The paper addresses a classic type of question in mathematics: among all polygons of a given type and fixed area, which shape is the "best" in some geometric sense? Here the relevant measure of quality is torsional rigidity, a quantity from physics that describes how strongly a solid bar resists twisting when torque is applied. The class of polygons considered are "tangential polygons," meaning polygons that have an inscribed circle touching every side. The main result is that, for any fixed number of sides, the regular polygon (the one with all sides and angles equal) has strictly greater torsional rigidity than any other tangential polygon with the same area. This extends and unifies earlier known results, including a classical theorem about triangles due to Polya and Szego.
The proof works by cutting each tangential polygon into right-triangular pieces, one for each side, all meeting at the center of the inscribed circle. Torsional rigidity is then computed by solving a partial differential equation on each piece, with boundary conditions that differ depending on whether a side of the triangle lies on the outer edge of the polygon or not. The key mathematical insight is that a certain function of the angle at the tip of each triangle is strictly concave, meaning the contribution to torsional rigidity is maximized when all these angles are equal, which happens precisely when the polygon is regular. The authors also derive an explicit formula showing exactly how much rigidity is lost when a polygon deviates from regularity, separating two sources of inefficiency: unequal angles and excess perimeter.
As bonus results, the paper uses the same framework to prove two additional facts. First, among regular polygons with equal areas, those with more sides always have greater torsional rigidity, with the circle being the limiting ideal shape. Second, the authors derive a concrete condition guaranteeing that a tangential polygon also has a larger "first Dirichlet eigenvalue" (related to the lowest vibrational frequency of a drum) than the equal-area regular polygon. This second application is somewhat surprising because optimizing eigenvalues is generally a harder problem, and the criterion here follows as a byproduct of the torsional rigidity analysis.