The process of dissolving polymers with organic solvents is crucial for the development of polymeric materials. This includes polymer synthesis, refining, painting, and coating. Recycling plastic waste has become increasingly important in reducing carbon emissions associated with materials development. However, the task of effectively dissolving plastics and plastic-like materials for recycling is not as simple as it seems.
Mitsubishi Chemical Group (MCG) has made significant strides in this field by utilizing a unique machine learning system. By leveraging their databank of quantum chemistry calculations, scientists have developed a predictive model to determine the miscibility of polymers with various solvent candidates. The key parameter used in this system is known as χ (chi). The integration of extensive data from computer experiments, using high-throughput quantum chemistry calculations, has successfully overcome the limitations posed by limited experimental data on polymer-solvent miscibility. The findings of this study were published in the journal Macromolecules.
With the application of multitask learning, the researchers have developed a highly accurate model to predict the miscibility of polymer-solvent mixtures. This model enables scientists to select and design solvent molecules that can selectively separate specific materials present in plastic waste consisting of different types of plastics. These specifically designed solvents, known as “miscibilizers,” can also contribute to the creation of high-performance polymer blends. The researchers emphasize the significance of these advancements in improving recycling rates and addressing the expectations for technological innovations in waste plastic resources.
One critical aspect of this research is the speed at which the predictive model can calculate the χ parameters. This model is approximately 40 times faster than conventional quantum chemistry calculations. With such rapid calculations, the model can efficiently screen millions of solvent molecules, significantly accelerating the screening process for potential candidates. The model’s accuracy in determining the requirements for transforming a polymer-solvent mixture into a homogenous substance suitable for recycling has simplified the trial and error process and reduced uncertainties.
While the model has demonstrated impressive capabilities, there are still opportunities for further enhancement. Currently, the model lacks the capacity to determine the impact of a polymer’s molecular weight or other compositional features on its miscibility. To tackle this limitation, the researchers have taken a progressive approach by making portions of the developed source code and data accessible to the public. This open innovation and crowd-sourcing approach can significantly expand the available data set. By incorporating a larger volume of diverse information, the model can learn and represent the true miscibility of polymers more effectively.
The ability to predict and understand polymer miscibility is a highly significant innovation in materials development and the ongoing pursuit of intelligent waste plastic recycling. As our society moves away from plastic-centric materials, advancements in polymer science become increasingly critical. With the continuous evolution of machine learning techniques and the adoption of open science practices, researchers aim to optimize the model’s capabilities further. Involving the public and encouraging contributions to the data set through open innovation and collaboration will undoubtedly accelerate the progress in materials informatics.
The development of a machine learning system for predicting polymer-solvent miscibility is a breakthrough in polymer science. Mitsubishi Chemical Group’s groundbreaking research, utilizing quantum chemistry calculations and integrating vast amounts of experimental data, has opened new possibilities in recycling plastic waste and advancing materials development. By embracing open innovation and open science, the researchers are actively seeking collaborations and contributions to improving the model. As this technology continues to evolve, it holds immense potential for driving sustainable practices and shaping the future of materials development in a decarbonized society.