Modeling the Complex Chemistry of Carbon Capture with AI

Carbon capture, utilization, and storage (CCUS) will be an important technology for limiting climate change. But certain carbon capture technologies can release their own emissions in the form of amines, ammonia derivatives that can be carcinogenic.

Now, engineers have developed a new way to model the emissions from real-world carbon sequestration facilities using machine learning. In the process, they discovered that a second-generation solvent, CESAR1, which has been proposed as a new benchmark to replace the most commonly used solvent in scrubbing processes, monoethanolamine (MEA), may actually require more complicated emissions mitigation measures. This raises questions about CESAR1’s economic feasibility.

“You need to take into consideration the cost of the countermeasures to evade emissions,” says Susana Garcia, a chemical engineer and professor at Heriot-Watt Univ. in Edinburgh, who co-led the new research with Berend Smit, a computational chemist at École Polytechnique Fédérale de Lausanne (EPFL).

The project began at the pilot carbon sequestration plant at Niederaußem — west of Cologne, Germany — which is connected to the country’s second-largest coal-fired power plant. The carbon capture plant, which houses a conventional amine scrubbing process, was designed to use MEA as a solvent, but had switched to the solvent mixture CESAR1 about a year prior to the study. Garcia was tasked with stressing the plant to see the effect of...