No Thumbnail Available
Files
Trinquet_28021700_2022.pdf
Open access - Adobe PDF
- 33.81 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Nonlinear optical materials have become vital components in modern optoelectronics. However, various regions of the electromagnetic spectrum are still lacking appropriate compounds. In an attempt to guide experimental research, this work investigates the second-harmonic generation phenomenon from both ab initio and machine learning approach. The first-principles computations are performed in the framework of characterizing six promising crystals for use in the mid-infrared region: MgSi(-Ge,-Sn)P2(As2). Both MgGeAs2 and MgSnAs2 are shown to display an important second-harmonic generation with the possibility of realizing angular phase matching. Moreover, the six materials constitute an ideal system to inspect the relations between the atomic arrangement, the electronic properties, and the second-harmonic generation response thanks to modifications of their crystalline structure and to the existence of binary analogs. This richness of possibilities allows to explore the interesting concept of quantifying the non-centrosymmetry and its influence on the second-harmonic generation and on Miller's coefficient. It is found that the latter seems to be more impacted by this "degree of centrosymmetry" than the second-harmonic response is. This coefficient relates linear to nonlinear optics by a simple relation which is termed Miller's rule. Its study is thus primordial since it holds the potential of greatly simplifying the determination of nonlinear optical properties. This last ambition is shared by a data-driven approach, made possible thanks to the recent emergence of a database with the relevant features. The present essay thus explores the possibility of predicting the second-harmonic generation response by a materials-oriented machine learning model, MODNet. Although the final results are mitigated, it paves the way for future improvements.