Curvy slicing proves that triple junctions locally minimize area (Q679312)

In 1873, the physicist Plateau published the observation that the singularities in soap films are of just two types: (1) three minimal surfaces meeting smoothly at angles of 120 degrees along a curve and (2) four such curves meeting smoothly at a point with angles of about 109 degrees as formed by the rays from the center of a regular tetrahedron to the vertices of the tetrahedron. In work culminating in 1976, \textit{J. E. Taylor} proved Plateau's observations in the mathematical context of \((\mathbb M,0,\delta)\)-minimal sets [Ann. of Math. 103, 489-539 (1976; Zbl 0335.49032)]. In the present paper, the authors prove conversely that any set of three minimal surfaces meeting smoothly, in the \(C^4\) sense, at 120 degree angles along a curve in \(\mathbb R^3\) form a locally area minimizing surface. The motivation for the authors' proof is provided by the proof that three half-planes meeting at 120 degree angles along a line are area minimizing: This proceeds by considering the slices obtained by intersecting with a plane perpendicular to the line where the half-planes meet. Each such slice is a ``Y'' shaped configuration of three rays forming 120 degree angles at a point, and this configuration can be shown by elementary means to be length minimizing. The area of any finite part of the surface formed by the three hyperplanes can be obtained by integrating the lengths of the intersection with the slicing plane. The area minimizing property of the three half-planes follows. For the general case of their theorem, the authors must slice by curvy discs that are perpendicular to the curve along which the minimal surfaces meet. Because the slicing discs are not flat, the area of any small enough part of the surface is the integral of the lengths of the slices in a weighted metric. Matters are delicate here because the weighted metric is \(C^{1,1},\) but not \(C^2,\) and a geodesic ``Y'' in a \(C^{1,1}\) metric is locally length minimizing, but this is not true in a \(C^{1,\alpha}\) metric if \(\alpha<1.\) Thus it is crucial that they are able to prove that the metric is \(C^{1,1}.\) The main theorem is also proved for \((k+1)\)-dimensional surfaces meeting along a \(k\)-dimensional surface in \(\mathbb R^{n}.\) Additional results concern boundary configurations.