Truly amazing: Autonomous computational chemistry calculations done by AI are no longer a distant future; it is a vibrant present. Soon, most such computations will be done by AI.

large_Weixin Image_20250821122627.png

medium_Weixin Image_20250821122627.png

small_Weixin Image_20250821122627.png

thumbnail_Weixin Image_20250821122627.png

Weixin Image_20250821122627.png

In our demo below, the user prompts our Aitomia on the Aitomistic Hub to calculate the reaction energy and thermodynamics of Diels-Alder reaction for ethene and 1,3-butadiene. Aitomia accomplishes these calculations from scratch, even without the provided initial coordinates of the reactants and products.

large_Weixin Image_20250821122619.png

medium_Weixin Image_20250821122619.png

small_Weixin Image_20250821122619.png

thumbnail_Weixin Image_20250821122619.png

Weixin Image_20250821122619.png

The time to solution was just 5 min on a single CPU, delivering accurate geometries and energies calculated with our state-of-the-art AIQM2 model.

The same exercise took our expert team member to do in 20+ minutes from scratch by setting up manually initial xyz files for all involved species, performing, and analyzing the calculations. The results were the same, but the human cost was much higher.

This Deep Thinking mode, enabling increasingly complex autonomous computational chemistry simulations, is already enabled for selected users on the

. Simpler, single-task workflows such as thermochemical and spectra calculations for a single molecule are available since May 11, 2025, for free for all registered users.

This Deep Thinking mode, enabling increasingly complex autonomous computational chemistry simulations, is already enabled for selected users on the Aitomistic Hub.

Aitomia: Reaction simulation in 5 minutes without human help

 got published in Chemical Science! This ML method’s high speed, competitive accuracy, and robustness enable organic reaction simulations beyond what is possible with the popular DFT methods. It can be used for TS structure search and reactive dynamics, often with chemical accuracy. It is one of the most frequently used methods on the 

 for online simulations via a web browser, including the AI assistant 

small_1755612058503.jpg

thumbnail_1755612058503.jpg

1755612058503.jpg

The accuracy is competitive with DFT and, in some cases, can even approach the gold-standard coupled cluster accuracy. The speed of AIQM2 is orders of magnitude faster, such that propagating 1k trajectories for a representative pericyclic reaction can be done overnight, when it takes 160 CPU-years with DFT. On top of that, AIQM2 provides error bars for its predictions, which provides a useful metric for researchers to judge how trustworthy the simulations are – a feature, which is absent with DFT approaches, where it is not clear how accurate the simulations are until their results are compared to a more accurate reference.

small_1755612072429.jpg

thumbnail_1755612072429.jpg

1755612072429.jpg

, where a set of neural networks corrects the predictions of the popular semi-empirical quantum mechanical method GFN2-xTB (its modified version to be precise). This allows us to elevate the accuracy of GFN2-xTB to nearly coupled cluster accuracy in many cases, while retaining the robustness of this semi-empirical approach. Simulations with coupled cluster accuracy are often unaffordable due to the high cost. Robustness is a big issue with pure universal neural network potentials, which are becoming nowadays more and more popular, but still suffer from breakdowns when they are used for applications far from their training set. The combination of the fast semi-empirical method and neural networks enables predictions which are much faster than typical DFT approaches, while our benchmarks show that the accuracy is comparable and sometimes even better.

Don't take our word for it: as one of the Reviewers of our paper said:

I used and tested a few popular MLPs and found their first-generation AIQM1 more general and reliable than the other MLPs. In addition, although a few general-purpose MLPs were developed by different groups, almost all (except very recent AIMNET2 and AIQM2) cannot apply to reaction processes and TS structures. Their MLP-predicted TS properties are very impressive.

The reviewer mentions AIQM1, which is the first generation of the successful AIQM approaches, which, however, had 

. This prompted us to develop AIQM2, addressing the issue.

This unique combination of speed, accuracy, and robustness for reaction simulations, enabled us performing large scale downhill quasi-classical trajectory simulations to revise the post-transition state product distribution in a representative reaction, previously explored in 

 by experts in the field using DFT B3LYP-D3/6-31G*.

thumbnail_p6.gif

p6.gif

small_1755612096869.jpg

thumbnail_1755612096869.jpg

1755612096869.jpg

We found that AIQM2 has much smaller errors in barrier heights and reaction energies compared to this DFT level: the errors of AIQM2 wrt coupled cluster reference are within 2 kcal/mol, while DFT has errors of ca 10 kcal/mol for the minima. GFN2-xTB performs even worse than DFT with the huge errors of ca 25 kcal/mol and even fails to find one of the transition states. This shows that the popular schemes where lower-level methods are used for initial reaction or conformer explorations might miss many key structures. AIQM2 removes the need for such initial explorations as it can be directly used to provide high-quality estimates.

In addition, high speed of AIQM2 enabled us to improve the precision of the simulations by propagating more trajectories (1k vs ca 100 in the original JACS study), as well as use the smaller time step (0.1 vs 1.0 fs), which is essential for total energy conservation:

small_1755612111577.jpg

thumbnail_1755612111577.jpg

1755612111577.jpg

If you are wondering about more extensive benchmarks, they are provided in the paper and below is a comparison for the reaction barrier heights:

small_1755612121195.jpg

thumbnail_1755612121195.jpg

1755612121195.jpg

AIQM2 performs well for neutral, closed-shell organic molecules with CHNO elements. It can also be applied to charged species and radicals. Extensions to more elements are available as part of the 

), where AIQM2 was a model UAIQMGFN2xTB*@cc versioned 20240106. Yes, the model existed since 6 January 2024, and just recently got published! As you can see from that overview, many more models already exist, which are awaiting their publication.

 and to predict other properties, such as 

Yuxinxin Chen, Pavlo O. Dral*. AIQM2: Organic Reaction Simulations Beyond DFT. Chem. Sci. 2025, accepted manuscript. DOI: 10.1039/D5SC02802G. Open access, under the CC-BY license. The figures and text are partially adopted from this article.

Preprint on ChemRxiv: https://doi.org/10.26434/chemrxiv-2024-j8pxp (2024.10.08).

large__ ChemSci-social-media.jpg

medium__ ChemSci-social-media.jpg

small__ ChemSci-social-media.jpg

thumbnail__ ChemSci-social-media.jpg

_ ChemSci-social-media.jpg

It can be used for TS structure search and reactive dynamics, often with chemical accuracy. 

Chemical Science: “AIQM2: Organic Reaction Simulations Beyond DFT”

This continuously updated course is created and maintained by an active, internationally recognized expert in the field. Learn at your own pace while staying current with the latest techniques and trends in aitomistic simulations.

large_Online Course_02(1).jpg

medium_Online Course_02(1).jpg

small_Online Course_02(1).jpg

thumbnail_Online Course_02(1).jpg

Online Course_02(1).jpg

: While the author will make a genuine effort to answer questions, responses may take time, and the number of questions handled will be reasonable to ensure quality. This is not a 24/7 unlimited Q&A service.

is limited — pre-register now to secure your spot.

This course is open for pre-registration!

Once we have enough pre-registered participants, we will launch it. 

large_living_5.jpg

medium_living_5.jpg

small_living_5.jpg

thumbnail_living_5.jpg

living_5.jpg

medium_living_6.png

small_living_6.png

thumbnail_living_6.png

living_6.png

medium_living_2.png

small_living_2.png

thumbnail_living_2.png

living_2.png

3. 📝 Select the course you want to pre-register for, and click ‘

large_living_3.png

medium_living_3.png

small_living_3.png

thumbnail_living_3.png

living_3.png

This continuously updated course is created and maintained by an active, internationally recognized expert in the field. 

Living Course on Aitomistic (AI + Atomistic) Computational Chemistry

Recently, we cooperated with Xiamen University and published a 

 in JPCA about how to do accurate and fast IR spectra simulations with AIQM models.

Specifically, it is about the accurate and affordable simulation of gas-phase infrared spectra with AIQM models. By comparing to more than 7000 experimental spectra, we found that AIQM models are much faster than DFT while achieving comparable accuracy.

Theoretical IR (infrared) spectroscopy is a powerful tool for assisting chemical structure identification. However, approaches based on quantum chemical calculations suffer from either high computational cost (e.g., density functional theory) or insufficient accuracy (semi-empirical methods). 

Hence, we introduce a new approach based on the universal machine learning models of AIQM series.

AIQM methods, particularly, newly introduced AIQM2, can provide IR spectra with accuracy close to DFT and the speed close to a semi-empirical GFN2-xTB method. 

large_slides20250716_04.jpg

medium_slides20250716_04.jpg

small_slides20250716_04.jpg

thumbnail_slides20250716_04.jpg

slides20250716_04.jpg

To ensure high speed and interpretability, our implementation is based on the harmonic approximation with the frequencies scaled by factors that we found empirically. 

large_slides20250716_05.jpg

medium_slides20250716_05.jpg

small_slides20250716_05.jpg

thumbnail_slides20250716_05.jpg

slides20250716_05.jpg

You can easily do such calculations yourself using MLatom, which comes with 

large_slides20250716_06.jpg

medium_slides20250716_06.jpg

small_slides20250716_06.jpg

thumbnail_slides20250716_06.jpg

slides20250716_06.jpg

You can even do the IR spectra simulation now online, on the 

. Opening your molecule visualizing it and then clicking on this infrared button, and then it will automatically do the AIQM2 calculations.

Here it shows the summary of calculations, which include the frequencies and intensities in raw format.

As you see, you can start from any geometry, even this ridiculously linear ethanol molecule, and then it will optimize the geometry with AIQM2 and do frequency calculations and produce here on the right infrared spectrum.

You can also attribute normal mode vibrations to specific bands in the infrared spectrum in the Jupyter Lab on the same website.

And, nowadays you can also do the simulations using Aitomia intelligent assistant. On the hub you can chat with it. Here's an example doing the same thing but for hexanol molecule by chatting in Aitomistic Hub. Aitomia produces a spectrum for hexanol with pretty decent accuracy compared to experiments.

If you have any questions about how to run these calculations, you can check our 

large_JPCA_cn_20250730-封面.jpg

medium_JPCA_cn_20250730-封面.jpg

small_JPCA_cn_20250730-封面.jpg

thumbnail_JPCA_cn_20250730-封面.jpg

JPCA_cn_20250730-封面.jpg

We published a paper in JPCA about the accurate and affordable simulation of gas-phase infrared spectra with AIQM models. 

JPCA: accurate and affordable IR spectra with AIQM

We’re glad to share that, following multiple requests, the July 2025 Edition of the hands-on online course Computational Chemistry & AI by Pavlo O. Dral is now available free of charge for academic users. 🎓

📚 The first edition (June 2024) was well received — one participant shared:

“Dr. Dral offers a fantastic introduction to the concepts around machine learning in chemistry!”

The course is based on the author’s experience teaching university courses and running hands-on workshops. Since the field is evolving rapidly, the course has already been updated several times.

2. 📂 Click the course icon in the left pan

large_WechatIMG1198.jpg

medium_WechatIMG1198.jpg

small_WechatIMG1198.jpg

thumbnail_WechatIMG1198.jpg

WechatIMG1198.jpg

large_WechatIMG1199.jpg

medium_WechatIMG1199.jpg

small_WechatIMG1199.jpg

thumbnail_WechatIMG1199.jpg

WechatIMG1199.jpg

✅ If you meet the eligibility criteria — especially having a verifiable academic affiliation (such as using your university email when registering) — your application is usually approved within 24 hours.

If you find the course helpful, we’d really appreciate your feedback, suggestions, and testimonials (contact@aitomistic.com).

And if you know colleagues, students, or friends who might benefit from it, please share this announcement with them.

Thank you for helping us support the academic community! 💬📨

The July 2025 Edition of the hands-on online course Computational Chemistry & AI by Pavlo O. Dral is now available free of charge for academic users.

Free Academic Access to Computational Chemistry & AI (July 2025 Edition)

Researchers at Aitomistic has published a 

 in npj Computational Materials presents an efficient ML protocol for accelerating trajectory surface hopping dynamics, while tackling many key issues making machine learning of excited states difficult. The protocol introduces a new machine learning interatomic potential based on multi-state learning and accelerated sampling of the critical regions near conical intersections. All the methods, together with the end-to-end active learning are available in 

Quantum chemical simulations of molecular excited states are important for understanding the photochemical and photophysical processes, but they are very computationally intensive, especially when used for nonadiabatic dynamics. Machine learning is an attractive tool to tackle this problem by replacing expensive quantum chemical calculations, however, the progress in ML for excited states is much slower than for ground states due to the complexity of the former (see, e.g., 

large_toc-graph-with_text.webp

medium_toc-graph-with_text.webp

small_toc-graph-with_text.webp

thumbnail_toc-graph-with_text.webp

toc-graph-with_text.webp

To address the key challenges in machine learning of excited states, this work has reported a comprehensive and robust protocol for ML based on an eight-year-long (!) in-depth research. This protocol is based on a multi-state learning model, MS-ANI, capable of learning excited states of various molecules with unprecedented accuracy. The core idea is to take the state number as an input feature, concatenated with the ANI atomic environment vectors, and passed to the neural network for the corresponding element type. Impressively, the accuracy of learning excited states is better or similar to the accuracy of learning ground-state energies, where information on excited-state energies improves the quality of ground-state predictions. 

small_image-1.webp

thumbnail_image-1.webp

image-1.webp

One of the key ingredients to train a robust potential for nonadiabatic dynamics is accelerated sampling of the regions near conical intersections. We have achieved this in this work by introducing gapMD, which ensures enhanced sampling of the low-gap regions by propagating trajectories with the gradient energy difference between adjacent surfaces when needed.

small_image-2.webp

thumbnail_image-2.webp

image-2.webp

Together, these advances enabled end-to-end active learning for generating machine learning models to perform nonadiabatic simulations.

small_image-3.webp

thumbnail_image-3.webp

image-3.webp

For simple cases such as fulvene, this active learning is very fast and converges just after three days on a single GPU card.

large_image-4.webp

medium_image-4.webp

small_image-4.webp

thumbnail_image-4.webp

image-4.webp

Active learning also works for more complex cases and, e.g., can be used to calculate photoreaction quantum yields. We have also discovered long-time-scale oscillations in cis-azobenzene photoisomerization.

small_image-5.webp

thumbnail_image-5.webp

image-5.webp

In this work, we have also shown for the first time that a single model can be used to perform the nonadiabatic dynamics of different molecules. This has laid the foundation for our follow-up study on the first-ever universal excited-state potential OMNI-P2x.

Video lecture about the theory and life-recorded tutorial:

Mikołaj Martyka, Lina Zhang, Fuchun Ge, Yi-Fan Hou, Joanna Jankowska*, Mario Barbatti*, 

 npj Comput. Mater. 2025, 11, 132. DOI: 10.1038/s41524-025-01636-z.

large_npj.png

medium_npj.png

small_npj.png

thumbnail_npj.png

npj.png

The protocol introduces a new machine learning interatomic potential based on multi-state learning and accelerated sampling of the critical regions near conical intersections. 

npj Comput. Mater.: Efficient Machine Learning Protocol For Accelerating Trajectory Surface Hopping Dynamics

Do you want to perform computational chemistry simulations but struggle with complex codes and lack computational resources? Are you conducting experiments without theoretical guidance and feel lost? No worries, the help is on the way!

We are pleased to announce Aitomia: a user-friendly intelligent assistant with expert knowledge in AI-driven atomistic (“aitomistic”) and quantum chemical simulations that is designed to help you in setting up, running, and analyzing atomistic simulations. It leverages the power of our existing software ecosystem 

, imparting Aitomia an impressive range of capabilities encompassing geometry optimizations, spectra simulation, molecular dynamics, and reaction modelling, and much more.

Aitomia is the first intelligent assistant publicly accessible online on a cloud computing platform for atomistic simulations of broad scope—

—where you can register and try it out for free! In a video example below, the user on Aitomistic Hub asks Aitomia in Chinese what the structure of hexanol is. The Aitomia returns the structure which the user can analyze in the Hub’s GUI. Later, the user asks to calculate the IR spectrum of hexanol, which Aitomia promptly calculates and delivers to the user in the form of a spectrum. Remarkably, the spectrum simulations are not just fast but also rather accurate compared to the experiment due to the state-of-the-art AIQM models.

If you are interested in installing Aitomia (and more!) locally to ensure full privacy of your calculations, it is also possible. Please contact us at contact@aitomistic.com, if you have any inquiries. You can also 

 to the news about Aitomia, its upcoming releases and subsequent upgrades, and other news by Aitomistic.

The details about Aitomia are described in a 

large_thumbnail_2_01.png

medium_thumbnail_2_01.png

small_thumbnail_2_01.png

thumbnail_2_01.png

A user-friendly intelligent assistant with expert knowledge in AI-driven atomistic (“aitomistic”) and quantum chemical simulations.

News

Aitomia: Reaction simulation in 5 minutes without human help

Chemical Science: “AIQM2: Organic Reaction Simulations Beyond DFT”

Living Course on Aitomistic (AI + Atomistic) Computational Chemistry

JPCA: accurate and affordable IR spectra with AIQM

Free Academic Access to Computational Chemistry & AI (July 2025 Edition)

npj Comput. Mater.: Efficient Machine Learning Protocol For Accelerating Trajectory Surface Hopping Dynamics

Aitomia: Your Intelligent Assistant for AI-Driven Atomistic and Quantum Chemical Simulations

Meet OMNI-P2x — the First Universal ML Potential for Excited States!

Aitomistic Hub Launched!