Large-Scale Condensed Matter DFT Simulations: Performance and Capabilities of the CRYSTAL Code.

Area of Science:

• Computational Materials Science
• Quantum Mechanics
• Solid-State Physics

Background:

• Efficient high-performance computing (HPC) is vital for first-principles calculations of large systems.
• Parallelization strategies are key to managing computational complexity in quantum-mechanical algorithms.
• Extending parallelization to all properties, not just basic functions, is challenging but necessary.

Purpose of the Study:

• To discuss the performance and capabilities of the massively parallel Crystal17 package.
• To present recent developments enhancing code scalability for HPC.
• To analyze the computational demands for large-scale solid-state calculations.

Main Methods:

• Implementation and testing of advanced parallelization strategies within the Crystal17 package.
• Quantitative analysis of code scalability and memory usage on HPC clusters.
• Validation of numerical size consistency for large atomic systems.
• Application to ab initio studies of diverse physical properties.

Main Results:

• Demonstrated significant improvements in Crystal17 code scalability, supporting up to 32,768 cores.
• Quantified scaling and memory requirements for calculations involving up to 14,000 atoms per cell.
• Documented high numerical size consistency, ensuring accuracy.
• Showcased successful ab initio investigations of various material properties.

Conclusions:

• The enhanced Crystal17 package effectively leverages HPC resources for large-scale first-principles solid-state calculations.
• The code demonstrates excellent scalability and accuracy for complex systems.
• This advancement facilitates deeper understanding of material properties through advanced computational modeling.