Hyderabad: A CSIR-IICT Hyderabad study uses Machine Learning (ML) to identify catalyst combinations that could help convert captured carbon dioxide into dimethyl ether, a fuel that can be used as a diesel alternative or blended with LPG.

Carbon dioxide is a major greenhouse gas responsible for global warming. But researchers at Hyderabad’s CSIR-Indian Institute of Chemical Technology (IICT) are exploring whether captured CO₂ can instead be used as a raw material to make fuel.

What has the study identified?

The team has developed an ML framework that can help identify catalysts and operating conditions for converting carbon dioxide into dimethyl ether (DME), also known as methoxymethane.

DME can be used as a transport fuel, blended with LPG and in chemical manufacturing. The researchers say their approach could help reduce the trial-and-error involved in finding catalysts that make the conversion more efficient.

How does CO₂ become DME?

Converting CO₂ into DME is not a single-step process. According to the researchers, it involves first converting carbon dioxide into methanol, followed by converting methanol into DME.

The process is chemically challenging because CO₂ is a stable molecule and does not readily react. Other reactions can also occur, while water formation and degradation of the catalyst can reduce the efficiency of the process.

This is where the researchers believe machine learning could help.

The study, authored by Ganesh Kumar Ramachandran, Banoth Upendar, Reddi Kamesh, Ashok Jangam, Sreepriya Vedantam and Venugopal Akula, analysed 330 experimental results from 39 peer-reviewed studies.

How was study conducted?

The team considered 16 factors related to catalyst properties and reaction conditions and used them to train and test several machine learning models.

A Gradient Boosted Regression Tree (GBRT) model performed best. On previously unseen data, it recorded scores of 0.92 for predicting CO₂ conversion and 0.94 for DME selectivity.

In simpler terms, the model was able to predict reasonably well how much CO₂ could be converted and how much of the resulting product would be DME.

What does the model actually do?

The model does not make DME itself. Instead, it can help researchers decide which catalyst combinations are worth testing in the laboratory.

The researchers used information available before a catalyst is produced, including the composition of active and promoter metals and properties of the catalyst support.

The model identified reaction temperature, pressure and the Si/Al ratio of the acidic catalyst among the factors that had the greatest influence on performance.

This could allow researchers to screen potential catalyst designs on a computer first, potentially reducing the number of experiments needed.

However, the predictions still need to be experimentally validated. A catalyst that performs well in a model or laboratory may not necessarily be economical, durable or effective when used at industrial scale.

Could it help India reduce emissions?

The researchers also point to the potential use of DME in LPG blends. Given India’s large LPG consumer base, even partial substitution could create a sizeable market for the fuel.

But converting CO₂ into fuel is not automatically a climate solution.

Its climate benefit would depend on several factors, including where the hydrogen used in the process comes from, how the CO₂ is captured, how much energy the conversion requires and the emissions associated with producing and using the fuel.

Ready-made replacement for fossil fuels

The Hyderabad study therefore represents an early research step rather than a ready-made replacement for fossil fuels.

Its significance lies in using machine learning to make catalyst research more targeted. If the predictions continue to hold up through laboratory and eventually larger-scale testing, the approach could help scientists explore CO₂-to-fuel technologies more quickly and efficiently.

The researchers are not using AI to directly turn atmospheric CO₂ into fuel. They are using it to identify the chemical conditions that could make such conversion work better.