Anti-affine algebraic groups (Q1014585)

A group scheme \(G\) of finite type over a field \(k\) is said to be \textit{anti-affine} if \(\mathcal{O}\left( G\right) =k,\) i.e. every regular function on \(G\) is constant. Any anti-affine algebraic group is smooth, connected, and commutative. The most well-known examples of anti-affine groups are abelian varieties, however this collection of group schemes includes universal vector extensions and some semi-abelian varieties. For any group \(G\) of finite type one has two subgroups \(G_{\text{ant}}\) and \(G_{\text{aff}},\) where \(G_{\text{ant}}\) is the smallest normal subgroup of \(G\) such than \(G/G_{\text{ant}}\) is affine (and, as the notation suggests, \(G_{\text{ant}}\) is anti-affine; furthermore it is central in \(G\)) and \(G_{\text{aff}}\) is the smallest connected affine subgroup such that \(G/G_{\text{aff}}\) is an abelian variety (hence anti-affine). One then gets the Rosenlicht decomposition: \(G=G_{\text{aff}}G_{\text{ant}}\), and \(\left( G_{\text{ant}}\right) _{\text{aff}}\) is a subgroup of \(G_{\text{aff}}\cap G_{\text{ant}}\) of finite index. The main objective in this paper is to classify anti-affine algebraic groups, and for certain choices of \(k\) one obtains a very nice classification. Before giving the classification, certain properties of anti-affine algebraic groups are given. It is shown that \(G\) (always assumed to be of finite type) is anti-affine over \(k\) if and only if it is anti-affine over any extension of \(k\) (an easy result that follows from the definition). Furthermore, \(G\) is anti-affine if and only if every finite-dimensional linear representation of \(G\) is trivial -- this further justifies the term ``anti-affine'' as affine groups admit a faithful linear representation. It is shown how the product of two anti-affine groups are anti-affine; and if an anti-affine group \(G\) decomposes as the product of two connected group schemes they must both be anti-affine. If \(f:G\rightarrow H\) is an isogeny of connected commutative algebraic groups, then \(G\) is anti-affine if and only if \(H\) is, thereby showing that ``anti-affineness'' is stable under isogenies. Finally, the multiplication map \(x\mapsto nx:G\rightarrow G\) is an isogeny when \(G\) is anti-affine. The classification begins by identifying all connected group schemes with an abelian variety. As \(G/G_{\text{aff}}\) is an abelian variety, say \(G/G_{\text{aff}}\cong A\) via an isomorphism \(\alpha,\) we say that \(G\) is a group scheme over \(A\). In the case where \(G\) is a semi-abelian variety it is proved that \(G\) corresponds to a sub-\(\Gamma_{k}\)-lattice of \(A^{\vee}\left( k_{s}\right) \), where \(k_{s}\) is the separable closure of \(k\) and \(\Gamma _{k}=\)Gal\(\left( k_{s}/k\right) ;\) furthermore this correspondence is bijective. As a consequence, if \(G\) is an anti-affine algebraic group over a field of positive characteristic then \(G\) is necessarily a semi-abelian variety; furthermore if \(k\) is a finite field then \(G\) is an abelian variety. Thus in characteristic \(p\) the anti-affine algebraic groups correspond to pairs \(\left( A,\Lambda\right) ,\) where \(\Lambda\) is the lattice described above. The classification is more complicated in characteristic zero -- for example one has universal vector extensions which are anti-affine. For \(G\) an \(A\)-group scheme and \(E\left( A\right) \) the universal vector extension of \(A\) there is a map \(E\left( A\right) \rightarrow G\) which induces a map \(\gamma:H^{1}\left( A,\mathcal{O}_{A}\right) ^{\ast}=E\left( A\right) _{\text{aff}}\rightarrow G_{\text{aff}}.\) It is proved that \(G\) is anti-affine if and only if \(\gamma\) is surjective, and that anti-affine groups over \(A\) obtained as vector extensions are classified by sub-\(k\)-vector spaces of \(H^{1}\left( A,\mathcal{O}_{A}\right) .\) Thus in characteristic zero the anti-affine algebraic groups correspond to triples \(\left( A,\Lambda ,V\right) ,\) where \(\Lambda\) is as above and \(V\) is a subspace of \(H^{1}\left( A,\mathcal{O}_{A}\right) .\) Some consequences of the work above are given, most notably some counterexamples to Hilbert's fourteenth problem. Hilbert's fourteenth problem asks if the coordinate ring of every quasi-affine variety is finitely generated. Here is it shown that \(X\), where \(X\rightarrow G\) is the \(\mathbb{G}_{m}\)-torsor associated to an ample invertible sheaf on a non-complete anti-affine algebraic group, is not noetherian.