Rapid and significant gains against climate change require the creation of novel, environmentally benign and energy efficient materials. One of the richest veins that researchers hope to tap into to create such useful compounds is a vast chemical space where molecular combinations that offer remarkable optical, conductive, magnetic, and heat transfer properties await discovery.
But finding these new materials has been slow.
“While computational modeling has allowed us to discover and predict the properties of new materials much faster than experimentation, these models are not always reliable,” says Heather J. Kulik PhD ’09, associate professor in the departments of Chemical Engineering and Chemistry. “To speed computational discovery of materials, we need better methods to remove uncertainty and make our predictions more accurate.”
A team from Kulik’s lab set out to address these challenges with a team that included Chenru Duan PhD ’22.
A tool to build trust
Kulik and his group focus on transition metal complexes, molecules composed of metals that lie in the middle of the periodic table and are surrounded by organic ligands. These complexes can be extremely reactive, giving them a central role in catalyzing natural and industrial processes. By altering the organic and metallic components of these molecules, scientists can generate materials with properties that can enhance applications such as artificial photosynthesis, solar energy absorption and storage, higher-efficiency OLEDS (organic light-emitting diodes), and device miniaturization. .
“The characterization of these complexes and the discovery of new materials currently happens slowly, often driven by a researcher’s intuition,” says Kulik. “And the process involves trade-offs: You might find a material that has good light-emitting properties, but the metal at the center might be something like iridium, which is extremely rare and toxic.”
Researchers trying to identify earth-abundant, non-toxic transition metal complexes with useful properties tend to look for a narrow set of features, with only modest assurance that they are on the right track. “People keep iterating on a particular ligand and get stuck in local opportunity areas, rather than making large-scale discoveries,” says Kulik.
To address these detection inefficiencies, Kulik’s team developed a new approach: a machine learning-based “recommender” that lets researchers know the optimal model to perform their search on. His description of this tool was the subject of an article in Nature Computer Science in December.
“This method outperforms all previous approaches and can tell people when to use methods and when they will be trusted,” says Kulik.
The team, led by Duan, began by investigating ways to improve upon the conventional detection approach, density functional theory (DFT), which is based on computational quantum mechanics. He created a machine learning platform to determine how accurate density functional models were in predicting the structure and behavior of transition metal molecules.
“This tool learned which density functionals were the most reliable for specific material complexes,” says Kulik. “We verified this by testing the tool with materials it had never encountered before, where it actually chose the most accurate density functionals to predict the material property.”
A pivotal breakthrough for the team was their decision to use electron density, a fundamental quantum mechanical property of atoms, as a machine learning input. This unique identifier, as well as the use of a neural network model to perform the mapping, creates a powerful and efficient aid for researchers who want to determine if they are using the appropriate density functional to characterize their target transition metal complex. “A calculation that would take days or weeks, making computational detection almost unfeasible, may take only hours to produce a reliable result.”
Kulik has incorporated this tool into molSimplify, an open source code on the lab’s website, which enables researchers anywhere in the world to predict properties and model transition metal complexes.
Optimization for multiple properties
In a related research push, which they showcased in a recent post on JACS AuKulik’s group demonstrated an approach to rapidly localize transition metal complexes with specific properties in a large chemical space.
His work began in a 2021 paper showing that agreement on the properties of a target molecule among a group of functionals of different density significantly reduced the uncertainty of a model’s predictions.
Kulik’s team capitalized on this idea by first demonstrating multi-objective optimization. In their study, they successfully identified molecules that were easy to synthesize, with significant light-absorbing properties, using earth-abundant metals. They searched 32 million candidate materials, one of the largest gaps ever searched for this application. “We disassembled complexes already found in known experimentally synthesized materials and recombined them in new ways, allowing us to maintain some synthetic realism,” says Kulik.
After collecting DFT results on 100 compounds in this giant chemical domain, the group trained machine learning models to make predictions across the space of 32 million compounds, with an eye toward achieving their specific design goals. They repeated this process generation after generation to select compounds with the explicit properties they wanted.
“In the end, we found nine of the most promising compounds and found that the specific compounds we chose through machine learning contained pieces (ligands) that had been experimentally synthesized for other applications that required optical properties, with favorable light absorption spectra.” , says. Kulik.
applications with impact
While Kulik’s overall goal is to overcome limitations in computational modelling, his lab takes full advantage of its own tools to optimize the discovery and design of potentially impactful new materials.
In one notable example, “We are actively working on optimizing metallic-organic frameworks for the direct conversion of methane to methanol,” says Kulik. “This is a holy grail reaction that people have wanted to catalyze for decades, but have not been able to do efficiently.”
The possibility of a fast path to transform a very potent greenhouse gas into a liquid that is easily transportable and could be used as a fuel or as a value-added chemical has great appeal to Kulik. “It represents one of those needle-in-a-haystack challenges that multi-objective optimization and screening for millions of candidate catalysts are well positioned to solve, an outstanding challenge that has been around for so long.”
