Quaternionic bundles on algebraic spheres (Q1815515)

It has remained an open question for many years whether there is a bijection between algebraic and topological vector bundles on spheres. If one denotes by \(\mathbb{F}\) one of the (skew) fields \(\mathbb{R}\), \(\mathbb{C}\) or \(\mathbb{H}\) and by \(A_n\) the coordinate ring \(\mathbb{R}[x_0,\dots,x_n]/(\sum x^2_1-1)\) of the sphere, then the more precise question is whether \(P_k(\mathbb{F}\otimes_\mathbb{R} A_n)\to VB^\mathbb{F}_k(S^n)\) is a bijection. Here \(P_k(\mathbb{F}\otimes_\mathbb{R} A_n)\) is the set of isomorphism classes of finitely generated \(\mathbb{F}\otimes_\mathbb{R} A_n\)-modules of rank \(k\) and \(VB^\mathbb{F}_k(S^n)\) the set of isomorphism classes of topological \(\mathbb{F}\)-vector bundles. Although some evidence was given in the past towards a positive answer of this conjecture, the present paper shows that the conjecture is actually false for \(\mathbb{F}=\mathbb{H}\), for \(n\geq 4\): there then exists a stably free, nonfree \(\mathbb{H}\otimes_\mathbb{R} A_n\)-module \(Q\) of rank 1, which is topologically trivial. (For \(n\geq 5\), the author even shows that there exist uncountably many such bundles.) A first proof uses results due to Ojanguren, Parimala, Sridharan and Wood, another one is more computational and allows for explicit examples.