MLatom 3: A Platform for Machine Learning-Enhanced Computational Chemistry Simulations and Workflows

Area of Science:

• Computational Chemistry
• Machine Learning Applications

Background:

• Machine learning (ML) is a growing tool in computational chemistry.
• Rapid ML development necessitates flexible software for custom workflows.

Purpose of the Study:

• Introduce MLatom 3, an open-source package for computational chemistry.
• Enable users to leverage ML for enhanced simulations and complex workflows.

Main Methods:

• MLatom 3 offers command-line, input file, and Python scripting options.
• Supports simulations on local machines and the XACS cloud computing service.
• Integrates with external software and libraries for flexibility.

Main Results:

• Enables calculations of energies, thermochemical properties, and geometry optimizations.
• Facilitates molecular dynamics, quantum dynamics, and spectral simulations (rovibrational, UV/vis, TPA).
• Provides pretrained ML models (e.g., AIQM1) and allows custom model development.

Conclusions:

• MLatom 3 provides a flexible, open-source framework for computational chemistry.
• It empowers users to perform diverse simulations using ML and quantum mechanical methods.
• Extensive interfaces enhance its adaptability and integration capabilities.