Attractive polymer models for two- and three-dimensional Brownian motion (Q1382514): Difference between revisions

Let us denote by \(\mathcal C\) the space of all continuous functions \(\omega : [0,1]\to \mathbb R^{n}\) with \(\omega (0)=0\). A polymer measure \(\nu _{g}\) on \(\mathcal C\) is defined by the density \(L^{-1}\exp (gJ(\omega ))\) with respect to the Wiener measure \(\mu \) on \(\mathcal C\). Here \(L\) is an appropriate norming factor and \(J(\omega ) = \int ^{1}_{0}\int ^{1} _{s}\delta (\omega (s)-\omega (t)) dt ds\) is the self-intersection local time of the Brownian motion. This definition of \(J\) (and hence of \(\nu _{g}\)) is purely formal, but it was shown by \textit{J. Rosen} [Commun. Math. Phys. 88, 327-338 (1983; Zbl 0534.60070)] that for \(\varepsilon >0\) the functional \(J^\varepsilon (\omega ) = \int ^{1-\varepsilon }_{0}\int ^{1}_{s+\varepsilon }\delta (\omega (s)-\omega (t)) dt ds\) may be given a good sense provided \(n\leq 3\). Hence the measures \(d\nu ^{\varepsilon }_{g} (\omega ) = \exp (gJ^\varepsilon (\omega )) d\mu (\omega )\) are well defined and the problem is to establish that \(\lim _{\varepsilon \to 0}\nu ^\varepsilon _{g}\) exists. In the case of the self-avoiding Brownian motion, which corresponds to non-positive coupling constants \(g\leq 0\), this was done by \textit{S. R. S. Varadhan} for \(n=2\) and by \textit{M. J. Westwater} [ibid. 72, 131-174 (1980; Zbl 0431.60100)] for \(n=3\). In the self-attractive case, that is, for \(g>0\), much less is known. For the planar Brownian motion, \textit{J.-F. Le Gall} [Séminaire de Probabilités XXVIII. Lect. Notes Math. 1583, 172-180 (1994; Zbl 0810.60078)] found that \(\mathbf E\exp (\lambda Y)< \infty \) if \(\lambda <\lambda _{0}\) for a certain constant \(\lambda _{0}\in [1,\infty [\), where \(Y=\lim _{\varepsilon \to 0} (J^\varepsilon -\mathbf E J^\varepsilon )\). The authors propose a different renormalization which should work also for \(n=3\) and for any \(g\in \mathbb R\). They define measures \[ d\nu ^{m,\varepsilon }_{g}(\omega ) = \frac {\sum ^{m}_{i=0} \frac {g^{i}}{i!}(J^\varepsilon )^{i}} {\mathbf E(\sum ^{m}_{i=0}\frac {g^{i}}{i!}(J^\varepsilon ) ^{i})} d\mu (\omega ) \] and they prove that for \(n=2,3\) and any \(g>0\) the family of measures \(\{\nu ^{m,\varepsilon }_{g}\}\) is tight. Moreover, they show that for any fixed \(g>0\) and \(m\) the measure \(\nu ^{m,\varepsilon }_{g}\) converges to the Wiener measure \(\mu \) as \(\varepsilon \to 0\). They conjecture, however, that for \(m=\varepsilon ^{-\gamma }\) with \(\gamma \) large enough the limit of \(\nu ^{m,\varepsilon }_{g}\) differs from \(\mu \).