An infinite family of strongly invertible L-space knots without Khovanov thin surgery

By Lemma 3.2, \varkappa^h(K_n,\Phi)\ncong 0 for h=-4 and 2n+4. In fact, \varkappa^{-4} has rank \geq 1, and it has a generator having \delta-grading q-2h=-8n-7, where q and h are normalized at N=0.

Similarly, \varkappa^{2n+4} has rank \ge 1 having a generator with \delta=-8n-9. Thus \varkappa(K_n,\Phi) is supported in at least two \delta-gradings.

It was already shown in [BKM] that K_1 has no thin slope. So in the following, we can assume that n>1. As explained at the beginning of this section, every thin slope of K_n has to be exceptional or symmetry-exceptional.

By Lemma 7.2, the only exceptional slope of K_n is 8n+6, yielding the small Seifert fibered space M \left(-1; \dfrac{1}{2} , \dfrac{2n+1}{4n+4} , \dfrac{2}{4n+5} \right). In Lemma 2.2 of [BK], it is shown that this Seifert fibered space is an L-space. Thus, Lemma 4.1 from [BKM] implies that we can write this Seifert fibered space uniquely as a double branched cover. This implies that the only possible branching set has to be the tangle filling T_n(8n+6), which was already shown above to be thick.

By Lemma 7.3, the only symmetry-exceptional slopes of K_n are 8n+6\pm2, which both yield manifolds with symmetry group (\mathbb Z_2)^2. Thus Lemma 2.4 of [BKM2] shows that we can write these manifolds in exactly three ways as double branched cover over links in 3-manifolds (possibly different from S^3). One of these is the tangle filling T_n(8n+6\pm2), which is already known to be thick.

However, the others have quotient manifolds different from S^3 as shown in Lemmas 7.4 and 7.5 below. Thus, we confirmed that K_n(8n+6\pm 2) has the unique description as the double branched cover of S^3.

In [BK] it was shown that the knots K_n are L-space knots, by identifying an exceptional surgery to an L-space.

From the surgery description along the link L in Figure ?, it follows directly that each K_n is strongly invertible. Using SnapPy [SnapPy], we show that this is the only symmetry of K_n.

First, we establish that L (which is called L12n1739 in the HTW census [HTW]) is hyperbolic and possesses a unique strong inversion. For that, we use SnapPy [SnapPy] to compute the symmetry group of orientation-preserving diffeomorphisms for the link L, and find that there exists exactly one non-trivial symmetry that extends to the link itself, which has to be the symmetry visible in Figure ?. This unique strong inversion inherently survives the Dehn filling that yields the knot K_n.

By Thurston’s hyperbolic Dehn filling theorem [Thurston], we know that any filling of L with all slopes sufficiently large is again hyperbolic and has the same symmetry group as L, and recent work of Futer–Purcell–Schleimer [FPS] gives a computable bound for that, cf. [BKM2]. We compute this bound and check that for n\geq9 the slopes are sufficiently large, so that we conclude that K_n has a unique strong inversion if n\geq9. For the finitely many small values of n, we build the knot K_n in SnapPy, verify that it is hyperbolic and that its symmetry group also has a unique strong inversion.

For the code and computations, we refer to [data].

This follows from Lemmas 4.4, 5.1 and 6.1 below.

First, we prove (3.1). Suppose N\ge 8n+7. Then the N-tangle and the box with the integer -(8n+6) are merged into an integer tangle corresponding to N-(8n+6)\ge 1. Let us call this diagram D. Then n_-(D)=n+7, where n_- denotes the number of negative crossings. Choose the leftmost crossing there, which is positive. Let D_- and D_\infty be the resulting diagrams after resolving the crossing horizontally or vertically. Then D_- corresponds to T_n(N-1), and D_\infty represents the unknot. In addition, n_-(D_\infty)=n+5+(N-8n-7). Thus c=n_-(D_\infty)-n_-(D)=N-8n-9, so Viro’s long exact sequence for a positive crossing (see for example [Wa0]) yields

For us, (h,q)=(-4, N-8n-15). Then \mathrm{Kh}_{q-3c-2}^{h-c-1}(D_\infty)\cong0, since h-c-1=0 and q-3c-2=0 cannot hold simultaneously. Also, \mathrm{Kh}_{q-3c-2}^{h-c}(D_\infty)\ncong 0 only when N=8n+5. Thus \mathrm{Kh}_q^h(D)\cong \mathrm{Kh}_{q-1}^h(D_-) under the assumption N\ge 8n+7. By Lemma 3.1, we have \mathrm{Kh}_{-9}^{-4}(T_n(8n+6))\cong\mathbb{Z}_2. Hence

Suppose N\le 8n+5. In this case, the merged tangle of the N-tangle and the box containing 8n+6 left half twists is an integer tangle corresponding to N-(8n+6)\le -1, which contains only negative crossings. Then n_-(D)=9n+13-N. As above, we choose the leftmost crossing there. Since n_-(D_\infty)=n+5, c=n_-(D_\infty)-n_-(D)=N-8n-8. Hence, the long exact sequence for a negative crossing yields

Recall (h,q)=(-4,N-8n-15). Then \mathrm{Kh}_{q-3c-1}^{h-c-1}(D_\infty)\cong 0, and \mathrm{Kh}^{h-c}_{q-3c-1}(D_\infty)\ncong 0 only when N=8n+4. This implies that

and

Finally, by using Lemma 3.1,

Thus

which implies (3.1).

The argument for (3.2) is similar to the argument for (3.1). Set (h,q)=(2n+4,N-4n-1). Suppose N\ge 8n+7. The long exact sequence (Proof) yields

Since \mathrm{Kh}_{20n+24-2N}^{10n+12-N}(D_\infty)\cong\mathrm{Kh}_{20n+24-2N}^{10n+13-N}(D_\infty)\cong0, if N\ne 10n+12, we have

and

When N=10n+12, (Proof) reduces to

by Lemma 3.1. Thus \mathrm{Kh}_{6n+11}^{2n+4}(T_n(10n+12))\cong\mathbb{Z}_2^2.

Finally, assume N\le 8n+5. As in (3.1), the long exact sequence (?) yields

Since \mathrm{Kh}_{20n+22-2N}^{10n+11-N}(D_\infty)\cong\mathrm{Kh}_{20n+22-2N}^{10n+12-N}(D_\infty)\cong0, we have

This completes the proof.

This easily follows from a direct calculation in the diagram D^i_j.

We perform this calculation using KnotJob [S].

We perform the calculation with KnotJob [S].

By resolving the topmost crossing in the upper box of D^{-(2n+3)}_{n+1}, we have the long exact sequence

First, set (h,q)=(-4,-9). By Lemma 4.2, \mathrm{Kh}^{-5}_{-10}(H_-)\cong0 and \mathrm{Kh}^{-4}_{-10}(H_-)\cong\mathbb{Z}_2. Since c=-4 by Lemma 4.1, we have

Hence, it suffices to show that \mathrm{Kh}^{-1}_{1}(D^{-(2n+2)}_{n+1})\cong\mathrm{Kh}_{1}^{0}(D^{-(2n+2)}_{n+1})\cong0. For the diagram D^{-(2n+2)}_{n+1}, we apply the resolution at the topmost crossing as above. Note that c=4 by Lemma 4.1. Then the long exact sequence reduces to

and

Thus, we need to know that \mathrm{Kh}^{-6}_{-13}(D^{-(2n+1)}_{n+1})\cong\mathrm{Kh}^{-5}_{-13}(D^{-(2n+1)}_{n+1})\cong0. By resolving the topmost crossing in the upper box of D^{-(2n+1)}_{n+1}, we obtain

where h\in \{-6,-5\}. Thus \mathrm{Kh}^{h+3}_{-3}(D^{-2n}_{n+1})\cong \mathrm{Kh}^{h}_{-13}(D^{-(2n+1)}_{n+1}). Then \mathrm{Kh}^{-3}_{-3}(D^{-2n}_{n+1})\cong\mathrm{Kh}^{-2}_{-3}(D^{-2n}_{n+1})\cong0 implies \mathrm{Kh}^{-1}_{1}(D^{-(2n+2)}_{n+1})\cong\mathrm{Kh}_{1}^{0}(D^{-(2n+2)}_{n+1})\cong0. Inductively, it is enough to see that

For D^0_{n+1}, we resolve the topmost negative crossing in the lower box. Then

for h \in \{-2n-3,-2n-2\}. The sign of F_\pm depends on the parity of n. If n is even (resp. odd), then the resulting diagram is F_+ (resp. F_-). By Lemma 4.3, \mathrm{Kh}_{-4n-2}^h(F_\pm) \cong\mathrm{Kh}_{-4n-2}^{h+1}(F_\pm)\cong0. Hence, \mathrm{Kh}^h_{-4n-3}(D^0_{n+1}) \cong \mathrm{Kh}^{h+1}_{-4n-1}(D^0_n) for h\in \{-2n-3,-2n-2\}. Inductively, it is enough to see that

However, D_0^0 is the Hopf link with linking number one. It has \mathrm{Kh}^h_q(D^0_0)\cong\mathbb{Z}_2 for (h,q)=(0,1), (2,5), and 0 otherwise. This finishes the computation of \mathrm{Kh}^{-4}_{-9}(D^{-(2n+3)}_{n+1})

Next, we set (h,q)=(2n+4,4n+5). Then (Proof) reduces to

By Lemma 4.2, \mathrm{Kh}^{2n+3}_{4n+4}(H_-)\cong\mathrm{Kh}^{2n+4}_{4n+4}(H_-)\cong0. Thus

For D_{n+1}^{-(2n+2)}, resolve the topmost crossing in the upper box as before. Then

By Lemma 4.2 again, \mathrm{Kh}^{2n+6}_{4n+14}(H_+)\cong \mathrm{Kh}^{2n+7}_{4n+14}(H_+)\cong0. So, we have

Inductively, we need to know that \mathrm{Kh}^{2}_1(D^{-1}_{n+1})\cong\mathbb{Z}_2. Resolving the remaining crossing in the upper twist, we have

For D^0_{n+1}, we resolve the topmost crossing in the lower box as above. Then

We calculate \mathrm{Kh}^h_{13}(D^0_{n}) for h\in \{5,6\}. By resolving the topmost crossing in the lower box,

Thus \mathrm{Kh}^h_{13}(D^0_n) \cong \mathrm{Kh}^{h+1}_{15}(D^0_{n-1}). Inductively, we have

for h\in \{5,6\}. As mentioned above, \mathrm{Kh}^{h+n}_{13+2n}(D^0_0)\cong0. Thus (?) reduces to

so \mathrm{Kh}^5_{11}(D^0_{n+1})\cong\mathbb{Z}_2 as desired.

Resolve the negative crossing marked with \ast in Figure Proof(a). Then the long exact sequence gives

where O denotes the unknot.

Hence we need to know that \mathrm{Kh}^{-3}_{-9}(T_n(8n+6))\cong0. With the notation from the previous section, this is \mathrm{Kh}^{-3}_{-9}(D^{-(2n+3)}_{n+1}). In (Proof), set (h,q)=(-3,-9). By Lemma 4.2, we have

In the proof of Lemma 4.4, we showed that \mathrm{Kh}^0_{1}(D^{-(2n+2)}_{n+1})\cong0. Thus we have \mathrm{Kh}^{-3}_{-9}(D^{-(2n+3)}_{n+1})\cong0.

Resolve the leftmost crossing in the box labeled with 2n+5. Since c=2n+2, we have the long exact sequence

where O denotes the unknot again. Thus \mathrm{Kh}^{2n+3}_{6n+10}(T_n(10n+11)) \cong \mathrm{Kh}^{2n+3}_{6n+9}(T_n(10n+10)). Inductively, it is enough to show that \mathrm{Kh}^{2n+3}_{4n+5} (T_n(8n+6))\cong0. Recall that T_n(8n+6) is represented by the diagram D^{-(2n+3)}_{n+1}.

Set (h,q)=(2n+3,4n+5) for (Proof). Then

Thus \mathrm{Kh}^{2n+6}_{4n+15}(D^{-(2n+2)}_{n+1}) \cong \mathrm{Kh}^{2n+3}_{4n+5}(D^{-(2n+3)}_{n+1}). For D_{n+1}^{-(2n+2)}, we have

as in the proof of Lemma 4.4. Hence \mathrm{Kh}^{2n+1}_{4n+1}(D^{-(2n+1)}_{n+1})\cong \mathrm{Kh}^{2n+6}_{4n+15}(D^{-(2n+2)}_{n+1}). Inductively, it is enough to show that \mathrm{Kh}^{-1}_{1}(D^{-1}_{n+1})\cong0. As in (?),

For D^0_{n+1}, as in (?), we have

Thus, it suffices to show that \mathrm{Kh}^3_{13}(D^0_{n})\cong0. As in (?),

Hence \mathrm{Kh}^{3}_{13}(D^0_{n}) \cong \mathrm{Kh}^{4}_{15}(D^0_{n-1}). Inductively, this is isomorphic to \mathrm{Kh}^{n+3}_{2n+13}(D^0_0). Recall that D^0_0 is the Hopf link with linking number one. It has \mathrm{Kh}^h_q(D^0_0)\cong\mathbb{Z}_2 for (h,q)=(0,1), (2,5), and 0 otherwise. Hence \mathrm{Kh}^{n+3}_{2n+13}(D^0_0)\cong0.

The proof follows the argument of Watson [W, Proposition 5.2]. The induction on the continued fraction expansion of r uses only the skein long exact sequence and the existence of non-trivial Khovanov homology in at least two \delta-gradings for the integer closures T(n) and T(n+1). Consequently, the same argument shows that every rational closure T(r) has Khovanov homology supported in at least two \delta-gradings, and hence is Khovanov thick.

In [BKM] it was already shown that 14 is the only exceptional slope of K_1, yielding the claimed Seifert fibered space. Thus, we can assume that n>1. We consider again the surgery description

from Figure ?, where the * denotes that the second cusp of L is left unfilled. Any exceptional filling r of K_n corresponds to the exceptional filling

of L. The 6-theorem of Agol [Agol] and Lackenby [Lackenby] says that if all slopes of L are longer than 6, the filled manifold is again hyperbolic and thus the filling is not exceptional.

Using SnapPy [SnapPy], we compute the short slopes on the first and third cusps. The slopes 1/(2n+3) and -1/(n+1) are not short for every n\geq 2. Thus, any exceptional filling of K_n must arise from a short slope on the second cusp of L. The set of short slopes on this cusp is \{\infty,-1,0,1,2,3,4,5\}. The slope \infty corresponds to the meridional filling and yields S^3. And it was already shown in Lemma 2.2 of [BK] that the slope (2,1) of L, which gets mapped to the slope 8n+6 of K_n is exceptional, yielding the Seifert fibered space M \left(-1; \dfrac{1}{2} , \dfrac{2n+1}{4n+4} , \dfrac{2}{4n+5} \right).

For each of the other six short fillings s\in \{-1,0,1,3,4,5\} we use SnapPy to verify that L(*,s,*) is hyperbolic. Since all other fillings of the remaining cusps yielding K_n are sufficiently long, this implies that all other fillings of K_n are hyperbolic.

We use the same strategy as in the proof of Lemma 7.2. But instead of the 6-theorem, we use the following. Let M be a cusped hyperbolic manifold. Then, a recent result of Futer–Purcell–Schleimer [FPS] implies that there exists a computable bound B(M) such that every filling of M with slopes of normalized lengths longer than B(M) has the same symmetry group as M, cf. [BKM2].

First, we use SnapPy to compute the bound B(L), where L is the link of the surgery description of the K_n. Then we check that the slopes 1/(2n+3) and -1/(n+1) on the first and third cusps of L have normalized length longer than B(L) for all n\geq 9. Hence, any symmetry-exceptional slope of K_n for n\geq9 must arise from a slope of normalized length less than B(L) on the middle cusp of L.

The slopes on this cusp, whose normalized lengths are less than the Futer–Purcell–Schleimer bound B(L), form a finite set. For all slopes s in this set except

we compute that L(*,s,*) has symmetry group isomorphic to \mathbb{Z}_2. Therefore, the only slopes that can give symmetry-exceptional fillings of K_n are the images of \{\infty,0,2,4\}.

The slope \infty gives the meridional filling of K_n and hence the manifold S^3, while the slope 2 is exceptional by the previous lemma and therefore does not contribute a symmetry-exceptional hyperbolic filling. It remains to consider the slopes 4 and 0. With SnapPy, we compute that the manifolds

both have symmetry group isomorphic to (\mathbb{Z}_2)^2. Since the normalized lengths of the slopes on the first and last cusp of L yielding K_n for n\geq9 are larger than B(L), it follows that the images of 4 and 0 are symmetry-exceptional slopes for K_n yielding manifolds with symmetry group (\mathbb{Z}_2)^2.

Finally, for n<9, we build the knots K_n, compute the Futer–Purcell–Schleimer bound B(K_n), and compute the symmetry groups of all fillings with length less than B(K_n). This yields that also for 2\le n< 9 the only symmetry exceptional slopes are 8n+4 and 8n+8.

For the surgery diagram as shown in Figure ?, it is straightforward to verify that the quotient of this diagram under the involution around the horizontal axis A_h shown there yields T_n(8n+4) as its branching set. Hence, this surgery diagram gives the manifold K_n(8n+4).

On the other hand, this diagram has other symmetries. If we take the quotient under the involution around the vertical axis A_v, then the ambient manifold is seen to be the lens space -L(2,1). See Figure ?(a), where we ignore the branching set, given by the surgery curves that intersect the symmetry axis.

Another symmetry is the involution around the axis perpendicular to the projection plane. Then Figure ?(b) shows that the ambient manifold after taking the quotient is the lens space -L(2n+1,n+1).

The argument is similar to that of Lemma 7.4. Figure ? shows the surgery diagram yielding K_n(8n+8). This is confirmed by the fact that taking the quotient under the involution around the axis A_h shown there yields the branching set T_n(8n+8).

As before, this surgery diagram has other symmetries. If we take the quotient around the axis perpendicular to the projection plane, the ambient manifold is L(2,1), as shown in Figure ?(a). Finally, taking the quotient around the vertical axis A_v yields L(4n+4,2n+1) as shown in Figure ?(b).