AI Achieves Solutions in Seconds Where Quantum Chemistry Requires Days


  • Transition states in chemical reactions are crucial but time-consuming to determine.
  • Quantum chemistry’s density functional theory is the traditional method, but it can take days for one transition state.
  • AI, specifically a generative AI tool called a diffusion model, has been employed to accelerate transition state calculations.
  • Researchers trained the AI using 9,000 chemical reactions, enabling it to generate transition states for new reactions in seconds.
  • This breakthrough could significantly speed up the development of new materials, fuels, pharmaceuticals, and the study of biological reactions.

Main AI News:

In the world of chemistry, the speed at which reactions occur can often be a limiting factor in the development of new materials, fuels, and pharmaceuticals. The key to unlocking these possibilities lies in understanding the elusive “transition state” – that critical moment when molecules collide and the fate of the reaction is sealed. However, grasping this fleeting instant has long been a time-consuming endeavor, relying heavily on the intricate complexities of quantum chemistry. But now, the landscape is evolving, with the emergence of a more expeditious ally: artificial intelligence.

Associate Professor Heather Kulik, hailing from MIT and senior author of a recent paper in Nature Computational Science, sheds light on the significance of the transition state: “The transition state helps to determine the likelihood of a chemical transformation happening. If we have a lot of something that we don’t want, like carbon dioxide, and we’d like to convert it to a useful fuel like methanol, the transition state and how favorable that determines how likely we are to get from the reactant to the product.

Currently, the gold standard for locating a transition state involves a quantum method known as density functional theory, which can sometimes demand days to calculate a single transition state. Seeking a swifter alternative, chemists have dabbled in the realm of machine learning. Yet, a significant challenge emerges as molecules collide at various angles and orientations during reactions, necessitating a fresh approach for each scenario.

Lead author Dr. Chenru Duan elaborates, “If the reactant molecules are rotated, then in principle, before and after this rotation, they can still undergo the same chemical reaction. However, in the traditional machine-learning approach, the model will see these as two different reactions. That makes the machine-learning training much harder, as well as less accurate.”

The researchers embarked on a mission to develop a program capable of calculating transition states for two molecules, regardless of their orientation. Employing a generative AI tool called a diffusion model, they trained it using a database of 9,000 chemical reactions, encompassing molecular structures of reactants, transition states, and products.

Duan explains, “Once the model learns the underlying distribution of how these three structures coexist, we can give it new reactants and products, and it will try to generate a transition state structure that pairs with those reactants and products.”

The breakthrough came when the model was tasked with generating 40 potential transition states for 1,000 new reactions, with the aid of a confidence model that determined the most likely candidates. Astonishingly, the program swiftly delivered accurate transition states for each reaction, completing the task in mere seconds.

Dr. Kulik envisions the transformative potential of this approach, stating, “You can imagine that really scales to thinking about generating thousands of transition states in the time that it would normally take you to generate just a handful with the conventional method.”

While the researchers primarily worked with small molecules containing up to 23 atoms for their experiments, they observed the model’s capability to accommodate larger, more complex molecules. In essence, this advancement heralds a new era, where the formidable complexities of quantum chemistry may be replaced by the agility and efficiency of a fast generative model.


The integration of AI in chemical reaction analysis presents a game-changing opportunity for various industries, including pharmaceuticals, materials science, and biochemistry. Faster and more accurate determination of transition states will accelerate research and innovation, potentially leading to the discovery of novel solutions and products in a shorter timeframe, ultimately impacting the market positively.