A virtual material structure

Discovering materials the logical way

For years, materials scientists have relied on their intuition to predict the properties of a novel material and then spent months constructing and characterising it – often only to find that it does not work the way they had hoped. Is there a better way?

When Professors Andre Geim and Kostya Novoselov won the 2010 Nobel Prize in Physics for isolating graphene, the world sat up and took notice. Graphene was hailed as a material of almost limitless potential but was, at the same time, incredibly simple in structure. So while scientists, the public and media raved about how it might be applied in flexible smartphones, as electronic wallpaper, in energy storage and myriad other applications, a select few started asking: “How did we not find this sooner?

”The UK-based Russian researchers’ breakthrough was as serendipitous as it was ingenious, coming about simply through playing with carbon graphite flakes to ascertain their electrical properties. While the playfulness of the experiment is commendable, as acknowledged by the Nobel committee, it does lead to questions over whether there are equally mind-blowing materials out there that we simply have not been playful or lucky enough to find.

Since graphene’s rise in the public’s consciousness, hybrid perovskites for improved solar cells, new topological insulators for future quantum computing applications, photocatalysts for various contaminant removal processes and a host of other materials have been discovered. However, the materials found so far only represent the tip of the iceberg: “We know about the existence of about 200,000 inorganic materials, but for many of them we don’t know their key properties,” explain materials scientists Claudia Draxl of the Humboldt University of Berlin and Matthias Scheffler of the Fritz Haber Institute of the Max Planck Society, who jointly lead the Novel Materials Discovery (NoMaD) project. “If we consider surfaces, nanostructures, organic materials and hybrid materials, the ‘chemical-compounds space’ is infinite and we are currently only looking at a very small fragment of this space.”

Two-hat researchers

To delve deeper into this vast chemical compounds space, materials experts have turned to computer science to extract and exploit information in big data, open data and the semantic web. Termed materials informatics, a completely new discipline has sprung up, with dedicated university departments and even its own Elsevier journal Materials Discovery founded in 2015.

The journal is actually designed to form a scholarly bridge between the materials sciences and information sciences. “The breadth of the audience makes it unique in that it provides the foundation for advancing materials science knowledge by extracting and exploiting information from big data and deep data analytics,” says Krishna Rajan, a materials informatics expert at the State University of New York at Buffalo and editor-in-chief of Materials Discovery.

Rajan is an active scientist in the field, having made important progress in materials informatics over the past 20 years. Like many materials informaticians, his work reflects that he has to wear two hats as a researcher: one when discovering and characterising materials, and the other when developing informatics techniques.

For example, last year his group discovered a completely new class of inorganic scintillators: materials that become luminous when excited by ionising radiation, with applications in the detection of nuclear substances. These materials are brighter and decay more slowly than any previously known compounds. Wearing his other hat, he and his group have also made inroads into addressing problems in materials discovery such as dealing with limited or uncertain information, and advancing multi-scale models and multi-scale characterisation of materials.

Similarly, Chris Pickard at the University of Cambridge has made headway in understanding both materials and new ways of discovering them and characterising their behaviour. Recently, he has predicted and elucidated phenomena relating to germanium anodes in lithium-ion batteries as well as the behaviour of known materials like carbon and water ice at extremely high pressures – work that could shed light on processes in the cores of Jupiter and Saturn.

Pickard’s predictions, however, were only made possible through his development of computational methods capable of illuminating material properties. By making key improvements to the very widely used Cambridge Serial Total Energy Package (CASTEP) code and also introducing the popular ‘ab initio random structure searching’ (AIRSS) method, he has not only made his own materials research simpler and much faster, but has also helped the entire materials community discover an eclectic mix of new materials with novel properties. “It is now relatively straightforward to predict the crystal structure (and even composition) of reasonably complex compounds,” he notes.

Another important member of this pioneering community is Gerbrand Ceder of the University of California, Berkeley. His recent work on solid-state batteries could make batteries in electronic devices last longer, while ensuring they are very safe. “We have developed a new solid-state conductor for electrolytes in solid-state batteries,” he says. “It essentially removes the dangerous liquid electrolyte from today’s batteries, creating the ultimate safe battery.”

Dealing with the data deluge

Ceder is cofounder of the Materials Project, a collaboration started in 2011 that provides a huge amount of computed materials properties data online. “Under the hood is large high-throughput computational machinery, and then property information is collected in a database and put on the website in very different forms, so you can look for a specific material’s attributes,” Ceder says.

The Materials Project is one of many platforms in the US, Europe and Asia providing big data to the materials community. For instance, on the other side of the Atlantic, NoMaD (Novel Materials Discovery), led by Draxl and Scheffler, stores hundreds of thousands of raw datasets from material properties calculations. Already saving researchers time, computing power and reducing the amount of needless repetition of experiments, these repositories are all working to enhance their offering, as a dataset is only as useful as the tools used to analyse it.

NoMaD has recently launched a European Centre of Excellence, backed by EU R&D funding, which aims to create a Materials Encyclopedia that will provide comprehensive characterisation of materials through their computed properties. Another role of the centre is to develop big-data analytics tools for materials science and engineering.

“The tools will be dedicated to extracting information and hidden correlations that may be invisible to the human eye,” say Draxl and Scheffler. “At present, more than 90 per cent of all computed materials data are thrown away – having tools in hand to exploit them will open new research perspectives.”

Ceder is also focusing on this issue. He is involved in an effort aimed at combining computed thermochemical data with natural language processing (NLP) techniques to extract information in the written literature about how materials are made. So, for example, if a group publishes research on graphene, Ceder and colleagues will use NLP to extract information from the paper about how the graphene was made and then add that to existing computed property information in their repository.

Intuitive machines

Ultimately, work in this field is aimed at making materials scientists’ lives easier. Many groups around the world are applying insights from various fields, including pattern recognition, machine learning, artificial intelligence and data topology, to take a lot of the guesswork out of the materials discovery process.

“The problem with first-principles calculations has always been that you cannot find structure if you don’t know where to start – it’s like trying to find a needle in a haystack,” says Ceder. “In some sense, by pursuing these computational techniques, we’re making a brain with great intuition that comes up with educated guesses. If we only have, say, five educated guesses it’s very easy and quick to verify them with first-principles calculations.”

With less time spent on characterisation and more coming up with ideas for new materials, many believe we are in a golden age of materials discovery, and with techniques to improve the discovery and design process being refined every day, be sure to expect to hear about the next wonder material that will improve your life soon.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them