The purely cosmetic surgery conjecture

The explanation

Take a knot, a closed loop tied in ordinary three-dimensional space. Cut out a thin tube around it. Then glue a solid tube back in a new way. This is Dehn surgery. The gluing is fixed by a fraction p/q called the slope. The result is a closed three-dimensional space, a three-manifold. Its first homology group is a simple count of loops. For slope p/q with p not zero, it is a cycle of |p| elementsSource 1. Slope 2 gives a cycle of 2 elements and slope 3 gives one of 3. So surgeries at slopes 2 and 3 give different spaces. But slopes +2 and −2 both give a cycle of 2 elements. This count cannot tell them apart.

The knot must be nontrivial, meaning it cannot be untied. For the unknot, every slope 1/q gives back ordinary space. Two different slopes giving the same oriented space are called purely cosmetic. Cameron Gordon posed the conjecture that this never happens for a nontrivial knot. Gordon and Luecke showed that nontrivial surgery on a nontrivial knot never gives back ordinary spaceSource 2. That handled the slope at infinity, called the meridian.

Many tools then cut down the remaining cases. Heegaard Floer homology is an algebraic invariant built from curves on surfaces. Using it, Hanselman reduced the possible pairs to ±2 or ±1/q. He showed the ±2 case needs a knot of genus twoSource 3. The genus is the fewest handles on a surface that the knot bounds. Futer, Purcell and Schleimer settled all knots up to 19 crossings by computerSource 4. Daemi, Lidman and Miller Eismeier excluded the pairs ±1/qSource 5. That left one case: slopes −2 and 2 on a genus-two knotSource 6.

The obstacle is that ordinary invariants agree on this last pair. The paper claims to rule it out for every nontrivial knotSource 7. It assumes a map that keeps orientation from the −2 surgery to the +2 surgery. It does not assume the map respects the knotSource 8. It then builds two counts of solutions to gauge-theory equations. These equations describe connections, which are rules for moving data around a space. The first count is a nonzero whole numberSource 9. The second count comes from a one-dimensional family of solutions. Its boundary forces a multiple of the first count to be zero. That contradiction ends the proofSource 10. Maps that reverse orientation are a separate questionSource 11.

How sure is this? There is no Lean check. A computer proof checker has not looked at any part of it. The paper is long and relies on heavy analysis of these equations. Each step depends on careful bounds and gluing estimates. No specialist review of the paper has been published.

What it doesn't show. This does not cover orientation-reversing coincidences, called chirally cosmetic surgeries, or knots in spaces other than the three-sphere.