Degree theory for Wiener maps (Q1114210): Difference between revisions

In this interesting paper, the author uses the framework of the Malliavin calculus to introduce a notion of degree for suitable maps, not necessarily even continuous, of a Banach space into itself. Let \(B^*\subset H\subset B\) be an abstract Wiener space in the sense of Gross, and let \(D^ s_ p(B;H)\) be the Sobolev space (of H-valued functions) of order s and exponent p, as defined by Malliavin; \(\mu\) is the Gaussian measure of B. Let \(\psi\) : \(B\to B\) be a Wiener map; that is, a function \(x\mapsto x+F(x)\), where \(F\in D^ 1(B;H)=\cap_{p>1}D^ 1_ p(B;H)\). The author proves first that, if F: \(B\to H\) is smooth in the Fréchet sense, its critical values are of \(\mu\)-measure zero, so that a signed Borel measure \(\psi^*\mu\) is defined (the set of critical points of \(\psi\) has \(\psi^*\mu\)-absolute variation zero, and elsewhere \(\psi\) is a local \(C^{\infty}\) diffeomorphism which preserves or reverses Fredholm orientation and pulls back \(\mu\) with the corresponding sign). Certainly \(\psi^*\mu\) is absolutely continuous with respect to \(\mu\), and \(d(\psi^*\mu)/d\mu\) is given by the (signed) formula of \textit{R. Ramer} [ibid. 15, 166-187 (1974; Zbl 0288.28011)]. The same formula may be used to define \(\psi^*\mu\) for any Wiener map \(\psi\), and it is natural to define by finite- dimensional analogy \(\deg (\psi)=\psi^*\mu (B).\) The author's main contribution is to prove that, if F satisfies a certain integrability condition (involving its derivative and divergence), then, for any bounded Borel function f, \(\int (f\circ \psi)d(\psi^*\mu)=\deg (\psi)\cdot \int fd\mu.\) Thus the degree has a local character, as in finite dimensions. Under a much stronger integrability condition, the degree is also shown to be an integer; and, when F: \(B\to H\) is smooth, the degree may be interpreted geometrically as the sum of the signs of the derivative over all points of a regular fibre. \{The prospective reader is warned that the paper contains a remarkable number of misprints and mistakes, far too numerous to list in full; most are baffingly obvious. 1.12 is misquoted; 2.1 seems unnecessary; in 2.2, c should be \(\epsilon\), and the conclusion \(cap_{p,k}(K^ c)\leq \epsilon\) (thus decreasing means increasing, and \(E_ n\), in assertions, means \(E^ c_ n)\); 2.3 is trivial; in 2.4, A must be of positive capacity (for instance open), and \(f=1-\phi (g)\)- there is a like confusion in the proofs of 2.5 and 4.1. And so on.\}