NAM29

Molecular Catalysts Design with Massively Parallel Physics-Based Computational Workflow

Speaker:
Croix J. Laconsay, Senior Scientist I, Schrödinger

Abstract:

Introduction
Molecular catalysts have traditionally been designed through experimental trial-and-error methods, which are resource-intensive, time-consuming, laborious, and costly. Historically, computational chemistry, particularly quantum chemical studies combined with domain expertise, has contributed indirectly to molecular catalyst design by elucidating reaction mechanisms. However, with the advent of advanced hardware architectures, improved theoretical methods, sophisticated algorithms, and increased automation, computational chemistry now offers the possibility for the direct design of molecular catalysts.

Materials and Methods
The results presented here were obtained using the Reaction Network Enumeration Profiler (RxnEnumProfiler)—previously Automated Reaction Workflow—panel of the Schrödinger Material Science Suite (Versions 24-3,4).1 The RxnEnumProfiler module is designed with a user-friendly graphics-user interface (GUI) or can be accessed through a command line interface (CLI). It can be used by either advanced computational catalysis researchers, students of all levels, or experimental homogeneous catalysis scientists with no prior background in atomic scale modeling.

Results and Discussion
We introduce an automated digital approach for predicting two key catalytic performance metrics for dynamically generated library of virtual molecular catalysts, leveraging quantum mechanics directly. Our workflow integrates three main components: 1) knowledge of the mechanism of the reference catalytic reaction; 2) a predefined in silico library for catalyst and/or substrate functionalization; and 3) a specified quantum-mechanical method. Proof-of-concept demonstration is presented across two representative homogeneous reactions, organocatalyzed asymmetric hydrogenation and Ni-catalyzed C-C cross-coupling.