Tagged particles of interacting Brownian motions with skew symmetric drifts (Q2732330)

Let \(d \geq 2\) and let \(\Theta\) be the set of all finite or infinite ``configurations'' \(\theta= \sum \varepsilon_{x_{i}}\) where \(x_{i} = (x_{i,m}) \in {\mathbb R}^{d}\) have no cluster points in \({\mathbb R}^{d}\), let \(\mu\) be a grand canonical Gibbs measure on \(\Theta\) with a pair interaction potential \(\Phi\), \(\mu\) being translation invariant and having locally bounded density and 1-correlation functions. Let \({\mathcal D}_{\infty}\) be the set of all smooth local functions on \(\Theta\), \(\nabla_{i}\) be the gradient with respect to the \(i\)-component and, for \(f,g \in {\mathcal D}_{\infty}\), \({\mathbb D}(f,g) : \Theta \rightarrow {\mathbb R}\) be defined as \((1/2) \sum_{i}\langle\nabla_{i} f,\nabla_{i} g\rangle_{{\mathbb R}^{d}}\). Suppose that \(({\mathcal E}^{\mu},{\mathcal D}_{\infty})\), where \({\mathcal E}^{\mu}(f,g) = \int D(f,g) d\mu\), is closable on \(L^{2}(\mu)\). Let \(\Psi\) be a skew symmetric \(d \times d\)-matrix valued function on \({\mathbb R}^{d}\). If \({\mathcal E} = {\mathcal E}^{\mu} + {\mathcal E}^{\Psi}\), where \({\mathcal E}^{\Psi}(f,g) = \int {\mathcal D}^{\Psi}(f,g) d \mu\), NEWLINE\[NEWLINE{\mathcal D}^{\Psi}(f,g) = \sum_{i \neq j} \sum_{m,n = 1}^{d} \Psi_{m,n}(x^{i}-x^{j})(\partial f/\partial x_{i,m})(\partial g/\partial x_{j,n}),NEWLINE\]NEWLINE satisfies \({\mathcal E}(f,g) \leq C {\mathcal E}(f,f)^{1/2} {\mathcal E}(g,g)^{1/2}\), it follows that \(({\mathcal E},{\mathcal D}_{\infty})\) is closable; let \(P_{\theta}^{\mu,\Psi}\) be the diffusion corresponding to its closure. Suppose that \(\sum_{m}(- \partial_{m} \Psi_{m,n} +(\partial_{m} \Phi) \Psi_{m,n})\) is in \(L^{\infty}(\mu_{0})\), where \(\mu_{x}\) is \(\mu\)-conditioned by \((\cdot(\{ x \}) = 1)\) and also that \({\mathcal E}^{^{\mu} 0} + {\mathcal E}^{\Psi}\) satisfies the inequality for all \(f,g \in {\mathcal D}_{\text{loc}}\). If the diffusion \(P_{\theta}^{\mu,\Psi}\) starts from a \(\theta\) with \(\theta(\{ x \}) = 1\), i.e. with a particle in \(x\), and if the diffusion never charges a \( \{ y \} \) with more than \(1\) (this happens for \(\mu_{x} = \mu_{0} \circ \tau_{x}^{- 1}\)-almost all \(\theta\), where \(\tau_{x}\) is the \(x\)-translation), then the particle in \(x\) has a trajectory \(X_{t}^{i(x,\theta)}\) under the diffusion; consider the distribution \(P_{\theta}^{\mu,\Psi,\varepsilon}\) of \(\varepsilon X_{\cdot \varepsilon^{- 2}}^{i(x,\theta)}\). NEWLINENEWLINENEWLINEThe main result is that, for \(\varepsilon \rightarrow 0\), it tends weakly on \(C([ 0,\infty),{\mathbb R}^{d})\) to the distribution \(P_{\theta}^{ \sim}\) of a continuous martingale \(((X_{t}^{ \sim i})_{i = 1,\dots,d})\), \(t\geq 0\), \(X_{0}^{ \sim} = 0\), defined by \(\langle X^{ \sim i},X^{ \sim j}\rangle_{t} = 2a_{i j}^{ \sim}(\tau_{x} \theta) t\) and this in \(\mu_{x}\)-measure in \(\theta\), where for the explicit expression of the \(a^{ \sim}\) and for other important details the reader is sent to the author's paper [Ann. Inst. Henri Poincaré, Probab. Stat. 34, No. 2, 217-248 (1998; Zbl 0914.60041)], the present result improving an analogous one there. The result is deduced from a more general one, in which, starting with a quasi-regular Dirichlet form \(({\mathcal E}_{Y},{\mathcal D}_{Y})\) on a general \(L^{2}(\Theta,\mu)\), with \({\mathcal E}_{Y} = {\mathcal E}_{Y}^{1}+{\mathcal E}_{Y}^{2}\), \({\mathcal E}_{Y}^{1}(g,g) = \int \sum_{i,j = 1}^{d} a_{ij}(D_{i} g)(D_{j} g) d\mu\), \(D_{i}\) being linear operators with \(D_{i}\)1= 1, \((a_{ij})\) being uniformly elliptic from \(\Theta\) to \(d \times d\)-matrices, considering NEWLINE\[NEWLINE{\mathcal E}_{X Y}(h,h) =\int \sum a_{ij}(x)(D_{i}-\partial_{i}) h(x)(D_{j}-\partial_{j}) h(x) dx d \mu + \sum \Bigl(\int f_{i}(x) f_{j}(x) dx\Bigr){\mathcal E}_{Y}^{2}(g_{i},g_{j})NEWLINE\]NEWLINE for \(h = \sum f_{i} \otimes g_{i} \in {\mathcal C}_{0}^{\infty}({\mathbb R}^{d}) \otimes {\mathcal D}_{Y}\) and \(\partial_{i} = \partial/\partial x_{i}\), the corresponding process \((X_{t},Y_{t})\), the distribution \(P_{\theta}^{\varepsilon}\) of \(\varepsilon X_{t \varepsilon^{- 2}}-\varepsilon x\) when starting from \((x,\theta)\) (which does not depend on \(x\)) and the distribution \(P_{\theta}^{ \sim}\) of a continuous martingale \((X_{t}^{ \sim i})\) with \(\langle X^{ \sim i},X^{ \sim j}\rangle_{t} = 2 a_{i j}^{ \sim}(\theta) t\), \(X_{0}^{ \sim} = 0\), the author shows, under analogous conditions, that \(\lim_{\varepsilon \rightarrow 0} P_{\theta}^{\varepsilon} = P_{\theta}^{ \sim}\) weakly. Examples are given, named ``hard core Brownian balls'', ``Lennard-Jones 6-12 potentials with hard core'' and ``vortices with hard core''.