Motivic construction of cohomological invariants (Q268288)

The author constructs a cohomological invariant \(u\) of degree 5 for groups of type \(E_8\) over a field \(k\) of characteristic 0. In the case \(k= \mathbb{Q}\) his result gives a positive answer to the following question, posed by Serre, for a group \(G\) of type \(E_8\) over \(\mathbb{Q}\), such that \(G_{\mathbb{R}}\) is a compact Lie group. Is it true that \(G_K\) is split if and only if \(q_K\) is hyperbolic? Here \(K\) is an extension of \(\mathbb{Q}\) and \(q=\langle\langle -1,-1,-1,-1, -1\rangle\rangle\) is a 5-fold Pfister form. If \(G\) is a simply connected simple algebraic group over a field \(k\) the existence of an invariant in \(H^3_{\text{ét}}(k,\mathbb{Q}/\mathbb{Z}(2))\) of \(G\)-torsors was conjectured by Serre and constructed by Rost. It is the Rost invariant of \(G\). Let \(G\) be an anisotropic group of type \(E_8\) over \(k\) \((\text{char\,}k=0)\) with trivial Rost invariant and let \(X\) be the variety of Borel subgroups of \(G\). Then an invariant \(u\) of degree 5 for \(E_8\) is obtained by constructing a smooth projective variety \(\widetilde Y\) over \(k\) of dimension 15, with no 0-cycles of odd degree, and such than the direct summands of the motives of \(X\) and \(\widetilde Y\) supporting the 0-cycles are isomorphic. The cohomological invariant \(u\in H^3_{\text{ét}}(k,\mathbb{Q}/\mathbb{Z}(2))\) for \(\widetilde Y\) is such that for any field extension \(K/k\) one has \(U_K= 0\) iff \(\widetilde Y_K\) has a 0-cycle of odd degree.