论文标题
部分可观测时空混沌系统的无模型预测
High throughput data-driven design of laser crystallized 2D MoS2 chemical sensors
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。
High throughput characterization and processing techniques are becoming increasingly necessary to navigate multivariable, data-driven design challenges for sensors and electronic devices. For two-dimensional materials, device performance is highly dependent upon a vast array of material properties including number of layers, lattice strain, carrier concentration, defect density, and grain structure. In this work, laser-crystallization was used to locally pattern and transform hundreds of regions of amorphous MoS2 thin films into 2D 2H-MoS2. A high throughput Raman spectroscopy approach was subsequently used to assess the process-dependent structural and compositional variations for each illuminated region, yielding over 5500 distinct non-resonant, resonant, and polarized Raman spectra. The rapid generation of a comprehensive library of structural and compositional data elucidated important trends between structure-property-processing relationships involving laser-crystallized MoS2, including the relationships between grain size, grain orientation, and intrinsic strain. Moreover, extensive analysis of structure/property relationships allowed for intelligent design, and evaluation of major contributions to, device performance in MoS2 chemical sensors. In particular, it is found that sensor performance is strongly dependent on the orientation of the MoS2 grains relative to the crystal plane.