BASF, Dow, Syensqo and 3M Speed R&D With AI, Robotics

For 18 months, Syensqo researchers parsed through a database of 4 million molecules to land on a single polymer that met their desired properties for a new product.

The researchers used artificial intelligence to accomplish the feat, which would have taken an estimated five to six years using traditional methods, said Ryan Murphy, the company's technology manager, in September at the CIEX conference in Indianapolis.

Chemical manufacturers have documented successful digital implementations at the operations level to deliver predictive maintenance insights, enhanced equipment performance and process safety.

Increasingly, chemical producers are extending digital capabilities to their research and development functions. CIEX, short for Chemical Innovation Exchange, brought together R&D leaders from some of the largest chemical producers in the world. Some of the results they cited in their presentations were staggering. In addition to Syensqo's new polymer discovery, presenters talked about supercharging R&D processes using AI, IoT and robotics.

The push toward smart R&D is happening simultaneously and perhaps not coincidentally with the drive for more sustainable materials. Manufacturers like 3M, for example, are scrambling to find alternatives to per- and polyfluoroalkyl substances, or PFAS, after bans and lawsuits have forced these companies to discontinue production.

They are pursuing parallel digital transformations across their operations — accelerating sustainable product development in research labs while separately optimizing manufacturing processes on plant floors. Here's a closer look at how four major chemical producers are using digitalization to develop new innovations faster than ever.

BASF Accelerates Catalyst Development

BASF is using AI to mine published literature and internal datasets to identify multi-metal catalysts, said Amit Gokhale, BASF's director of process and chemical engineering R&D. The company has an internal division focused on these efforts in collaboration with partners like Halo, an AI-powered matching platform that connects researchers with industry partners, facilitating collaboration opportunities.

"BASF is the largest catalyst manufacturer. We are trying to see how we can use some of these tools for developing catalysts faster and increasing the timeline for deployment of catalysts into our plants," Gokhale told Chemical Processing.

The researchers are applying AI to various modeling methods like density functional theory, a quantum mechanics calculation, to come up with new catalyst formulations. The company had used these types of tools in the past but typically applied them after the fact to explain why a catalyst worked well.

"But now all that data that's there in the literature, we can harvest that, and we can use that for developing new formulations," Gokhale said. "We have actually a few success stories from that as well, which are currently patented."

Looking ahead, Gokhale envisions a future in which AI tools will eventually suggest process reconfigurations by mining patent literature, potentially recommending changes like switching distillation techniques.

"I have no doubt that in four or five years down the line we will have created tools like that, which are going to reconfigure existing plant processes," he said.

3M’s Quest for New Materials 

3M is in the midst of a multiyear digital R&D transformation involving investments in computational chemistry, which spans from predicting molecular properties and designing small molecules to modeling the performance of complex, fully formulated adhesive systems, Guy Joly, 3M's vice president of R&D told conference attendees.

Last year, 3M launched its Materials Data Hub, which comprises a library of material data cards that enable customers to predict adhesive performance and reach material selection decisions faster, said Joly. These data cards are created through extreme mechanics — extensive mechanical characterization of adhesives combined with sophisticated mathematical models of adhesive performance.

"As we develop new materials, we're generating more material data cards and kind of feeding this virtual cycle," Joly said.

3M also has begun investing in lab automation to advance its efforts around new adhesives. One major focus has been addressing bottlenecks in rheology testing — the study of how materials flow under stress. The company invested in high-throughput capabilities that allow researchers to load data into the cloud, run experiments overnight and find processed data ready for analysis the next morning.

3M researchers are using machine learning to collect rheology data two to four times faster than they could using traditional methods with about 10 times the data density, Joly said.

These digital R&D capabilities are also being applied to 3M's search for alternatives to PFAS, synthetic compounds often called "forever chemicals" due to their environmental persistence.

In December 2022, 3M announced it would exit all PFAS manufacturing and discontinue its use across its product portfolio by the end of 2025. The PFAS product line represented approximately $1.3 billion in net sales, and the company said it would work to facilitate an orderly transition for customers while fulfilling current contractual obligations.

Joly told Chemical Processing that computational chemistry is playing an important role in the PFAS alternatives effort.

"Certainly, computational chemistry plays a big role there … in terms of informing design," he said, though he declined to elaborate on specifics.

Syensqo Supercharges Polymer Discovery

Syensqo (pronounced Science-Co.), spun off from Solvay in 2023, is using AI to help lower its product carbon footprint in an increasingly uncertain geopolitical climate, Murphy said.

The company launched an AI-focused division called Syensqo AI about two years ago to implement AI initiatives across the organization. This centralized team handles everything from negotiating partnerships with companies like Microsoft to building internal tools and managing backend infrastructure. Part of this process involved education and training to ensure employees understood and embraced the new technology, Murphy said.

Murphy said the company adopted a strategy from the marketing world: moving employees from awareness to advocacy. The company created a network of AI advocates spread across different geographical sites who could answer questions and demonstrate practical applications. Syensqo also launched "AI office hours" where employees could gather with their computers to ask questions with trusted peers.

For the polymer discovery project, Murphy emphasized that starting with AI tools from the beginning proved critical to success. "When you come in with these tools midway, they're not as effective," he said.

The team assembled data scientists, modeling experts and experimental chemists at the project's outset. They leveraged internal databases the company had built, accessed external databases, and tapped into open-source and third-party simulation tools.

Syensqo has also integrated robotics into its R&D workflows, particularly in formulation work. In another example, the company used AI to comb through thousands of surfactant chemistries, examine the different combinations and then use robotics to conduct experiments, Murphy said.

The company operates a laboratory dedicated to automation and robotics, inherited from its Solvay legacy. The robotic systems handle the mixing and downstream testing of promising candidates identified through AI analysis.

"It's possible to do with humans and scientists in the lab, but this just speeds everything up," Murphy said. "It's more tedious work that was eliminated and really getting to the more interesting results."

Dow’s R&D Robots

Dow employs robotics to enable high-throughput research. This process allows the company to bring discoveries to market twice as fast, said A.N. Sreeram, Dow's chief technology officer.

Sreeram cited an example of how Dow uses robotics to accelerate product development in consumer applications like surface cleaners and detergents. The robots dispense various chemical precursors to create initial formulations for primary screening tests. In secondary screening phases, the robots handle tasks like making polymer samples and precisely weighing materials before and after reactions.

The company's automated systems can produce over 10,000 samples per week — a scale impossible to achieve with traditional manual methods.

"You cannot hire unskilled labor in Timbuktu to be lower cost," he said. "It's not about the cost."

The automated process allows researchers to precisely control formulation variables and use digital line scans to quantify properties like foaming characteristics in soaps.

Sreeram also described Dow's work developing materials to prevent thermal runaway in lithium-ion batteries — a dangerous condition where one overheating battery cell can cause neighboring cells to catch fire, potentially destroying an entire battery pack.

The company uses high-throughput robotics to formulate silicone foam materials that can contain battery fires. Their testing involves deliberately triggering thermal runaway in one cell to see if their materials can prevent the fire from spreading to adjacent cells.

Sreeram cited additional examples of high-throughput R&D, including the use of quantum computing and digitized research reports to advance next-generation technologies. Dow has digitized all the company's research reports dating back to its early years.

"Along with all the data and literature, we run a lot of deep learning and ML work," Sreeram said.

Navigating IP, Security and Workforce Concerns

As AI capabilities advance, companies are grappling with emerging challenges around patent protection, data security and responsible implementation. To protect proprietary data, BASF conducts its AI work on internal servers rather than using external cloud platforms. The company imports data into its own systems and keeps all information in-house, Gokhale said.

Gokhale cautioned that advancing AI capabilities raise complex questions about patentability. One concern is whether patent examiners might consider AI-generated innovations "obvious" if the technology can combine existing knowledge in ways humans hadn't.

"If you have an AI tool which is now picking things from literature, would you consider that as someone skilled in the art or not?" Gokhale said. "Those things are interesting challenges which would come up, I think, in the very near future."

Murphy said Syensqo has prioritized responsible AI adoption as the technology evolves rapidly. The company has signed AI responsibility policies with employee resource groups worldwide to foster dialogue and build trust around the technology's deployment.

"We're very excited about this technology and the tools, but really one of the more important parts about this: We can't just imagine we're going to throw these technologies, generative AI and all these tools, over the fence and it's going to fit us perfectly,” Murphy said. ⊕