More about signatures and approximation (Q1332619)

Let \(M\) be a \(d\)-dimensional non-singular real algebraic variety and \(Y \subset M\) a codimension 1 semialgebraic set. Let \(\sigma\) be a sign distribution on \(M \setminus Y\), in other words, a continuous assignment of \(+ 1\) or \(- 1\) to every point of \(M \setminus Y\). The question addressed by this paper is: When does \(\sigma\) correspond to the sign of a regular function \(f\) on \(M\) vanishing on \(Y\)? For \(M\) compact, the authors give a proof that \(\sigma\) corresponds to the sign of a regular function if and only if the set \(\overline {\sigma^{-1} (-1)} \cap \overline {\sigma^{-1} (1)}\) where \(\sigma\) changes sign is the union of the \((d - 1)\)-dimensional parts of some irreducible components of \(Y\). As is pointed out in the paper, this follows from the work of L. Bröcker using fans, but the proof given here uses more traditional methods from algebraic geometry (general position, algebraic homology) and rests on the factoriality of the ring of regular functions on the complement of the set of points containing an open neighborhood in \(M\) on which \(\sigma\) is constant. The authors go on to consider a smooth function \(\varphi\) on \(M\) whose zero-set \(Y\) is that of a regular function. They show that such a function can be approximated in \(C^ 0 (M)\) by a regular function if the sign distribution of \(\varphi\) corresponds to that of a regular function. Furthermore, if \(Y\) is almost regular (the germs of a complexification are the complexifications of the germs), then the approximation can be done with smooth functions in the compact-open topology of \(C^ \infty (M)\). If we assume simply that the ring of regular functions on \(M\) is factorial, than all of the above results follow more simply, and a recipe for a general example is given which shows that this assumption is necessary if \(M\) is not compact. In the analytic case, one considers a non-singular real-analytic manifold \(M\) and a codimension 1 subset \(Y\) which is the common zeros of a finite number of global analytic functions. The authors prove that the same results hold with the notion of a component of \(Y\) replaced by a minimal proper closed subspace which is itself the common zeros of a finite number of global analytic functions and under the additional hypothesis that, for every such component \(Z\), either the ideal of functions vanishing on \(Z\) is principal or \(Z\) is analytically irreducible. The topology used for approximation is the Whitney topology. -- As an application the authors prove the following theorem: Let \(M\) be a \(d\)-dimensional real analytic manifold, let \(I\) be a coherent ideal sheaf in the sheaf of analytic functions on \(M\), and let \(Y\) be the support of the quotient sheaf. Suppose that \(I\) is locally principal and that \([Y] = 0\) in \(H^ \infty_{n - 1} (M, \mathbb{Z}_ 2)\). Then \(I\) is principal.