FHI-aims

Our big overview paper "Roadmap on Advancements of the FHI-aims Software Package" is now on ArXiV: arxiv.org/abs/2505.00125 200+ authors detailing 20 years of progress in FHI-aims. Thank you to our INCREDIBLE community of people who have contributed to this!

Registered users can find the new release available to download here: fhi-aims.org/get-the-code-m… Not yet a user of FHI-aims? Then you can see here how to become part of the community! fhi-aims.org/get-the-code-m…

🎉 Happy new FHI-aims release day 🎉 Version 260331 is now available, with a huge amount of changes inside, and over 25 contributors putting new code into this version. A huge thank you to everyone! You can see the list of changes & contributors here: fhi-aims.org/uploads/docume…

📣 A new FHI-aims patch version 250822_1 is now available! If you're a registered user, head over to the aimsclub to grab the newest version: fhi-aims.org/get-the-code-m… 📖 The changelog can be seen here: fhi-aims.org/uploads/docume… As always, thank you to all contributors!

We once again have two great speakers lined up, with Andrew J. Logsdail introducing QM/MM approaches for multiscale simulations of solid-state systems and Gabriel A. Bramley showcasing the recently released EmbASI framework for QM/QM embedding.

📣 We are happy to announce the next FHI-aims Webinar on Multiscale Modeling with Explicit Embedding Approaches in FHI-aims! For registration and more details, click here: lnkd.in/dEM52S3q

What a great week in Shanghai for the Hands-on Workshop on Electronic-Structure Theory and Artificial Intelligence for Materials Science! A huge thank you to our speakers, tutorial preparers, organizers, and participants for making it an excellent event to learn about FHI-aims.

Check out the full program here: indico.ms1p.org/event/12/page/… It promises to be a very interesting few days of science!

We are looking forward to the Hands-on Workshop on Electronic-Structure Theory and Artificial Intelligence for Materials Science tomorrow! In case you aren't in Shanghai, you can find livestreams of all the lectures, beginning at 10.00 CST 5 November here: koushare.com/live/details/4…

The recordings of our webinar on "Explainable AI models for Materials Science with the SISSO Approach" are now available on our YouTube channel: youtube.com/playlist?list=… Thank you to everyone who attended and to our great speakers, Thomas Purcell and Lucas Foppa! 👏

Excited to share one of my last PhD projects! We tackle finite-size effects in systems with high dof (e.g. HOIPs) using ML potentials + structural descriptors to analyze dynamic SOC in large supercells. Grateful to Kanai Lab and my mentee! 🙌 🔗 pubs.aip.org/aip/cpr/articl…

You can find the model and tutorial on its use in the HuggingFace repository: huggingface.co/nitbha007/MACE… Huge congratulations to the authors in this work for addressing the lack of a universal MLIP for IR spectroscopy!

MACE4IR is trained on 10 million geometries and corresponding FHI-aims DFT energies, forces and dipoles from the @GoogleDeepMind QCML database. It can predict IR spectra for organic, inorganic, metal complexes and biologically relevant species at a fraction of the cost of DFT.

Interested in rapid and reliable prediction of IR spectra? Check out this pre-print on the foundational model, MACE4IR, by Nitik Bhatia, Ondrej Krejci, Silvana Botti, Patrick Rinke, and Miguel A. L. Marques : arxiv.org/abs/2508.19118

Registered users can find the new release available to download here: fhi-aims.org/get-the-code-m… Not yet a user of FHI-aims? Then you can see here how to become part of the community! fhi-aims.org/get-the-code-m…

You can see the full list of changes and the people that made them possible here: fhi-aims.org/uploads/docume…

📢 New FHI-aims release 250822 📢 Lots of new things including: QM-in-QM embedding with EmbASI, superconductivity, implicit solvation with Environ, Q4C up to 1500 atoms, faster range separated hybrids, and a whole lot more. A huge thank you to everyone who contributed!

"aims-PAX is a flexible, fully automated, open-source software package for performing active learning for machine learning force fields using a parallelized algorithm that enables efficient resource management." You can find the open-source code here: github.com/tohenkes/aims-… 2/2

Interested in the automated construction of Machine Learning Force Fields? Check out this exciting new pre-print describing aims-PAX, which utilizes FHI-aims, from Tobias Henkes, Shubham Sharma, Alexandre Tkatchenko, Mariana Rossi, and Igor Poltavsky: arxiv.org/abs/2508.12888 1/2

My group is on🔥,2nd @ChemRxiv preprint in a week! Efficient Molecular Crystal Structure Prediction and Stability Assessment with AIMNet2 Neural Network Potentials. #compchem collaboration with Marom lab @cmu_mse @CmuScience chemrxiv.org/engage/chemrxi…

📣 We are happy to announce the next FHI-aims webinar in September. Find out about explainable AI models for materials science using the SISSO approach with Thomas Purcell and Lucas Foppa! For more details/registration, click here: indico.ms1p.org/e/webinar_sisso