Hairer's reconstruction theorem without regularity structures (Q2031493): Difference between revisions

Let \(\gamma\in\mathbb R\), let \(F:\mathbb R^d\to\mathscr D^\prime(\mathbb R^d)\) be a mapping such that \(x\mapsto F_x(\psi)\) is measurable for every \(\psi\in\mathscr D(\mathbb R^d)\), let there exist \(\varphi\in\mathscr D(\mathbb R^d)\) such that \(\int_{\mathbb R^d}\varphi\ne 0\) and \[ |(F_z-F_y)(\varphi^\varepsilon_y)|\lesssim\varepsilon^{\alpha_K}(|z-y|+\varepsilon)^{\gamma-\alpha_K},\qquad|F_x(\varphi_x^\varepsilon)|\lesssim\varepsilon^{\beta_K} \] hold uniformly for \(z,y,x\in K\) and \(\varepsilon\in(0,1]\) for every compact \(K\) in \(\mathbb R^d\), where \(\alpha_K\) and \(\beta_K\) are some real numbers satisfying \(\alpha_K\le\min\,\{0,\gamma\}\) and \(\beta_K<\gamma\), and \[ \varphi_y^\varepsilon(z):=\varepsilon^{-d}\varphi(\varepsilon^{-1}(z-y)). \] The main result of the paper asserts that, under the above assumptions, there exists a distribution \(f\in\mathscr D^\prime(\mathbb R^d)\) such that, for every compact \(K\) in \(\mathbb R^d\) and any integer \(r>\max\,\{-\alpha_{\overline{K}_2},-\beta_{\overline K_2}\}\), for \(\alpha=\alpha_{\overline K_2}\) and \(K_2:=K+\overline{B}(0,2)\), the estimate \[ |(f-F_x)(\psi^\lambda_x)|\le C\cdot\begin{cases}\lambda^\gamma&\text{if }\gamma\ne 0,\\ (1+|\log\lambda|)&\text{if }\gamma=0\end{cases} \] holds uniformly for \(x\in K\), \(\lambda\in(0,1]\) and \(\psi\in\{\phi\in\mathscr D(B(0,1)):\|\phi\|_{C^r}\le 1\}\), where \(C\) depends on \(\alpha\), \(\gamma\), \(r\), \(d\), \(\varphi\) and \(K\). If \(\gamma>0\), then such \(f\) is unique and the dependence \(F\mapsto f\) is linear. If \(\gamma\le 0\), then \(f\) is not unique but one can choose \(f\) in such a way that the dependence \(F\mapsto f\) is linear. This result is subsequently used to characterize the negative Hölder spaces and applied to the theory of multi-dimensional Young integration where a product between Hölder functions and Hölder distributions is defined.