← Back to arXiv
arXivAnalysis / PDEsarXiv:2608.17081

Counterexamples to Magnanini's conjecture concerning critical points of the torsion function

The torsion function is a mathematical object defined on a region (domain) in the plane by solving a specific differential equation. Physically, it describes how a rod with that cross-sectional shape resists twisting. A natural question is how the shape of the domain controls the behavior of this function, particularly where its peaks (local maxima) occur. A related and more elementary geometric object is the distance function, which at any interior point simply measures how far that point is from the boundary of the domain. Mathematicians had observed that these two functions often behave similarly, and Magnanini conjectured a precise relationship: the number of peaks of the torsion function should never exceed the number of peaks of the distance function.

The authors disprove this conjecture by constructing explicit counterexamples. They build planar domains that are simply connected (meaning they have no holes) where the distance function has only a single interior peak, but the torsion function has as many peaks as one desires. This shows that the torsion function can be far more complicated than the geometry of the domain might suggest, and that the two functions can behave in fundamentally different ways despite their superficial similarities. The construction is clever because simple, convex-looking domains tend not to produce this behavior, so the authors had to carefully design shapes that force the torsion function to develop multiple peaks.

The paper also resolves a separate question posed by Steinerberger about the shape of the level sets of the torsion function near its global maximum. Level sets are the curves along which the function takes a constant value, and near a peak these curves close in around it like contour lines around a hilltop. Steinerberger asked whether these contour curves must become increasingly round (specifically, elliptical) as they approach the peak. The authors answer this question for simply connected domains, adding another piece to the broader puzzle of understanding how geometry and these physically meaningful functions are related.

Read original →