Kevlar, Teflon, and nylon. These are only a few examples of well-known polymers, or large-molecule chemicals, that have had a profound impact on society. Polymers are essential for building the systems that improve the way the world works, from 3D printing to Teflon-coated frying pans.
It’s never easy to find the next revolutionary polymer, but Georgia Tech scientists are utilizing artificial intelligence (AI) to mold and reshape the industry’s future. The team led by Rampi Ramprasad creates and modifies AI algorithms to speed up the discovery of new materials.
Two articles that were released this summer in the Nature family of journals showcase the noteworthy developments and triumphs that have resulted from years of AI-driven polymer informatics research.
The first, published in Nature Reviews Materials, highlights current developments in polymer design in three important and modern application domains: recyclable plastics, energy storage, and filtering technologies. The second, which was published in Nature Communications, focuses on using AI algorithms to find a subclass of polymers that may be used for electrostatic energy storage. The materials that are intended for this purpose have successfully undergone testing and synthesis in the lab.
According to Ramprasad, a professor in the School of Materials Science and Engineering, study in this area was primarily motivated by curiosity in the early days of artificial intelligence in materials science, which was sparked by the White House’s Materials Genome Initiative more than ten years ago.
We have just recently started to witness concrete, real-world examples of AI-driven rapid polymer discovery success. These achievements are currently driving important changes in the field of industrial materials R&D. This is the reason this review is so important and relevant.
AI prospects
Innovative algorithms that can quickly forecast polymer properties and formulations before they are physically formed have been developed by Ramprasad’s team. Defining application-specific target properties or performance requirements is the first step in the process.
To forecast these intended results, machine learning (ML) models are trained using material-property data that already exists. The group can also produce new polymers, the properties of which are predicted using machine learning models.
Following laboratory synthesis and testing, the best candidates who satisfy the desired property requirements are chosen for validation in the real world. Through an ongoing, iterative process of integration between the outcomes of these new tests and the original data, the predictive models are further refined.
AI offers particular difficulties even if it can hasten the search for novel polymers. High-quality initial data is crucial since the precision of AI predictions relies on the availability of rich, varied, and large starting data sets. Moreover, creating algorithms that can produce synthetic and chemically realistic polymers is a challenging task.
Once the algorithms produce their predictions, the actual work begins: establishing that the proposed materials can be produced in the laboratory and perform as anticipated, as well as proving that they can be scaled up for usage in the real world.
These materials are designed by Ramprasad’s team, and partners at multiple universities, including Georgia Tech, handle their fabrication, processing, and testing. Co-author of the research published in Nature Reviews Materials, Professor Ryan Lively of the School of Chemical and Biomolecular Engineering works closely with Ramprasad’s group on a regular basis.
They make substantial use of the machine learning models produced by Rampi’s team in their daily research, according to Lively.
These technologies let them work more quickly and explore new ideas more quickly. Because they may make model-guided decisions prior to devoting time and resources to laboratory exploration of the concepts, this epitomizes the promise of machine learning and artificial intelligence.
Ramprasad’s group and associates have used AI to make major strides in a variety of areas, such as energy storage, filtration technologies, additive manufacturing, and recyclable materials.
Polymer progress
One significant accomplishment is the creation of novel polymers for capacitors, which are devices that store electrostatic energy. This work is detailed in the Nature Communications report. These gadgets are essential parts of hybrid and electric cars, among other things. University of Connecticut researchers collaborated with Ramprasad’s group.
Currently available capacitor polymers provide thermal stability or high energy density, but not both. The researchers discovered that insulating materials composed of polynorbornene and polyimide polymers can achieve high energy density and good thermal stability at the same time by utilizing AI tools.
To work in harsh conditions, such aircraft applications, the polymers can be further improved while yet being environmentally sustainable.
According to Ramprasad, one of the most obvious instances of how AI may direct the development of novel materials is the new class of polymers with high energy density and great thermal stability. It is also the outcome of years of multidisciplinary collaboration with the Office of Naval Research, as well as support from Greg Sotzing and Yang Cao at the University of Connecticut.
Industry potential
Industry participation in the Nature Reviews Materials publication highlights the potential for practical application of AI-assisted materials development. Scientists from General Electric and Toyota Research Institute are also co-authors of this research.
Recently split out of Georgia Tech, Ramprasad co-founded Matmerize Inc., a software firm, to further accelerate the use of AI-driven materials creation in industry. Companies in a variety of industries, including energy, electronics, consumer goods, chemical processing, and sustainable materials, already employ their cloud-based polymer informatics software.
Matmerize has translated its study into a strong, adaptable, and industry-ready technology, allowing customers to digitally create materials with increased efficiency and lower costs, according to Ramprasad.
What began as a curiosity has grown into a serious movement, and we are entering an exciting new era of material design.
