Maths Olympiad Prep

Library / /5 of 7

, 2010

Geometry Difficulty 6.1 National Olympiad Prove it Romania

Determine all integer numbers n3n \ge 3 such that the regular nn-gon can be decomposed into isosceles triangles by noncrossing diagonals.

Solution

The required numbers are of the form n=2r(2s+1)n = 2^r(2^s+1), where rr and ss are nonnegative integer numbers which do not vanish simultaneously. Clearly, any such nn works.

To establish the converse, let KK be a regular nn-gon, n4n \ge 4, which can be decomposed into isosceles triangles by noncrossing diagonals. Begin by noticing that each edge ee of KK must be an edge of a unique isosceles triangle TeT_e in the decomposition. Two cases are possible: either ee is opposite the apex of TeT_e or ee and one of the adjacent edges of KK are the edges of TeT_e issuing from the apex. (Since n4n \ge 4, TeT_e cannot be equilateral, so the apex is well defined.)

If nn is even, no vertex of KK lies on the perpendicular bisector of an edge of KK, so the first case is ruled out. Consequently, the decomposition must contain exactly one of the two bracelets of n/2n/2 isosceles triangles clipped off by short diagonals joining consecutive vertices of KK of likewise parity. These short diagonals are the edges of a regular n/2n/2-gon which is also decomposed into isosceles triangles by noncrossing diagonals and the conclusion follows by induction.

If nn is odd, then KK has a unique edge ee opposite the apex of TeT_e: Since nn is odd and each short diagonal clips off two edges of KK, at least one such ee exists. The apex of TeT_e lies on the perpendicular bisector of ee, so it must be the vertex of KK opposite ee. Uniqueness of ee should now be clear: were there another such ee', the interiors of TeT_e and TeT_{e'} would overlap. Consequently, ee is unique and KK splits into TeT_e and two polygons LL and LL' which are reflections of one another in the perpendicular bisector of ee.

To complete the proof, it is sufficient to show that the number of vertices of LL is one plus a power of 2. Begin by noticing that LL inherits by restriction a decomposition into isosceles triangles by noncrossing diagonals. Let x0,,xmx_0, \dots, x_m be a circular labelling of the vertices of LL around the boundary, where x0x_0 is the vertex of KK opposite ee and xmx_m is a vertex of ee. Since dist(xi,xj)<dist(x0,xm)\text{dist}(x_i, x_j) < \text{dist}(x_0, x_m) if {i,j}{0,m}\{i, j\} \neq \{0, m\}, it follows that x0xmx_0x_m is an edge of an isosceles triangle with apex at some xkx_k, 0<k<m0 < k < m. Notice that dist(x0,xi)<dist(xi,xm)\text{dist}(x_0, x_i) < \text{dist}(x_i, x_m) if 0<i<m/20 < i < m/2 and dist(x0,xi)>dist(xi,xm)\text{dist}(x_0, x_i) > \text{dist}(x_i, x_m) if m/2<i<mm/2 < i < m, to deduce that mm must be even, k=m/2k = m/2, and LL splits into an isosceles triangle, x0xm/2xmx_0x_{m/2}x_m, and two polygons, x0xm/2x_0 \cdots x_{m/2} and xm/2xmx_{m/2} \cdots x_m, which are reflections of one another in the perpendicular bisector of the segment x0xmx_0x_m. Now we are essentially back in the situation that arose above. Repeat the same argument verbatim to infer that m/2m/2 must be even and so on all the way down to conclude that mm must be a power of 2.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.