Research guided by AI reveals new molecules for stronger and more durable plastics

According to researchers from MIT and Duke University.

Using automatic learning, researchers have identified retication molecules that can be added to polymer materials, allowing them to resist more strength before tearing. These reticlations belong to a class of molecules called mechanophores, which modify their shape or other properties in response to the mechanical force.

“These molecules can be useful for making polymers which would be stronger in response to force. You apply them a certain stress, and rather than cracking or breaking, you see something that has higher resilience,” explains Heather Kulik, the Lammot teacher from the chemical genius bridge to MIT, who is also a teacher of chemistry and the senior author of the study.

The reticulators that the researchers identified in this study are compounds containing iron called ferroces, which so far had not been widely explored for their potential as mechanophores. The experimental assessment of a single mechanophore can take weeks, but researchers have shown that they could use an automatic learning model to considerably speed up this process.

MIT Postdoc Ilia Kevishvili is the main author of the newspaper in free access, which appeared in ACS Central Science.

The other authors include Jafer Vakil, a student graduated from Duke; David Kastner and Xiao Huang, both MIT graduate students; And Stephen Craig, professor of chemistry in Duke.

The weakest link

Mechanophores are molecules that react to force in a unique way, generally by modifying their color, structure or other properties. In the new study, the MIT and Duke team wanted to determine if they could be used to help make the polymers more resilient for damage.

The new work is based on a study in 2023 by Craig and Jeremiah Johnson, the chemistry teacher A. Thomas Guerin at MIT and their colleagues.

In this work, the researchers found that, surprisingly, incorporating low reticulators into a polymer network can strengthen the greater overall material.

When materials with these weak reticlations are stretched at the break point, all the cracks spread through the material try to avoid the strongest links and to go through the lowest links. This means that the crack should break more links than it would do if all the connections were of the same force.

To find new ways to exploit this phenomenon, Craig and Kulik have united their strengths to try to identify the mechanophores that could be used as weak reticulators.

“We had this new mechanistic insight and this new opportunity, but it came with a great challenge: from all the possible compositions of matter, how can we focus on those with the greatest potential?” Feig says.

“Total credit to Heather and Ilia to identify this challenge and design an approach to meet.”

The discovery and characterization of mechanophores is a difficult task that requires long experiences or intense simulations in calculating molecular interactions. Most known mechanophores are organic compounds, such as cyclobutane, which has been used as retication in the 2023 study.

In the new study, the researchers wanted to focus on molecules called ferolans, which hold potential as mechanophores. Ferrocenes are organometallic compounds that have an iron atom sandwiched between two rings containing carbon. These rings can have different chemical groups, which changes their chemical and mechanical properties.

Many ferrocenes are used as pharmaceuticals or catalysts, and a handle are known to be good mechanophores, but most have not been assessed for this use. Experimental tests on a single potential mechanophore can take several weeks, and calculation simulations, although faster, take a few more days. Evaluating thousands of candidates using these strategies is an intimidating task.

Realizing that an automatic learning approach could considerably accelerate the characterization of these molecules, the MIT team and Duke decided to use a neural network to identify ferrocenes that could be promising mechanophores.

They started with information from a database known as the Cambridge structural database, which contains the structures of 5,000 different ferrocenes which have already been synthesized.

“We knew that we did not have to worry about the question of synthesis, at least from the point of view of the mechanophore itself.

First, the researchers carried out calculation simulations for around 400 of these compounds, which allows them to calculate the force necessary to separate the atoms in each molecule. For this application, they sought molecules that would break quickly, because these weak links could make polymer materials more resistant to tear.

Then, they used this data, as well as information on the structure of each compound, to form an automatic learning model. This model was able to predict the force necessary to activate the mechanophore, which in turn influences the resistance to tearing, for the 4,500 compounds remaining in the database, plus 7,000 additional compounds similar to those of the database but reorganized atoms.

The researchers discovered two main characteristics which seemed likely to increase resistance to tear. One was the interactions between the chemical groups which are attached to the ferrocène rings. In addition, the presence of large bulky molecules attached to the two rings of the ferrocene made the molecule more likely to separate in response to the forces applied.

Although the first of these characteristics was not surprising, the second line was not something that a chemist would have predicted beforehand and could not have been detected without AI, according to the researchers. “It was something really surprising,” said Kulik.

More difficult plastics

Once the researchers have identified around 100 promising candidates, the Craig laboratory in Duke synthesized a polymer material incorporating one of them, known as M-TMS-FC. In the material, M-TMS-FC acts as a retication, connecting the polymer strands that make up the polyacrylate, a type of plastic.

By applying a force to each polymer until he has torn, the researchers found that the Lieur M-TMS-FC weak produced a strong and resistant polymer. This polymer has proven to be about four times harder than the polymers made with standard ferrocène as retication.

“This really has major implications, because if we think of all the plastics we use and all the accumulation of plastic waste, if you make the materials more difficult, it means that their lifespan will be longer. They will be usable for longer, which could reduce long -term plastic production,” explains Kevishvili.

Researchers are now hoping to use their automatic learning approach to identify mechanophores with other desirable properties, such as the ability to change color or become catalystically active in response to force. These materials could be used as a constraint sensors or switched catalysts, and they could also be useful for biomedical applications such as the administration of drugs.

In these studies, researchers plan to focus on ferrocpes and other mechanophores containing metals that have already been synthesized but whose properties are not fully understood.

“Transition metal mechanics are relatively under-explored, and they are probably a little more difficult to do,” said Kulik. “This calculation workflow can be widely used to expand the space of the mechanophores that people have studied.”

This story is republished with the kind authorization of Mit News (Web.mit.edu/newsoffice/), a popular site that covers news of research, innovation and MIT teaching.