Using Artificial Intelligence to increase the rate of materials discovery tenfold

Using Artificial Intelligence to increase the rate of materials discovery tenfold

Thursday 25 Jan 18


Tejs Vegge
Professor, Head of Section
DTU Energy
+45 45 25 82 01


The White Paper on Mission Innovation Clean Energy Materials Innovation Challenge is presented today January 25 at 10 am (CST).

Follow this link to the live streaming and to read the white paper

The amount of CO2 and methane in the atmosphere is rising and smog clouds hang heavily over the world's big cities. Top researchers from all over the world aim to do something about it using supercomputers and Artificial Intelligences as part of the Mission Innovation initiative.

Mission Innovation (MI) is a global initiative with 22 member countries and the European Union to dramatically accelerate global clean energy innovation. The MI initiative was launched on the first day of climate summit COP21 in Paris in 2015. It identified seven key Innovation Challenges essential to limiting the rise in global temperatures to 2 ˚C. One of these is the Clean Energy Materials Innovation Challenge, because, as Professor Tejs Vegge from DTU Energy puts it: “The green change to cleaner energy is happening fast, but not fast enough!”

“Before the sustainable energy technologies can come in place, the building blocks, the clean energy materials must also be in place, and materials discovery takes time”, explains the DTU professor and continues; "If we want to limit global warming, development of new clean energy materials must take place at a much faster rate than today.”

Tejs Vegge is one of Europe's leading experts within computer-assisted research in next-generation battery materials, and he was one of 55 international top researchers, who – together with 75 observers from universities and governments all over the world – met in the Fall of 2017 in Mexico City for the Mission Innovation workshop on the Clean Energy Materials Innovation Challenge.

"If we want to stop Global Warming, the development of clean energy materials must take place at a much faster rate than today"
Professor Tejs Vegge, DTU Energy

"One goal of Mission Innovation is to be able to discover and produce new clean energy materials ten times faster than we do today. The traditional trial-and-error way to do materials development is simply too slow," insists Tejs Vegge. "If we are going to reach the goal of limiting global warming in time, we have to combine existing knowledge in the fields of clean energy materials and automated synthesis with the use of artificial intelligence."

Human research intuition needs a boost

Artificial intelligence has to be utilized because human-driven materials innovation will likely work too slowly to stop Global Warming in time.

Despite widespread use of gunpowder in the medieval wars, scholars of the time spent 300 years to find the optimal mixing ratio between nitrates, sulfur and charcoal. Despite the sale of millions of cars, it has taken some 100 years of car engine research to improve fuel consumption from Ford T's 5-9 km/liter in 1908 to today's 33 km/liter in the most fuel-efficient everyday cars. And it took 25 years of research from IBM's Simon mobile phone weighing 500 grams which only made calls, to the development of ultra-light smartphones which combine phone, camera, multimedia player and game console in the same device.

Researchers are constantly getting faster at developing new materials, e.g. using supercomputers to boost the speed of discovery, but the speed may still be too slow for the challenges at hand. Today, most materials discovery is still based on traditional approaches, i.e. at first you put up a theory, do some practical tests and make a new theory based on the results of the experiment and so it goes on and on before getting a breakthrough.

Historically, this way of doing research has worked fine, but the pace has to speed up.

"Efficient use of artificial intelligence and supercomputers can speed up the research and materials development by up to an order of magnitude, in part because they are able reduce the number of laboratory tests needed," says Tejs Vegge.

DTU Energy uses similar approaches when developing new materials for clean energy conversion and storage, e.g. new battery materials and electrocatalysts for conversion of CO2 into sustainable fuels and chemicals: computer models are used to test and evaluate thousands of hypothetical material combinations and structures before the researchers move on with the most promising combinations for synthesis, characterization and testing.

Fully autonomous research
Artificial Intelligence (AI) is already a part of our daily lives. We are familiar with the artificial personal assistents built into our mobile phones, e.g. Apple’s Siri. Siri uses speech recognition and a natural language user interface to try to answer questions, make recommendations and perform actions by delegating requests to a set of internet services. And the world's major pharmaceutical companies already use synthesis robots to make syntheses and test them.

"We need to develop a new research infrastructure and methodology before we can fully automate the discovery and production of clean energy materials," explains Tejs Vegge, describing how artificial intelligence will connect the big data volumes with synthesis robots. The method is easier to describe than to perform.

Robots as synthesis cooks
The aim is to combine supercomputers with an AI specialized in processing large amounts of data, obtained from trawling the scientific literature for papers on synthesis recipes. The system can be given important keywords, like temperature and material hardness, and is then asked to suggest relevant syntheses.

"At first, the AI can be asked to deliver, say, ten different suggestions for synthesis based on analyzing existing data, then program the automated synthesis robot to produce these and based on the subsequent characterization and testing, go back and propose new supercomputer calculations and finally propose new synthesis conditions. The loop is then repeated until new and improved clean energy materials are produced", explains Tejs Vegge.

Politicians have to pitch in
When scientists choose two or three promising material compositions in advance, they may already have restricted themselves too much. An AI system can identify trends and correlations in multiple dimensions, and they can, for example, combine structures and set up new sequences of molecules or atoms in materials in ways we would never have imagined.

Tejs Vegge’s vast expertise in computational modeling of clean energy materials made him part of the Mission Innovation Clean Energy Materials Innovation Challenge and co-author of the White Paper, which is now being sent to, among other recipients, the Danish government (click on the link to follow the live stream on January 25 at 10 am (CST). Follow this link to the live streaming and the white paper

"The White Paper contains a number of strong recommendations from leading international experts on what to invest in, in order to enable use artificial intelligence and automated synthesis and characterization to accelerate the discovery of next-generation of clean energy materials. Some of these recommendations may appear futuristic, but I hope the decision makes are brave enough to follow them, so we can meet the objectives of the Paris Accord." says Tejs Vegge.

News and filters

Get updated on news that match your filter.