A nontrivial knot remembers its surgery slope

← All digests

Family 306 · Topology
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-22

1. The question

Remove a tubular neighborhood of a knot from the three-sphere. The remaining space has a torus boundary. Glue back a solid torus in a different way, and the result is a new closed three-manifold. This operation is Dehn surgery.

Could two genuinely different gluing slopes produce the same oriented three-manifold? If so, the surgery would be purely cosmetic: the changed slope would leave no detectable change in the oriented homeomorphism type.

The manuscript claims that this never happens for a nontrivial smooth knot in $S^3$. Earlier results had reduced the possible exception to surgery slopes $-2$ and $2$ on a genus-two knot. The new argument targets that remaining case using two compatible gauge-theoretic counts, one forced to be nonzero and the other forced to have total boundary count zero.

2. Background and essential definitions

Slopes on a boundary torus

A smooth knot is an embedded circle $K\subset S^3$. Its exterior is

$$E_K=S^3\setminus\operatorname{int}(\nu K),$$

where $\nu K$ is a tubular neighborhood, homeomorphic to a solid torus.

The boundary torus has a meridian $\mu$, going once around the knot, and a preferred longitude $\lambda$, running parallel to the knot with linking number zero. A slope is an unoriented primitive curve class

$$p\mu+q\lambda,\qquad \gcd(p,q)=1,$$

considered up to changing both signs. It is denoted $p/q$, with $q=0$ written $\infty$. Dehn filling attaches a solid torus so that its meridian becomes this curve. Write the result as $S^3_{p/q}(K)$.

The meridional filling $\infty$ restores $S^3$. Integer slopes have $q=1$, so the remaining candidates $\pm2$ represent $\pm2\mu+\lambda$.

What “the same” means

Purely cosmetic surgery requires an orientation-preserving homeomorphism between the two filled manifolds. A homeomorphism is a continuous bijection with continuous inverse; it need not respect the decomposition into a knot exterior and a solid torus.

In particular, the putative homeomorphism need not carry one surgery core to the other. Excluding only maps that preserve those cores or the exterior would be a weaker statement.

Orientation-reversing coincidences are called chirally cosmetic and form a different problem. The orientation condition in this theorem cannot be dropped.

Genus and basic invariants

A Seifert surface is a compact connected oriented surface in $S^3$ with boundary $K$. The knot genus $g(K)$ is the smallest genus of such a surface. Genus two means that a minimal surface has two handles.

First homology is an elementary invariant of a filled manifold:

$$H_1(S^3_{p/q}(K);\mathbb Z)\cong \begin{cases} \mathbb Z/|p|\mathbb Z,&p\ne0,\\ \mathbb Z,&p=0. \end{cases}$$

It immediately distinguishes zero surgery from nonzero finite slopes, but it cannot separate $-2$ from $2$: both have first homology of order two. The unresolved case therefore needs finer information.

3. Development history and previous partial results

Cosmetic filling questions were formulated in work of Gordon and in Kirby’s problem list, with Bleiler, Hodgson and Weeks developing the terminology and distinctions between different kinds of coincidence.

Gordon and Luecke proved that nontrivial surgery on a nontrivial knot cannot give $S^3$. This settles pairs involving the meridional slope. Boyer and Lines obtained obstructions from classical three-manifold invariants and the Alexander polynomial.

Heegaard Floer theory supplied stronger restrictions. The rational surgery formula of Ozsváth and Szabó, Wang’s genus-one result, and Wu’s work on same-sign surgeries progressively narrowed the possibilities. Ni and Wu showed that a purely cosmetic finite pair must have opposite slopes.

Hanselman used graded Floer information expressed through immersed curves to restrict possible pairs to $\pm2$ or $\pm1/q$, with the $\pm2$ case requiring genus two. Daemi, Lidman and Miller Eismeier then used filtered instanton homology to exclude the $\pm1/q$ cases. Their result, together with the earlier work, leaves precisely the genus-two $\pm2$ possibility, with further polynomial restrictions that the new paper does not need for its reduction. Filtered instanton result

Other approaches settle many classes or finite collections. Polynomial and quantum invariants give additional obstructions; results cover nontrivial cable, composite and pretzel knots. An equivariant theorem for strongly invertible knots requires the comparison map to commute with specified involutions, a restriction absent from the present claim.

Futer, Purcell and Schleimer combined invariants, hyperbolic estimates and verified computations to settle the conjecture through 19 crossings and in a one-cusped census. Their procedure does not guarantee a decision for every possible pair. These computational and special-class results provide substantial evidence and coverage, while the manuscript claims a uniform final argument for the remaining case.

4. What the new manuscripts claim

The September 23, 2026 manuscript states that for every smooth nontrivial knot $K\subset S^3$ and every pair of slopes $r,s\in\mathbb Q\cup\{\infty\}$,

$$S^3_r(K)\cong_+ S^3_s(K) \quad\Longrightarrow\quad r=s.$$

The symbol $\cong_+$ denotes an orientation-preserving homeomorphism of the entire filled manifolds. No condition on surgery cores, peripheral curves, or exterior preservation is imposed.

The proof uses the earlier reduction rather than rederiving every slope restriction. It assumes a genus-two knot and an orientation-preserving identification between its $-2$ and $2$ surgeries, then constructs a contradiction. Three-dimensional smoothing results allow that identification to be replaced by an orientation-preserving diffeomorphism for the gauge-theoretic construction.

The manuscript’s analytic work concerns specially constructed four-dimensional cobordisms and parameter families. It does not claim a general new spinor Floer theory on every knot-surgery manifold as a prerequisite or consequence.

5. Why the advance matters

Importance: 5/5, provisional. If correct, the theorem resolves the purely cosmetic surgery conjecture for all nontrivial knots in the three-sphere.

Dehn surgery is one of the main ways to construct closed three-manifolds. The claim says that for a fixed nontrivial knot, the oriented manifold remembers which slope was used, even when a comparison homeomorphism is free to disregard the visible surgery description.

The result also completes a long sequence of increasingly sharp restrictions. Its novelty is concentrated in the last possible configuration left by earlier Floer and instanton results. That concentration does not make the endpoint narrow: eliminating that case gives the full all-slopes theorem.

The two-count method is a further feature of the manuscript. It tries to arrange a nonzero ordinary instanton count and a one-dimensional spinor family within one compatible set of geometric and analytic choices, so that their boundary contributions can be compared directly.

6. The main idea

Count solutions on four-dimensional cobordisms

A cobordism is a four-dimensional space whose boundary contains the three-manifolds being compared. In gauge theory, one studies connections on bundles over such spaces, modulo changes of bundle coordinates called gauge transformations.

Instantons satisfy a curvature equation. Their solution spaces can be organized into Floer chain complexes at the boundary and maps associated with cobordisms. Suitable zero-dimensional solution spaces have signed counts; these counts can encode information invisible to ordinary homology.

Assuming the cosmetic identification, the paper assembles surgery cobordisms into a repeated closed arrangement. Ordinary surgery families provide maps that behave as units modulo two. The negative direction is more delicate: it uses an interval of metrics, a degree-two sphere test and a signed sum over two determinant lifts. The signs cancel two unwanted minimal lens-space contributions.

Symplectic caps and established Donaldson nonvanishing results provide a nonzero cap pairing. The paper moves the probing conditions into fixed final caps, away from the variable blocks. Repeating the construction on the free part of an integral Floer lattice, with congruences modulo a power of two, yields a nonzero integer count

$$\Omega\ne0.$$

This is a count with signs and compatible determinant data, not just a claim that some solution exists.

Enlarge the equations by adding spinors

A spinor is an auxiliary field coupled to the connection through a Dirac operator. The manuscript repeats the arrangement so that the complex index $n_D$ of that operator is positive while preserving the nonzero ordinary count.

Multiplication of a spinor by a complex phase gives a circle symmetry. Impose $\eta=n_D-1$ complex phase conditions and divide by the effective circle action. The expected real dimension becomes

$$2n_D-2\eta-1=1.$$

The desired output is therefore a compact oriented one-dimensional solution space, after handling its allowed boundary phenomena. An ordinary compact oriented one-manifold has total signed boundary zero: every interval contributes one positive and one negative endpoint. The gauge-theoretic argument aims to use the same conservation principle with the appropriate regularized counts.

Near an ordinary instanton, the phase directions form a complex projective space $\mathbb{CP}^{n_D-1}$. Each phase condition represents twice its hyperplane class. The resulting instanton-link boundary contribution is consequently

$$2^\eta\Omega$$

up to one common orientation sign. Since $\Omega$ is a nonzero integer, this contribution cannot vanish over $\mathbb Q$.

Account for every other possible end

The hard part is ensuring that other ends do not cancel this nonzero contribution. Connections may become reducible, meaning that they preserve a splitting into line bundles. Solutions may also break along long regions or concentrate curvature into bubbles.

The paper constructs simultaneous families of metrics using toric geometric pieces, together with a long spherical tube. Period and curvature estimates constrain possible line-bundle reductions. Index inequalities then exclude unwanted fixed and mixed degenerations, including parameter-family corners.

The analytic choices are made compatibly with the already established ordinary count. Energy bounds and compactness come before the regularization and dimension exclusions; the argument does not simply invoke a generic perturbation and assume that all boundary phenomena disappear.

The remaining ends involve one minimal cap separating along a lens-space boundary, without bubbles. Cylindrical gluing identifies their local neighborhoods. Their two determinant sectors cancel with the same orientation convention used in constructing the negative ordinary unit.

With those lens contributions cancelled and the other ends excluded, the total boundary relation would read

$$0=\pm2^\eta\Omega,$$

contradicting the nonzero ordinary count. This excludes the genus-two $\pm2$ pair, and the earlier reductions supply the full theorem.

7. Scope, limitations, and open questions

The knot is smooth, nontrivial, and lies in $S^3$. The theorem is not a blanket assertion about arbitrary torus-boundary manifolds, links, or knots in other ambient three-manifolds.

The conclusion is explicitly orientation-preserving. Chirally cosmetic surgeries remain a different phenomenon and are not excluded by this statement.

The comparison map is unrestricted with respect to the surgery decomposition. This is stronger than an equivariant or core-preserving theorem, and it is why the ordinary manifold invariants and four-dimensional constructions are needed.

The proof’s intermediate necks have different roles. Some are long but finite surgery seams; the actual infinite boundary faces are specified sphere and lens-space cuts. The manuscript does not require general gluing at positive bubble levels or a general regular spinor Floer theory for the knot surgeries.

Finally, the theorem supplies a qualitative uniqueness result. It does not give a practical algorithm reconstructing the slope from a presentation of the filled manifold, an efficient complexity bound, or a classification of orientation-reversing coincidences.

8. Sources and further reading

Daemi–Lidman–Miller Eismeier’s filtered instanton paper gives the decisive prior reduction used here. The manuscript also cites Ni–Wu’s Cosmetic surgeries on knots in S3 and Hanselman’s Heegaard Floer homology and cosmetic surgeries in S3 for the preceding slope restrictions.

For the gauge-theoretic context, its references include Donaldson’s Floer Homology Groups in Yang–Mills Theory, Kronheimer–Mrowka’s Witten’s conjecture and Property P, and the nonabelian-monopole work of Feehan and Leness. The source below provides the particular compatible counts, metric families and boundary analysis summarized in this digest.

Source manuscripts

• M1. OpenAI, Purely cosmetic surgery on knots in the three-sphere (September 23, 2026).

Sources consulted on 22 October 2026 at repository revision 1581d4c. Download the manuscript citations.