On Castelnuovo's equivalence defect. (Q3320511)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3851261
Language Label Description Also known as
default for all languages
No label defined
    English
    On Castelnuovo's equivalence defect.
    scientific article; zbMATH DE number 3851261

      Statements

      On Castelnuovo's equivalence defect. (English)
      0 references
      0 references
      1984
      0 references
      algebraic family of divisors on a curve of genus g
      0 references
      Castelnuovo-Severi inequality
      0 references
      finite characteristic
      0 references
      linear series
      0 references
      equivalence defect
      0 references
      theta-divisor on the Jacobian variety
      0 references
      Let C be a curve of genus g defined over an algebraically closed field K, and let \(D=\{D_ t\}_{t\in T}\) be an algebraic family of divisors on C which is parametrized by a curve T. In order to be able to decide whether such a family is ''contained in a linear series'', i.e. whether all the divisors \(D_ t\) are mutually linearly equivalent, \textit{G. Castelnuovo} [R. Accad. Lincei, Rend., Ser. V 15, 337-344 (1906)] introduced a remarkable invariant \((D)\geq 0\) which has the property that \(z(D)=0 \Leftrightarrow\) all the \(D_ t\) are linearly equivalent ( \(\Leftrightarrow D\equiv 0\), i.e. \(D\sim a\times C+T\times A,\) with \(a\in Div(T)\) and \(A\in Div(C));\) for this reason, \textit{C. Torelli} [Rend. Circ. Mat. Palermo 37, 27-46 (1914)] later called z(D) the ''equivalence defect'' of D. Castelnuovo's principal result is that this invariant may be calculated from the fundamental invariants attached to the family by the formula \[ (*)\quad z(D)=frac{1}{2}\sigma(D,D), \] where \(\sigma(D,D)=2n(D)\nu(D)-\gamma(D)\) is the ''Severi-Weil metric'' on Di\(v(T\times C)\), and \(n(D)=(D\cdot t\times C)\) (for \(t\in T)\), \(\nu(D)=(D\cdot T\times P)\) (for \(P\in C)\), and \(\gamma(D)=(D.D).\) From (*) one obtains the inequality \(\sigma(D,D)\geq 0,\) which is often referred to as the ''Inequality of Castelnuovo-Severi'' [cf. \textit{A. Mattuck} and \textit{J. Tate}, Abh. Math. Semin. Univ. Hamburg 22, 295-299 (1958; Zbl 0081.376)]. The main purpose of the paper under review is to investigate the formula (*). As it turns out, if one uses Castelnuovo's original definition of z(D), then (*) is no longer valid if char(K)\(\neq 0\); one therefore has to redefine the invariant z(D) in a suitable way. Two such methods are known: (1) via intersection numbers involving the \(\theta\)-divisor on the Jacobian variety \(J_ C\) of C [due to \textit{G. Castelnuovo}, Atti R. Accad. Lincei, Rend., Cl. Sci. Fis. Mat. Nat. Ser. V 30, 50-55, 99-103, 195-200, 355-359 (1921)] and also \textit{A. Weil} [''Variétés abéliennes et courbes algébriques'' (Paris 1948; Zbl 0037.162)]; and (2) via intersection theory on symmetric products of curves and Schubert's formula (in char(K)\(\neq 0\), due to \textit{J. Igusa} [J. Math. Soc. Japan 1, 147-197 (1949; Zbl 0039.032)]). - In this paper another method is presented, one that does not leave the surface \(X=T\times C\); it is as follows. For any family \(D=\{D_ t\}_{t\in T}\) let \(\delta(D)=\min_{t} \ell(D_ t)\) denote the ''generic dimension'' of D (here, \(\ell(D_ t)=\dim H^ 0(C,{\mathcal L}(D_ t))),\) and let \(var(D)=\sum(\ell(D_ t)-\delta(D_ t))\) denote the ''total variation'' of the family. Then one has: Theorem: Let \(D\in Div(X/T)^+,\) i.e. let D be an effective divisor on \(X=T\times C\) which is flat over T. If \(\delta(D)=1\), then \[ (**)\quad \nu(D)+var(D)\leq frac{1}{2}\delta(D,D). \] Moreover, if in addition we have \(n(D)=g\) and \(var(D)=0,\) then equality holds in (**); i.e. we have \(\nu(D)=frac{1}{2}\sigma(D,D).\) Thus, if we put \(z'(D)=\max(\nu(D')+var(D')),\) where the maximum extends over all divisors \(D'\in Div(X/T)^+\) with \(\delta(D')=1\) and \(D'\equiv D,\) then on the one hand it is easy to see that \(z'(D)=0 \Leftrightarrow\) \(D\equiv 0\) (of course one has \(z'(D)\geq 0),\) and on the other hand we have ''Castelnuovo's formula'': \(z'(D)=frac{1}{2}\sigma(D,D).\) (Also, it is easy to see that in characteristic 0 one has \(z'(D)=z(D)\).) Moreover, from the above theorem it is possible to deduce ''Castelnuovo's formula on Jacobians'', viz. \[ (***)\quad(T\cdot \alpha_ D^*(\Theta))=frac{1}{2}\sigma(D,D), \] where \(\Theta\) denotes the theta- divisor on the Jacobian variety \(J_ C\) of C, \(\alpha_ D:J_ T\to J_ C\) denotes the homomorphism induced by D, and ( \(\cdot)\) denotes the intersection number of the Jacobian \(J_ T\) of T. By a more careful analysis of the above proof one also obtains a formula for the local intersection number \((T\cdot \alpha_ D^*(\Theta))_ x\) (provided that this intersection number is defined), which after summation over \(x\in J_ T\) yields the ''global formula'' (***).
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references