Speaker
Description
FHI-aims (Fritz Haber Institute ab initio materials simulations) [1-3] is a versatile electronic-structure software package developed for computational studies in molecular and materials science. Widely used by a global network of developers, researchers at the Fritz Haber Institute, academic institutions, and industry, FHI-aims leverages numeric atom-centered basis sets to deliver computational precision on par with leading benchmark codes for density functional theory (DFT) and many-body methods. Notably, it achieves this high level of precision [4] while retaining computational efficiency similar to plane-wave pseudopotential methods. The code demonstrates remarkable applicability, routinely handling systems comprising thousands of atoms with semi-local and hybrid density functionals – recently demonstrated up to 30,000 atoms for hybrid density functionals [5]. Additionally, it exhibits excellent scalability on modern high-performance computing platforms. FHI-aims boasts advanced electronic-structure capabilities for both molecules and solids and seamlessly integrates into complex simulation environments. This integration includes the ability to serve as a parallel library accessible through Python or via internet sockets or through its graphical user interface GIMS [6], making it a powerful tool for a wide range of scientific investigations. Recent updates include integration with high-throughput simulation frameworks like atomate2 (based on pymatgen) [7] and Taskblaster (based on ASE) [8], enhancing its capability to explore material properties efficiently. The FHI-aims community actively develops new features and interfaces with new external frameworks, such as dispersion correction models (XDM [9] and D3 [10]), band unfolding, the Kubo-Greenwood formula [11], crystal orbital overlap population analysis [12], and improvements to periodic GW calculations and to DFPT functionality, making FHI-aims a continuously evolving tool. As an outlook, we present a framework for active learning with SISSO [13], which is being tightly integrated with FHI-aims with the goal of broadening the usability of AI materials discovery.
References
[1] V. Blum et al., Comp. Phys. Commun. 180, 2105 (2009)
[2] V. Havu, et al., J. Comp. Phys. 228, 8367 (2009)
[3] V. Gavini, et al., Model. Simul. Mater. Sci. Eng. 31.6, 063301 (2023)
[4] K. Lejaeghere, et al. Science 351.6280, aad3000 (2016)
[5] S. Kokott, et al., J. Comp. Phys. 161, 024112 (2024)
[6] Visit: https://gims.ms1p.org
[7] Visit: https://materialsproject.github.io/atomate2/
[8] Visit: https://taskblaster.readthedocs.io/en/latest/
[9] A. J. A. Price, A. Otero-de-la-Roza, and E. R. Johnson, Chem. Sci. 14, 1252 (2023)
[10] S. Grimme, S. Ehrlich, and L. Goerigk, J. Comp. Chem., 32, 1456 (2011)
[11] J. Quan, C. Carbogno, and M. Scheffler, arXiv:2408.12908 (2024)
[12] I.Takahara et al, Modelling Simul. Mater. Sci. Eng. 32, 055028 (2024)
[13] T. A. R. Purcell, M. Scheffler, and L. M. Ghiringhelli, J. Chem. Phys. 159, 114110 (2023)
