Minimal generation of basic open semianalytic sets (Q1109181)

The question of how many analytic functions are needed to describe basic open semianalytic sets i.e. sets of the form \[ Z=\{z\in {\mathbb{R}}^ n;\quad f_ 1(z)>0,...,f_ r(z)>0\} \] has been open for a long time. So far all efforts to obtain any result concerning this problem by the means of elementary calculus haveasure on the product space \({\mathcal M}et\times {\mathcal E}\), where \({\mathcal M}et\) is the space of metrics on M and \({\mathcal E}\) is the space of maps from M into \({\mathbb{R}}^ D\). S(g,x) is invariant under the group of motions of \({\mathbb{R}}^ D\), the group of diffeomorphisms of M and the conformal group of M. The integral over x actually reduces to a Gaußian integral because the action is quadratic in x, and so one is lead to construct a natural measure on \({\mathcal M}et\), conformally invariant and invariant under the group of diffeomorphisms, in other words a natural measure on \({\mathcal M}_ p\), the space of moduli of Riemann surfaces of genus p. As the integral above doesn't make sense, one considers first a finite dimensional model, giving a formula for a measure on a finite dimensional \({\mathcal C}^{\infty}\)-manifold, satisfying corresponding invariance requirements. Then, transplanting formally this formula to the Fréchet manifold \({\mathcal M}et\) one finds a measure on \({\mathcal M}et\) which descends to the Teichmüller space \({\mathcal T}_ p\) (and actually \({\mathcal M}_ p)\) if and only if a certain expression involving infinite dimensional determinants of the Laplacian \(\Delta_ g\) (or equivalently \({\bar \partial}^*{\bar \partial})\) of M and the positive differential operator \({\bar \partial}^*_ T{\bar \partial}_ T\) on the holomorphic tangent bundle T on M, depends only on the image of g in \({\mathcal T}_ p\). To give a rigorous meaning to these determinants one uses so-called \(\zeta\)-function regularization. The measure is conformally invariant if and only if \(D=26\) (formulas of conformal anomalies). The formulas of conformal anomalies can be derived from a local Riemann- Roch-Grothendieck (RRG) theorem for a holomorphic family of compact Riemann surfaces \(\pi\) : \(X\to S\) and tensor powers of the relative tangent bundle \(T=T_{X/S}\) equipped with a \({\mathcal C}^{\infty}\) hermitian metric. To have a globally defined \({\mathcal C}^{\infty}\)-metric one uses the Quillen metric \(\| \|_ Q\) on the determinant bundle DET \({\bar \partial}_ T\otimes n\). The RRG theorem then gives an expression on the level of differential forms for the first Chern class \(c_ 1(DET {\bar \partial}_ T\otimes n\), \(\| \|_ Q)\). The critical dimension \(D=26\) appears as an avatar of the coefficient 1/12 of the Todd genus occurring in the expression for \(c_ 1.\) A basic result for the rest of the underlying article is a theorem of Mumford which says that for a family of connected Riemann surfaces as above, one has an isomorphism: \(M_{\pi}: DET {\bar \partial}_{T_{X/S}}\cong (DET {\bar \partial}_{{\mathcal O}})^{\otimes 13}\), compatible with base change and such that if \(\pi\) is an algebraic morphism of quasi-projective complex varieties, then \(M_{\pi}\) is algebraic. A result of Beilinson and Drinfeld, using Mumford's theorem, implies that the Polyakov measure on \({\mathcal T}_ p\) coincides (up to a multiplicative constant) with the one defined by the metric on the canonical line bundle on \({\mathcal T}_ p\), thus establishing the ``algebraic'' nature of Polyakov's measure. Mumford's theorem can also be applied to obtain a more or less explicit description of the Polyakov measure in the vecinity of the divisor at infinity \(\Delta =\bar {\mathcal M}_ p-{\mathcal M}_ p\), where \(\bar {\mathcal M}_ p\) is the compactification of \({\mathcal M}_ p\) by stable curves. Explicit formulas for the Polyakov measure on \({\mathcal T}_ p\) for \(p=1,2,3\) are derived. On \({\mathcal T}_ 1\) Dedekind's \(\eta\)-function is involved, whereas on \({\mathcal T}_ 2\) and \({\mathcal T}_ 3\) \(\vartheta\)- functions appear.