First-principles prediction of electronic transport in fabricated semiconductor heterostructures via physics-aware machine learning

Abstract

First-principles techniques for electronic transport property prediction have seen rapid progress in recent years. However, it remains a challenge to predict properties of heterostructures incorporating fabrication-dependent variability. Machine-learning (ML) approaches are increasingly being used to accelerate design and discovery of new materials with targeted properties, and extend the applicability of first-principles techniques to larger systems. However, few studies exploited ML techniques to characterize relationships between local atomic structures and global electronic transport coefficients. In this work, we propose an electronic-transport-informatics (ETI) framework that trains on ab initio models of small systems and predicts thermopower of fabricated silicon/germanium heterostructures, matching measured data. We demonstrate application of ML approaches to extract important physics that determines electronic transport in semiconductor heterostructures, and bridge the gap between ab initio accessible models and fabricated systems. We anticipate that ETI framework would have broad applicability to diverse materials classes.

The effect of electron–phonon and electron-impurity scattering on the electronic transport properties of silicon/germanium superlattices

“…36 We illustrated in our previous publications that the layer compositions and external substrate induced strains prompt non-uniform monolayer separations, and modulate the Seebeck coefficients of Si/Ge SLs, within the CRTA. [35][36][37][38] Here, we highlight the effect of EPS and IMS processes on the Seebeck coefficients, resistivities, and power-factors of these Si/Ge SLs in diverse strain environments. We also discuss how these predictions compare with the DOS scattering approximation and CRTA (ESI, † Fig.…”

“…Previous first-principles studies did not include the effect of energy dependent electron relaxation times on the electronic transport properties of SLs. 33,[36][37][38] We also discuss how these predictions change when different scattering approximations, such as constant relaxation time and DOS scattering approximations, are used. We investigate two classes of short-period Si/Ge SLs in diverse lattice strain environments: SLs with varied layer compositions grown on identical substrates and SLs with identical compositions but grown on different substrates.…”

“…Such substrate strain modulated non-monotonic Seebeck coefficients have been observed in other short-period SLs. [35][36][37][38]100 The r of Si 2 Ge 2 SLs with a 3.98% in-plane strain also displays a nonmonotonic behavior, due to the nature of the D 8 and D > bands in the strained SL (Fig. 5(d)).…”

“…The structural variability introduced by non-uniform strain environments, dislocations and other growth dependent parameters is often too difficult to represent in first-principles models due to large computational expenses, making the prediction unreliable. As a consequence, only a few first-principles studies analyzed the electronic structure [29][30][31][32] or transport properties [33][34][35][36][37][38] of even planar Si/Ge SLs. The electronic transport properties of nonpolar semiconductors are significantly affected by electron scattering due to lattice vibrations, ionized impurities, neutral impurities, dislocations, vacancies, and interstitials.…”

“…The role of lattice strain and interface roughness on the electronic properties of p-type Ge/Si 0.5 Ge 0.5 55,56 and n-type SiGe/Si SLs 8,57 has been highlighted by experimental studies. In our recent first-principles studies, [35][36][37][38] we showed that the cross-plane electronic transport properties of n-type[001]-Si/Ge SLs can be tuned by varying the SL layer compositions, periods, and growth substrates. We noted that such nanostructuring strategies positively impact the PF in the high-doping regime.…”