Tightness property for symmetric diffusion processes (Q1117592)

Consider a sequence \({\mathcal E}^ n\), \(n\in {\mathbb{N}}\), of closable forms on \(L^ 2({\mathbb{R}}^ d;m_ n)\) of the following type \[ {\mathcal E}^ n(f,g)=2^{-1}\sum^{d}_{i,j=1}\int_{{\mathbb{R}}^ d}(\partial f/\partial x_ i)(x)(\partial g/\partial x_ j)(x)a^ n_{ij}(x)m_ n(dx),\quad f,g\in D({\mathcal E}^ n):=C_ 0^{\infty}({\mathbb{R}}^ d), \] where \(m_ n\) are positive Radon measures on \(R^ d\) with supp \(m_ n={\mathbb{R}}^ d\) and \((a^ n_{ij}(x))_{i,j}\), \(x\in {\mathbb{R}}^ d\), are symmetric, non-negative matrices with greatest eigenvalues uniformly bounded in x and n by some constant \(c>0\). Let \(M^ n=(P_ x^ n,X_ t)\) be the symmetric diffusion processes associated with the closures of (\({\mathcal E}^ n,D({\mathcal E}^ n))\) in the sense of \textit{M. Fukushima} [Dirichlet forms and Markov processes. (1980; Zbl 0422.31007)]. For a probability measure \(\mu\) on \({\mathbb{R}}^ d\) define a probability measure \(P^ n_{\mu}(\cdot):=\int P^ n_ x(\cdot)\mu (dx)\) on \(C([0,\infty),{\mathbb{R}}^ d)\), i.e. the space of all continuous functions on \(R^ d\) equipped with the topology of locally uniform convergence. The main theorem of the paper states that if all diffusions \(M^ n\) are conservative, \(\sup_{n\in {\mathbb{N}}}m_ n(K)<\infty\) for any compact set \(K\subset {\mathbb{R}}^ d\), \(\mu_ n=\Phi_ ndm_ n\) with \(\sup_{n\in {\mathbb{N}}} \| \Phi_ n\|_{\infty}<\infty\), \(\{\mu_ n\}\) is tight and \[ \sup_{n\in {\mathbb{N}}}\sum^{\infty}_{k=0}m_ n(T_{R+k})\ell^{1/2}[k/\sqrt{dcT}]<\infty \quad for\quad any\quad T>0,\quad R>0, \] where \(T_ p:=\{x\in {\mathbb{R}}^ d:p\leq | x| <p+1\}\) and \(\ell (a)=(2\pi)^{-1/2}\int^{\infty}_{a}e^{-u^ 2/2}du\), then the sequence \(\{P^ n_{\mu_ n}\}_{n\in {\mathbb{N}}}\) is tight on \(C([0,\infty),{\mathbb{R}}^ d).\) The proof is partly based on a result in \textit{T. Lyons} and \textit{W. Zheng}, Les processus stochastiques, Coll. Paul Lévy, Palaiseau/Fr. 1987, Asterisque 157-158, 249-271 (1988), implying that the component processes \((X^ i_ t-X^ i_ 0)_{t\geq 0}\), \(1\leq i\leq d\), can be written as the sum of a forward and a backward martingale.