OMNI-P2x is a universal neural network potential for excited-state simulations. It approaches TD-DFT quality predictions, e.g., for UV/vis absorption spectra at a fraction of cost. OMNI-P2x is faster and more accurate than semiempirical QM methods such as TD-DFTB.
Up-to-date online tutorials are provided at http://mlatom.com/docs/tutorial_omnip2x.html.
You can use OMNI-P2x in MLatom (open-source under the MIT license) locally and on online platforms. See MLatom documentation on how to use it locally.
It is convenient to do calculations with OMNI-P2x via a web browser on online platforms with convenient GUIs and AI agents. Currently, two options are available:
- Aitomistic Hub - provides a convenient web interface for performing simulations, with limited free resources mainly for demo purposes.
- Aitomistic Lab@XMU - similar to the Hub above, but is available only for the academic users registered with their academic emails, for whom it provides more resources.
You can request a ready-to-use implementation in the Aitomic package (which builds upon MLatom by providing add-ons for unpublished features like this one), it is free for academic users subject to a license agreement. The source code is planned to be openly released in MLatom under the MIT license upon the acceptance of the manuscript in the journal of our choice.
- The model weights are available in this repository under the folder
model_weights. - Sample scrips for archival purposes on how the published simulations were performed. Instead, we recommend to refer to the up-to-date tutorials for the latest MLatom version.
- Mikolaj Martyka, Xin-Yu Tong, Joanna Jankowska, and Pavlo O. Dral. OMNI-P2x: A Universal Neural Network Potential for Excited-State Simulations. 2025. Preprint on ChemRxiv: https://doi.org/10.26434/chemrxiv-2025-j207x (2025-04-16)