Rational cuspidal curves and symplectic fillings

To set notation, we begin with some basic facts about homology and cohomology. H_2(N;\mathbb{Z})\cong \mathbb{Z} with intersection form \langle n \rangle and is generated by [C]. H^2(N;\mathbb{Z})\cong \mathbb{Z} and is generated by a class G where \langle G, [C]\rangle =1. On the boundary, H^2(-Y;\mathbb{Z})\cong H_1(-Y;\mathbb{Z})\cong \mathbb{Z}/n\mathbb{Z}, generated by the meridian \mu of the surgery knot. The restriction map i^*: H^2(N;\mathbb{Z})\to H^2(-Y;\mathbb{Z}) sends G to the Poincaré dual of \mu.

Since, spin^c structures on a manifold M are affinely identified with H^2(M;\mathbb{Z}), given any two spin^c  structures \mathfrak{s} and \mathfrak{s}' on N, their difference \mathfrak{s}'-\mathfrak{s} represents a class in H^2(N;\mathbb{Z}). Under this identification, c_1(\mathfrak{s}')-c_1(\mathfrak{s}) = 2(\mathfrak{s}'-\mathfrak{s}).

We will first fix a reference spin^c  structure, \mathfrak{s}_0 on N, defined by the property that

One important property of this reference spin^c  structure is that its restriction to \partial N=-Y is self-conjugate. To see this, observe that

so \langle c_1(\mathfrak{s}_0)-c_1(\overline{\mathfrak{s}}_0), [C]\rangle = 2n. Therefore \mathfrak{s}_0-\overline{\mathfrak{s}}_0 = n G \in H^2(N;\mathbb{Z}). Let i^*: H^2(N;\mathbb{Z})\to H^2(-Y;\mathbb{Z}) denote the restriction map to the boundary. Then i^*(\mathfrak{s}_0-\overline{\mathfrak{s}}_0) = n\operatorname{PD}(\mu) = 0\in H^2(-Y;\mathbb{Z}). Therefore, i^*(\mathfrak{s}_0) is self-conjugate so it corresponds to a spin structure s_0 on -Y.

Recall the correspondence described above between spin structures on -Y and characteristic sublinks of the surgery diagram for -Y given by n-surgery on K. To see that the empty sublink does not correspond to s_0 when n is even, consider the unique spin^c  structure \mathfrak{t} on N which satisfies \langle c_1(\mathfrak{t}), [C]\rangle =0. This spin^c  structure is self-conjugate on all of N, since \mathfrak{t}-\overline{\mathfrak{t}}=0\in H^2(N;\mathbb{Z}). Therefore \mathfrak{t} is the image of the unique spin structure on N under the corresponding Bockstein homomorphism \beta: H^1(\operatorname{Fr}(N);\mathbb{Z}/2\mathbb{Z})\to H^2(\operatorname{Fr}(N);\mathbb{Z}). Next, observe that \langle c_1(\mathfrak{s}_0)-c_1(\mathfrak{t}), [C]\rangle =n, so \mathfrak{s}_0-\mathfrak{t}=\frac{n}{2}G\in H^2(N;\mathbb{Z}), and the restriction to -Y, i^*(\mathfrak{s}_0-\mathfrak{t})=\frac{n}{2}\operatorname{PD}(\mu)\in H^2(-Y;\mathbb{Z}). Since \frac{n}{2}\operatorname{PD}(\mu) is a non-zero element in H^2(-Y;\mathbb{Z}), i^*(\mathfrak{s}_0)\neq i^*(\mathfrak{t}). Therefore these two self-conjugate spin^c  structures on -Y correspond to distinct spin structures on -Y. Since i^*(\mathfrak{t}) corresponds to the empty sublink (by definition, the empty sublink of an integral surgery diagram corresponds to the spin structure which extends over the corresponding 4-manifold described as that 2-handlebody), s_0 must correspond to the non-empty sublink.

Now, we consider two spin^c  structures on N, our reference spin^c  structure \mathfrak{s}_0 and the spin^c  structure \mathfrak{s}_J, coming from an almost complex structure on N compatible with the symplectic structure, such that the complex tangencies of J along Y give the contact structure \xi. The difference of these two spin^c  structures corresponds to a class \mathfrak{s}_J-\mathfrak{s}_0\in H^2(N;\mathbb{Z}). We know that 2(\mathfrak{s}_J-\mathfrak{s}_0) = c_1(\mathfrak{s}_J)-c_1(\mathfrak{s}_0), so since H^2(N;\mathbb{Z})\cong \mathbb{Z} has no 2-torsion, we can determine the class \mathfrak{s}_J-\mathfrak{s}_0 uniquely from computing the Chern classes.

By definition, we chose \mathfrak{s}_0 to have c_1(\mathfrak{s}_0)=nG (where G is the generator dual to [C]). We can use the adjunction formula to compute

(This can be seen by replacing C by a curve which smooths its singularities symplectically, thus increasing the genus of C to its arithmetic genus without changing the homology class it represents.)

Therefore c_1(\mathfrak{s}_J)-c_1(\mathfrak{s}_0) = (2-2g+n-n)[G], so

Restricting this class to the boundary, \partial N = -Y, we have i^*(\mathfrak{s}_J)-i^*(\mathfrak{s}_0) = (1-g)\operatorname{PD}(\mu).

Now recall the definition of \Gamma(\xi,s_0) from the beginning of this subsection. Since \xi is given by the complex tangencies to -Y under the almost complex structure J, \mathfrak{s}_J restricts to the spin^c  structure on -Y determined by \xi. If \tau is a trivialization that induces the spin structure s_0, then \Gamma(\xi,s_0) is the Poincaré dual to the class in H^2(-Y;\mathbb{Z}) corresponding to the spin^c  structure i^*(\mathfrak{s}_J) (where the correspondence depends on \tau). Note that under this correspondence the spin^c  structure i^*(\mathfrak{s}_0) corresponds to the 0 element in H^2(-Y;\mathbb{Z}). Therefore, \Gamma(\xi,s_0)=\operatorname{PD}(i^*(\mathfrak{s}_J)-i^*(\mathfrak{s}_0)), and the result follows.

Tosun [Tosun, Theorem 1.1(b)] classified tight contact structures on small Seifert fibered spaces with e_0=-2 and r_1+r_2+r_3<2, showing that all such tight contact structures all arise from a Legendrian surgery picture associated to the plumbing (?), as in Figure ?. Since (Y_C,\xi_C) is fillable, \xi_C is tight [EliashbergGromov]. As explained above Y_C is the small Seifert fibered space with (e_0;r_1,r_2,r_3)=(-2;1/2,1/3,15/22). Since, r_1+r_2+r_3 = \frac{50}{33} < 2, Tosun’s classification result applies.

We will prove that the canonical contact structure is the unique tight contact structure (on the Seifert fibered space at hand) whose associated spin^c structure \mathfrak{t} extends to a rational homology ball. We do this by computing the Heegaard Floer correction term associated to the spin^c  classes of all tight contact structures, using Tosun’s classification.

We begin by observing that, since (Y_C,\xi_C) has a rational homology ball filling, then the correction term d(Y_C,\mathfrak{t}) vanishes. We set out to prove that this characterises the canonical contact structure (up to conjugation).

Since torus knots are L-space knots (in fact, lens space knots [Moser]), and the surgery coefficient is 64, which is larger than 2g(T(3,22))-2 = 40, Y is an L-space [LiscaStipsicz-I, Proposition 4.1]. (Now we could use Ghiggini’s classification result from [Ghiggini-Lclassification] instead of Tosun’s, as mentioned in the introduction.)

Since every tight contact structure on Y_C arises as the boundary of a Stein handlebody as in Figure ?, we can use this Stein filling to compute the degree of the contact invariant for each contact structure. The contact invariant is also the generator of the associated tower in \operatorname{HF}^+(Y), so its degree is the correction term associated to the corresponding contact structure.

The possible realizations of the link L are distinguished by the rotation numbers s, t of the components with framing -3 and -8: these can take values \pm1 and \pm6, \pm4, \pm2, or 0, respectively. All other rotation numbers vanish, since the framing is -2. Moreover, reversing both signs corresponds to conjugation. Let J_{s,t} be the complex structure on the Stein plumbing P associated to Legendrian realization of the link with rotation numbers s and t, and \xi_{s,t} the associated contact structure on Y_C.

Since c_1(J_{s,t}) evaluates on a sphere in the plumbing as the rotation number of the corresponding unlink, we can compute c_1^2(J_{s,t}) from the intersection matrix Q of the plumbing P. Q and its inverse Q^{-1} are:

Then c_1^2(J_{s,t}) = vQ^{-1}v^t, where v = (0,0,s,0,0,t).

For convenience, call P^* = P\setminus B^4, a cobordism from S^3 to Y. Then,

and it is easily verified that the minimum of this function, with s = \pm1 and t\in\{0,\pm2,\pm4,\pm6\} is 0, attained only at (s,t) = \pm(1,6); that is, only at the canonical contact structure and its conjugate.

Since the proof is very similar to that of Proposition 3.1, we omit the details that would be repeated and only include the calculations where the two cases differ.

Again, in this case we have r_1+r_2+r_3 = \frac{194}{129} < 2, so Tosun’s classification result still applies. Moreover, the surgery coefficient is again sufficiently large, so Y_C is again an L-space.

In the plumbing graph there are still only two rotation numbers s and t that vary, corresponding to the vertices of weight -6 and -8, respectively. s takes values in \{\pm2, 0\}, while t takes values in \{\pm4, \pm2, 0\}.

The intersection matrix Q associated to the plumbing is:

and its inverse is

Therefore, in the same notation as in the proof of Proposition 3.1,

and the minimum of this function as s and t vary in the corresponding ranges is 0, attained only at (s,t) = \pm(4,6); that is, only at the canonical contact structure and its conjugate.

It is apparent from the plumbing description of V_6, that there is a linear plumbing of symplectic spheres with self-intersections (-8,-2,-2,-2,-2). This is one of Fintushel–Stern’s original rational blowdowns L(p^2,p-1) with p=6. Rationally blowing down this chain yields a symplectic filling with b_2=1, so it is necessarily V_1.

Since V_6 itself is a plumbing, its replacement by the rational homology ball filling V_0 is itself a rational blow-down relation, and indeed this plumbing coincides with the plumbing of Figure 1(f) with q=2 in [BhupalStipsicz].

We now show that there is no rational blow-down from V_1 to V_0. There is a unique plumbing with b_2=1 which can be rationally blown down which is a -4-sphere. The non-torsion part of H_2(V_1) is generated by the orthogonal complement of the classes in the corresponding embedding, namely 3e_0 - e_1 - \dots - e_7. Therefore the intersection form of V_1 is \langle -16 \rangle, so no homology class has self-intersection -4. Thus there can be no rational blow-down from V_1 to V_0.

First, we observe the existence of the rational blow-downs from W_9 to W_6 and W_4, and from W_6 and W_4 to W_1. Using the fact that W_9 is necessarily the plumbing (?) (by Proposition 3.3), we can visibly see sub-plumbings of the form (-8,-2,-2,-2,-2) and (-6,-2,-2). Note that these symplectic embeddings are not unique, but it is possible to realize disjoint embeddings of both the (-8,-2,-2,-2,-2) and (-6,-2,-2) chains as disjoint sub-plumbings. Rationally blowing down a (-8,-2,-2,-2,-2) chain yields a symplectic filling with b_2=4 which is necessarily W_4. Rationally blowing down a (-6,-2,-2) chain yields symplectic fillings with b_2=6, which must be either W_6 or Z_6. Rationally blowing down disjoint plumbings (-8,-2,-2,-2,-2) and (-6,-2,-2) yields a symplectic filling with b_2=1, so it must be W_1. Using the homological properties that define Z_6, we will show below that Z_6 does not admit any further rational blow-downs. Therefore, rationally blowing down a (-6,-2,-2) plumbing in W_9 which is disjoint from a (-8,-2,-2,-2,-2) plumbing, necessarily results in W_6. This shows as well that there is a further rational blow-down from W_6 to W_1. If we exchange the order of which of the two disjoint plumbings we rationally blow-down first, we see the sequence of rational blow-downs from W_9 to W_4 to W_1.

Next, we will show there is a rational blow-down from W_9 to Z_6. Consider the (-6,-2,-2) chain embedded as a sub-plumbing of W_9 given by turning right instead of left at the 3-valent vertex. This sub-plumbing is not disjoint from the (-8,-2,-2,-2,-2) chain, and thus potentially yields a different result from the rational blow-down of the other embedding of the (-6,-2,-2) chain (where you turn left at the 3-valent vertex). We will now verify that rationally blowing down this embedding of the (-6,-2,-2) plumbing is not W_6, and thus must be Z_6.

First, we argue that the result of this rational blow-down is simply connected. Since W_9 is a tree plumbing of spheres, it is simply connected. The fundamental group of the rational homology ball which replaces the (-6,-2,-2) plumbing is normally generated by the meridian of the last -2-sphere in the chain, so it suffices to show this curve is null-homotopic in the complement of the plumbing in W_9. This meridian can be realized as the equator of the next -2-sphere S in the W_9 plumbing, so it bounds a disk (the other half of the -2-sphere S) in the complement of the (-6,-2,-2) chain. Consequently the result of this rational blow-down is a simply-connected filling, and in particular it has trivial H_1.

We will next show that W_6 necessarily has non-trivial H_1. To see this, consider the long exact sequence of the pair (W_6,\partial W_6):

Here we identify H_2(W_6,\partial W_6)\to H^2(W_6) by Alexander–Lefschetz duality. Under this identification, the map H_2(W_6)\to H^2(W_6) in the sequence is described by the intersection form Q_{W_6} of W_6, by sending a\in H_2(W_6) to Q_{W_6}(a,\cdot). The size of the cokernel of this map is the determinant of Q_{W_6}. In [bigGS, Proof of Theorem 6.5, p. 52], we computed the intersection forms for W_6 and Z_6, by finding an integral homology basis for the orthogonal complement in H_2(\mathbb{C}\mathbb{P}^2\#16\overline{\mathbb{C}\mathbb{P}}\,\!^2) of the corresponding embeddings of the resolutions for W_6 and Z_6 (which are listed above). We calculated that \det(Q_{W_6})=64 and \det(Q_{Z_6})=256. (Note, this distinguishes W_6 from Z_6.) On the other hand, H_1(\partial W_6) \cong \mathbb{Z}/{256}\mathbb{Z} since \partial W_6 = -S^3_{256}(T(6,43)). Since the co-kernel of the map H_2(W_6)\to H^2(W_6) has order 64, the map from H^2(W_6) to H_1(\partial W_6)\cong \mathbb{Z}/{256}\mathbb{Z} cannot be surjective. Therefore, there are some elements which are not in the kernel of the map H_1(\partial W_6)\to H_1(W_6), so H_1(W_6)\neq 0. This shows that the rational blow-down of this embedding of the (-6,-2,-2) chain is necessarily Z_6.

Finally, the entirety of the W_9 plumbing can be rationally blown down in a non-linear way (this plumbing is [BhupalStipsicz, Figure 1(j)] for q=4) to W_0.

Next, we will show that fillings which are not connected by edges are not related by a symplectic rational blow-down. Note that because rational blow-down strictly decreases b_2, we only need to obstruct rational blow-downs from larger fillings to smaller fillings.

We will next show that Z_6 cannot be symplectically rationally blown down to any other filling. To do this, we start by looking at the classes in H_2(Z_6) which could be represented by a symplectic sphere. Classes in H_2(Z_6) must be in the orthogonal complement of the classes listed above for the Z_6 embedding of the resolution in H_2(\mathbb{C}\mathbb{P}^2\#16\overline{\mathbb{C}\mathbb{P}}\,\!^2). Classes in this orthogonal complement are classes of the form \sum_{i=0}^{15} a_ie_i where

where, using the equalities above, the last equation can equivalently be written as

If the class represents a symplectic sphere, by Lemma 4.3, there exists an index i_0\in \{0,\dots, 15\} such that a_{i_0}=1 and a_i\in \{-1,0\} for all i\neq i_0. Combining this with the forced equalities of coefficients above, we see that i_0\notin\{1,6,7,8,10,\dots, 15\}. Furthermore i_0\neq 9 because if a_9=1, a_6=a_7=a_8=-1, but then there are no solutions to Equation (?) if a_0,\dots,a_5\in \{0,-1\}. Therefore the only possibilities are that i_0\in \{0,2,3,4,5\}, and we get the following possibilities:

Note that classes of the first type have square -9, the second type have square -3 and the third type have square -8, and the last have square -2. In particular, any symplectic plumbing of spheres which embeds in Z_6 can only include spheres with self-intersection numbers in the set \{-2,-3,-8,-9\}.

We start with a general observation: any plumbing of spheres which can be symplectically rationally blown down contains at least one sphere with self-intersection strictly less than -3. For linear plumbings this follows from the fact that they are all obtained as 2-expansions of (-4), and for non-linear plumbings this follows from inspection of the families in [BhupalStipsicz] (note the parameters p,q,r must be non-negative).

Therefore, the only possible plumbings that may be symplectically rationally blown down which can embed in Z_6 must include at least one sphere of self-intersection -8 or -9. Because b_2(Z_6)=6, any embeddable plumbing must have b_2\leq 6. The only 2-expansion of (-4) of length \leq 6 which includes a -9-sphere is (-9,-2,-2,-2,-2,-2). Those that include a -8-sphere are (-2,-8,-2,-2,-2,-3) and (-8,-2,-2,-2,-2). For the non-linear plumbings in [BhupalStipsicz], we can immediately rule out embeddings of plumbings in Figure 1(a),(e),(h),(i),(j) and 2(a),(b),(c) because they contain either -4- or -6-spheres. We rule out 1(b),(c),(d),(g) because any plumbing in these families has b_2>6 when self-intersections are restricted to lie in \{-2,-3,-8,-9\}. Thus the only non-linear plumbing which could be rationally blown down with b_2\leq 6, and self-intersection numbers in this class is the case of [BhupalStipsicz, Figure 1(f) with q=2], which happens to be the plumbing (?). We will rule out these remaining cases now using the pairwise intersections of these classes (we will freely use the lemmas of Section ?). If there were an embedding of the linear plumbing (-9,-2,-2,-2,-2,-2) in Z_6, up to permuting the indices \{2,3,4,5\}, the first sphere in the chain would represent e_0-e_1-e_{10}-\dots-e_{15}-e_2, and the next three must represent e_2-e_3, e_3-e_4, e_4-e_5 (in order for these spheres to have pairwise intersections according to the linear chain). This leaves no possibility e_i-e_j, i,j\in\{2,3,4,5\} for the fourth -2-sphere in the chain which has intersection 0 with the first two -2-spheres and 1 with the third. An embedding of (-8,-2,-2,-2,-2) would necessarily (up to a permutation of \{2,3,4,5\}) have the first four spheres in homology classes e_2-e_1-e_{10}-\cdots-e_{15}, e_3-e_2, e_4-e_3, e_5-e_4, leaving no possible i,j\in \{2,3,4,5\} such that e_i-e_j could have intersection +1 with e_5-e_4 and intersection 0 with the first three classes. Note that the non-linear plumbing (?) which is case 1(f) on the Bhupal–Stipsicz list, contains the linear chain (-8,-2,-2,-2,-2) as a sub-plumbing so this non-linear plumbing is also obstructed from symplectically embedding into Z_6. Similarly, an embedding of the linear chain (-2,-8,-2,-2,-2,-3) would (up to permutation of \{2,3,4,5\}) have the first four spheres in classes, e_2-e_3, e_3-e_1-e_{10}-\cdots - e_{15}, e_4-e_3, e_5-e_4, leaving no options for the fifth sphere in the chain with the correct intersection number. Thus there is no embedding into Z_6 of any plumbing which can be symplectically rationally blown down.

Note at this point, we have established that Z_6 cannot be the filling which results from rationally blowing down the (-6,-2,-2) chain in W_9 which is disjoint from the (-8,-2,-2,-2,-2) chain (since Z_6 admits no further rational blow-downs). Because W_6 is the only other symplectic filling with b_2=6, W_6 is necessarily the result of the rational blow-down of this disjoint (-6,-2,-2) chain in W_9. This establishes that W_6 can be rationally blown down to W_1, and gives another way to see that W_6 and Z_6 are not symplectomorphic.

Next, we will similarly obstruct rational blow-downs from W_6 to W_4 and W_0, though the obstruction to a rational blow-down to W_0 will be significantly more subtle. In this case, the orthogonality relations imply classes in H_2(W_6;\mathbb{Z}) are precisely those of the form \sum_{i=0}^{15}a_ie_i where

Proceeding as before, the classes satisfying these constraints which are represented by symplectic spheres are

Note that the squares of these classes are \{-9,-3,-8,-2\}. Since the only plumbing with b_2=2 which can be rationally blown down is (-5,-2) and there are no symplectic -5-spheres in W_6, there can be no symplectic rational blow-down from W_6 to W_4. We can obstruct the embeddings into W_6 of symplectic linear plumbings of with b_2=6 as before. The only two possibilities are (-9,-2,-2,-2,-2,-2) and (-2,-8,-2,-2,-2,-3). In the former case, the first five spheres in the chain would necessarily (up to permuting indices \{2,\dots, 6\}) represent e_0-e_1-e_{10}-\cdots-e_{15}-e_2, e_2-e_3, e_3-e_4, e_4-e_5, e_5-e_6, leaving no option for the sixth sphere in the chain. In the latter case, the chain would need to take the form e_3-e_2, e_2-e_1-e_{10}-\cdots-e_{15}, e_4-e_2, e_5-e_4, e_6-e_5, e_0-e_6-e_j for some j\in\{2,\dots, 5\}, but any value of j would result in a non-zero intersection of the last sphere in the chain with one of the first four spheres.

Next we consider potential non-linear plumbings which can be rationally blown down. In fact, there is a unique non-linear plumbing from the Bhupal-Stipsicz list with b_2=6 which we cannot rule out with homology classes alone. This is the plumbing [BhupalStipsicz, Figure 1(f)] for q=2, which happens to be the plumbing (?) (the filling of the other exceptional cuspidal contact manifold). This consists of a linear chain (-8,-2,-2,-2,-2) with an additional -3-sphere intersecting the second to last -2-sphere. Up to permutation of indices \{2,\dots, 6\}, the spheres in the linear chain necessarily represent e_2-e_1-e_{10}-\dots-e_{15}, e_3-e_2, e_4-e_3, e_5-e_4, e_6-e_5 and the additional -3-sphere would represent e_0-e_5-e_6. Since all of these classes could be represented by symplectic spheres in W_6, obstructing this rational blowdown is a little more subtle.

Suppose there exists such a symplectic plumbing in W_6. By gluing the concave neighborhood of the resolution, this is equivalent to assuming that in \mathbb{C}\mathbb{P}^2\#16\overline{\mathbb{C}\mathbb{P}}\,\!^2, there exists an embedding of the resolution configuration (with homology classes as specified for W_6) and a disjoint embedding of the plumbing (?) representing homology classes as described above. Observe that the classes e_7 and e_{15} intersect all of the components in these two configurations non-negatively. Therefore, by Lemma 4.2, there exist embedded exceptional spheres representing e_7 and e_{15} which intersect all curve components in these configurations non-negatively (so the geometric and algebraic intersection numbers match). Consequently, we have an embedding into \mathbb{C}\mathbb{P}^2\#16\overline{\mathbb{C}\mathbb{P}}\,\!^2 of a configuration of curves as in Figure ?. Blowing down symplectic -1-spheres in this configuration representing e_i classes repeatedly as in Figure ?, we eventually reach a configuration of curves in \mathbb{C}\mathbb{P}^2 consisting of the following components: a cubic C with a simple cusp at a point r, a line L which tangentially intersects C at an inflection point p (with multiplicity 3), and a line T which intersects C and L transversally at p, and intersects C tangentially (with multiplicity 2) at an additional point q. (Here r,p,q are all distinct points on C.) In fact, such a configuration of symplectic curves cannot exist, by the Riemann–Hurwitz formula. We fix an almost complex structure J such that C, L, and T are J-holomorphic. Then, using the pencil of J-holomorphic lines through p gives a degree-2 map \pi\colon\mathbb{C}\mathbb{P}^1\to\mathbb{C}\mathbb{P}^1 with at least two ramification points (corresponding to the inflection line and the cusp respectively). Therefore Riemann–Hurwitz reads: 2 = 2\cdot 2 - \sum (e_\pi(p)-1), which implies that these are the only two ramification points, from which we deduce that there is no other tangent drawn to the cubic from the inflection point. (In fact, this exact argument appears in [bigGS, p. 51].) Thus, we reach a contradiction, and the plumbing (?) cannot embed symplectically in W_6, so we conclude there is no rational blow-down from W_6 to W_0.

To see there is no rational blow-down from W_4 to W_0, observe that H_2(W_4) consists of classes \sum_{i=0}^{13} a_ie_i where

The classes of this form which can be represented by symplectic spheres are

These classes have squares -6 or -2. The only plumbing with b_2\leq 4 which can be symplectically rationally blown down and involves only -6- and -2-spheres is the linear chain (-6,-2,-2) which connects W_4 to W_1. There is no such plumbing of length 4 so there is no symplectic rational blow-down from W_4 to W_0.

The classes in H_2(W_1) have the form \sum_{i=0}^{10}a_ie_i such that

These are all integer multiples of 3e_0-e_1-\dots-e_6-e_{10} which has square -16, so there is no -4-sphere in W_1 to rationally blow down.

Suppose (Y_C,\xi_C) has a strong symplectic filling W. Let N_C denote a standard concave neighborhood of C, and let (X,\omega) denote the closed symplectic manifold which results from symplectically gluing N_C to W along their common contact boundary. Symplectically blow up X at n:=C\cdot C- C'\cdot C' smooth points of C\subset X. The proper transform of C yields a symplectic embedding of C' in X\# n\overline{\mathbb{C}\mathbb{P}}\,\!^2. Since C'\cdot C'>0, there exists a concave neighborhood of C' [bigGS, Theorem 2.13] in X\#n\overline{\mathbb{C}\mathbb{P}}\,\!^2 whose contact boundary is (Y_{C'},\xi_{C'}). The complement of this neighborhood is a strong symplectic filling of (Y_{C'},\xi_{C'}).

Without loss of generality, suppose that p_1 is a singular cusp with \ell(p_1) = \ell = \min \ell(p_i). By Lemma 6.1 above, it suffices to prove the statement for C\cdot C = \sum M(p_i) + \sum M(q_i) + 2\ell + 2.

Let (W,\omega_W) be a strong symplectic filling of (Y_C,\xi_C), and (X',\omega') be the closed symplectic manifold obtained by gluing a standard concave neighborhood of C to (W,\omega_W). We now view C as a symplectic curve in X'.

Blow up C at all its singular points until we obtain the minimal smooth resolution of C, i.e. the proper transform \widetilde{C} of C is a smooth sphere of self-intersection \widetilde{C}\cdot\widetilde{C} = C\cdot C - \sum M(p_i) - \sum M(q_j) = 2\ell + 2. Note that, by definition, at p_1 the last blow-up creates an exceptional divisor that intersects \widetilde{C} at a single point with multiplicity \ell.

Blow up \ell more times at p_1, so that at p_1 we get to the normal crossing resolution of the singularity (C,p_1), and then \ell+1 more times at the intersection between (the proper transform of) C and the last exceptional divisor. Let (X,\omega) denote the corresponding blow-up of (X',\omega').

The 4-manifold X contains the following configuration of symplectic spheres: a +1-sphere C', the proper transform of C, an exceptional divisor intersecting C transversely once, and a string of -2-curves departing from it; this is depicted in (?).

The left leg contains exactly \ell-1 vertices of weight -2, while the right leg contains \ell of them. In total, the chain of -2-vertices has length 2\ell.

By Theorem 4.1, X is symplectomorphic to a blow-up of \mathbb{C}\mathbb{P}^2, and C' can be identified with a line in \mathbb{C}\mathbb{P}^2. Using the standard basis for H_2(\mathbb{C}\mathbb{P}^2\#N\overline{\mathbb{C}\mathbb{P}}\,\!^2), by Lemma 4.3 the -1-sphere adjacent to C' is in the homology class h-e_0-e_1. Since the chain of -2s is of length 2\ell > 2, by Lemma 4.6, up to relabeling the e_i, the homology classes in the chain are e_1-e_2, \dots, e_{2\ell}-e_{2\ell+1}. The central vertex in the plumbing is in the homology class e_{\ell+1}-e_{\ell+2}.

The (-\ell-1)-sphere in (?) is disjoint from the line C', so by Lemma 4.3 its homology class is of the form e_i - e_{j_1} - \cdots - e_{j_\ell} for some i, j_1, \dots, j_\ell. It is also disjoint from the two halves of the chain of -2s, so the coefficients appearing in the homology class of the (-\ell-1)-sphere of e_1,\dots,e_{\ell+1} are all equal—namely they are either all 0 or all -1. Since the (-\ell-1)-sphere intersects the class e_{\ell+1}-e_{\ell+2} once positively, either i=\ell+2 or j_k=\ell+1 for some k. The former possibility is ruled out by Lemma 4.5, therefore e_1,\dots, e_{\ell+1} must all appear with coefficient -1 in the homology class for the (-\ell-1)-sphere. However, the self-intersection of this class is at most -\ell-2, a contradiction. This proves the first assertion.

Note that when C is a rational cuspidal curve, i.e. if \nu = 0, the 3-manifold Y_C, that is the boundary of a regular neighborhood of C, is a rational homology sphere; in particular, since every weak symplectic filling of \xi_C can be deformed to a strong symplectic filling [OhtaOno2, Lemma 1.1], the last assertion follows.

Let \ell = \ell(p_1) = \ell(p_2). Again, by Lemma 6.1, we can assume that s(C) = \sum M(p_i) + 2\min \ell(p_i) + 1. As above, suppose that there is a filling (W,\omega_W) of (Y_C,\xi_C) and glue it to a standard concave neighborhood of C to obtain the closed manifold (X',\omega'). Blow up \ell times at each of the points p_1 and p_2 to get to the normal crossing divisor resolution at p_1 and p_2, in the blown-up manifold (X,\omega). By assumption, the proper transform of C is a symplectic +1-sphere, and by Theorem 4.1 we can identify it with a line in a blow-up of \mathbb{C}\mathbb{P}^2. However, in X we see the following configuration.

By Lemma 4.3 and 4.5, the vertices e, v, and w must be in the homology classes h-e_1-e_2, e_2-e_3-\dots-e_{\ell+2}, and e_1-e_0, respectively. The class e' can either be in the homology class h-e_1-e_j or in the homology class h-e_2-e_j, for some j>2. In either case, we get a contradiction, because neither of the classes of v' or w' can be in a class e_2-\sum e_k or e_1-\sum e_k, by Lemma 4.5.

This proves the first assertion. The second assertion follows verbatim as in the case of Proposition 6.2.

Since the multiplicity sequence for the unique singularity is [2], following the notation from Section ?, we have M(p)=4 and \ell(p)=2. Therefore, by Proposition 6.2, when s\geq 10, (Y_s,\xi_s) has no symplectic fillings.

For any minimal symplectic fillings W of (Y_s,\xi_s), we can glue W to a concave neighborhood of C to obtain a closed symplectic manifold X such that the pair (X,C) is relatively minimal. Conversely, for any relatively minimal symplectic embedding of C into a closed symplectic manifold X, there is a concave neighborhood of C whose complement is a minimal symplectic filling of X. Therefore, to analyze the symplectic fillings when 5\leq s\leq 9, we classify relatively minimal symplectic embeddings of C into a closed symplectic manifold X. If C embeds in X, then we can blow-up X, once at the singular point p to resolve the singularity of C, and s-5 times at another point q, so that the resulting proper transform \tilde C has self-intersection 1. See Figure ?.

By Theorem 4.1, there exists a symplectomorphism of X\#(s-4)\overline{\mathbb{C}\mathbb{P}}\,\!^2 to \mathbb{C}\mathbb{P}^2\#N\overline{\mathbb{C}\mathbb{P}}\,\!^2 identifying the proper transform \tilde C, a smooth symplectic sphere of self-intersection +1 with \mathbb{C}\mathbb{P}^1\subset\mathbb{C}\mathbb{P}^2. Using the standard basis \{h,e_1,\cdots, e_N\} for H_2(\mathbb{C}\mathbb{P}^2\#N \overline{\mathbb{C}\mathbb{P}}\,\!^2), [C]=h. We determine the possible homology classes for the other components using the lemmas from section ?. By Lemma 4.3, the tangent exceptional sphere E represents 2h-e_1-\cdots - e_5. When s>5, there is a chain of s-5 additional exceptional divisors. The first sphere in the chain is a -1-sphere intersecting C transversally and disjoint from E, and thus represents the class h-e_1-e_2 (up to relabeling). By Lemma 4.6, the remaining spheres in the chain represent (a truncation of) e_2-e_3, e_3-e_4, e_4-e_5 or in the case that s=8 (so there are exactly two -2-spheres in the chain) they can represent e_2-e_3, e_1-e_2. By 4.2, we can blow down exceptional divisors representing e_1,\dots, e_5 which either intersect the configuration positively or are components of the image of the configuration after some blow-downs. After blowing down all exceptional classes, the image of the total transform of C descends to a conic (the image of E) with one tangent line (the image of \tilde C), and, when s>5 another line (the image of the first exceptional sphere in the chain) in \mathbb{C}\mathbb{P}^2.

By [bigGS, Theorem 1.5], such a configuration has a unique symplectic isotopy class in \mathbb{C}\mathbb{P}^2. Therefore there is exactly one relatively minimal symplectic embedding (up to symplectic isotopy) of the total transform of C into X\#{(s-4)}\overline{\mathbb{C}\mathbb{P}}\,\!^2 \cong \mathbb{C}\mathbb{P}^2\#5\overline{\mathbb{C}\mathbb{P}}\,\!^2 for each homological embedding. By deleting a concave regular neighborhood of the total transform of C, we get exactly one minimal symplectic filling of (Y_s,\xi_s) for each homological embedding. When s=5,6,7,9, this shows that there is a unique minimal symplectic filling, corresponding to the unique relatively minimal symplectic embedding of C into \mathbb{C}\mathbb{P}^2\#(5-(s-4))\overline{\mathbb{C}\mathbb{P}}\,\!^2. Since (Y_s,\xi_s) is a rational homology sphere, the Mayer–Vietoris long exact sequence implies the complementary filling has b_2=9-s. By additivity of the signature, it is also negative definite.

When s=8, there are two homological embeddings, (where the final -2-sphere in the chain can either represent e_3-e_4 or e_1-e_2). To see that the two corresponding fillings W_A and W_B are different, we look at their intersection forms. The torsion-free part of the former is generated by the class e_1-e_2-e_3-e_4+2e_5, while that of the latter is generated by the class e_4-e_5. In particular, the intersection forms of W_A and W_B are non-isomorphic, so W_A and W_B are non-diffeomorphic.

We proceed with the same methods as used in the genus-1 case. The resolutions we embed into X\#(s-7)\overline{\mathbb{C}\mathbb{P}}\,\!^2 \cong \mathbb{C}\mathbb{P}^2\#N\overline{\mathbb{C}\mathbb{P}}\,\!^2 are shown in Figures ? and ? for the two types of singularity configurations. In each case, we blow up a sufficient number of times so that the proper transform \tilde C has self-intersection 1, thus identifying \tilde C with \mathbb{C}\mathbb{P}^1\subset \mathbb{C}\mathbb{P}^2 using Theorem 4.1.

In the case of two simple cusps, there are two tangent exceptional divisors of self-intersection -1 representing classes 2h-e_1-\cdots -e_4-e_5 and 2h-e_1-\cdots-e_4-e_6 by Lemma 4.3. By Lemmas 4.3 and 4.6, the exceptional divisors in the chain of length s-9 represent classes h-e_1-e_2, e_2-e_3, e_3-e_4, or h-e_1-e_2, e_2-e_3, e_1-e_2 (when s<12, the sequence is truncated and there is a unique option). Using Lemma 4.2, we blow down exceptional spheres in the e_i classes which either intersect the configuration positively or are contained in the image of the configuration after blow-down. In the first case, the embedding descends to two conics intersecting transversally with one line tangent to both, and another line passing through two of the intersection points of the conics. In the second case, our two conics intersect at one simple tangency and two transverse intersection points, and the last line is tangent to the two conics at the point where they are tangent to each other. In both cases, without the last line, the configuration has a unique symplectic isotopy class by [bigGS, Theorem 1.5], so adding in the last line, the entire configuration has a unique symplectic isotopy class by [bigGS, Proposition 5.1]. Therefore, there is a unique relatively minimal symplectic embedding when 9\leq s\leq 11 and there are two relatively minimal symplectic embeddings when s=12.

When 9\leq s\leq 11, the unique relatively minimal symplectic embeddings correspond to unique minimal symplectic fillings of (Y_s,\xi_s) and the Betti number calculation follows from the Mayer–Vietoris long exact sequence and the fact that Y_s is a rational homology sphere. To distinguish the two fillings when s=12, we can again look at the intersection form. For the homological embedding where the last sphere represents e_3-e_4, the generator of the non-torsion part of the homology of the complement is e_1-e_2-e_3-e_4+2e_5+2e_6 which has self-intersection -12. For the homological embedding where the last sphere represents e_1-e_2, the generator of the non-torsion part of the homology of the complement is e_4-e_5-e_6 which has self-intersection -3.

We observe that it is not possible to find a homological embedding if there are more than two -2-spheres in the chain (i.e. when s>12), since such a (-2)-sphere must represent a class e_i-e_j which intersects e_3-e_4 (or e_1-e_2) once positively, and has intersection zero with all other curves in the configuration. Therefore, there are no symplectic embeddings of C into a closed symplectic manifold when s\geq 13, so (Y_s,\xi_s) has no symplectic fillings when s\geq 13. Note that Proposition 6.2 would only imply this for s\geq 14.

In the case of one cusp of type (2,5), there is one exceptional divisor E of self-intersection -1 tangent to the proper transform \tilde C, and another exceptional divisor E' of self-intersection -2 which intersects E once transversally and is disjoint from all other curves. Up to relabeling, [E]=2h-e_1-\cdots -e_5 and [E']=e_5-e_6. The exceptional divisors in the chain of length s-9 again represent classes h-e_1-e_2, e_2-e_3, e_3-e_4 or h-e_1-e_2, e_2-e_3, e_1-e_2. Blowing down exceptional spheres, this configuration descends to a single conic with two lines, which has a unique symplectic isotopy class by [bigGS, Theorem 1.5]. Thus this yields a unique symplectic filling when 9\leq s\leq 11. The two symplectic fillings corresponding to the two homological embeddings when s=12 can be distinguished by their intersection forms, as their non-torsion homology is generated by e_1-e_2-e_3-e_4+2e_5+2e_6 or e_4-e_5-e_6 respectively as in the previous case. Similarly, we obtain no possible homological embeddings when there are more than two -2-spheres in the chain so when s\geq 13, there are no symplectic fillings of (Y_s,\xi_s).

That there are no fillings when s\geq 17 follows from Proposition 6.2.

For 10\leq s\leq 16, we proceed as in the lower genus cases. Blow up an embedding of C into X once at the cusp to get the minimal smooth resolution so that the exceptional sphere E is tangent with multiplicity 3 to the proper transform of C. We blow-up s-10 additional times at a different point so that \tilde C has self-intersection 1, yielding a chain of s-10 exceptional spheres as in Figure Proof. Identifying \tilde C with \mathbb{C}\mathbb{P}^1 by Theorem 4.1 so [\tilde C]=h, by Lemma 4.3, we see that [E]=3h-2e_1-e_2-\cdots-e_7, and the chain of s-8 exceptional spheres represent the classes in (a truncation of) the sequence h-e_1-e_2, e_2-e_3, e_3-e_4, e_4-e_5, e_5-e_6, e_6-e_7. Blowing down exceptional spheres representing e_1,\cdots, e_7 using Lemma 4.2, E descends to a cubic curve with either a node or a cusp, \tilde C descends to an inflection line, and the first divisor in the chain descends to a line passing through the singular point of the cubic and one other point. Without the last line, this configuration has a unique symplectic isotopy class by [bigGS, Proposition 5.11]. We get a unique symplectic isotopy class for the configuration which includes the last line by [bigGS, Proposition 5.1]. Thus there is a unique relatively minimal symplectic embedding of this blow-up of C into a closed manifold and that manifold must be \mathbb{C}\mathbb{P}^2\#7\overline{\mathbb{C}\mathbb{P}}\,\!^2. We conclude that (Y_s,\xi_s) has a unique minimal symplectic filling with b_2 = 16-s.

Proceeding as in prior cases, we blow up an embedding of C to its minimal resolution, then blow up s-13 additional times at another point until \tilde C has self-intersection 1 and can be identified with \mathbb{C}\mathbb{P}^1. The three exceptional divisors from the minimal resolution must have homology classes as specified by the options shown in Figures Proof and ? (depending on whether there are 1 or 2 cusps). Note that in both cases, there are two options for the homology class of the exceptional divisor E'': e_6-e_7 or e_4-e_5. Also for both cases, the homology classes of the divisors in the extra chain must represent either h-e_1-e_2,e_2-e_3,e_3-e_4 or h-e_1-e_2,e_2-e_3,e_1-e_2. If [E'']=e_4-e_5, only the second option is possible. When [E'']=e_6-e_7, both options are possible, but note that they only differ when all three components are included corresponding to the s=16 case. Thus, we get three cases when s=16 and two cases when 13\leq s\leq 15. Note, there are no allowed homological embeddings if the chain is longer, so (Y_s,\xi_s) has no fillings when s\geq 17. Each configuration descends under blow-downs to a configuration in \mathbb{C}\mathbb{P}^2 with a unique symplectic isotopy class by [bigGS, Theorem 1.5 and Proposition 5.1], so each homological embedding corresponds to a unique minimal filling. If [E'']=e_6-e_7, we have a relatively minimal symplectic embedding into \mathbb{C}\mathbb{P}^2\#7\overline{\mathbb{C}\mathbb{P}}\,\!^2. If [E'']=e_4-e_5 the relatively minimal embedding is into \mathbb{C}\mathbb{P}^2\#6\overline{\mathbb{C}\mathbb{P}}\,\!^2. Symplectic fillings complementary an embedding in \mathbb{C}\mathbb{P}^2\#7\overline{\mathbb{C}\mathbb{P}}\,\!^2 will have b_2=17-s, and those which are complementary to an embedding in \mathbb{C}\mathbb{P}^2\#6\overline{\mathbb{C}\mathbb{P}}\,\!^2 will have b_2=16-s. Thus, when 13\leq s\leq 15, the two fillings have different b_2 so they are clearly not diffeomorphic. When s=16, there are two potential fillings with b_2=17-s. To distinguish these, we look at the square of the generator for the homology of the complement. In one case the generator is e_1-e_2-e_3-e_4+2e_5+2e_6+2e_7 of square -16 and in the other case the generator is e_4-e_5-e_6-e_7 of square -4.

The determination of homological embeddings has a similar classification to the previous case. What is different in this situation is that not all of these homological embeddings correspond to actual geometric embeddings, and we need to prove a symplectic isotopy result for the one which does. The possible homological embeddings are shown in Figure ?, where the additional chain of s-13 exceptional divisors represent classes h-e_1-e_2, e_2-e_3, and either e_3-e_4 or e_1-e_2. Note that when [E'']=2h-e_1-e_2-e_3-e_5-e_6, and s=16, the last divisor in the chain must represent e_1-e_2 to be disjoint from E''. Therefore we have three homological embeddings.

For the two homological embeddings where [E'']=2h-e_1-\cdots-e_4-e_7, if we suppose this can be realized symplectically in \mathbb{C}\mathbb{P}^2\#7\overline{\mathbb{C}\mathbb{P}}\,\!^2, then we can blow down exceptional spheres so that the image in \mathbb{C}\mathbb{P}^2 of the resolution becomes three symplectic conics in a pencil (all three intersecting at 4 points transversally or all three intersecting at one point tangentially and two points transversally) with a symplectic line tangent to all three conics. This configuration is obstructed by [bigGS, Proposition 5.23] (note that the statement refers to a pencil with four transverse intersections, but the same proof applies to the case where the pencil has one base point of multiplicity two–in the birational derivation blow up once at each base point in the pencil). Therefore these homological embeddings cannot be symplectically realized, so there are no corresponding symplectic fillings.

For the last remaining homological embedding where [E'']=2h-e_1-e_2-e_3-e_5-e_6 and the chain represents a (truncation of) h-e_1-e_2, e_2-e_3, e_1-e_2, for any symplectic realization of this embedding into \mathbb{C}\mathbb{P}^2\# 6\overline{\mathbb{C}\mathbb{P}}\,\!^2, there will exist exceptional spheres so that the result blows down to a configuration in \mathbb{C}\mathbb{P}^2 of three symplectic conics intersecting at one common triple tangency (the image of e_1 and e_2), one transverse triple point (the image of e_3) and a transverse double point for each of the three pairs (the images of e_4,e_5,e_6), together with one symplectic line tangent at the common tangency of the conics and another a symplectic line tangent to all three conics at non-intersection points. To verify this configuration has a unique symplectic isotopy class in \mathbb{C}\mathbb{P}^2, we add a symplectic line through the two points which pass through all three conics–this does not change the symplectic isotopy classification by [bigGS, Proposition 5.1]. Then we find a birational equivalence as in Figure ? by (1) blowing up twice at the tangency of the conics and once at the triple intersection of the conics, then (2) blowing down the proper transforms of the two lines which passed through intersection points of the conics (which became -1-spheres after step (1)), and (3) blowing down the remaining -1-sphere which is the image under the blow-downs of step (2) of an exceptional -2-sphere introduced in step (1). The resulting configuration contains a single conic (the image of the line which had three tangencies to the original conics), and five lines—four of which are tangent to the conic (the images of the three original conics and the image of one of the exceptional divisors), and one of which passes transversally through the tangency point of the fourth line with the conic (and otherwise has generic transverse double point intersections). This configuration has a unique symplectic isotopy class because it can be built from a single conic by iteratively applying [bigGS, Proposition 5.1]. Therefore, there is a unique minimal symplectic filling corresponding to this homological embedding.