Small ball probability estimates for the Hölder semi-norm of the stochastic heat equation (Q6085099): Difference between revisions

The aim of the work is to compute upper and lower bounds for the probabilities of the existence of the solution of the stochastic heat equation in Hölder semi-norm small balls. More precisely, the authors considered the problem: \[ \left\{ \begin{array}{lr} \partial_t u(t, x)= \frac12 \partial_x^2u(t, x) + \sigma(t, x, u(t, x)). \dot{W}(t, x), \;\; t\in [0, T] \subset \mathbb{R}_+, \; x\in \mathbb{T}:= \mathbb{R}/\mathbb{Z}, \\ u(t, 0)=u(t, 1), \; t\in [0, T] \subset \mathbb{R}_+,\\ u(0, .)=u_0(x), x\in \mathbb{T}, \end{array} \right.\tag{1} \] where \(\dot{W} \) is a space-time white noise, \(\sigma\) is globally Lipschitz with respect to the third variable with constant \(\mathcal{D}\) and uniformly elliptic with lower respectively upper constants \( \mathcal{C}_1, \; \mathcal{C}_2\). The main result is quoted as: Let \(0<\theta <\frac12\) and \(0<\epsilon \leq \frac12\). Suppose that \( \mathcal{H}_t^{(\theta)}(u)\backslash_{t=0}:= \sup_{x\neq y}\frac{|u(t,x)- u(t,y)|}{|x-y|^{\frac12-\theta}}\backslash_{t=0} \leq \frac{\epsilon}{2} (1\wedge (\pi^{1/2} 2^{\theta-1/2}\Gamma^{-1}(1-\theta))\), with \(\Gamma\) being the Gamma function. Then for any \(\eta>0\), there exist positive constants \(C_1, C_2, C_3, C_4 >0\) dependent on \( \mathcal{C}_1, \; \mathcal{C}_2, \; \mathcal{D}, \; \theta, \; \eta\), such that \[ C_1e^{-C_2 T \epsilon^{-\eta-3/\theta}} \leq P\left( \sup_{(t,x)\neq(s,y) \in [0, T]\times \mathbb{T}}\frac{|u(t,x)- u(s,y)|}{|t-s|^{\frac14-\frac\theta2}+|x-y|^{\frac12-\theta}}\leq \epsilon\right) \leq C_3e^{-C_4 T (\epsilon^{1/\theta} |\log\epsilon|^{1/2})^{-3}}.\tag{2} \] The proof is built via two categories of approaches. One approach is to study first the additive case, i.e. (\(\sigma\) is constant), where the Gaussian theory is applied, and then to consider the general case as a perturbation of the additive one. The other approach is to establish intermediary estimates for the probabilities on small balls for \( \mathit{H}_x^{(\theta)}(u):= \sup_{0\leq t\neq s\leq T}\frac{|u(t,x)- u(s,x)|}{|t-s|^{\frac14-\frac\theta2}}\) and \( \mathcal{H}_t^{(\theta)}(u)\), such as the result quoted below: Under the same hypothesis above and with \( \mathcal{H}_0^{(\theta)}(u)\leq \frac{\epsilon}{2}\), then there is a \( \mathcal{D}_0>0\) such that for all \( \mathcal{D}< \mathcal{D}_0\), there exist positive constants \(C_1, C_2, C_3, C_4 >0\) dependent only on \( \mathcal{C}_1, \; \mathcal{C}_2, \; \theta\), such that: \[ C_1e^{-C_2 T \epsilon^{-3/\theta}} \leq P\left( \sup_{0\leq t\leq T} \mathcal{H}_t^{(\theta)}(u)\leq \epsilon\right) \leq C_3e^{-C_4 T (\epsilon^{1/\theta} |\log\epsilon|^{1/2})^{-3}}. \tag{3} \] The reviewer is confused about the coherence of some estimates, such as (2) and (3), because for \(\epsilon\) close to \(1\) and the constants \(C_1, C_2, C_3, C_4\) don't dependent on \(\epsilon\), the fact that \( C_1e^{-C_2 T \epsilon^{-3/\theta}} \leq C_3e^{-C_4 T (\epsilon^{1/\theta} |\log\epsilon|^{1/2})^{-3}}\) implies \( e^{C_4 T (\epsilon^{1/\theta} |\log\epsilon|^{1/2})^{-3}-C_2 T \epsilon^{-3/\theta}} \leq \frac{ C_3}{ C_1}. \)