In past decades, there has been a huge improvement in the standard of living of humans all across the globe. This improvement however has come with its own cost- Damage to Environment and Climate Change issues. Energy constitutes one of the major contributors to pollution and climate change be it by burning gasoline in cars or burning coal in power plants. A promising solution to the ever-increasing energy demands is the movement from fossil fuels to cleaner and renewable fuels like Hydrogen. While a Hydrogen based economy has been in talks for a lot of time, the movement is fraught with its own challenges. These include issues such as developing low-cost catalysts which drive the necessary chemical reactions at high speeds. Discovering new catalysts is a time and resource-intensive process which makes it difficult to discover new catalysts. To address these issues, Meta AI(formerly Facebook AI) has released a new dataset focused on Oxygen Evolution Reaction(OER) to accelerate the research and development of Hydrogen fuel.
Meta AI and Carnegie Mellon University’s (CMU) Department of Chemical Engineering have been working together on Open Catalyst Project. The aim of the Open Catalyst Project is to build machine learning models that simulate chemical reactions and accelerate the discovery of low-cost catalysts. A shortage of training datasets has been the roadblock for such projects now. But that is set to change with the Open Catalyst Project by Meta AI and CMU’s Department of Chemical Engineering.
As part of the project, the OC20 dataset which is the world’s largest training dataset of materials for renewable energy is also open-sourced.
The OC20(Open Catalyst 2020) Dataset by Meta AI and CMU focuses on oxide catalysts for the Oxygen Evolution Reaction (OER) which is a critical chemical reaction used in green hydrogen fuel production via wind and solar energy. The OER data set contains about 8M data points from 40K unique simulations which makes it the largest data set for oxide catalysis while spanning a swath of oxide materials across 52 elements. The data set and baseline models will be open-sourced in the next few months to help the global scientific community advance renewable energy technologies.
Identification and discovery of catalysts is a time-consuming and resource-intensive process but Machine Learning can accelerate the process. By replacing the Density Functional Theory(or DFT) which is currently used for the simulations with Machine Learning models, the time required for simulations can be brought down from hours or days to seconds.
Other Articles you might be Interested In
Google Launches a new game to educate people about Quantum Computing
Trends in Artificial Intelligence in 2021
Alphabet spins out SandBox AQ, an AI and Quantum Computing Company