Fermi.jl: A Modern Design for Quantum Chemistry.

Area of Science:

• Computational chemistry
• High-performance computing
• Programming language development

Background:

• Quantum chemistry calculations require significant numerical resources, traditionally relying on low-level languages like C++ and Fortran.
• Developing new computational methods is often slowed by the complexity of these efficient but cumbersome languages.
• Higher-level languages like Python are sometimes used for non-performance-critical parts, creating a "two-language problem".

Purpose of the Study:

• Introduce Fermi.jl, a novel quantum chemistry package implemented in the Julia programming language.
• Demonstrate Julia's capability to address the "two-language problem" in computational chemistry by offering both ease of development and high performance.
• Provide the first implementations of post-Hartree-Fock methods in Julia.

Main Methods:

• Leveraging Julia's core features such as multiple dispatch, metaprogramming, and interactive capabilities.
• Designing Fermi.jl as a modular package to facilitate the addition of new methods and implementations.
• Emphasizing code reusability through general functions and specialized methods.

Main Results:

• Fermi.jl successfully implements various post-Hartree-Fock methods.
• Performance evaluations of popular *ab initio* methods within Fermi.jl demonstrate its feasibility and efficiency.
• The package showcases the potential of Julia for demanding scientific computing tasks.

Conclusions:

• Fermi.jl offers a promising solution to the "two-language problem" in quantum chemistry.
• The package's modular design and reliance on Julia's features enable efficient development and code reusability.
• Encourages wider adoption of Julia in the scientific community for computational chemistry research.