Which algebraic groups are Picard varieties? (Q2341343)

Let \(k\) be a field, \(X\) a proper scheme over \(k.\) Then \(\mathbf{Pic}_{X/k} ^{0},\) the neutral component of the Picard scheme \(\mathbf{Pic}_{X/k},\) parameterizes algebraically trivial invertible sheaves on \(X.\) If \(k\) is perfect, then the reduced neutral component is the Picard variety Pic\(^{0}\left( X\right) .\) In the work under review, the author investigates the following: given a connected commutative algebraic group \(G\), is there a proper scheme \(X\) such that \(G\cong\)Pic\(^{0}\left( X\right) \)? Generally, for a locally noetherian scheme \(S\), let \(0\rightarrow V\times T\rightarrow G\rightarrow A\rightarrow0\) be an exact sequence of commutative \(S\)-group schemes, where \(V\) is a vector group (that is, the additive group of a locally free sheaf of finite rank), \(T\) a quasi-split torus, and \(A\) an abelian scheme. Then there exists an \(S\)-scheme \(X\) which is proper and flat and whose geometric fibers are integral such that \(G=\mathbf{Pic}_{X/k} ^{0}\left( S\right)\). Applying this to the case \(S=\text{Spec}k\), \(k\) an algebraically closed field gives an affirmative answer to the question above since (a) every connected commutative unipotent group is a vector group and every torus is split (hence quasi-split), and (b)\ \(G\) is known to be an extension of an abelian variety by the product of a unipotent group and a torus. On the other hand, if char \(k=p>0\) then the above approach no longer holds, as proved by an example. For any integer \(n\geq3,\) let \(W_{n}\) be the Witt group of dimension \(n\). The second main result (shown in slightly weaker form here) of this paper is that \(W_{n}\) is not isogenous to the Picard variety of any projective variety with finite non-normal locus.