Aitomia: An Agentic Framework for AI-Driven Atomistic and Quantum Chemical Simulations

Area of Science:

• Computational Chemistry
• Artificial Intelligence in Science

Background:

• Atomistic and quantum chemical (QC) simulations are crucial for understanding molecular behavior but often require specialized expertise and significant computational resources.
• Existing workflows can be complex, posing barriers for both experts and nonexperts in accessing and utilizing these powerful simulation tools.

Purpose of the Study:

• To introduce Aitomia, an agentic framework designed to democratize AI-driven atomistic and quantum chemical simulations.
• To enable experts and nonexperts to efficiently set up, run, analyze, and summarize complex chemical calculations through natural language interaction.

Main Methods:

• Aitomia is built upon the MLatom software ecosystem, integrating with established quantum chemistry programs (Gaussian, ORCA, PySCF, xtb).
• It supports a range of methods including Density Functional Theory (DFT), semiempirical methods (GFN2-xTB), and high-level wave function methods.
• The framework autonomously executes computational workflows for tasks such as geometry optimization, thermochemistry, and spectra simulations.

Main Results:

• Aitomia delivers results for infrared spectra in seconds and reaction thermochemistry in minutes.
• The simulation outcomes demonstrate high accuracy, comparable to experimental data or high-level theoretical references.
• Significant reduction in manual user effort is achieved through automated workflow execution.

Conclusions:

• Aitomia effectively lowers the barrier to entry for performing advanced atomistic and quantum chemical simulations.
• The framework accelerates research and development by democratizing access to powerful computational tools.
• Aitomia empowers a broader range of users to leverage AI for chemical discovery and analysis.