The constants in the central limit theorem for the one-dimensional Edwards model (Q1284851): Difference between revisions

The author derives rigorous bounds for certain constants describing the central limit theorem for the following model from statistical mechanics. Fix \(T>0\) and let \((L(T,x))_{x\in\mathbb R}\) denote the local times of one-dimensional Brownian motion \((B_t)_{t\in[0,T]}\). Define a transformed path measure \(\hat P_T\) by means of the Radon-Nikodym derivative \(\exp\{-\int_{\mathbb R}L(T,x)^2 dx\}/Z_T\). This is the so-called Edwards model, which is a model for a polymer chain as it suppresses self-intersections. It was proved by \textit{R. van der Hofstad, F. den Hollander} and \textit{W. König} [Ann. Probab. 25, No. 2, 573-597 (1997; Zbl 0873.60009)] that, as \(T\to\infty\), the normalization constant behaves like \(Z_T=e^{-T(a^*+o(1))}\), and that the normalized endpoint, \(\frac{|B_T|-b^*T}{c^* \sqrt T}\), converges in \(\hat P_T\)-distribution towards a standard normal variable. Here \(a^*\), \(b^*\) and \(c^*\) are constants that are characterized in terms of the largest eigenvalue of a certain one-parameter family of Sturm-Liouville differential operators on \(L^2(\mathbb R^+)\). (These constants also appear in scaling limits for the mean and variance in an analogous central limit theorem for the analogous discrete version of the Edwards model as the strength of the self-repellence decreases to zero.) Roughly speaking, \(a^*\) is characterized as the unique constant \(a\in\mathbb R\) that makes a certain variational formula (depending on \(a\)) equal to zero, and \(b^*\) and \(c^*\) appear as the values of related variational problems with additional conditions resp.\ are obtained by differentiating w.r.t.\ \(a\) at \(a^*\). The results of the paper are the rigorous numerical bounds \(2.188\leq a^*\leq 2.189\), \(1.104\leq b^*\leq 1.124\) and \(0.60\leq c^*\leq 0.66\) which are derived by means of power series approximations, spectral analysis and differential inequalities. Part of the proof is computer-assisted. As an interesting feature of the result on \(c^*\), the asymptotic variance in the above central limit theorem turns out to be strictly smaller than one (the value for free Brownian motion).