 in HFL−(s_1,s_2+1) and HFL−(s_1+1,s_2) respectively. So the image of a under the differential d_1 is nonzero and a does not survive in E2-page of the spectral sequence. Thus d_2 is trivial in both Case (1a) and Case (1b).
In Case 2, HFL−(s_1+1,s_2+1)=F[−2h]. In order to have nontrivial d_2, HFL−(s_1,s_2) must contain a generator with homological degree [−2h−3]. So the h-function of HFL−(s_1,s_2) must have the pattern as in Case 3. Then HFL−(s_1,s_2)≅F[−2h−2]⊕F[−2h−3]. Corresponding to this case, there are four possibilities of the h-function for HFL(s_1,s_2):
We will use the symmetry of Heegaard Floer link homology to compute HFL(s_1,s_2). Let h∗=h(−s_1,−s_2). By Remark 3.6, h(−s_1,−s_2−1)−h(−s_1,−s_2)=1−(h(s_1,s_2)−h(s_1,s_2+1)) and h(−s_1−1,−s_2)−h(−s_1,−s_2)=1−(h(s_1,s_2)−h(s_1+1,s_2)). So the h-function for HFL(−s_1,−s_2) corresponding to these four subcases are :
Observe that in all these four cases for HFL(−s_1,−s_2), HFL−(−s_1+1,−s_2+1)=0, so d_2=0 in the spectral sequence corresponding to HFL(−s_1,−s_2). Now the computation of HFL(−s_1,−s_2) is quite straightforward.
In Dual-h (2a),
[TABLE]
By the grading reason, d_2=U_1=U_2=0. Then it is easy to obtain HFL(−s_1,−s_2)≅F[−2h∗]⊕F[−2h∗]⊕F[−2h∗] and the Euler characteristic χ=3. By symmetry, HFL(s_1,s_2) should contain 3 generators with same total gradings. Observe that HFL−(s_1,s_2)=F[−2h−2]⊕F[−2h−3]. By grading reason, the generator with total grading −2h−2 survives in HFL(s_1,s_2). Thus HFL(s_1,s_2)≅F[−2h−2]⊕F[−2h−2]⊕F[−2h−2]. The Euler characteristic χ=3
In Dual-h (2b),
[TABLE]
In this case, HFL(−s_1,−s_2)≅F[−2h∗]. By the similar argument in Dual-h (2a), HFL(L)(s_1,s_2)≅F[−2h−2]. The Euler Characteristic χ=1.
In Dual-h (2c),
[TABLE]
By degree reason, d_2=U_1=U_2=0.Then HFL(−s_1,−s_2)≅F[−2h∗]⊕F[−2h∗]. So HFL(s_1,s_2)≅F[−2h−2]⊕F[−2h−2] and the Euler Characteristic is χ=2.
In Dual-h (2d),
[TABLE]
So HFL(L)(s_1,s_2)≅F[−2h−2]⊕F[−2h−2] by similar argument above and the Euler Characteristic is χ=2.
Now we can consider Case 3. In this case, HFL−(s_1+1,s_2+1)≅F[−2h]⊕F[−2h−1]. Then there are three possibilities for HFL−(s_1,s_2) if d_2 may be nontrivial. HFL−(s_1,s_2) is either F[−2h−4] or F[−2h−4]⊕F[−2h−5] or F[−2h−3]. If HFL−(s_1,s_2)=F[−2h−4], its h-function is shown in Case (3a) and if HFL−(s_1,s_2)≅F[−2h−4]⊕F[−2h−5], its h-function is shown in Case (3b) :
In Case (3a) and Case (3b), we observe that both generators in HFL−(s_1+1,s_2+1) have nontrivial images in HFL−(s_1,s_2+1) under U_1 and in HFL−(s_1+1,s_2) under U_2 by Lemma 3.3. So the two generators have nontrivial images under the differential d_1=U and cannot survive in E2-page. Thus d_2 is trivial in both cases.
If HFL−(s_1,s_2)≅F[−2h−3], there are four possibilities for the h-function corresponding to HFL(s_1,s_2):
Let h∗=h(−s_1,−s_2)=h(s_1,s_2)+s_1+s_2. By similar argument above, we find the h-function of HFL(−s_1,−s_2) corresponding to each case:
Observe that in all four cases, HFL−(−s_1,−s_2)=0. So d_2 is trivial in the spectral sequence corresponding to HFL(−s_1,−s_2). We compute HFL(−s_1,−s_2) and therefore HFL(s_1,s_2).
In Dual-h(3c), HFL(−s_1,−s_2)≅F[−2h∗+1]. By symmetry, HFL(s_1,s_2)≅F[−2h−3] with Euler Characteristic χ=−1.
In Case(3d), HFL(L)(s_1,s_2)≅F[−2h−3]⊕F[−2h−3]⊕F[−2h−3] and the Euler Characteristic is χ=−3 by similar computation above.
In Case(3e), HFL(L)(s_1,s_2)≅F[−2h−3]⊕F[−2h−3] and the Euler Characteristic is χ=−2.
In Case(3f), HFL(L)(s_1,s_2)≅F[−2h−3]⊕F[−2h−3] and the Euler Characteristic is χ=−2.
Thus we conclude that for any L-space link L=L_1∪L_2 with two components, once the h-function is determined, we can compute HFL(s_1,s_2) with any (s_1,s_2)∈H. By the equations in Section 2.2, the h-function is determined by Alexander polynomials Δ_L(x_1,x_2), Δ_L_1(t), Δ_L_2(t) and the linking number lk(L_1,L_2).
∎
Furthermore, We also get a bound for rank_F(HFL(s_1,s_2)) and the Euler characteristic χ(HFL(s_1,s_2)) with any (s_1,s_2)∈H.
Proof of Corollary 1.4:
Consider the following short exact sequence:
[TABLE]
where C_1(s_1,s_2+1) is the quotient complex with (s_1,s_2+1)∈H. By the computation of HFL(s_1,s_2), we have
[TABLE]
Now we claim that rank_F(H_∗(C_1(s_1,s_2+1)))≤2 with any (s_1,s_2)∈H. From the short exact sequence (3.1), we know that
[TABLE]
If rank_F(H_∗(C_1(s_1,s_2+1)))≥3, then at least one of HFL−(s_1+1,s_2+1) and HFL−(s_1,s_2+1) should have rank at least 2, and the other one should have rank at least 1. By the computation in Section 2.1, the possible h-functions corresponding to HFL−(s_1+1,s_2+1) and HFL−(s_1,s_2+1) are as follows:
Here we assume that the unique generator of H_∗(A−(s_1,s_2+1)) has homological grading −2h. In Case 1, we have U_1:F[−2h+2]⊕F[−2h+1]→F[−2h]. Let α denote the generator of F[−2h+2]⊆HFL−(s_1+1,s_2+1) and let β denote the generator of F[−2h]≅HFL−(s_1,s_2+1). By Lemma 3.3, U(α)=β. Then H_∗(C_1(s_1))≅F[−2h+1]. Its rank in Case 1 is 1. In Case 2, we have U_1:F[−2h+1]→F[−2h]⊕F[−2h−1]. By the similar argument, H_∗(C_1(s_1,s_2+1))≅F[−2h]. Then it has rank 1 in this case. In Case 3, we have U_1:F[−2h+2]⊕F[−2h+1]→F[−2h]⊕F[−2h−1]. The image of the generator of F[−2h+2] is the generator of F[−2h] and the image of the generator of F[−2h+1] is the generator of F[−2h−1]. So H_∗(C_1(s_1,s_2+1))=0. Thus with any (s_1,s_2)∈H, rank_F(H_∗(C_1(s_1,s_2+1)))≤2. By Equation (3.2), rank_F(HFL(s_1,s_2))≤rank_F(H_∗(C_1(s_1,s_2+1)))+rank_F(H_∗(C_1(s_1,s_2)))≤2+2=4 with any (s_1,s_2)∈H. Therefore −4≤χ(HFL(L,s_1,s_2))≤4.
∎
In fact, for the first example given in the proof of Theorem 1.2 where d_2=0, we have χ(HFL(L,s_1,s_2))=−4. Similarly, we can construct an example so that χ(HFL(L,s_1,s_2))=4.
Example 3.7**.**
Assume that the h-function corresponding to HFL(s_1,s_2) is as follows:
In this case, HFL(s_1,s_2)≅F[−2h−2]⊕F[−2h−2]⊕F[−2h−2]⊕F[−2h−2]. Thus χ(HFL(s_1,s_2))=4.
Example 3.8**.**
The Heegaard Floer link homology HFL of the two-bridge link b(20,−3).
Yajing Liu proved that the two-bridge link L=b(20,−3) is an L-space link [5, Theorem 3.8]. Its two link components are both unknots with linking number 2. Its normalized multi-variable Alexander polynomial is [2]:
[TABLE]
Let L_1 and L_2 denote the unknot components. We obtain the normalized Alexander polynomials of L_1 and L_2:
[TABLE]
Using results of Section 2.2, we compute the h-function for HFL(s_1,s_2) with any (s_1,s_2)∈H by the above Alexander polynomials. The h-function is shown in Figure 4. The numbers denote h(s_1,s_2) for any (s_1,s_2)∈H. For example h(0,0)=h(−1,0)=2 and the black dot ∙ denotes the lattice point (s_1,s_2)∈H where HFL(s_1,s_2) is nonzero. By explicit computation, the link Floer homology HFL(s_1,s_2) is shown in Figure 5 for any (s_1,s_2)∈H.
We observe that ∣χ(s_1,s_2)∣=rank_F(HFL(s_1,s_2)) for any (s_1,s_2)∈H and the rank of HFL(s_1,s_2) ranges from [math] to 4 for this link L. This indicates that the bound for the rank in Corollary 1.4 can be realized by some L-space link with some (s_1,s_2)∈H. Here rank_F(HFL(2,2))=1, rank_F(HFL(2,1))=2, rank_F(HFL(1,0))=3 and rank_F(HFL(0,0))=4, rank_F(HFL(3,0))=0.
4. Application to Thurston norm
P.Ozsváth and Z.Szabó showed that Heegaard Floer link homology detects the Thurston norm of the link complement [12]. In Section 3, for any L-space link L=L_1∪L_2 with two components and any s∈H, we computed HFL(L,s) by using the Alexander polynomials Δ_L(t_1,t_2), Δ_L_1(t), Δ_L_2(t) and the linking number lk(L_1,L_2). So we can compute the link Floer homology polytope for the link L and also compute its dual Thurston polytope and the Thurston (semi-)norm [12, Theorem 1.1].
In Section 1, we introduced complexity χ_−(F) for any compact oriented surface F with boundary. For any link L⊆S3, and any homology class h∈H_2(S3,L), we can assign a function:
[TABLE]
This function can be naturally extended to a semi-norm, the Thurston semi-norm, denoted by x:H_2(S3,L;R)→R.
Theorem 4.1**.**
[14, Theorem 1]**
The function x:H_2(S3,L;R)→R is a semi-norm that vanishes exactly on the subspace spanned by embedded surfaces of non-negative Euler characteristic.
Assume that L⊆S3 is a link with l components on S3. Let u_i denote the meridian of the ith-component L_i of L. Recall that every lattice point s∈H can be written as
[TABLE]
where s_i∈Q satisfies the property that
[TABLE]
is an even integer for i=1,⋯,l.
In [12] the Heegaard Floer link homology provides a function y:H1(S3−L;R)→R defined by the formula
[TABLE]
A trivial component of a link L is an unknot component which is also unlinked from the rest of the link.
For an oriented link L⊆S3 with no trivial components, its Heegaard Floer link homology detects the Thurston (semi-)norm of the link complement. For each h∈H1(S3−L;R), we have*
[TABLE]
where u_i is the meridian of the i-th component of L, and ∣⟨h,u_i⟩∣ denotes the absolute value of the Kronecker paring of h∈H1(S3−L;R) and u_i∈H_1(S3−L;R).
The unit ball for the norm x is called Thurston polytope, and the unit ball for the norm y is called link Floer homology polytope, which is also the convex hull of those s∈H for which HFL(L,s)=0. The unit ball for the dual norm x∗ of x in H_1(S3−L;R) is called dual Thurston polytope. By Theorem 4.2, twice link Floer homology polytope can be written as the sum of the dual Thurston polytope and an element of the symmetric hypercube in H1(S3−L) with edge-length two [12]. Next, we will give some examples of L-space links with two components and compute their link Floer homology polytopes by using the Alexander polynomials and linking numbers in detail. Moreover, we will compute the dual Thurston polytopes and Thurston norms of link complements by Theorem 4.2. We also compare the link Floer homology polytope and the convex hull of those s∈H for which χ(HFL(L,s))=0.
Example 4.3**.**
Dual Thurston polytope for L-space link L=L7n1.
The link L7n1 in Figure 6 is an L-space link [5, Example 3.17]. The link component L_1 is an unknot and the other link component L_2 is a right-handed trefoil. The linking number is 2 and the multi-variable Alexander polynomial is:
[TABLE]
The normalized Alexander polynomials of L_1 and L_2 are:
[TABLE]
[TABLE]
The h-function corresponding to HFL(s_1,s_2) with any (s_1,s_2)∈H is shown in Figure 7. In this figure, the numbers denote the h-function, and ∙ denotes the lattice point (s_1,s_2)∈H where HFL(s_1,s_2) is nonzero. By explicit computation, the link Floer homology HFL(s_1,s_2) is shown in Figure 8.
Moreover, HFL(0,0)≅F[−2]⊕F[−3], so χ(HFL(0,0)) is zero. For any other lattice point (s_1,s_2) denoted by ∙ except (0,0), HFL(s_1,s_2) has rank one and χ(HFL(s_1,s_2)) is also nonzero. Thus in this example, the link Floer homology polytope is the same with the convex hull of those (s_1,s_2)∈H for which χ(HFL(s_1,s_2)) are nonzero. By Theorem 4.2, the dual Thurston polytope on H_1(S3−L;R) is in Figure 9.
Here the thick red line is the dual Thurston polytope for L7n1. Observe that the dual Thurston polytope is the same with the Newton polytope of the Alexander polynomial Δ_L(t_1,t_2). The unknot component of L7n1 bounds a surface F_L_1 with Euler characteristic −1 and the right-handed trefoil link component L_2 bounds a surface F_L_2 with Euler characteristic −3. The surfaces F_L_1 and F_L_2 have maximal Euler characteristic in their respective homology classes.
Example 4.4**.**
Dual Thurston polytope for pretzel link L=b(−2,3,8)
We claim that the pretzel link b(−2,3,8) is an L-space link. It has two components. The link component L_1 is an unknot and the other link component L_2 is a right-handed trefoil shown in Figure 10. The linking number of this pretzel link is 5. Let P_1 be the knot obtained from b(−2,3,8) by 1-Dehn surgery on unknotted component L_1. The knot P_1 is the T(5,6;2,1) twisted torus knot [13, Proposition 3.1]. This twisted torus knot is an L-space knot proved by F. Vafaee [15, Theorem 1]. Suppose that d is sufficiently large. Then S_1,d3(L)=S_d−253(P_1) is an L-space. The link components L_1 and L_2 are L-space knots, so S_13(L_1) and S_d3(L_2) are both L-spaces. We also have d−25>0, so the pretzel link b(−2,3,8) is an L-space link by L-space surgery criterion [5, Lemma 2.6]. The symmetrized Alexander polynomial of b(−2,3,8) is:
[TABLE]
The h-function corresponding to HFL(s_1,s_2) with any (s_1,s_2)∈H is shown in Figure 13. By explicit computation, the link Floer homology HFL(s_1,s_2) is shown in Figure 11. In Figure 11, rank_F(HFL(1/2,1/2))=χ(HFL(1/2,1/2))=2 and rank_F(HFL(−1/2,−1/2))=χ(HFL(−1/2,−1/2))=2. The link Floer homology polytope is the same with the convex hull of those (s_1,s_2)∈H for which χ(HFL(s_1,s_2)) are nonzero. By Theorem 4.2, the dual Thurston polytope for b(−2,3,8) is the shaded area in Figure 12.
Remark 4.5**.**
For the L-space links L7n1 and b(−2,3,8), the Thurston polytopes are both dual to the Newton polytopes of their symmetrized Alexander polynomials Δ_L(t_1,t_2). P. Ozsváth and Z.Szabó point out that the Thurston polytope of an alternating link is dual to the Newton polytope of its multi-variable Alexander polynomial [12]. This is also true for L-space knot. A natural question is whether the Thurston polytope of an L-space link with two components (which is not a split union of two L-space knots) is dual to the Newton polytope of its symmetrized Alexander polynomial.
5. Two-component L-space links with vanishing Alexander polynomials
In Section 4, we have given examples of L-space links where the Thurston polytopes are dual to the Newton polytopes of their symmetrized Alexander polynomials. In this section, we mainly discuss 2-component L-space links with vanishing Alexander polynomials, especially split L-space links with two components. Recall that the multi-variable Alexander polynomials for split links are [math]. So the Newton polytopes for split L-space links are empty, but the link Floer homology polytopes may be nontrivial. To see this in detail, we need some lemmas first.
Lemma 5.1**.**
[5, Example 1.13(A)]**
Split disjoint unions of L-space knots are L-space links.
For a split L-space link L=L_1⊔L_2 with two components which are both L-space knots and any (s_1,s_2)∈H, the h-function h(s_1,s_2) satisfies that*
[TABLE]
where h_1(s_1) and h_2(s_2) denote the h-functions of knots L_1 and L_2 respectively.
Remark 5.3**.**
L-space knots can be regarded as special L-space links with just one component. For any L-space knot K⊆S3, we can associate a chain complex A−(s_1) filtered by Alexander grading and H_∗(A−(s_1)) has a unique generator for any s_1. Let −2h(s_1) be the homological grading of this unique generator.
Proposition 5.4**.**
Let L=L_1⊔L_2 be the split union of two L-space knots L_1 and L_2. Then HFL(L,s_1,s_2)≅HFL(L_1,s_1)⊗HFL(L_2,s_2)⊗(F⊕F_(−1)) with any (s_1,s_2)∈H.
Proof.
The proof is quite straightforward by using our computation of HFL(s_1,s_2) in Section 3. For any (s_1,s_2)∈H, the possible values for the h-function corresponding to HFK−(L_1,s_1) are like:
Here h_1(s_1)=a and a is any positive integer. Observe that
[TABLE]
[TABLE]
The long exact sequence is induced by the short exact sequence:
[TABLE]
By the long exact sequence above,
Case (1) HFK(L_1,s_1)≅0
Case (2) HFK(L_1,s_1)≅F[−2a]
Case (3) HFK(L_1,s_1)≅F[−2a+1]
Case (4) HFK(L_1,s_1)≅0
Similarly, for the link component L_2, we assume that h_2(s_2)=b. There are also four possibilities for the h-function corresponding to HFK(L_2,s_2). By Lemma 5.2, h(s_1,s_2)=h_1(s_1)+h_2(s_2). We find that there are only four possibilities for the h-function such that HFL(L,s_1,s_2)=0
In Case a, the h-functions for link components L_1 and L_2 are both like Case (2): (a+1)aa and (b+1)bb. Then HFL(s_1,s_2)≅F[−2(a+b)]⊕F[−2(a+b)−1] and HFK(L_1,s_1)≅F[−2a] and HFK(L_2,s_2)≅F[−2b]. So HFL(L,s_1,s_2)≅HFK(L_1,s_1)⊗HFK(L_2,s_2)⊗(F⊕F_(−1)).
In Case b, the h-function for link component L_1 is like Case (2): (a+1)aa and the h-function for L_2 is like Case (3): bbb−1. In Case c, the h-function for L_1 is like Case (3) and for L_2, h-function is like Case (2). In Case d, the h-functions for both components are like Case (3). Thus we can use similar argument above to prove that HFL(L,s_1,s_2)≅HFK(L_1,s_1)⊗HFK(L_2,s_2)⊗(F⊕F_(−1)) in these cases.
If the h-function corresponding to HFL(s_1,s_2) is not in the above four cases, HFL(s_1,s_2)=0, and at least one of HFK(L_1,s_1) and HFK(L_2,s_2) is zero. Thus the conclusion also holds.
∎
Proof of Theorem 1.8:
Let L=L_1∪L_2 be an L-space link with vanishing Alexander polynomial. The linking number of L_1 and L_2 is [math] by Equation (2.1). By Theorem 1.2, the Heegaard Floer link homology HFL(s_1,s_2) is determined by Δ_L(t_1,t_2), Δ_L_1(t) and Δ_L_2(t). So HFL(L,s_1,s_2)≅HFL(L_1⊔L_2,s_1,s_2)≅HFK(L_1,s_1)⊗HFK(L_2,s_2)⊗(F⊕F_(−1)) with any (s_1,s_2)∈H.
Example 5.5**.**
The link Floer homology polytope for the split disjoint union of two right-handed trefoils.
Let L=L_1⊔L_2 be the split disjoint union of two right-handed trefoils. Recall that the right-handed trefoil is an L-space knot with Alexander polynomial Δ_L_1(t)=t−1+t−1, and
[TABLE]
Observe the short exact sequence 0→A−(s_1−1)→A−1(s_1)→CFK−(s_1)→0. We have HFK−(L_1,s_1)=H_∗(A−(s_1)/A−(s_1−1)), and χ(HFK−(L_1,s_1))=h_1(s_1−1)−h_\textbullet(s_1) which is also the coefficient of ts_1 in 1−t−1Δ_L_1(t). Since L_1 is an L-space knot. h_1(s_1)=0 for sufficiently large s_1≫0. So the h-function h_1(s_1) can be determined as follows:
[TABLE]
where h_1(0)=h_1(−1)=1, h_1(s)=0 if s≥1 and h_1(s)=−s if s≤−1. Similarly, for another right-handed trefoil L_2, the h-function h_2(s_2) is the same with h_1(s_1). By Proposition 5.4, we can find all (s_1,s_2)∈H where HFL(L,s_1,s_2) are nonzero. So HFL(L,1,1)=F[0]⊕F[−1], HFL(L,0,1)=HFL(L,1,0)=F[−1]⊕F[−2], HFL(L,−1,1)=HFL(L,0,0)=HFL(L,1,−1)=F[−2]⊕F[−3] and HFL(L,−1,0)=HFL(L,0,−1)=F[−3]⊕F[−4], HFL(L,−1,−1)=F[−4]⊕F[−5]. For other lattice points (s_1,s_2)∈H, HFL(L,s_1,s_2)=0. Thus the link Floer homology polytope is a square in Figure 14.
Remark 5.6**.**
In general, let L=L_1⊔L_2 be the split union of any two L-space knots. Recall that the genus of a knot K is defined as:
[TABLE]
Here L_1 and L_2 are both L-space knots, so HFK(L_1,g(L_1))≅Z,HFK(L_2,g(L_2))≅Z [9, Theorem 1.2] and g(L_i)=max{s≥0∣HFK_∗(L_i,s)=0} for i=1 and i=2 [8, Theorem 1.2]. The link Floer homology polytope of L_i is the interval [−g(L_i),g(L_i)] where i=1 or 2. By Proposition 5.4, the link Floer homology polytope for L=L_1⊔L_2 is a rectangle with vertices (g(L_1),g(L_2)),(g(L_1),−g(L_2)),(−g(L_1),g(L_2)) and (−g(L_1),−g(L_2)) (see Figure 14).
Bibliography15
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] M.Borodzik, E. Gorsky. Immersed concordances of links and Heegaard Floer homology. ar Xiv e-prints, 2016.
2[2] N. Dawra. On the link Floer homology of L 𝐿 L -space links. ar Xiv: 1505. 01100 v 1, 2015.
3[3] E. Gorsky, A. Némethi. Lattice and Heegaard-Floer homologies of algebraic links. International Mathematics Research Notices, 2015, doi: 10.1093/imrn/rnv 075
4[4] E. Gorsky, A. Némethi. Links of plane curve singularities are L 𝐿 L -space link. Algebr. Geom. Topol. 16 (2016), no. 4, 1905-1912.
5[5] Y. Liu. L 𝐿 L -space surgeries on links. To appear in Quantum Topology.
6[6] C. Manolescu and P. Ozsváth. Heegaard Floer homology and integer surgeries on links. ar Xiv: 1011. 1317, 2010.
7[7] C. T. Mc Mullen. The Alexander polynomial of a 3-manifold of a 3-manifold and the Thurston norm on cohomology. Ann. Sci. de l’Ecole Norm. Sup., 35(2): 153-171, 2002.
8[8] P. Ozsváth, Z. Szabó. Holomorphic disks and genus bounds, Geom. Topol. 8 (2004), 311-334.