仿生資料庫

Database of Energy and Electronic States in Au_n/WSe₂ Heterostructure (Au_n / WSe₂異質結構能量與電子能態資料庫)

Introduction

WSe₂ monolayer is a two-dimensional material semiconductor with a direct band gap. Due to its tunable and controllable electronic features, it is widely used in electronic and optoelectronic devices. Doping can shape the electronic natures of the material. The adsorption of gold nanoparticles will affect the electronic structure and conductive behavior of WSe₂, the shape, size and coverage of gold nanoparticles on the surface of the material are the keys to modifying the function of two-dimensional materials. The adsorption of gold nanoparticles of various sizes also provides different adsorption distance and adsorption energy for WSe₂.  In this database, gold nanoparticles (n = 12-70) at different adsorption distances (d = 4-10Å) are calculated by first-principles calculation combined with machining learning algorithms to predict the adsorption energy values and density of states at the Fermi level of the extended Au_n / WSe₂ system properties, where n is the number of atoms of gold nanoparticles. Mechanical learning is achieved through neural networks and genetic algorithms. The learning process is 10,000 times, the root mean square (RMS) error is less than 0.01, and the error rate is less than 1%.

單層WSe₂為直接能隙的二維材料半導體，由於其可調整和控制的電子特性在電子和光電設備中的應用極為廣泛，而摻雜可以改變材料的電子性質，金奈米粒子的吸附會影響WSe₂的電子結構與導電行為，金奈米粒子在材料表面上的形狀、大小和覆蓋率是修飾二維材料功能的關鍵，各種尺寸的金奈米粒子的吸附對於WSe₂也提供不同的吸附距離和吸附能。在本資料庫中，金奈米粒子(n=12-70)於不同的吸附距離(d=4-10Å)透過第一原理計算結合機械學習演算法預測擴展Au_n/WSe₂系統的吸附能和電子性質，其中n是金奈米粒子的原子數。機械學習則透過類神經網絡與基因演算法得以實現，其學習迴圈為10,000次，均方根（RMS）誤差小於0.01，誤率則低於1%。