## A stochastic evolution equation arising from the fluctuations of a class of interacting particle systems.(English)Zbl 1087.60050

A (finite) measure-valued process $$V$$ in $${\mathbb R}^d$$ is decribed by a stochastic partial differential equation whose weak (and thus rigorous) form is the following:
\begin{align*}{ \langle\phi,V(t)\rangle- \langle\phi,V(0)\rangle &=\int_0^t\langle\phi d(\cdot, V(s))+ L(V(s))\phi,V(s)\rangle ds\cr &\phantom{=\int_0^t}+\int_{U\times[0,t]}\langle\phi\beta(\cdot,V(s),u)+ \nabla\phi^T\alpha(\cdot,V(s),u),V(s)\rangle W(duds),}\end{align*}
where $$L(v)\phi(x)= \frac12 \sum_{i,j} a_{ij}(x,v) \partial_{x_i}\partial_{x_j} \phi(x)+\sum_ib_i(x,v) \partial_{x_i}\phi(x)$$ and $$a=(a_{ij})$$, $$\alpha=(\alpha_i)$$, $$b=(b_i)$$, $$a_{ij}, \alpha_i, b_i$$ ($$1\leq i,j\leq d$$), $$d,\beta$$ are real functions on appropriate spaces, $$U$$ is a Polish space, and $$W$$ is a space-time Gaussian white noise on $$U\times[0,\infty)$$ with covariance measure $$\mu(du)dt$$, $$\mu$$ being a $$\sigma$$-finite measure on $$U$$.
In an earlier paper by the same authors [in: Stochastics in finite and infinite dimensions; 233–258 (2000; Zbl 0991.60053); see also Stochast. Processes Appl. 83, No. 1, 103–126 (1999; Zbl 0996.60071)], it was shown that under suitable assumptions on the coefficients the process $$V$$ can be approximated by the weighted empirical measure process
$V^n(t)= {1\over n}\sum_{k=1}^nA_k^n(t)\delta_{X_k^n(t)},$
of a finite particle system satisfying
\begin{align*}{ X_k^n(t)=&X_k(0)+\int_0^t\sigma(X_k^n(s),V^n(s))dB_k(s)+ \int_0^tc(X_k^n(s),V^n(s))ds\cr &+\int_{U\times[0,t]}\alpha(X_k^n(s),V^n(s),u)W(duds),}\end{align*}
\begin{align*}{ A_k^n(t)=&A_k(0)+\int_0^tA_k^n(s)\gamma^T(X_k^n(s),V^n(s))dB_k(s)+ \int_0^tA_k^n(s)d(X_k^n(s),V^n(s))ds\cr &+ \int_{U\times[0,t]}A_k^n(s)\beta(X_k^n(s),V^n(s),u)W(duds),}\end{align*}
for $$k=1,2,\dots,n$$, where $$B_k$$ are independent standard $${\mathbb R}^d$$-valued Brownian motions, independent of $$W$$, $$(X_k(0),A_k(0))$$ are exchangeable random variables in $${\mathbb R}^d\times{\mathbb R}$$ independent of $$\{B_k\}$$ and $$W$$, and $$\sigma, c$$ and $$\gamma$$ are determined by $$a,b,\alpha,\beta$$.
In the present paper a fluctuation result is obtained. It is proved that $$S_n=\sqrt{n}(V^n-V)$$ converges in law in the space $$C_{\Phi'}[0,\infty)$$, where $$\Phi$$ is a Fréchet nuclear (Schwartz modified) space. The limit $$S$$ is a distribution-valued process which is the unique solution of a stochastic evolution equation
$\langle\phi,S(t)\rangle=\langle\phi,S(0)\rangle +\langle\phi,M(t)\rangle +\int_0^t\langle F_1(V(s))\phi,S(s)\rangle ds +\int_{U\times[0,t]}\langle F_2(V(s),u)\phi,S(s)\rangle W(duds),$
where $$M$$ is a distribution-valued martingale and $$F_1,F_2$$ are linear in $$\phi$$. This interesting but rather involved theorem is related to old results of M. Hitsuda and I. Mitoma [J. Multivariate Anal. 19, 311–328 (1986; Zbl 0604.60059)], where a much simpler setting was investigated and the fluctuation limit was a Gaussian process.

### MSC:

 60H15 Stochastic partial differential equations (aspects of stochastic analysis) 60H35 Computational methods for stochastic equations (aspects of stochastic analysis) 60B12 Limit theorems for vector-valued random variables (infinite-dimensional case) 60F17 Functional limit theorems; invariance principles 60F25 $$L^p$$-limit theorems 60H10 Stochastic ordinary differential equations (aspects of stochastic analysis) 93E11 Filtering in stochastic control theory

### Citations:

Zbl 0991.60053; Zbl 0996.60071; Zbl 0604.60059
Full Text: