François Cornet

Deep generative models are a powerful tool for exploring the chemical space within inverse-design workflows; however, their effectiveness relies on sufficient training data and effective mechanisms for guiding the model to optimize specific properties. We demonstrate that designing an expert-informed data representation and training procedure allows leveraging data augmentation while maintaining the required sampling controllability. We focus our discussion on a specific class of compounds (transition metal complexes), and a popular class of generative models (equivariant diffusion models), although we envision that the approach could be extended to other chemical spaces and model types. Through experiments, we demonstrate that augmenting the training database with generic but related unlabeled data enables a practical level of performance to be reached.

This work presents OM-Diff, an inverse-design framework based on a diffusion generative model for in silico design of organometallic complexes.

Organometallic complexes are ubiquitous in homogeneous catalysis and other technological applications. Optimization of such complexes for specific applications is challenging due to the large variety of possible metal-ligand combinations and ligand-ligand interactions. Here we present OM-Diff, an inverse design framework based on a diffusion generative model for in-silico design of such complexes from scratch. Given the importance of the spatial structure of a catalyst, the model directly operates on all-atom (including hydrogen) representations in 3D space. To handle the symmetries inherent to that data representation, OM-Diff combines an equivariant diffusion model and an equivariant property predictor to drive sampling at inference time. The model can conditionally generate novel ligands beyond those in the training dataset. We demonstrate the potential of the proposed approach by designing catalysts for a family of cross-coupling reactions, and validating a selection of novel proposed compounds with DFT calculations.

To enable fast, resource efficient development and broad scale deployment of of high accuracy Machine-Learned Interatomic Potentials (MLIPs) with minimum expert involvement, we introduce CURATOR, an autonomous batch active learning workflow for constructing MLIPs. CURATOR integrates state of the art models, uncertainty quantification techniques, batch selection algorithms with user defined labeling and chemical-structure space exploration methods for data and compute efficient active learning. We also developed a novel efficient gradient computation method that calculates forces and stress based on the energy derivative with respect to accelerate CURATOR. Our evaluation across different chemical systems demonstrates that CURATOR considerably reduces the computational resources and time required to develop reliable MLIPs. In practical applications in novel complex materials and interfaces, CURATOR shows promising results, underscoring its potential in accelerating materials discovery. The flexibility and efficiency of CURATOR mark a significant advancement in the field of computational materials science, paving the way for more efficient and larger time-length scale atomistic simulations.