Hyderabad CSIR-IICT Scientists Develop ML Tool to Turn CO2 Into Fuel

Scientists at the CSIR-Indian Institute of Chemical Technology in Hyderabad have developed a machine learning tool to assist in converting carbon dioxide into dimethyl ether, an alternative fuel that can replace diesel or be blended with liquefied petroleum gas.

The framework is designed to predict how different catalyst compositions and operating conditions affect carbon dioxide conversion rates and dimethyl ether output, potentially cutting down years of trial-and-error laboratory experiments.

The research was conducted by Ganesh Kumar Ramachandran, Banoth Upendar, Reddi Kamesh, Ashok Jangam, Sreepriya Vedantam, and Venugopal Akula of CSIR-IICT. To build the system, the team compiled 330 experimental results across 39 peer-reviewed studies and evaluated 16 parameters, including catalyst characteristics and reaction conditions, across multiple machine learning models.

A Gradient Boosted Regression Tree model delivered the strongest performance. In testing on unseen data, the model achieved accuracy scores of 0.92 for predicting carbon dioxide conversion and 0.94 for predicting dimethyl ether selectivity. The researchers identified reaction temperature, pressure, and the silicon-to-aluminium ratio of the acid catalyst as the primary factors influencing performance.

The model relies on parameters available before a catalyst is synthesised, such as the elemental composition of active and promoter metals and support properties. This allows researchers to screen potential catalyst formulations on a computer prior to conducting physical experiments.

Dimethyl ether, also known as methoxymethane, can be produced from captured carbon dioxide and hydrogen. While currently used in India primarily within pharmaceutical and specialty sectors through imports from China and Japan, blending it with LPG at even a 20 percent ratio could help address fuel demand and import reliance.

Converting carbon dioxide directly into dimethyl ether involves a complex two-step reaction: generating methanol from carbon dioxide and subsequently converting methanol into dimethyl ether. The chemical stability of carbon dioxide, catalyst degradation, water formation, and competing reactions have historically limited process efficiency. The researchers noted that while the model helps optimise and design new catalysts, predictions must still undergo rigorous laboratory testing before any industrial deployment.