Strong law of large numbers for supercritical superprocesses under second moment condition (Q2355246)

The aim of the paper under review is to prove that for a large class of suitable functions the strong law of large numbers for supercritical superprocesses under a second moment condition holds. The spatial motion of a supercritical superprocess \(X_t\) is also a Hunt process. For \(f\in L^2(m) \cap L^4(m)\) of the measure \(m\), the authors assume that the semigroup \(\{T_t\}_{t\geq 0}\) is compact, where \(T_tf(x)= P_{\delta_x} \langle f, X_t\rangle\). \textit{Z.-Q. Chen} et al. [J. Funct. Anal. 254, No. 7, 1988--2019 (2008; Zbl 1138.60054)] first established the law of large numbers for a symmetric Hunt process with the smallest condition at ``infinity'' of the linear term in the branching mechanism of the superprocess. The paper under review provides a different setup, which can cover spatial motions. A spatial motion may be discontinuous and non-symmetric. The basic proof follows two steps. The first step is to prove an almost sure limit result along discrete times. The essential difficulty lies in the second step: to prove that the result holds for continuous times. Chen et al. [loc. cit.] found ways to overcome the difficulty of transition from discrete time to continuous times with a highly non-trivial application of the martingale formulation of superprocesses, but the symmetry and the condition at ``infinity'' are too restrictive to extend into spatial motions. Ideas and ingredients from [\textit{M. A. Kouritzin} and \textit{Y.-X. Ren}, Stochastic Processes Appl. 124, No. 1, 505--521 (2014; Zbl 1304.60090)] and [\textit{L. Wang}, J. Theor. Probab. 23, No. 2, 401--416 (2010; Zbl 1198.60037)] are adapted in this paper for the proof of the strong law of large numbers. Section 1 gives an introduction and the basic setup for the paper as well as the definitions of spatial processes, superprocesses and Hunt realizations. The key result is stated in Theorem 2.1 that, under those assumptions, for any finite measure \(\mu\) on \(E\), \(f\in L^2(m) \cap L^4(m)\), \[ \lim_{t\to \infty} e^{-\lambda_0 t}\langle U_qf, X_t\rangle = \langle U_qf, \hat{\phi}_0\rangle_m W_{\infty} \;\;\;P_{\mu}-a.s., \] where \(\lambda_0\) is assumed to be the largest real part of eigenvalues with \(T_t \phi_0(x) = e^{\lambda_0 t}\phi_0 (x)\), \(q>\max\{K, \lambda_0\}\) with \(\sup_{x\in E}(|\alpha (x)|+|A(x)|)\leq K\) in the Hunt process \(X_t\), and \(W_{\infty} =\lim_{t\to \infty}e^{-\lambda_0 t}\langle \phi_0, X_0\rangle\). The main result of this paper is Theorem 1.4, stating that, under some additional assumptions on the semigroups and the bounded Borel function \(f\), \[ \lim_{t\to \infty} e^{-\lambda_0 t}\langle f, X_t\rangle (\omega) = W_{\infty} (\omega) \langle f, \hat{\phi}_0\rangle_m. \] Section 2 states moment estimates from the last three authors' work to get the second moment upper bounds for Hunt processes and recalls that the martingale measure for the superprocess \(X_t\) satisfies \[ \langle U_qf, X_t\rangle = \langle T_t(U_qf), \mu \rangle +e^{qt}\int_t^{\infty} e^{-qu}M_t\, du \] for the cadlag square-integrable martingale \(M_t\) under the measure \(\mu\). Section 3 gives the proofs of the key result Theorem 1.2 and the main result Theorem 1.4. The authors first verify certain upper bounds under the assumptions in Lemma 3.1, and express \(U_qf\) in terms of \(\phi_0\) and its complement to limit the complement term by the Borel-Cantelli lemma for the completion of the proof of the key result. By using the separability of the compact metric space \(E\), there is a countable dense subset in the separable space of bounded continuous functions, and there are related probability measures \(\nu_t, \nu\) on the separable space \(E\). Over the intersection of events needed for the main result to hold, the probability measure \(\nu_t\) converges weakly to \(\nu\). Hence the weak limit holds for the infinity measure \(W_{\infty}\) and Theorem 1.4 follows. Section 4 lists super inward/outward Ornstein-Uhlenbeck processes and pure jump subordinate Brownian motions with or without Gaussian components to illustrate the key result and the main result of the paper.