Atomic neural networks (ANNs) constitute a class of machine learning methods for predicting potential energy surfaces and physicochemical properties of molecules and materials. Despite many successes, developing interpretable ANN architectures and implementing existing ones efficiently are still challenging. This calls for reliable, general-purpose and open-source codes. Here, we present a python library named PiNN as a solution toward this goal. In PiNN, we implemented an interpretable graph convolutional neural network variant, PiNet, as well as the established Behler-Parrinello high-dimensional neural network potential. These implementations were tested using datasets of isolated small molecules, crystalline materials, liquid water and an aqueous alkaline electrolyte. PiNN comes with a visualizer called PiNNBoard to extract chemical insight "learned" by ANNs, provides analytical stress tensor calculations and interfaces to both the Atomic Simulation Environment and a development version of the Amsterdam Modeling Suite. Moreover, PiNN is highly modularized which makes it useful not only as a standalone package but also as a chain of tools to develop and to implement novel ANNs. The code is distributed under a permissive BSD license and is freely accessible at https://github.com/Teoroo-CMC/PiNN/ with full documentation and tutorials.
Batteries and supercapacitors are electrochemical energy storage systems which involve multiple time‐scales and length‐scales. In terms of the electrolyte which serves as the ionic conductor, a molecular‐level understanding of the corresponding transport phenomena, electrochemical (thermal) stability and interfacial properties is crucial for optimizing the device performance and achieving safety requirements. To this end, atomistic machine learning is a promising technology for bridging microscopic models and macroscopic phenomena. Here, we provide a timely snapshot of recent advances in this area. This includes technical considerations that are particularly relevant for modelling electrolytes as well as specific examples of both bulk electrolytes and associated interfaces. A perspective on methodological challenges and new applications is also discussed.
Response of the electronic density at the electrode-electrolyte interface to the external field (potential) is fundamental in electrochemistry. In density-functional theory, this is captured by the so-called charge response kernel (CRK). Projecting the CRK to its atom-condensed form is an essential step for obtaining the response charge of atoms. In this work, the atom-condensed CRK is learnt from the molecular polarizability using machine learning (ML) models and subsequently used for the response-charge prediction under an external field (potential). As the machine-learnt CRK shows a physical scaling of polarizability over the molecular size and does not (necessarily) require the matrix-inversion operation in practice, this opens up a viable and efficient route for introducing finite-field coupling in the atomistic simulation of electrochemical systems powered by ML models.
Molecular dipole moment in liquid water is an intriguing property, partly due to the fact that there is no unique way to partition the total electron density into individual molecular contributions. The prevailing method to circumvent this problem is to use maximally localized Wannier functions, which perform a unitary transformation of the occupied molecular orbitals by minimizing the spread function of Boys. Here we revisit this problem using a data-driven approach satisfying two physical constraints, namely: (a) The displacement of the atomic charges is proportional to the Berry phase polarization; (b) Each water molecule has a formal charge of zero. It turns out that the distribution of molecular dipole moments in liquid water inferred from latent variables is surprisingly similar to that obtained from maximally localized Wannier functions. Apart from putting a maximum-likelihood footnote to the established method, this work highlights the capability of graph convolution based charge models and the importance of physical constraints on improving the model interpretability.
Electrochemical energy storage always involves the capacitive process. The prevailing electrode model used in the molecular simulation of polarizable electrode–electrolyte systems is the Siepmann–Sprik model developed for perfect metal electrodes. This model has been recently extended to study the metallicity in the electrode by including the Thomas–Fermi screening length. Nevertheless, a further extension to heterogeneous electrode models requires introducing chemical specificity, which does not have any analytical recipes. Here, we address this challenge by integrating the atomistic machine learning code (PiNN) for generating the base charge and response kernel and the classical molecular dynamics code (MetalWalls) dedicated to the modeling of electrochemical systems, and this leads to the development of the PiNNwall interface. Apart from the cases of chemically doped graphene and graphene oxide electrodes as shown in this study, the PiNNwall interface also allows us to probe polarized oxide surfaces in which both the proton charge and the electronic charge can coexist. Therefore, this work opens the door for modeling heterogeneous and complex electrode materials often found in energy storage systems.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.