Tetrahedra on deformed spheres and integral group cohomology (Q2380318)

The authors prove that any topological two-sphere \(\Sigma\) in \({\mathbb R}^3\) contains the vertices of a (possibly degenerate) tetrahedron with \(D_4\)-symmetry, under any metric on \({\mathbb R}^3\). Equivalently, there are points \(v_1, v_2, v_3,v_4 \in\Sigma,\) with \(v_1\neq v_3\) and \(v_2\neq v_4,\) such that \(d(v_1,v_2)=d(v_2,v_3)=d(v_3,v_4)=d(v_4,v_1)\) and \(d(v_1,v_3)=d(v_2,v_4).\) This result would follow from the still-open ``square-peg problem'', that any Jordan curve in \({\mathbb R}^2\) contains the vertices of a square. The result is proved using tools from equivariant algebraic topology, applied to the \(D_4\)-equivariant map \(f : (S^2)^4 \to {\mathbb R}^4 \times {\mathbb R}^2\) given by the deviations from the means of the distances above. The required points then correspond to zeros \((x_1,x_2,x_3,x_4)\) of \(f\) satisfying \(x_1\neq x_3\) or \(x_2\neq x_4.\) Fadell-Husseini index theory provides a nontrivial cohomological obstruction to the existence of such a map having no zeros, thus yielding the result. The argument requires some group cohomology and explicit spectral sequence calculations of the integral cohomology of certain fiber spaces as \({\mathbb Z}_4\)-modules. The obstruction vanishes if one uses coefficients in \({\mathbb Z}_2,\) so this problem gives an example where index theory over fields is not sufficient. The same approach to solve the topological square-peg problem also fails because the obstruction vanishes.