To facilitate chemical space exploration for material screening or to accelerate computationally expensive first-principles simulations, inexpensive surrogate models that capture electronic, atomistic, or macroscopic materials properties have become an increasingly popular tool over the last decade. The most fundamental quantity common across all such machine learning (ML)-based methods is the f ingerprint used to numerically represent a material or its structure. To increase the learning capability of the ML methods, the common practice is to construct fingerprints that satisfy the same symmetry relations as displayed by the target material property of interest (for which the ML model is being developed). Thus, in this work, we present a general, simple, and elegant fingerprint that can be used to learn different electronic/atomistic/structural properties, irrespective of their scalar, vector, or tensorial nature. This fingerprint is based on the concept of multipole terms and can be systematically increased in sophistication to achieve a desired level of accuracy. Using the examples of Al, C, and hafnia (HfO 2 ), we demonstrate the applicability of this fingerprint to easily classify different atomistic environments, such as phases, surfaces, point defects, and so forth. Furthermore, we demonstrate the generality and effectiveness of this fingerprint by building an accurate, yet inexpensive, ML-based potential energy model for the case of Al using a reference data set that is obtained from density functional theory computations. Finally, we note that the fingerprint definition presented here has applications in fields beyond materials informatics, such as structure prediction, identification of defects, and detection of new crystal phases.

One of the key bottlenecks in the development of high voltage electrical systems is the identification of suitable insulating materials capable of supporting high voltages. Under high voltage scenarios, conventional polymer based insulators, which are one of the popular choices of insulators, suffer from the drawback of space charge accumulation, which leads to degradation in desirable electronic properties and facilitates dielectric breakdown. In this work, we aid the development of novel polymers for high voltage insulation applications by enabling the rapid prediction of properties that are correlated with dielectric breakdown, i.e.,the bandgap (Egap) of the polymer and electron injection barrier (Φe) at the electrode–insulator interface. To accomplish this, density functional theory based methods are used to develop large, chemically diverse datasets of Φe and Egap. The deviation of the computed properties from experimental observations is addressed using a statistical technique called Bayesian calibration. Furthermore, to enable rapid estimation of these properties for a large set of polymers, machine learning models are developed using the created dataset. These models are further used to predict Egap and Φe for a set of 13k previously known polymers. Polymers with high values of these properties are selected as potential high voltage insulators and are recommended for synthesis. Finally, the models developed here are deployed at www.polymergenome.org to enable the community use.

Dielectric polymers are widely used in electronics and energy technologies, but their performance is severely limited by the electrical breakdown under a high electric field. Dielectric breakdown is commonly understood as an avalanche of processes such as carrier multiplication and defect generation that are triggered by field-accelerated hot electrons and holes. However, how these processes are initiated remains elusive. Here, nonadiabatic quantum molecular dynamics simulations reveal microscopic processes induced by hot electrons and holes in a slab of an archetypal dielectric polymer, polyethylene, under an electric field of 600 MV/m. We found that electronic-excitation energy is rapidly dissipated within tens of femtoseconds because of strong electron–phonon scattering, which is consistent with quantum-mechanical perturbation calculations. This in turn excites other electron–hole pairs to cause carrier multiplication. We also found that the key to chemical damage is localization of holes that travel to a slab surface and weaken carbon–carbon bonds on the surface. Such quantitative information can be incorporated into first-principles-informed, predictive modeling of dielectric breakdown.

Doping conjugated polymers, which are potential candidates for the next generation of organic electronics, is an effective strategy for manipulating their electrical conductivity. However, selecting a suitable polymer–dopant combination is exceptionally challenging because of the vastness of the chemical, configurational, and morphological spaces one needs to search. In this work, high-performance surrogate models, trained on available experimentally measured data, are developed to predict the p-type electrical conductivity and are used to screen a large candidate hypothetical data set of more than 800 000 polymer–dopant combinations. Promising candidates are identified for synthesis and device fabrication. Additionally, new design guidelines are extracted that verify and extend knowledge on important molecular fragments that correlate to high conductivity. Conductivity prediction models are also deployed at for broader open-access community use.