�鶹��Ƶ

About 100 million metric tons of high-density polyethylene (HDPE), one of the world’s most commonly used plastics, are produced annually, using more than 15 times the energy needed to power New York City for a year and adding enormous amounts of plastic waste to landfills and oceans.

Cornell chemistry researchers have found ways to reduce the environmental impact of this ubiquitous polymer – found in milk jugs, shampoo bottles, playground equipment and many other things – by developing a machine-learning model that enables manufacturers to customize and improve HDPE materials, decreasing the amount of material needed for various applications. It can also be used to boost the quality of recycled HDPE to rival new, making recycling a more practical process.

“Implementation of this approach will facilitate the design of next-generation commodity materials and enable more efficient polymer recycling, lowering the overall impact of HDPE on the environment,” said ., associate professor of chemistry and chemical biology in the �鶹��Ƶ and �鶹��Ƶ (A&S).

“,” published March 14 in the Journal of the American Chemical Society, is a collaboration between DiStasio and polymer experts , the Tisch University Professor in the Department of Chemistry and Chemical Biology (A&S) and , the Frank and Robert Laughlin Professor of Physical Chemistry (A&S).

Jenny Hu, doctoral student, is the first author. From the DiStasio group, Zachary Sparrow, postdoctoral researcher; Brian Ernst, former postdoctoral researcher; and Spencer Mattes, doctoral student, contributed.

HDPE requires so much energy because it’s made on a huge scale, said Fors, whose lab focuses on sustainable polymers. There are also challenges to recycling it. 

“It’s more expensive to recycle polyethylene than it is to make virgin plastic,” he said. “Another problem is when you mechanically recycle it, you start breaking polymer chains, which causes the properties to degrade.”

HDPE materials lose quality every time they are recycled, Coates said. “You can’t just take these plastics and melt them down. It’s not like aluminum that’s perfect every time. You have to work hard to valorize it and make the plastics useful.”

Recyclers have about five cents to spend on valorizing – or boosting the quality – for each pound of recycled plastic, Coates said.

Currently, recycling facilities improve the quality of recycled output by adding a small amount of virgin plastic. However, the mix of recycled material varies day by day, making how much new plastic to add uncertain.

The key to using less material (and energy) manufacturing polyethylene – as well as controlling the quality and physical properties of recycled material – lies in understanding how the various lengths of polymer chains in a sample, called its molecular weight distribution, influences its properties. The key factors: how viscous it is during manufacturing and its strength and toughness as a finished product.

DiStasio and members of his lab trained their machine-learning model, called PEPPr (PolyEthylene Property PRedictor) using a library of more than 150 polyethylene samples synthesized and characterized by Coates, Fors and members of their labs.

“We needed a library of polymers with different molecular weight distributions,” DiStasio said. “We also wanted to have polymers with a diverse set of both processability and mechanical properties.”

Machine-learning power is necessary for the complex task of understanding the relationship between the composition of these materials and their properties, the researchers wrote. 

PEPPr solves two problems, DiStasio said. If the molecular weight distribution of an HDPE sample is known, the model can predict its properties: melt viscosity, toughness and strength. It can also be used for the inverse; if a user has a set of targeted properties in mind, the model can tell them what polymer sample would have those properties.

“If you want to make a plastic bag, you will need different properties in the melt than if you want to make a kayak,” Fors said.

The PEPPr approach is a first step toward smarter, more specific polymer design, as well as more effective and sustainable recycling processes, the researchers said. They plan to expand the scope of properties that can be predicted and add processing methods, which can be quite influential, to the model. They also hope to expand the model to include other polymer classes.

“We should be able to develop these types of models for any type of commercial polymer,” Fors said. “It should be a general way to tune properties and recycle other materials, as well.” 

This research was supported by the National Science Foundation Center for Sustainable Polymers and the Cornell Center for Materials Research with funding from the Research Experience for Undergraduates program.