Themed collection Digital Discovery – Editors Choice Collection 2023

The capability to replicate the predictions by machine learning (ML) or artificial intelligence (AI) models and the results in scientific workflows that incorporate such ML/AI predictions is driven by a variety of factors.

We report the findings of a hackathon focused on exploring the diverse applications of large language models in molecular and materials science.

We describe the current state of the SELFIES library (version 2.1.1), and, in particular, the advances and improvements we have made in its underlying algorithms, design, and API.

Graph representations of hierarchical knowledge, including experiment provenances, will help usher in a new era of data-driven materials science.

Advances in robotic automation, high-performance computing, and artificial intelligence encourage us to propose large, general-purpose science factories with the scale needed to tackle large discovery problems and to support thousands of scientists.

An automated solid dispenser was developed using a dual-arm robot and fuzzy logic controller, mimicking the operations of human researchers.

Self-driving labs benefit from occasional and asynchronous human interventions. We present a heuristic framework for how self-driving lab operators can interpret progress and make changes during a campaign.

We used synthetically generated crystals to train ResNet-like models to enhance the prediction of space groups from ICSD powder X-ray diffractograms. The results show improved generalization to unseen structure types compared to previous approaches.

A sonochemical Materials Acceleration Platform was implemented to synthesize CdSe nanocrystals under 625 unique conditions (in triplicate) in less than 6 weeks. The modularity of the workflow is adaptable to a variety of applications.

Ising machines are used to create molecules with desired properties. GPU-based Ising machines are shown to outperform qubit-based ones in terms of scalability.

Automated patent mining creates domain-specific datasets of molecular structures for generative modeling with limited human intervention.

We present ODACell, an automated electrolyte formulation and coin cell assembly system for accelerated battery research.

Variational-autoencoders with an additional predictor neural-network and gradient-based optimization allow us to generate new Suzuki-catalysts and predict the binding energies.

Group SELFIES is a molecular string representation that incorporates tokens which represent substructures while maintaining robustness, which improves the performance of molecular generative models.

The use of robotic instrumentation and Python scripts allows for fully automated and robust combinatorial polymer synthesis.

We present an interpretable uncertainty-aware machine learning model to predict battery degradation trajectories. Using LSTM Recurrent Neural Networks, we reach an RMSE of 106 and MAPE of 10.6%.