Birational models for varieties of Poncelet curves (Q1914852)

A pair \((S,C)\) consisting of a smooth conic \(S\) and a curve \(C\) of degree \(c\) in the complex projective plane will be called a Poncelet pair if there exist \(c+1\) tangents to \(S\) whose \({c+1 \choose 2}\) intersection points lie on \(C\). \(C\) will then be called Poncelet related to \(S\) or simply a Poncelet curve. A theorem of Darboux states that in this case there is an infinity of such sets of \(c+1\) tangents. The Poncelet curves reappeared in a natural way in the study of stable vector bundles on projective spaces. -- In this paper we consider the variety \(Pon_c\) of Poncelet curves of degree \(c\). It is the image of the space of Poncelet pairs \((S,C)\) through the second projection. We show that for \(c\geq 5\), this projection is birational. This fails for \(c\leq 3\), simply because of dimension reasons, and has been proved for \(c=4\) by Le Potier via the study of divisors on the corresponding moduli space of rank-2 semi-stable sheaves on \(\mathbb{P}^2\). As a corollary, we get the rationality of \(Pon_c\) for every \(c\). The generic injectivity of the above mentioned projection brings some evidence for (and maybe could help proving) its generalizations to: Barth's morphism which associates to a stable rank-2 vector bundle of type \((0,c)\) on \(\mathbb{P}^2\) its curve of jumping lines, and the restriction of the previous to Hulsbergen bundles. In the last paragraph we describe a projective birational model of \(Pon_c\) when \(c\) is odd.