Revolutionary Autonomous Platform Streamlining Polymer Blend Discovery for Enhanced Material Performance

In the realm of material science, the quest for innovative polymer-based materials is ongoing. Rather than embarking on a lengthy process of developing new polymers from scratch, researchers can save both time and resources by blending existing polymers to achieve specific desired properties.

However, finding the optimal blend presents significant challenges. The sheer number of possible combinations is vast, and the interactions between polymers are complex, making it difficult to predict the characteristics of a new blend.

To expedite the search for new materials, researchers at MIT have pioneered a fully automated experimental platform designed to efficiently identify the best polymer blends. This closed-loop workflow employs an advanced algorithm to explore a diverse array of potential polymer combinations, which are then processed by a robotic system for mixing and testing.

As experiments progress, the algorithm analyzes the results and determines subsequent experiments, iterating this process until the resulting polymer meets predefined user criteria.

During their investigations, the system autonomously discovered hundreds of blends that surpassed the performance of their individual components. Remarkably, the top-performing blends did not always rely on the highest quality components, underscoring the algorithm’s capacity to identify superior combinations.

This validates the importance of using an optimization algorithm that evaluates the entire design space simultaneously,” states Connor Coley, an assistant professor at MIT. “By considering the full formulation space, we can uncover new or improved properties that might otherwise go unnoticed.

This innovative workflow holds great promise for discovering polymer blends that could lead to significant advancements in various applications, such as enhanced battery electrolytes, cost-effective solar panels, and specialized nanoparticles for safer drug delivery.

The research team also included Guangqi Wu, a former MIT postdoctoral researcher now at Oxford University, along with graduate student Tianyi Jin and professor Alfredo Alexander-Katz from MIT’s Department of Materials Science and Engineering. Their findings are published in Matter.

When developing new polymer blends, scientists face an overwhelming number of starting materials. After selecting a few candidates for mixing, they must also determine the precise composition and concentration of each polymer within the blend.

The extensive design space necessitates algorithmic strategies and high-throughput workflows since exhaustive testing of all combinations is impractical,” Coley adds.

While autonomous workflows for single polymers have been explored, fewer studies have focused on polymer blends due to their significantly larger design spaces.

In this study, the MIT researchers aimed to develop random heteropolymer blends by combining two or more polymers with varying structural features. These adaptable polymers exhibit potential for high-temperature enzymatic catalysis, a process that accelerates chemical reactions.

Their closed-loop workflow starts with an algorithm that autonomously identifies several promising polymer blends based on targeted properties.

Initially, they experimented with a machine-learning model to forecast the performance of new blends; however, accurate predictions proved challenging given the vast possibilities. Consequently, they turned to a genetic algorithm inspired by biological evolution processes such as selection and mutation to seek optimal solutions.

This system encodes the composition of each polymer blend into a digital format akin to a chromosome, allowing the genetic algorithm to iteratively refine and identify promising combinations.

While the algorithm itself isn’t novel, we had to adapt it for our specific needs, such as restricting the number of polymers in a single blend to enhance discovery efficiency,” Wu explains.

Additionally, given the expansive search space, they calibrated the algorithm to effectively balance exploration—searching for random polymers—and exploitation—optimizing previously tested blends.

The algorithm simultaneously sends 96 polymer blends to an autonomous robotic platform for mixing and property measurement. The experiments primarily focused on enhancing enzyme thermal stability by optimizing retained enzymatic activity (REA), indicating how stable enzymes remain after exposure to high temperatures in combination with polymer blends.

Results from these experiments are relayed back to the algorithm, which continues generating new sets of polymer combinations until achieving optimal results.

Robotic System Challenges

• Developing methods for evenly heating polymers
• Optimizing pipette movement speed

This suggests that instead of solely creating new polymers , we may often enhance performance by strategically blending existing materials ,” Wu states .

Moreover , their autonomous platform can generate and assess up to 700 new polymer blends daily while requiring human intervention only for chemical refills and replacements .< p > Although this research concentrated on polymers for protein stabilization , the platform ‘s adaptability could extend to various applications such as developing new plastics or battery electrolytes .< p > Looking ahead , researchers aim to explore additional polymer properties while leveraging experimental data to refine their algorithms and improve operational efficiency of the autonomous liquid handling system .< blockquote > < p > ” With pressing demands for enhanced thermal stability in proteins and enzymes , these findings are promising . Given rapid advancements in machine learning and AI within material science , there is immense potential for optimizing random heteropolymer performance or tailoring designs based on specific applications ,” notes Ting Xu from UC Berkeley , who was not involved in this study . < p > This research was partially supported by grants from the U.S . Department of Energy , National Science Foundation , and Class of 1947 Career Development Chair .