The world is coming to grips with human-caused contributions to climate change. IBM researchers are turning to AI, quantum computing, and other emerging technologies to discover new solutions faster. Their goal is to help mitigate the effects of climate change through the development of more environmentally friendly materials and ways to curb carbon emissions faster.
Researchers are using AI to help make important breakthroughs in the fight against climate change with carbon capture. Led by Mathias Steiner, scientists at IBM Research’s lab in Rio de Janeiro are applying AI to the accelerated discovery of a new polymer for CO2 filtering membranes.
In Steiner and his colleagues’ quest to find a new material that suited their needs, they first outlined the material’s desired properties, such as its permeability, chemical selectivity for specific gases, and durability, by creating data sets from scratch. They then turned to deep search to rapidly sift through large bodies of scientific papers, patents, and databases.
Using this data, Steiner’s team created predictive, generative AI models to construct a new molecule candidate for a polymer that could more efficiently separate CO2. The next steps involved using a high-performance computing (HPC) cluster within a hybrid cloud infrastructure to simulate the new molecule and its behavior under realistic gas separation conditions.
“We’ve decided to combine materials science, cloud computing and AI,” Steiner says. “The idea is to accelerate the discovery of low-cost, high-efficiency materials and methods for carbon dioxide capture, separation, and storage. We are not just creating the technology and data — we are also sharing it all with the world. After all, collaboration is really crucial in research.”
More recently, Steiner and team turned to HPC and hybrid cloud to help develop a new algorithm for approximating tiny empty spaces — capillary networks — in porous rocks where CO2 captured from flue gas or other emission sources in energy production could be safely stored in liquid or solid form.
These results could help cut the time for lab-scale rock analysis from months to days, while driving down costs, potentially increasing efficiency, and reducing the risks of geological carbon storage. The prototype simulation environment developed in the project, called FlowDiscovery, is now available for joint research and development with partners in academia and industry.
Steiner's team isn't the only one exploring AI solutions for climate change at IBM Research. Teo Laino's work revolves around applying machine learning to chemistry and materials science.
“We’re trying to identify more biodegradable molecules that perform equal to or better than the chemicals they replace,” Laino says. A distinguished research scientist, he is the main developer of the team’s latest cutting-edge tech — an autonomous chemical lab dubbed IBM RoboRXN is a cloud-based AI-driven robotic lab that efficiently automates a majority of the initial groundwork in materials synthesis, making it possible for scientists to synthesize materials remotely. How it works.RoboRXN aimed at creating molecules for materials that don’t yet exist.
Laino and his colleagues rely on a hybrid-cloud environment, as well as quantum computing and AI-based lab automation to accelerate the molecular discovery process. The method has already produced successful results: Over the last few months, IBM researchers have relied on AI and RoboRXN to create new photoacid generators, or (PAGs). These are molecules that semiconductor makers rely on for photolithography, a fabrication process that scales down processors while at the same time enabling more transistors to fit onto them.
“AI plays a key role in accelerating our work,” Laino says. “We need to rapidly find more biodegradable materials whose byproducts won’t damage the environment if they end up in the ocean and accumulate in the tissue of living things,” he adds. “It’s expensive to operate with existing chemicals because, as government regulations are passed, you have strong measures in place to make sure you’re not contaminating the environment.” The European Commission, for example, is now revising its regulations to phase out and ban certain harmful chemicals.
Much like with Steiner's team, AI and deep search have enabled Laino and his team to quickly collect and analyze all available information, including patents, papers and other literature, to identify new candidate molecules for creating less toxic chemical compounds.
The scientists also rely on AI to augment available data with simulations and in silico models to make new calculations, and use generative machine learning to create new molecules based on that information. These models involve discovering patterns in input data that can be used to create new examples which could plausibly been derived from the original dataset.
Once possible molecular configurations are determined, Laino's team uses RoboRXN to develop customized AI models to test a hypothesis against reality. “If the performance isn’t what the model or the hypothesis predicts, you can use the results to augment your initial knowledge and try again,” Laino says. “It’s an iterative process.”
Beyond creating new materials, it's important to create better approaches to manage technology, such as the hybrid-cloud resources Steiner employs in his research.
The IT ecosystem itself has contributed greatly to problem of excess of carbon emissions in the atmosphere, accounting for around 2% of global emissions. That’s on par with the aviation industry’s emissions from fuel, according to a 2018 study published in Nature.1 Some models predict IT electricity use could exceed 20% of the global total over the next decade or so, with data centers alone accounting for more than one-third of that.
The gravity of climate change really hit IBM Fellow Tamar Eilam, a 21-year veteran of IBM Research, at the 2019 ACM/IEEE International Conference on Software Engineering (ICSE). “I started asking questions around IBM about what the company was doing, and what more the company could do,” Eilam says. "My passion was about mitigation, how you could reduce CO2 emissions. That’s the number one thing we can do to slow climate change."
My passion was about mitigation, how you could reduce CO2 emissions. That’s the number one thing we can do to slow climate change.
The following January, Eilam joined IBM’s newly launched Future of Climate initiative, to accelerate the discovery of climate-change solutions. The initiative emphasizes the use of deep search, quantum computer-based simulation, generative machine learning models, and cloud-based autonomous labs like RoboRXN to accelerate the discovery of new materials, like complex polymers and materials for carbon CO2 capture and separation.2
Hybrid cloud unites these disparate technologies to provide researchers with the necessary computing resources. But hybrid cloud environments must do more than simply support groundbreaking discoveries — they must also play an active role in cutting data center energy use and emissions, Eilam says.
Data centers and enterprises can cut their energy use through dynamic mapping, resource allocation, placement, scheduling, configuration management, and other hybrid-cloud administration tools that dictate where and when different workloads run. AI helps find the best configuration that minimizes a cloud system’s carbon footprint. “You collect data and create models to assess what configurations will give you the best energy savings and pair particular workloads,” Eilam says. “Then, through reinforcement learning, you keep improving.”
Hybrid cloud management allows servers in one location to be able to tap into renewable energy sources during certain hours of the day, or at the very least, do as much of their data and workload processing as possible during off-peak hours. And serverless computing can take this concept even further: It ensures companies only use the computing resources they need, rather than having severs running all the time. “In that way, you optimize utilization by creating a hybrid cloud that expands and contracts as needed,” Eilam says.
While AI is already speeding up our fight against climate change, quantum computing is poised to potentially have a massive impact in the future. Flaviu Cipcigan is an IBM researcher working in quantum; he is inspired by nature and is a global technical leader for Future of Climate.
As an undergraduate student of computational physics at the University of Edinburgh, Cipcigan was recognized for his research into whether a certain type of quantum gate could be simulated using a classical computer. “I tried to figure out how a calculation could be changed so it could be done by a classical computer, which helps us better understand how to make quantum computers more powerful,” he says. For the past four years at IBM, his research has focused on determining how well AI and quantum computing can help with global challenges such as climate change or antimicrobial resistance.
Cipcigan is now particularly interested in designing new molecules for capturing carbon dioxide, an is currently supervising an Oxford graduate student investigating how quantum computers can help us better understand the shapes of certain molecules. Cipcigan wants to determine when quantum computing can help researchers more fully simulate the properties of new materials.
Practical quantum computers are still years away, but the technology has already shown the potential for accelerating discoveries in chemistry that could stimulate the development of more energy-efficient batteries and electronics. Mercedes-Benz, for example, is using IBM’s quantum computers to design quantum chemistry algorithms that could help develop new and better batteries.
Mitsubishi Chemical has also partnered with IBM Research to improve battery technology. In collaboration with the IBM Quantum Hub at Japan’s Keio University, the company is studying how quantum computers could create accurate simulations of what’s happening inside key chemical reactions at the molecular level.
More recently, E.ON became the first energy utility in Europe to join the network, exploring how the technology could optimize the world’s rapidly decentralized energy infrastructure. From E.ON’s perspective, distribution grids would soon have to fulfill a much wider range of tasks, as many smaller companies and households feed energy into the grid using their own photovoltaic solar energy systems, or their electric vehicles. Quantum computing could play an integral role in controlling these processes more efficiently, even as the increasing number of electric cars leads to more complex charging processes.
We are not alone in fighting climate change. Our goal at IBM is to reach net zero greenhouse gas emissions by the end of the decade. We will try our hardest to get there — while also helping the world become a much more sustainable place to live.
Watch the replay: The Future for AI & Quantum for Accelerated Discovery
To learn more about the ways that IBM Research is developing cutting-edge technology, watch a replay of the November 23 discussion from The Future for AI & Quantum for Accelerated Discovery event. The session explored the different ways RoboRXN and new quantum software tools such as the Qiskit application modules for chemistry, are putting materials discovery on the fast track.
- Dr. Pauline Ollitrault, researcher, quantum, IBM Research Europe-Zurich
- Dr. Teodoro Laino, distinguished researcher, manager, IBM Research Europe-Zurich
Materials Discovery: It can take over 10 years to come up with new materials. At IBM Research, we’re looking to accelerate the discovery process using new AI methods, robotics, the hybrid cloud, and quantum computers.
Climate and Sustainability: To respond to the global climate emergency, we're using AI and hybrid cloud to accelerate discovery of climate mitigation and adaption solutions.
Date22 Nov 2021
Jones, N. How to stop data centres from gobbling up the world’s electricity. Nature. 561, 163-166 (2018) ↩
Park, N., Zubarev, D., Hedrick, J., et al. A Recommender System for Inverse Design of Polycarbonates and Polyesters Macromolecules. 53 (24), 10847-10854. (2020). ↩