Weighted Active Space Protocol for Multireference Machine-Learned Potentials
arXiv:2505.10505 · doi:10.1073/pnas.2513693122
Abstract
Multireference methods such as multiconfiguration pair-density functional theory (MC-PDFT) offer an effective means of capturing electronic correlation in systems with significant multiconfigurational character. However, their application to train machine learning-based interatomic potentials (MLPs) for catalytic dynamics has been challenging due to the sensitivity of multireference calculations to the underlying active space, which complicates achieving consistent energies and gradients across diverse nuclear configurations. To overcome this limitation, we introduce the Weighted Active-Space Protocol (WASP), a systematic approach to assign a consistent active space for a given system across uncorrelated configurations. By integrating WASP with MLPs and enhanced sampling techniques, we propose a data-efficient active learning cycle that enables the training of an MLP on multireference data. We demonstrate the method on the TiC+-catalyzed C-H activation of methane, a reaction that poses challenges for Kohn-Sham density functional theory due to its significant multireference character. This framework enables accurate and efficient modeling of catalytic dynamics, establishing a new paradigm for simulating complex reactive processes beyond the limits of conventional electronic-structure methods.
References in corpus (22)
- Canonical sampling through velocity-rescaling
- Escaping free-energy minima
- PLUMED 2: New feathers for an old bird
- Recent developments in the PySCF program package
- Less is more: sampling chemical space with active learning
- Phase transitions of hybrid perovskites simulated by machine-learning force fields trained on-the-fly with Bayesian inference
- Libxc: a library of exchange and correlation functionals for density functional theory
- Machine learning for electronically excited states of molecules
- Rethinking Metadynamics: from bias potentials to probability distributions
- Symmetry-Adapted Machine-Learning for Tensorial Properties of Atomistic Systems
- MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields
- Automated construction of molecular active spaces from atomic valence orbitals
- Silicon liquid structure and crystal nucleation from ab-initio deep Metadynamics
- autoCAS: a program for fully automated multi-configurational calculations
- Rare Event Kinetics from Adaptive Bias Enhanced Sampling
- Metadynamics for Training Neural Network Model Chemistries: a Competitive Assessment
- Unraveling the Crystallization Kinetics of the GeSbTe Phase Change Compound with a Machine-Learned Interatomic Potential
- Steering Orbital Optimization out of Local Minima and Saddle Points Toward Lower Energy
- Excited states, symmetry breaking, and unphysical solutions in state-specific CASSCF theory
- ArcaNN: automated enhanced sampling generation of training sets for chemically reactive machine learning interatomic potentials
- Exploring the design space of machine-learning models for quantum chemistry with a fully differentiable framework
- Semiclassical Nonadiabatic Molecular Dynamics Using Linearized Pair-Density Functional Theory