Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

---

In this article, the energy gap of 300 nitride compounds was analyzed and investigated using quantum simulations based on Density Functional Theory (DFT). The energy gaps of the compounds were calculated using two approximations, GGA-PBE and HSE06. The parameters considered in the machine learning studies were categorized into two groups: atomic and crystalline parameters. The atomic parameters include covalent radius, electronegativity, the number of valence electrons, and the first ionization energy. After collecting data on atomic features, a multiple linear regression model was fitted to the data. Subsequently, variable selection was performed using the stepwise regression method with the AICc criterion, and the effect size of various features was calculated. Additionally, to improve the accuracy of the model in predicting HSE06, three crystal features were incorporated into the model, and eight different models were fitted, each including one or more of these crystal features. The findings indicate that the model without any crystal variables has an adjusted coefficient of determination (R²) of 75.45%. However, with the inclusion of crystal features, the results improve significantly. Specifically, adding the PBE energy gap as a variable increases the R² to 99.03% (from 75.45% to 99.03%).

Keywords: Energy Gap, Density Functional Theory, Semiconductor, Machine Learning, Linear Regression

Full-Text [PDF 1453 kb]   (80 Downloads)    

Add your comments about this article : Your username or Email:

---

---