Chemistry

Clean technologies from molecular design

 

Tackling global challenges in sustainable energy production calls for the rational design of new generations of clean materials, coupling efficiency with low environmental impact.

 

Professor Rajan has established a research center at the Department of Materials Design and Innovation, University at Buffalo, New York that explores how to develop ‘quantum signatures’ at a fundamental molecular scale that can guide the rational design of clean materials.

 

Read their original article: https://doi.org/10.1038/s41597-021-00798-x

 

Read more in Research Outreach

 

Image credit: HQuality/Shutterstock

 

 

Transcript:

Hello and welcome to Research Pod! Thank you for listening and joining us today.

 

In this episode we will be looking at the research of Professor Krishna Rajan at the Department of Materials Design and Innovation, University at Buffalo, New York. Using materials informatics, Professor Rajan takes an a priori approach to the discovery of new materials with a heavy focus on minimising environmental impacts without compromising engineering functionality.

 

Tackling global challenges in sustainable energy production calls for the rational design of new generations of clean materials, coupling efficiency with low environmental impact. Materials discovery offers a powerful means to predict structure-property relationships in novel materials, but often lacks the ability to account for the complex interplay of length- and time-scales that ultimately determine the potential of a material to exhibit desired functionalities.

 

The Department of Materials Design and Innovation, or MDI, at University at Buffalo, New York, pioneers the development and application of materials informatics. Through this unique approach to materials science, the MDI is solving problems in environmental sustainability, renewable energy, and climate change.

 

One of the fundamental issues in materials science, and in the design and synthesis of technologically relevant materials, concerns the so-called structure-property relationship. Given the structure of a molecule or crystal, how can we infer which properties the system will exhibit?  Importantly, how will it behave in realistic conditions, and how will it interact with its environment?

 

In recent years, the development of advanced quantum mechanical methods has provided a huge amount of information concerning the properties of wide classes of materials. These methods can account for a system’s stability, chemical bonding patterns, and even electronic or optical properties, along with the use of sophisticated experimental techniques to probe the structure and properties of materials. These approaches can often be used to guide the discovery of new materials with enhanced capabilities, for instance for the clean production or storage of energy, or for the removal of toxic agents from the environment.

 

However, using large amounts of information for property prediction in a robust and systematic way is highly problematic, even when complex statistical or machine learning based approaches are used to explore the available data. The problem becomes even more complex when the properties of completely new materials, bearing little resemblance to known ones, have to be investigated and predicted.

 

Professor Rajan explains that “The challenge is that structure-property relationships are often non-linear, and the goal is to seek patterns among multiple length scales and timescales. Rarely can a single multiscale theory meaningfully and accurately capture such information.”

 

To address this challenge, we therefore not only need an efficient strategy for searching large data sets, but also techniques to learn from those data, by generating information that can be used outside the models on which the data themselves are based.

 

Professor Rajan has established a research center, within the Department of Materials Design and Innovation, Collaboratory for a Regenerative Economy, that explores how to develop ‘quantum signatures’ at a fundamental molecular scale that can guide the rational design of clean materials.

Using advanced machine learning techniques, they can bridge the gaps in knowledge that theory and heuristics alone cannot capture. His approach marks a change of paradigm in materials discovery by shifting the emphasis from simply searching among large volumes of data to identifying the key materials physics and chemistry (or “inorganic genes”) that can be integrated in statistical and machine learning methods.

This can provide design rules that govern targeted functionalities in new materials. These inorganic genes encode the fundamental information that collectively characterises the stability and properties of a material. They allow us to map the complexity of a material description of how these fundamental pieces of information self-organise and self-assemble, ultimately giving a material its characteristic properties.

 

The materials informatics approach aims to design materials with targeted properties. It involves choosing a number of structural, chemical and processing parameters on the basis of experimental and theoretical evidence. During this parameter selection, it is important not to miss critical descriptors that can hold clues for the prediction of the material’s properties.

 

“We have harnessed and enhanced machine-readable representation of the molecular structure,” explains Professor Rajan, “allowing us to rapidly explore and successfully predict multi-scale properties of large numbers of materials.” The information encoded in this representation allows one to target those properties of a material that are influenced by metrics other than just its chemistry. These macroscopic properties are the result of collective phenomena extending far beyond the microscopic scales that are described by quantum atomistic theories.

 

Despite the tremendous potential of materials informatics in property prediction, there is one important aspect of materials design that needs to be addressed: the impact that new materials or their preparation techniques can have on the environment, and the hazard they can pose to human health. The focus of Professor Rajan’s work is on the chemical design of materials that considers the potential environmental impacts in all phases of a material’s life cycle – from synthesis, to manufacturing, and end-of-life replacement. It seeks, in addition, to account for the material’s engineering functionality.

 

This is a daunting task, which requires careful use of machine learning techniques coupled with computational chemistry. In the approach taken at MDI, environmental and health considerations are included from the outset in the material’s design space. These are treated on the same footing as the technological and engineering aspects. This adds to the complexity of the overall materials discovery process, but it also offers a pathway to create new materials that exhibit new or improved functionalities alongside environmental safety.

 

Professor Rajan has developed new accelerated computational methods to establish sets of molecular “fingerprints.” These can serve as strategic markers to rapidly identify new materials’ chemistries and to provide robust guidelines for assessing the environmental hazard of organic chemicals used during manufacturing.

 

An important example of the application of this approach has been in the case of per- and polyfluoroalkyl substances, or PFAS, which are man-made chemicals used in a variety of industries around the globe. Once adsorbed, these chemicals act as very persistent and dangerous contaminants, both in the environment and in the human body. Exposure to PFAS can lead to adverse human health effects, including liver, kidney, and immunological conditions, as well as cancer.

 

A major challenge in addressing the issue of environmental and health hazards of chemicals is the sparsity of data compared with the vast chemical space that we need to explore. Using informatics, modelling, and machine learning techniques, Rajan’s group has developed a framework for estimating the likelihood of toxicity impact of the rapidly increasing amount of PFAS related compounds being introduced into the environment for which hazard data does not exist.

 

Using a robust machine learning method for the visualisation of large sets of imbalanced and sparse data, Professor Rajan has been able to develop an accelerated approach for estimating previously unmeasured PFAS properties. This enables him to quantify the potential toxicity of alternatives to existing PFASs and even to provide guidance on what the most important safety concerns for these systems should be focusing on.

 

Although legacy PFASs have been banned worldwide, the method developed by Professor Rajan provides guidelines for classifying new PFAS compounds and their toxicities. This powerful approach will also help with developing similarity metrics to guide the selection of alternative and safer chemicals; providing an informatics framework for environmentally conscious selection of safer chemicals in materials design.

 

The MDI researchers have applied this approach to material property prediction to several important compounds which might be of major relevance in sustainable energy production. An example of the success of his methods has been the ability to predict structural properties of cubic perovskite crystals. These materials, which are important in solar cell development, are cheap to produce and easy to manufacture.

 

The researchers have developed a unique machine learning framework that captures the quantum mechanical based descriptions of interactions between chemical bonding and bond geometry using a calculated molecular property. One form of the output of such calculations is visualised in terms of a complex 3-dimensional shape – known as Hirshfeld surfaces – that encodes information concerning properties such as molecular packing, molecular shape, close contact points and inter-molecular interactions in crystals. They are typically used to rationalise the ability of molecular building blocks to assemble and form flexible or rigid structures in multicomponent systems.

 

Combined with machine learning methods, the Hirshfeld surface provides a fingerprint to reliably predict lattice parameters (and potentially other properties) of perovskite crystals. A machine learning approach based on Hirshfeld surfaces has also been used to study metal-organic frameworks, or MOFs – an extremely rich class of mesoporous materials – which can be synthesised with a variety of pore sizes, geometries, and network structures. This can have application in several fields, including membrane technologies and chemical sequestration.

 

The method proposed by Professor Rajan allows for the mapping of high-dimensional correlations between the chemical and crystallographic properties of MOFs and to shed a light onto how the interplay between chemical bonding and network geometry governs relationships between diverse families of MOFs.

 

This correlative analysis provides solid guidelines on how to engineer new MOF structures, and to optimise their ability to act as materials for applications in renewable energy technologies, like carbon capture.

 

That’s all for this episode – thanks for listening, and stay subscribed to Research Pod for more of the latest science. See you again soon.

Leave a Reply

Your email address will not be published.

Subscribe and stay updated with our latest content.

If you have any questions about how we handle your data, please review our privacy notice.

You can unsubscribe at any time by contacting us at admin@email.researchpod.org.

We use MailChimp and Salesforce Marketing Cloud as our marketing automation platforms.

By submitting this form, you acknowledge that the information you provide will be transferred to MailChimp for processing in accordance with their Privacy notice and Terms, and to Salesforce Marketing Cloud in accordance with their Privacy notice and Terms.

subscribe to our newsletter

Researchpod Let's Talk

Share This

Copy Link to Clipboard

Copy