cp's OEIS Frontend

A convex lattice n-gon is a polygon whose n vertices are points on the integer lattice Z^2 and whose interior angles are strictly less than Pi.

For the even-indexed values, see A089187. The precisely known odd values were a(3)=1 (trivial), a(5)=5 and a(7)=13 (Arkinstall), a(9)=21 (Rabinowitz), a(11)=43 (Olszewska), a(13)=65 (Simpson), and a(15)=103 (Castryck). Additional values up to a(25) were first obtained as upper bounds "by massive calculations with several independent search programs" by Hugo Pfoertner. Pfoertner has made nice pictures of the best polygons he has found. See his link below. - Jamie Simpson, Dec 08 2022, adapted by Günter Rote, Sep 18 2023

From Günter Rote, Sep 18 2023: (Start)

The values a(n) can be computed in time polynomial in n by an algorithm of Eppstein, Overmars, Rote, and Woeginger from 1992: They showed how to compute the smallest convex n-gon in a set of N points in O(nN^3) time. The N points can be taken as an O(n^2) X O(n^2) grid: Each dimension of the bounding rectangle must be at least n/2, because a horizontal or vertical line can contain at most two vertices; since the area is known to be bounded by n^3, the other dimension cannot exceed 4n^2. In our case, the runtime can be reduced to O(nN^2), since the lowest vertex can be assumed to be a fixed point, say, the origin. By considering the lattice width, the grid can be reduced to size N=O(n^2) X O(n^1.5). Overall, this yields a theoretical runtime bound of O(n^8), for reporting all k-gons up to size n. This estimate agrees roughly with the observed runtimes in practice.

I have implemented the algorithm in Python and uploaded the program to the Code Golf Stackexchange site. It runs up to a(40) in a couple of minutes and produces some smallest polygon for each n. The values up to a(102) have been computed on a workstation in 31 hours. (End)

Consider convex lattice n-gons, that is, polygons whose n vertices are points on the integer lattice Z^2 and whose interior angles are strictly less than Pi. a(n) is the least possible number of lattice points in the interior of such an n-gon.

The result a(5) = 1 seems to be due to Ehrhart. Using Pick's formula, it is not difficult to prove that the determination of a(k) is equivalent to the determination of the minimal area of a convex k-gon whose vertices are lattice points.

Results before 2018 for odd n came from the following authors: a(3) (trivial), a(5) (Arkinstall), a(7) and a(9) (Rabinowitz), a(11) (Olszewska), a(13) (Simpson) and a(15) (Castryck). - Jamie Simpson, Oct 18 2022