论文标题
部分可观测时空混沌系统的无模型预测
Multislice Electron Tomography using 4D-STEM
论文作者
论文摘要
储层计算是预测湍流的有力工具，其简单的架构具有处理大型系统的计算效率。然而，其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间，然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用，该系统模拟了湍流的若干特征。我们表明，使用径向基函数作为非线性投影器，即使只有部分观测并且不知道控制方程，也能稳健地捕捉复杂的系统非线性。最后，我们表明，当测量稀疏、不完整且带有噪声，甚至控制方程变得不准确时，我们的网络仍然可以产生相当准确的预测，从而为实际湍流系统的无模型预测铺平了道路。
Electron tomography offers important three-dimensional (3D) structural information which cannot be observed by two-dimensional imaging. By combining annular dark field scanning transmission electron microscopy (ADF-STEM) with aberration correction, the resolution of electron tomography has reached atomic resolution. However, tomography based on ADF-STEM inherently suffers from several issues, including a high electron dose requirement, poor contrast for light elements, and artifacts from image contrast nonlinearity. Here, we developed a new method called MultiSlice Electron Tomography (MSET) based on 4D-STEM tilt series. Our simulations show that multislice-based 3D reconstruction can effectively reduce undesirable reconstruction artifacts from the nonlinear contrast, allowing precise determination of atomic structures with improved sensitivity for low-Z elements, at considerably low electron dose conditions. We expect that the MSET method can be applied to a wide variety of materials, including radiation-sensitive samples and materials containing light elements whose 3D atomic structures have never been fully elucidated due to electron dose limitations or nonlinear imaging contrast.
