Scientists are working hard to reduce plastic waste and make earth cleaner. They have found a new way to dissolve plastics using special liquids. Mitsubishi Chemical Group (MCG) has been a big help in this process. They have used fancy calculations to create a smart computer system that can figure out which liquids work best with different kinds of plastics. This helps us recycle plastics more effectively and be kinder to the environment.
Traditional methods mainly faced challenges due to a scarcity of experimental data on polymer-solvent miscibility. The new machine learning system addresses such limitations by integrating vast datasets from computer experiments using high quantum chemistry calculations.
The new study has been published in the Macromolecules journal. It is learned the researchers adopted a methodological framework called multitask learning. By amalgamating data from quantum chemistry calculations and real experiments, they have come up with a model that predicts the miscibility of any polymer-solvent combination with exceptional accuracy.
This predictive model allows for the selection and design of solvent molecules for recycling plastic waste. It can selectively separate different types of plastics within a mixture and create high-performance polymer blends.
Ryo Yoshida, author and researcher of the study, said that the new development of miscibilizers for various types of polymers will play important role in improving the recycling rates of waste plastics.
The efficiency of the developed model is noteworthy. It can calculate χ parameters approximately 40 times faster than conventional quantum chemistry calculations. This high speed facilitates rapid screening of millions of candidate solvent molecules.
Yoshida acknowledged the continuous need for improvement in machine learning techniques. To foster open innovation and science, the researchers made a portion of the source code and data available on the public domain.
