Skip to main content
Stock photo of gel
July 7, 2020

Developing new models to revolutionize design of soft materials

Soft matter, such as a polymer or gel, holds promise for creating new kinds of multifunctional materials that could change their shape and properties in response to external stimuli.

Such materials would be useful for a variety of applications, including building soft robots that can adapt to their environment, squeeze through tight spaces and perform many different tasks.

“For materials made from soft matter, the functionality of the material begins to manifest itself at the mesoscale,” says Wenxiao Pan, an assistant professor of mechanical engineering at UW-Madison. “Therefore, the performance of materials—such as the photovoltaic performance of nanocrystal assemblies—can be dramatically altered by tuning their mesoscale features.”

 Wenxiao Pan
Wenxiao Pan

But in order to effectively design high-performance soft materials with tailored properties, researchers need a better understanding of the underlying mesoscale structures and processes. The mesoscale, which bridges the atomic and continuum scales, includes structures ranging in size from nanometers to micrometers.

To address this challenge, Pan is developing a theoretically sound and computationally efficient framework for modeling soft matter systems at the mesoscale. “This modeling framework will enable accurate analysis and prediction of mesoscale structures and dynamics in soft matter, and it could facilitate computer-aided design of multifunctional materials from soft matter,” she says.

The U.S. Department of Defense awarded Pan a $599,983 grant for the project through its Defense Established Program to Stimulate Competitive Research (DEPSCoR) competition, which selected a total of six projects for funding this year. The DEPSCoR funds teams to pursue science and engineering research in areas relevant to DOD initiatives supporting the national defense strategy.

In awarding her the grant, the DOD recognized Pan’s innovative ideas for this high-impact research.

“If successful, this project has the potential to revolutionize soft materials design, and may even lead to new data-driven soft matter informatics, akin to the promising field of hard materials informatics,” says Evan Runnerstrom, program manager at the Army Research Office, an element of the U.S. Army Combat Capabilities Development Command’s Army Research Laboratory. “This is important to the Army because soft matter is expected to form the basis for many highly adaptable technologies that can be deployed in multiple environments, like soft robotics.”

When researchers work on designing new materials starting from the atomic scale, the process usually involves running simulations by tracking individual atoms. “However, it is very expensive and requires a significant amount of computational resources to do atomistic simulations that will reach up to the mesoscale or the continuum scale,” Pan says.

The modeling framework Pan is developing promises much more efficient mesoscale simulations due to a new data-driven “coarse-graining” method. Essentially, her approach starts from atomistic descriptions and, by proper coarse-graining of the atomistic details, arrives at the mesoscale.

Pan likens the process of coarse-graining a complex system to reducing the dimension of the system. “So, you can think of it like reducing the dimension of an object from 3D to 2D, but we capture some of the features that this 3D object has, and then we focus on those features that the 2D plane also can capture,” Pan says.

Pan says this new data-driven method, which will leverage modern machine learning techniques, could overcome the limitations of existing course-graining methods.

Kaibo Liu, an associate professor of industrial and systems engineering, is a co-principal investigator on this project. He is contributing a novel machine learning technique, which will render the coarse-grained model constructed for a system transferable among different conditions. In addition, Liu will contribute to fundamental data science related to transfer learning of deep neural networks.