The affine Tverberg theorem revisited
The Bárány–Kalai–Tverberg conjecture asks whether the affine Tverberg theorem extends from the simplex to arbitrary convex polytopes and simplicial balls of dimension N=(d+1)(r-1). Standard equivariant methods fail when r is not a prime power.
The affine map is lifted to a map from the r-fold join to R^N, and the problem is reduced to showing that 0 lies in the image of the deleted join. The proof combines a Leray–Vietoris–Begle type lemma, a bad-vertex triangulation, and acyclicity arguments over shellable polytopal subdivisions arising from the upper hull of an auxiliary polytope. The proof was formalized in Lean 4.
The conjecture is proved for all r ≥ 2. For any affine map from an N-dimensional polytope, or a face-wise affine map from a simplicial ball, there are r pairwise disjoint boundary faces whose images intersect. The formalization is available at github.com/savvakumov/AffineTverberg.
