A new massively parallel version of CRYSTAL for large systems on high performance computing architectures.

Area of Science:

• Computational materials science
• Solid-state physics
• Quantum chemistry

Background:

• Accurate simulation of complex solid systems requires efficient computational software leveraging high-performance computing (HPC).
• Existing ab initio codes face challenges with large-scale systems and modern HPC architectures, particularly regarding memory constraints.

Purpose of the Study:

• To evaluate the performance and scalability of the CRYSTAL code on advanced HPC architectures.
• To demonstrate the feasibility of large-scale ab initio calculations for complex crystalline systems.

Main Methods:

• Utilized the CRYSTAL code, a periodic ab initio method employing Gaussian basis sets.
• Performed density functional calculations using hybrid functionals on low-symmetry systems.
• Conducted performance analysis on state-of-the-art HPC architectures, assessing running time and memory usage.

Main Results:

• CRYSTAL demonstrated high parallel efficiency in both computation time and memory usage on HPC systems.
• Simulations involving up to 100,000 atomic orbitals and 8000 atoms in low-symmetry systems were successfully executed.
• The software efficiently utilizes thousands of processors on advanced European HPC resources.

Conclusions:

• CRYSTAL is capable of efficiently handling very large unit cells for crystalline systems on modern HPC architectures.
• The software's performance characteristics are well-suited for current HPC trends, including many-core processors with limited memory.
• Ab initio calculations of complex solid systems at an unprecedented scale are now feasible for researchers.