Explainer
7 minute read

Accelerating discovery for societal and economic impact

Much of the world’s ability to mitigate the effects of climate change will come down to our ability to quickly identify new materials that can be created, consumed, and recycled with minimal environmental impact.

Accelerating discovery of more sustainable materials through AI and automation

Much of the world’s ability to mitigate the effects of climate change will come down to our ability to quickly identify new materials that can be created, consumed, and recycled with minimal environmental impact.

Urgent problems call for urgent solutions. Consider materials, for example. We urgently need to design new materials to tackle societal challenges from fostering good health and clean energy to bolstering sustainability, climate action, responsible production — and so much more.

But there’s a challenge: the process of material design and discovery is traditionally long and complex. It takes roughly 10 years and upwards of $10 to $100 million on average to discover one new material with specific properties. And, there are more possible chemical element combinations than there are atoms in the universe.

Fortunately, advancements in computing can compress the time to discovery. The convergence of revolutions in AI, quantum computing and hybrid cloud is enabling us to consider a wider range of questions once thought out of reach.

At IBM Research, we are supercharging the scientific method by developing novel capabilities in areas like deep search and AI-enabled simulations that will empower us to formulate and test hypotheses with unprecedented speed and scale.

We are also promoting the concept of communities of discovery — bringing our tech and expertise together with industry, government, and academic partners, who want to tackle big problems. Our goal is to create global networks of open communities of discovery that brings together specialized resources from across the globe to prepare for and augment response to complex and urgent global crises.

Today, at the Aspen Ideas Festival, IBM Research held a forum on accelerating discovery of sustainable materials, and more generally across the spectrum of science. Remarks were made from IBM’s Dr. Darío Gil, senior vice president and director of IBM Research, Dr. Erwin Gianchandani, assistant director of the Directorate for Technology, Innovation and Partnerships (TIP) at the U.S. National Science Foundation (NSF), IBM’s Dr. Zaira Nazario, principal research scientist and quantum theorist, and Scott Dadich, founder and Co-CEO of Godfrey Dadich Partners and director of The Fastest Path to Progress.

The event convened individuals from in industry, science, technology, media, and policy to discuss the urgency and possibilities to accelerate the pace of discovery.

In support of this bold vision, Gianchandani spoke about NSF’s decision to establish a new directorate, the agency’s first in more than three decades.

“NSF supports researchers across all areas of science and engineering to follow their curiosity and discover something revolutionary,” Gianchandani said.

“It’s time to speed discovery to impact at speed and scale. Through the new TIP Directorate, they can take the next step and work across disciplines, partner with industry, develop their discovery into a product or launch a startup. The nation and the world are relying on these discoveries and new technologies to make an impact on climate change, improve our way of life, grow the economy and create new jobs.”

One new NSF initiative in the TIP Directorate, Gianchandani described, is the NSF Regional Innovation Engines, or NSF Engines for short. Designed to advance research inspired by pressing societal challenges, entrepreneurship, and inclusive workforce development, these NSF Engines could usher a revolution of business and economic growth.

NSF hopes that the NSF Engines tap into a diversity of demography, geography, and socioeconomic statuses to impact bottom-up, middle-out growth in industries and communities across America and create opportunities for loosely connected stakeholders — including industry, academia, government, potential beneficiaries and nonprofits — to unite as tightly integrated regional ecosystems of innovation, opportunity and prosperity.

IBM Research has sought to expedite the discovery of more sustainable materials by augmenting traditional manual, serial, and human-intensive work with automated, parallel, and iterative processes driven by artificial intelligence, simulation, and automation.

This work has begun to bear fruit in several key areas.

Capturing carbon emissions

Capturing CO₂ at the point of origin is thought to be one of the most effective ways to limit its release into the environment. Once captured, the gas could then be sequestered and stored for centuries. But capturing and separating CO₂ from exhaust gases in energy production and transportation is tricky, and moving CO₂ to a storage site so it doesn’t enter the atmosphere again is also far from trivial.

Typically, researchers go through a complicated, time consuming, and computationally expensive process to validate the properties of novel molecules to determine if they are strong candidates for use in polymer membrane materials. IBM Research has turned to an area of AI called generative modeling to accelerate the discovery and design of industrial-scale polymer membranes to more efficiently separate and capture CO₂ at its point of emission.

IBM researchers have used an AI inverse molecular design workflow to design and test new molecular building blocks. The researchers then performed molecular dynamics simulations of CO₂ separation performance to determine the suitability of those AI generated monomers as components for carbon capture polymer membranes.

Recently, the team also reduced the discovery-to-validation turnaround time per polymer candidate to several days in a standard computing environment, now enabling an accelerated discovery alternative to lab screening that traditionally takes months to complete. In addition, IBM researchers are currently applying and adapting the IBM Research GT4SD is an open-source library to accelerate hypothesis generation in the scientific discovery process that eases the adoption of state-of-the-art generative models.Generative Toolkit for Scientific Discovery (GT4SD) to cover advanced polymer classes.

Once the CO₂ is captured, it needs to be stored for the long term. Injection of CO₂ into rock formations for geological storage is a promising route to scale. The researchers’ latest work follows a series of advances in simulating and validating injection, trapping, and storage of CO₂ at pore scale. This includes the development of a Si/SiO₂ lab-on-chip1 platform for testing porosity models that represent reservoir rock on well-defined geometries at the microscale.

Toward a greener cloud

Semiconductors are core to much of the technology we use today, including the computers used to develop more sustainable new materials. However, various classes of chemicals used in semiconductor manufacturing — including photoacid generators (PAGs) — have become the subject of sustainable regulatory scrutiny over the years.

Industry research efforts have helped identify potential persistence, bioaccumulation, and toxicity (PBT) risks associated with certain PAGs, as well as relevant photodecomposition products. As a result, it remains extremely challenging to design, synthesize, test, and bring to market new PAGs with improved sustainability profiles in a timely manner.

IBM researchers are developing an end-to-end AI-powered workflow to accelerate the design and synthesis of PAG candidates with lower PBT risk profiles by incorporating environmental, health and safety endpoints into the design process. Part of our Project Photoresist, this work serves as a prototype for future Accelerated Discovery workflows.

Better batteries for renewable energy

Battery-powered alternatives are being developed to lessen the environmental impact of electric vehicles and utility grid-scale energy storage, to name a few key areas. Despite years of work, however, many question the effectiveness of available battery technologies, the sustainability of the materials and processes used in their manufacture, and their post-lifespan recyclability.

Lithium-ion batteries, the most advanced battery technology on the market today, utilize heavy metals that pose environmental and health concerns. They are difficult to recycle and their supply chains in Central Africa and elsewhere have come under scrutiny for careless and exploitative extraction practices.

IBM Research has sought to accelerate scientific discovery of more sustainable materials by augmenting traditionally manual, serial, and human-intensive work with automated, parallel, and iterative processes driven by AI.

IBM researchers are developing EV batteries free of cobalt, nickel, or other heavy metals. The researchers are using AI-assisted workflow to discover safer and higher-performing electrolyte materials for high power and fast charging new batteries using more sustainable cathode materials, including iodine extracted from brine.

IBM Research’s electrolyte discovery workflow constitutes high throughput screening of component materials based on their physico-chemical properties obtained by AI-assisted quantum chemical simulations. Novel deep learning models then map the structure-composition-performance relationship of new electrolyte formulations. Implementation of such AI workflow reduces the required volume of experiments needed to find safer (less flammable) and higher performing electrolyte systems.

A faster path to biodegradable plastics

Plastics are ubiquitous and useful modern materials, but their widespread utility and indiscriminate disposal has also left an adverse and enduring environmental legacy. Unless we change how we produce, use, and dispose of plastic, the amount of plastic waste entering fragile aquatic ecosystems could nearly triple from 9 to 14 million tons per year in 2016 to a projected 23 to 37 million tons per year by 2040.

Researchers are looking to develop materials and chemical processes that reduce or circumvent the negative environmental impacts of pollution and hazardous compounds. Growing demand for tailored polymers made from more biodegradable stocks and easier to recycle requires innovative synthetic design thinking.

IBM researchers see continuous flow reactors — as opposed to applying batch processes — as a faster, more efficient way to drive research for developing more environmentally friendly biocompatible and biodegradable polymers. Continuous flow chemistry conducts chemical reactions in a pipe or tube, facilitating the mixing of reactants and the control of reaction parameters.

IBM researchers have automated continuous flow reactor configurations and developed new catalytic systems for improving step-growth and chain-growth polymerizations, shortening conventional processes from hours to minutes, and to seconds and milliseconds, respectively, while conducting the reactions at room temperature.

AI for the chemical industry’s green transition

IBM Research is likewise working to make the chemical industry cleaner and more sustainable through AI and the customization of tools such as IBM RXN for Chemistry is the free AI tool in the cloud for digital chemistry. Start a project, here.IBM RXN for Chemistry — available on IBM Cloud — on sustainable processes. One of our goals is to use enzyme-based biochemical processes to clean up traditional chemical processes that use toxic solvents, high temperatures and pressures, and generate a lot of waste.

Enzymes are one of the key enablers of sustainable chemical processes because of their extremely high catalytic activity, chemo-, regio-, and stereo-selectivity, reusability, and ease of product recovery. Enzymes are still not widely used, however, as chemists work to bridge the knowledge gap between large corpora of enzymatic chemical reaction data and our understanding of structure-activity relationships. Such relationships ultimately impact the decision of which enzymes can be used to replace traditional processes.

With the newly developed models, IBM RXN for Chemistry was able to predict the product of a given enzyme-mediated reaction nearly half of the time on the first try.

Using AI to capture hidden patterns in large enzymatic data sets is a powerful strategy, and IBM researchers have integrated enzymes into retrosynthetic chemical reaction planning tools for the first time using machine learning. The IBM Research Europe team in Switzerland, for example, taught its RXN for Chemistry tool to predict reaction outcomes and plan chemical syntheses using enzymatic data from public biochemical reaction databases.

With the newly developed models, IBM RXN for Chemistry was able to predict the product of a given enzyme-mediated reaction nearly half of the time on the first try. The technology could also devise green retrosynthetic routes based on enzymes that produce a desired target chemical nearly 40% of the time.

As of June 2022, IBM RXN for Chemistry has been used to generate more than 8 million new predictions by users and businesses worldwide including Thieme. Coming soon as a full IBM service, we invite you to learn more about this autonomous lab in the cloud by watching this video.

Joining together to advance discovery

Looking ahead, IBM Research is committed to continuing the exploration and application of technologies and techniques that advance material science and can ultimately create a better planet for tomorrow.

The fastest path to progress

Notes

  1. Note 1GT4SD is an open-source library to accelerate hypothesis generation in the scientific discovery process that eases the adoption of state-of-the-art generative models. ↩︎
  2. Note 2IBM RXN for Chemistry is the free AI tool in the cloud for digital chemistry. Start a project, here. ↩︎

References

  1. Jaione Tirapu-Azpiroz, Matheus Esteves Ferreira, Ademir Ferreira Silva, Ricardo Luis Ohta, Rodrigo Neumann Barros Ferreira, Ronaldo Giro, Benjamin Wunsch, Mathias B. Steiner, "Advanced optical on-chip analysis of fluid flow for applications in carbon dioxide trapping," Proc. SPIE 11955, Microfluidics, BioMEMS, and Medical Microsystems XX, 1195507 (2 March 2022); https://doi.org/10.1117/12.2610336