Project

Back to overview

Fourth-Generation Neural Network Potentials for Molecular Chemistry

English title Fourth-Generation Neural Network Potentials for Molecular Chemistry
Applicant Goedecker Stefan
Number 206936
Funding scheme Project funding
Research institution Departement Physik Universität Basel
Institution of higher education University of Basel - BS
Main discipline Physical Chemistry
Start/End 01.04.2022 - 31.03.2025
Approved amount 356'933.00
Show all

All Disciplines (2)

Discipline
Physical Chemistry
Condensed Matter Physics

Keywords (2)

machine learning; solvation effects

Lay Summary (German)

Lead
In dem Projekt werden neue auf maschinenellem Lernen basierende Methoden entwickelt, die es erlauben atomistische Simulationen in der Chemie wesentlich schneller durchzufuehren.
Lay summary


Maschinegelernte Potenziale (MLPs) haben sich zu einem wichtigen Werkzeug für die Durchführung atomistischer Simulationen mit der Genauigkeit elektronischer Strukturmethoden entwickelt, jedoch zu einem Bruchteil der Rechenkosten. Bisher wurden die meisten Anwendungen in den Materialwissenschaften gemacht, während organische Moleküle in erster Linie im Vakuum untersucht wurden. Obwohl die meisten chemischen Reaktionen in der flüssigen Phase ablaufen, sind Anwendungen von MLPs auf Solvatation und Molekularchemie in Lösung noch sehr selten. Abgesehen von der Komplexität des beteiligten Konfigurationsraums besteht eine große Herausforderung bei der Untersuchung dieser Systeme in der Notwendigkeit einer äußerst
genauen Beschreibung der intra- und intermolekularen Wechselwirkungen, von starken kovalenten
Bindungen über Wasserstoffbrückenbindungen bis hin zu elektrostatischen und Dispersionswechselwirkungen. Ein besonders wichtiger Aspekt ist die Ladungsverteilung in den beteiligten Spezies, die von den meisten aktuellen MLPs, die auf lokalen Eigenschaften wie umgebungsabhängigen Atomenergien basieren, nicht korrekt erfasst werden kann.

Ziel dieses Projekts ist es, die Anwendbarkeit von neuen MLP's auf die Molekularchemie in Lösung zu erforschen, indem es sich auf zwei Hauptaspekte konzentriert: die Qualität der Referenzdaten der Dichtefunktionaltheorie (DFT) und die Verallgemeinerung der 4G-HDNNP-Methode des maschinellen Lernens.

Die entwicklten Methoden koennen helfen viele gesellschaftlich relevante Probleme z.B. auf dem Gebiet der Katalyse zu erforschen.


Direct link to Lay Summary Last update: 04.02.2022

Responsible applicant and co-applicants

Employees

Name Institute

Associated projects

Number Title Start Funding scheme
182877 Development of a Neural Network Potential with Accurate Electrostatic Interactions 01.04.2019 Project funding

Abstract

Machine learning potentials (MLP) have become an important tool for performing atomistic simulations with the accuracy of electronic structure methods but at a small fractionof the computational cost. To date, most applications have been reported in materials science, while organic molecules have been primarily studied for benchmark purposes in vacuum. Although most chemical reactions occur in the liquid phase, applications of MLPs to solvation and molecular chemistry in solution are still very rare. Apart from the complexity of the involved configuration space, a major challenge for studying these systems is the need for a highly accurate description of intra- as well as intermolecular interactions, from strong covalent bonds via hydrogen bonding to electrostatic and dispersion interactions. A particularly crucial aspect is the charge distribution in the involved species, which cannot be captured correctly by most current MLPs based on localproperties like environment-dependent atomic energies and charges. Recently, we have developed a fourth-generation high-dimensional neural network potential (4G-HDNNP), which combines the accurate description of local bonding and reactivity with long-rangeinteractions based on the global charge distribution in the system. This global description is not only essential for molecules containing delocalized electrons, e.g. in aromatic groups or conjugated $\pi$-systems, but also if the molecular charge is changing, like in (de)protonation, which is a key step in many types of reactions in organic chemistry. All these systems can in principle be studied by 4G-HDNNPs, which explicitly take into account the global charge distribution resulting from reactions, different functional groups and varying total charges, making this method a promising approach for molecular chemistry. The goal of this project is to explore the applicability of 4G-HDNNPs to molecular chemistry in solution by focusing on two major aspects, the quality of the density functional theory (DFT) reference data and the generalization of the 4G-HDNNP method. High-quality reference data will be obtained by benchmarking the accuracy of exchange correlation functionals beyond the Generalized Gradient Approximation (GGA) level to Quantum Monte Carlo and Coupled Cluster calculations, and by including dispersion and self-interaction corrections (SIC). The 4G-HDNNP will be extended by employing new descriptor types for structural discrimination that are applicableeven to difficult situations like conical intersections and by the introduction of charge constraints, which, along with SIC and constrained DFT calculations, will allow to overcome the integer charge problem in both, DFT and the 4G-HDNNP, in a consistent approach. This new set of computational tools will be implemented in the open-source software RuNNer and applied to representative solute-solvent model systems covering important scenarios in synthetic organic chemistry.
-