What Metal-Organic Frameworks Are and Why AI Matters
Metal-organic frameworks are crystalline materials built from metal ions and organic linkers that form highly porous, tunable structures suitable for gas storage, separation, and chemical capture across many industrial applications. For years, this promise stayed locked in the lab because every new framework configuration demanded slow trial-and-error testing. Atomis is changing that by pairing metal-organic frameworks with AI material simulation, turning a traditionally linear research process into a loop of rapid digital experiments. Instead of synthesizing and testing each candidate step by step, the company screens large databases of potential structures in silico, guided by both performance and cost. This shift compresses development timelines for CO2 capture technology, refrigerant recycling, and advanced coatings, while trimming the number of physical experiments. It also points to a broader pattern: AI-accelerated material science is becoming a practical route to commercializing MOFs at scale.
Compressing MOF Development with Matlantis
Atomis has integrated Matlantis, a cloud-based atomistic simulation platform, into its early-stage workflow to speed the evaluation of new metal-organic frameworks. Matlantis uses a universal machine learning interatomic potential, known as Preferred Potential (PFP), trained on millions of density functional theory calculations. By treating atoms as nodes and their interactions as edges in a graph neural network, the system can model complex structures and reactions much faster than traditional methods. “If you want to handle a few hundred atoms, that can take days or weeks on the supercomputer, but with the help of machine learning methodology, we can essentially accelerate the simulation speed,” said Taku Watanabe of Matlantis. For Atomis, this means simulations that once waited in a shared supercomputer queue can run on demand, allowing researchers to iterate without interrupting experiments and to reserve high-accuracy DFT work for final verification of the most promising candidates.
From Lab Discovery to Commercial-Ready CO2 Capture and Refrigerants
Atomis is focused on turning metal-organic frameworks into practical products for CO2 capture technology, refrigerant recycling, deodorization, and coatings. The company aims at mass-market applications such as ton-scale CO2 separation, where MOFs can selectively trap greenhouse gases or refrigerant molecules within their porous structures. To reach that scale, Atomis shifts attention from pure performance to a balance of durability, cost, and manufacturability. As CEO Daisuke Asari notes, the team considers target price ranges and customer needs from the very first design stages instead of chasing exotic reagents alone. AI material simulation helps narrow down the millions of possible frameworks to those that can be produced efficiently and in large quantities. By building a database that combines structural information with cost data, Atomis is able to rank candidates not only by theoretical performance but also by commercial feasibility and potential for large-scale deployment.
Cutting Experimental Cycles and Communicating with Customers
AI-accelerated material science reduces experimental iteration costs for Atomis by moving much of the guesswork into the digital realm. Researchers use Matlantis to pre-screen MOF candidates, simulate gas adsorption behavior, and understand structural stability before they synthesize any material in the lab. They then use density functional theory as a final, high-precision filter on a narrower set of options. According to Atomis researcher YuhChyuan Chang, results from Matlantis matched those from a university supercomputer closely enough to allow continuous simulation without workflow interruptions. Python scripting on the platform lets the team run complex, condition-based calculation series with a single script, building datasets over time. The same simulations also support customer engagement: Atomis produces molecular videos showing gas molecules entering adsorption sites, giving clients a visual, evidence-based picture of how candidate MOFs will behave in real-world CO2 capture, refrigerant recovery, or coating applications.
