[Paper Review] Navigating chemical reaction space with a steering wheel

Miguel Steiner, Markus Reiher|arXiv (Cornell University)|Aug 31, 2023

Machine Learning in Materials ScienceMaterials Science181 references3 citations

TL;DR

This paper introduces the Steering Wheel, a flexible, interactive algorithm that enables intuitive, on-the-fly guidance of automated first-principles exploration of chemical reaction networks (CRNs). By integrating with the Chemoton framework and the Heron GUI, it allows researchers to dynamically steer exploration toward specific reaction pathways, intermediates, or catalysts—balancing depth and breadth in a way that overcomes the combinatorial explosion of brute-force methods, with demonstrated success in transition metal catalysis.

ABSTRACT

Autonomous reaction network exploration algorithms offer a systematic approach to explore mechanisms of complex chemical processes. However, the resulting reaction networks are so vast that an exploration of all potentially accessible intermediates is computationally too demanding. This renders brute-force explorations unfeasible, while explorations with completely pre-defined intermediates or hard-wired chemical constraints, such as element-specific coordination numbers, are not flexible enough for complex chemical systems. Here, we introduce a Steering Wheel to guide an otherwise unbiased automated exploration. The Steering Wheel algorithm is intuitive, generally applicable, and enables one to focus on specific regions of an emerging network. It also allows for guiding automated data generation in the context of mechanism exploration, catalyst design, and other chemical optimization challenges. The algorithm is demonstrated for reaction mechanism elucidation of transition metal catalysts. We highlight how to explore catalytic cycles in a systematic and reproducible way. The exploration objectives are fully adjustable, allowing one to harness the Steering Wheel for both structure-specific (accurate) calculations as well as for broad high-throughput screening of possible reaction intermediates.

Motivation & Objective

• To address the combinatorial explosion in automated reaction network exploration by enabling real-time, user-directed steering of unbiased first-principles exploration.
• To overcome limitations of pre-defined intermediates or hard-coded chemical constraints in mechanism elucidation and catalyst design.
• To provide a general, extensible, and intuitive framework for guiding automated CRN exploration in complex systems such as transition metal catalysts.
• To enable both high-throughput screening and accurate, structure-specific calculations within a single, unified exploration workflow.

Proposed method

• The Steering Wheel integrates with the Chemoton framework and the Heron graphical user interface to allow real-time filtering and selection of reaction intermediates and reactive sites during automated exploration.
• It operates through alternating Network Expansion and Selection Steps, where expansion generates new compounds and reactions, and selection applies filters based on chemical structure, reactivity, or graph-based rules.
• Filters are implemented via user-defined methods that evaluate molecular aggregates or reactive site coordinates, enabling custom logic for reactivity and selectivity.
• The system supports logical combinations of filters (AND/OR) and allows dynamic rule changes during exploration, enabling focus shifts across diverse reactivity types.
• The framework is extensible: new filters, selection steps, and expansion protocols can be added by defining custom methods without modifying core code.
• All exploration steps are encoded in standardized data structures, ensuring compatibility between successive stages of the workflow.

Experimental results

Research questions

• RQ1 How can automated reaction network exploration be made both systematic and interactive, without pre-defining all intermediates or imposing rigid chemical constraints?
• RQ2 Can a user-guided steering mechanism effectively balance breadth and depth in exploring complex chemical space, especially in transition metal catalysis?
• RQ3 To what extent can the Steering Wheel algorithm maintain computational efficiency while enabling flexible, on-the-fly redirection of exploration toward specific reaction pathways?
• RQ4 How can the framework support both high-throughput screening and accurate, targeted calculations within the same exploration workflow?

Key findings

• The Steering Wheel enables intuitive, real-time guidance of automated reaction network exploration through a graphical interface, significantly improving usability and control over otherwise opaque exploration processes.
• The algorithm successfully guided exploration of complex catalytic systems, including Wilkinson’s catalyst, Ziegler–Natta catalyst, and the Monsanto process, demonstrating its applicability across diverse reaction mechanisms.
• The framework supports both breadth-first and depth-first exploration strategies, allowing users to systematically navigate vast chemical spaces while maintaining computational feasibility.
• The integration of dynamic filters and user-defined reactivity rules allows for targeted exploration of specific reaction types—such as hydrolysis, condensation, and organometallic transformations—within a single network.
• The system was successfully applied to a gallium single-site catalyst, requiring non-covalent complex handling, which was enabled by the extensibility of the framework in Scine Chemoton version 3.2.
• All reaction networks and exploration protocols are publicly available on Zenodo with full reproducibility via Apptainer containers and open-source software, ensuring transparency and reusability.