How cloud computing could help store carbon dioxide in tiny rock pores

Scientists have made great progress to capture carbon dioxide (CO₂) and store it safely—there is no question about that. But CO₂ levels continue to rise.

To help find a way for global industries to achieve carbon capture and storage (CCS) goals, our IBM Research team has decided to combine high-performance computing (HPC) and hybrid cloud.

In a recent paper¹ published in Nature's Scientific Reports, we present a new, best-in-class algorithm for approximating tiny empty spaces—capillary networks—in porous rocks that naturally occur in geological formations. That’s where CO₂ captured from flue gas or other point emission sources in energy production could be safely stored in liquid or solid form.

Together with researchers from the São Carlos Institute of Physics-University of São Paulo, one of Brazil's top universities, and Petrobras, a Brazilian multinational corporation in the petroleum industry, we show that our algorithm performs a core analysis of a rock formation faster and more accurately than using lab tests alone. Our results could help cut the time for completing rock analysis from months to days, drive down costs, potentially increase efficiency and reduce risks of geological carbon storage.

Rock stars for carbon storage

The idea is to first compress the captured CO₂ gas into a liquid while it is injected into the rock’s pore network. Once injected, the liquid would be mineralized—turned into a solid—and could be stored efficiently, safely and for decades or even centuries.

While this is one of the most-promising solutions to our carbon storage problem, there are obstacles: How effective is this process given the existing reservoir conditions? What should the physics and chemistry be based on to account for the specifics of the rock? How can we accelerate CO₂ mineralization in the rock and ensure that geological storage is sustainable?

To address these questions, we developed algorithms that allow us to examine rock image data at high resolution. Using x-ray microtomography, we created a series of images that then formed a three-dimensional, digital representation of a rock sample. Next, we extrapolated our findings to predict what would happen at larger scale, in the same rock sample with more than 3,000 times the volume. This is the scale at which the lab characterization and validation of a rock sample’s permeability typically takes place. The high accuracy of the algorithm helps us to better estimate how much CO₂ these microscopic pores—with diameters ranging from nanometers to millimeters—could hold.

Opting for cloud and HPC

Once the spatial distribution of the connected pore space in the rock sample is known, we apply a flow model to computationally predict the fluid permeability based on the capillary network geometric boundaries. The ability to accurately approximate a rock’s capillary networks is key to quantifying fluid flow and CO₂ processes within the rock’s pores.

This is where cloud and high-performance computing comes in.

To create 3D computational rock representations, our simulator uses advanced image processing algorithms. These algorithms digitally segment a rock sample by analyzing its cross-section images with the pore spaces and other features represented in varying levels of gray. Once that segmentation is done, our new capillary network algorithm creates a highly accurate capillary network representation. It analyzes the void space that has been identified and approximates the volume of the connected pore space available in the rock.

In our paper, we show that the accuracy improvements achieved with the new algorithm allow us to extract scaling laws that connect a rock’s porosity with its permeability without the need for correction or calibration. Also, we demonstrate how rock permeability at lab scale emerges from the microscopic flow properties in the rock’s capillary network. These better predictions could be made routinely available at much shorter times compared to standard lab characterization.

The new numerical algorithm and digital rock simulation capability is built on our cloud-based FlowDiscovery Simulator. Presented in March at the 2021 Meeting of the American Physical Society, this tool helps to evaluate CO₂ trapping and, eventually, conversion scenarios at pore scale. The simulator could enable researchers and engineers to perform more rapid analysis and optimization of the rock-specific requirements for mineralizing and storing CO₂ efficiently, safely and long-term.

Maximizing impact with all the carbon-capturing tech

Although we have significantly improved the accuracy of pore scale flow simulations, we are not done yet. Our key challenge now is to extend the rock network simulations to include chemical processes and optimized materials that lead to a better mineralization of CO₂. Our ultimate goal is to discover highly efficient, low-cost materials and scalable methods for safe, long-term storage of carbon dioxide in our fight against climate change.

Looking ahead, IBM’s ability to build effective AI-based simulation tools and deliver them using hybrid cloud could help industries to better implement CCS and incorporate our work as a way to reduce the impact of CO₂ emissions. In FlowDiscovery we plan to use AI for shortlisting and optimizing materials leading to enhanced mineralization of CO₂. This way, AI could help us select the best options from a large catalog of available candidate materials.

Our results with researchers from São Carlos Institute of Physics and Petrobras show that such a model could work, and we’ve made the micro-tomography data of rock samples and benchmark network extraction algorithms publicly available to encourage other researchers to use and test them. Up to now, our digital rock data set has been downloaded more than 1,500 times.

This work is part of IBM’s larger Future of Climate initiative, launched in early 2020 to pool materials discovery technology and scientific know-how across the company’s worldwide network of research labs. The broader portfolio includes the research and development of strategies to reduce the carbon footprint of cloud computing and within the supply chain, as well as techniques to model the impact of climate change. In January, IBM also became an inaugural member the MIT Climate and Sustainability Consortium, along with other enterprises including Apple, Boeing, Cargill, Dow, PepsiCo and Verizon.