ΑΙhub.org
 

Improving the understanding of metal-organic frameworks


by
15 March 2023



share this:

Scanning electron microscope image of MOF crystalsScanning electron microscope image of MOF crystals. Image credit: CSIRO. Reproduced under a CC BY 3.0 licence.

By Nik Papageorgiou

How does an iPhone predict the next word you’re going to type in your messages? The technology behind this, and also at the core of many AI applications, is called a transformer; a deep-learning model that handles sequences of data in parallel, and can be fine-tuned for specific tasks.

Now, researchers at EPFL and KAIST have created a transformer for Metal-Organic Frameworks (MOFs), a class of porous crystalline materials whose potential applications include energy storage and gas separation. MOFs are composed of thousands of tunable molecular building blocks (metal nodes and organic linkers), and, considering all possible configurations, a vast number of MOFs could potentially be synthesised. Given this vast space, it is a challenge to find the material that has the characteristics you are looking for. One option is to use machine learning techniques to search the property-structure space.

The “MOFtransformer” developed by the researchers is based on the transformer architecture that forms the core of popular language models such as GPT-3, the predecessor to ChatGPT. The central idea behind these models is that they are pre-trained on a large amount of text, so when we start typing on an iPhone, for example, models like this autocomplete the most likely next word.

“We wanted to explore this idea for MOFs, but instead of giving a word suggestion, we wanted to have it suggest a property,” says Professor Berend Smit, who led the EPFL side of the project. “We pre-trained the MOFTransformer with a million hypothetical MOFs to learn their essential characteristics, which we represented as a sentence. The model was then trained to complete these sentences to give the MOF’s correct characteristics.”

The researchers then fine-tuned the MOFTransformer for tasks related to hydrogen storage, such as the storage capacity of hydrogen, its diffusion coefficient, and the band gap of the MOF (an “energy barrier” that determines how electrons can move through a material).

The approach showed that the MOFTransformer could get results using far less data compared to conventional machine-learning methods, which require much more data. “Because of the pre-training, the MOFTtransformer knows already many of the general properties of MOFs; and because of this knowledge, we need less data to train for another property,” says Smit. Moreover, the same model could be used for all properties, while in conventional machine learning, a separate model must be developed for each application.

The researchers hope that the MOFTransformer will pave the way for the development of new MOFs with improved properties for hydrogen storage and other applications.


The MOFTransformer library is available here.

Read the article: A Multi-modal Pre-training Transformer for Universal Transfer Learning in Metal-Organic Frameworks.




EPFL




            AIhub is supported by:


Related posts :



Livestream of RoboCup2025

  18 Jul 2025
Watch the competition live from Salvador!

A behaviour monitoring dataset of wild mammals in the Swiss Alps

  17 Jul 2025
Scientists at EPFL have created MammAlps, a multi-view, multi-modal video dataset that captures how wild mammals behave in the Swiss Alps.

#ICML2025 social media round-up 1

  16 Jul 2025
Find out what participants have been getting up to during the first couple of days of the conference.

Congratulations to the #ICML2025 award winners!

  16 Jul 2025
Find out which articles have won the outstanding paper, outstanding position paper, and the test-of-time awards.

Tackling the 3D Simulation League: an interview with Klaus Dorer and Stefan Glaser

  15 Jul 2025
With RoboCup2025 starting today, we found out more about the 3D simulation league, and the new simulator they have in the works.

What’s coming up at #RoboCup2025?

  10 Jul 2025
Find out when the different leagues competitions and the symposium are taking place.

Wildlife researchers train AI to better identify animal species in trail camera photos

  09 Jul 2025
Scientists are working on improving AI performance in wildlife monitoring through species and environment-specific training.

What’s on the programme at #ICML2025?

  07 Jul 2025
Find out what the International Conference on Machine Learning has in store.



 

AIhub is supported by:






©2025.05 - Association for the Understanding of Artificial Intelligence


 












©2025.05 - Association for the Understanding of Artificial Intelligence