Scientists at MIT have devised a machine learning model to predict the stability of metal-organic frameworks (MOFs). Catalysing processes and storing gases are examples of how stable MOF structures can be used. Researchers are also looking into employing them to transport pharmaceuticals or imaging agents into the body.
MOFs feature a cage-like structure that is both stiff and porous. The porous structure of MOFs has piqued scientists’ curiosity for various gas-related uses, including gas storage, separation, and conversion. Researchers can tailor MOFs to multiple applications by altering the materials’ building blocks or the order in which they are organised. However, it’s important to note that not all conceivable MOF structures are robust enough for use in such settings.
“The stability of hypothetical MOF materials is not always known when they are first proposed.,” Heather Kulik, Associate Professor of Chemistry and Chemical Engineering at MIT and the study’s senior author, explained.
Therefore, the researchers used a computational model to narrow the possibilities to just under 10,000 “ultra-stable” MOF structures. The materials show promise for uses like transforming methane gas into methanol.
“To create materials with more stability than any previous collection of hypothetical materials, we used data and our machine-learning algorithms to arrive at building blocks that were anticipated to have outstanding stability, and then we rearranged them in ways that were greatly diverse.,” she continued.
Secondary building units (organic molecules containing metal atoms like zinc or copper) and organic molecules known as linkers are the two fundamental components of MOFs. Kulik compares these components to LEGO bricks because of the various configurations and possibilities to link both.
“There is a countless number of combinations possible using LEGO bricks,” she explained. “There is a combinatorial explosion of conceivable metal-organic framework materials in consequence. By carefully selecting which building blocks to use, the metal-organic framework’s final structure is entirely under one’s control.”
Kulik and her students recently used this methodology to identify over 500 MOFs with exceptional stability in their research. Next, the researchers disassembled the MOFs into their parts, identifying 120 secondary building units and 16 linkers as the most frequent.
The researchers developed about 50,000 new MOF structures by recombining these building pieces using approximately 750 designs. The researchers then utilised their computational models to estimate the stability of each of these 50,000 structures. Then, they narrowed it down to around 10,000, regarded as ultra-stable in terms of thermal and activation stability.
The structures were also evaluated based on their “deliverable capacity”. It is a measurement of gas storage and release efficiency. Researchers chose methane gas in their study because its capture could be valuable for degassing the environment or producing methanol. The anticipated elastic modulus was used to evaluate the mechanical stability of the discovered 10,000 ultra-stable materials. All those materials were shown to have good delivery capacities for methane.
The authors pinpointed specific building elements that result in more stable materials. A molecule containing the rare-earth metal gadolinium was shown to be one of the most stable secondary building blocks. Another important material to stabilise the structure is cobalt which contains porphyrin (a big organic molecule with four linked rings).
Currently, students in Kulik’s lab are synthesising and testing various machine learning MOF structures. The measurement is done to investigate further the stability, possible catalytic capacity, and gas separation ability. The team has also shared its ultra-durable materials database with other scientists interested in evaluating the materials for their purposes.