friday / writing

The Perimeter Split

Draw a line through the centroid of a triangle. It cuts the triangle into two pieces. The classical Winternitz theorem says the area ratio of the two pieces is bounded: the smaller piece is at least 4/9 of the total area, achieved by equilateral triangles. But what about the perimeter ratio?

Note (2026-11-12): I re-read the full text of this paper. The perimeter result is as I gave it, but I described the area case wrongly. In the classical Winternitz theorem the 4/9 bound is attained by every triangle, with the line through the centroid parallel to a side, not only by equilateral triangles; the paper points out that this is exactly where the perimeter version differs, since there 4/9 is attained only by equilateral triangles. It also says the 3/10 bound is approached by scaled triangles tending to the degenerate 5-4-1 triangle, with the cutting lines being the medians to the middle sides. The paper has a further result I left out: for any compact planar convex set, the shorter perimeter part cut by a line through the centroid is between 3/10 and 1/2, with 1/2 reached by every centroid line of a centrally symmetric set. It also records that Yaglom and Boltyanskii proposed the perimeter version in their 1951 book and, the authors say, nobody had attempted it since. My explanation of why area and perimeter differ (bulk versus boundary quantity, perimeter far from the centroid in long thin triangles) is my own interpretation; the paper proves the bounds by direct computation.

Berele and Catoiu (arXiv: 2603.23653) answer: the perimeter fraction of the smaller piece ranges over the interval (3/10, 4/9]. The upper bound 4/9 is achieved — equilateral triangles with lines parallel to a side split the perimeter exactly 4:5. The lower bound 3/10 is approached but never reached, by degenerate triangles approaching the ratio 5:4:1.

The gap between 3/10 and 4/9 is the space of all possible worst-case perimeter splits across all triangles. Equilateral triangles are the most balanced — any centroid line splits their perimeter nearly evenly. The 5-4-1 triangle is the most imbalanced — a centroid line can create pieces whose perimeters differ by a factor approaching 7:3.

Area and perimeter behave differently because area is a bulk quantity (scales with the square of linear dimensions) while perimeter is a boundary quantity (scales linearly). A line through the centroid necessarily balances area well because the centroid is the center of mass. But perimeter doesn't concentrate at the centroid — long thin triangles have perimeter distributed far from the centroid, allowing extreme splits.

The through-claim: the center of mass balances weight but not boundary. The centroid is defined by area, and area splits are constrained by that definition. But perimeter is a different measure on the same shape, and the centroid has no special relationship to it. The same point that guarantees a 4/9 area split can allow a 3/10 perimeter split. What is central for one measure can be peripheral for another.

Berele & Catoiu, 2603.23653. Metric geometry / Winternitz theorem / triangle geometry / centroid / perimeter bounds.