Tensor product theorem for Hitchin pairs -- an algebraic approach (Q424843)

Let \(X\) be a smooth projective curve over an algebraically closed field \(k\). When \(k=\mathbb{C}\), the celebrated Narasimhan-Seshadri theorem [\textit{M. S. Narasimhan} and \textit{C. S. Seshadri}, ``Stable and unitary vector bundles on a compact Riemann surface'', Ann. Math. (2), 82, 540--567 (1965; Zbl 0171.04803)] states that a rank \(n\) and degree \(0\) vector bundle over \(X\) is semistable if and only if it carries a flat unitary connection, i.e., if and only if it is associated to a representation of \(\pi_1X\) in \(\mathrm{U}(n)\). This correspondence allows to prove that the tensor product of two semistable vector bundles is again semistable. In the 80's, \textit{N. J. Hitchin} in [``The self-duality equations on a Riemann surface'', Proc. Lond. Math. Soc., III. 55, 59--126 (1987; Zbl 0634.53045)] introduced Higgs bundles. These are pairs \((V,\varphi)\) composed by a vector bundle \(V\) and by a section \(\varphi\) of \(\mathrm{End}(V)\otimes\Omega_X\), where \(\Omega_X\) is the canonical line bundle of \(X\). Hitchin proved that a rank \(n\) and degree \(0\) Higgs bundle is semistable if and only if it arises from a representation of \(\pi_1X\) in \(\mathrm{GL}(n,\mathbb{C})\) (and also from a solution of the so-called Hitchin equations). Again this yields a proof that the tensor product of two semistable Higgs bundles bundles is again semistable. These proofs of the tensor product theorem are not, however, completely algebraic. The first purely algebraic proofs of the tensor product theorem for vector bundles is due to \textit{F. A. Bogomolov} [``Holomorphic tensors and vector bundles on projective manifolds'', Math. USSR, Izv. 13, 499--555 (1979); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 42, 1227--1287 (1978; Zbl 0439.14002)] and then by \textit{D. Gieseker} [``On a theorem of Bogomolov on Chern classes of stable bundles'', Am. J. Math. 101, 77--85 (1979; Zbl 0431.14005)]. The aim of the present paper is to provide a purely algebraic proof for the tensor product theorem for Hitchin pairs over \(X\), where \(k\) is any algebraically closed field of zero or positive characteristic (satisfying some natural bounds). Fix a locally free \(\mathcal{O}_X\)-module over \(X\), \(U\). Let \(\mathcal{F}\) be a locally free \(\mathcal{O}_X\)-module. A Higgs structure on \(\mathcal{F}\) is a map \(\theta:\mathcal{F}\to\mathcal{F}\otimes U\) such that \(\theta\wedge\theta=0\). A Hitchin pair is a pair \((\mathcal{F},\theta)\) composed by a locally free \(\mathcal{O}_X\)-module \(\mathcal{F}\) together with a Higgs structure \(\theta\). If \((\mathcal{E},\phi)\) and \((\mathcal{F},\theta)\) are two Hitchin pairs, then their tensor product is the Hitchin pair \((\mathcal{E}\otimes\mathcal{F},\phi\otimes1+1\otimes\theta)\). The main result of the paper is hence an algebraic proof of the fact that if both \((\mathcal{E},\phi)\) and \((\mathcal{F},\theta)\) are are such that \(\det(\mathcal{E})\cong\mathcal{O}_X\cong\det(\mathcal{F})\) and are semistable, then so is \((\mathcal{E}\otimes\mathcal{F},\phi\otimes1+1\otimes\theta)\), under the assumption that \(\mathrm{rk}(\mathcal{E})+\mathrm{rk}(\mathcal{F})<\mathrm{char}(k)+2\). As the authors say, apart from the aesthetics of a purely algebraic proof, the new result which emerges by this approach are variations in positive characteristics for the notions of principal Hitchin pairs. It is noticed that the Frobenius pull-back is an inconsequential operation for Hitchin pairs, so notions such as strong semistability do not yield anything new in the Higgs setting. It is also shown that the bounds developed in [\textit{S. Ilangovan, V.B Mehta} and \textit{A. J. Parameswaran}, Semistability and semisimplicity in representations of low height in positive characteristic'', A tribute to C. S. Seshadri. A collection of articles on geometry and representation theory. Basel: Birkhäuser. Trends in Mathematics, 271-282 (2003; Zbl 1067.20061)] and in [\textit{V. Balaji} and \textit{A. J. Parameswaran}, ``Semistable principal bundles. II: Positive characteristics'', Transform. Groups 8, No. 1, 3--36 (2003; Zbl 1084.14013)] are immediately applicable. In the context of the recent work of \textit{Ngô Bao Châu} [``Hitchin fibration and endoscopy. (Fibration de Hitchin et endoscopie.)'', Invent. Math. 164, No. 2, 399--453 (2006; Zbl 1098.14023)] this approach could be of interest in positive characteristics.