Construction of an Edwards' probability measure on \(\mathcal C(\mathbb R_+, \mathbb R)\) (Q606634): Difference between revisions

The Edwards measure is a transform of the \(d\)-dimensional Brownian motion suppressing self-inter\-sec\-tions. It would like to have the density, on the finite time interval \([0,T]\), equal to \(\exp(-\beta H_t)/Z_{T,\beta}\), where \(H_T\) is the self-intersection local time and \(Z_{T,\beta}\) the normalisation constant, and \(\beta\in(0,\infty)\) is a parameter. This measure is interesting only in dimensions \(\leq 3\). A rigorous definition is difficult (but possible) in dimensions two and three, which is by now classical, and it is simple in terms of the local times in dimension one. Using Ray-Knight theorems, in earlier work by \textit{R. van der Hofstad, F. den Hollander} and \textit{W. König} [Ann. Probab. 25, No.~2, 573--597 (1997; Zbl 0873.60009); Ann. Probab. 31, No.~4, 2003--2039 (2003; Zbl 1052.60019)], a central limit theorem and a large-deviation principle for the endpoint of the path (i.e., at time \(T\)) was proved. The present paper works in dimension one; its main result is the construction of the Edwards measure on the entire time interval \([0,\infty)\), for any \(\beta\in(0,\infty)\). More precisely, it is shown that there is a probability measure on paths \([0,\infty)\to\mathbb R\) such that the above measure converges towards this measure as \(T\to\infty\) on paths of any finite length. The construction is in terms of a martingale density with respect to the free Brownian motion with an explicit martingale; it follows ideas of Roynette, Vallois and Yor about constructing penalized and/or perturbed Brownian motions. The main approach uses Ray-Knight theorems; the martingale is constructed in terms of Girsanov transforms of two-and zero-dimensional squared Bessel processes. The author gives conjectures about the large-time behaviour of the path under the infinite-time Edwards measure constructed by him -- of course, the behaviour should be as for the finite-time Edwards measure as revealed by van der Hofstad, den Hollander and König [loc. cit.].