仿生資料庫

Database of first principle calculation of TiO₂ absorption spectrum (氧化鈦吸收光譜之第一原理計算參數資料庫)

Introduction

For materials containing transition metal elements such as Ti, the first principle calculation on its outer electron orbital will result in incorrect answer due to the overlapping with O-p orbitals that induced by the strong interaction of the local orbitals. To improve the accuracy of the simulation, DFT-1/2 method is applied to improve the prediction accuracy of absorption spectrum of titanium oxide structure. In the DFT-1/2 method, there are two parameters (cutoff radii and power index ) that can be set to modify the outer orbitals. This database will import the absorption peak distance and related DFT-1/2 parameters calculated by first principles into the machine learning algorithm which combined the neural network algorithm and the genetic algorithm to provide the required DFT-1/2 parameters for ideal adsorption spectrum of TiO₂ structure. The learning process is up to 1,000,000 times, the root mean square (RMS) error is less than 0.0062, and the false rate is less than 0.01.

對於含有過渡金屬元素的材料如TiO₂，當以第一原理模擬鈦原子外層電子軌域時，會因為局部軌域的強交互作用而使其與O-p軌域重疊，導致在計算上的不準確。為了提升模擬精準度，在計算中會以DFT-1/2方法進行修正，使氧化鈦結構之吸收光譜預測與實驗結果相符。於DFT-1/2方法中，可設置cutoff radii及power index兩項參數，來修正外層軌域之結構。 本資料庫將以第一原理計算得到之吸收波峰距離及相關DFT-1/2參數導入機械學習演算法中，結合類神經網路演算法和基因演算法，提供在不同波峰距離下所需設置之DFT-1/2參數數值。學習程序為1,000,000次，均方根（RMS）誤差小於0.0062，誤率則小於0.01。