Supporting Data and Code for "QUASAR: A Universal Autonomous System for Atomistic Simulation and a Benchmark of Its Capabilities"

This repository contains the source code, documentation, and complete benchmark outputs for QUASAR (Universal Autonomous System for Atomistic Simulation), an open-source agentic AI system designed for production-grade computational chemistry and materials science research. QUASAR autonomously orchestrates complex multi-scale atomistic workflows across diverse computational methods, including: Density Functional Theory (DFT) Machine Learning Potentials Molecular Dynamics (MD) Monte Carlo simulations The system incorporates adaptive planning, context-efficient memory management, and hybrid knowledge retrieval to navigate real-world research scenarios without human intervention. Repository Contents quasar-source.zip: Complete QUASAR source code, documentation, and Docker configuration benchmark-outputs.tar.gz: Full simulation inputs, outputs, and analysis from the three-tiered benchmark suite: Tier I (Task Execution): K-point convergence, NPT equilibration, helium void-fraction analysis Tier II (Workflow Orchestration): NiO band gap calculation, CO₂ adsorption isotherms, aluminum melting point determination Tier III (Frontier Research): Photocatalyst screening, COF gas separation selectivity, novel material assessment