In order to understand the properties of materials, predict their performance and invent or discover new ones it is necessary to understand the complex behavior that keeps matter together, starting from its tiniest components. Nicola Marzari, who joined EPFL in 2011, does just that, by developing and applying to materials science powerful computational modeling techniques based on the fundamental principles of quantum mechanics. These "first-principles" calculations allow one to solve, rather than model, the exact behavior of materials starting from the atomic scale up, and thus do not require any prior knowledge or experimental input.
Such a change of paradigm in the process of invention and discovery means that we can now explore the properties of a new material even before it has been created in a laboratory, or we can rapidly screen on a computer thousands of them to find the ones that have the best performance for the job. We can study materials under conditions that could never be replicated here, such as those at the center of the Earth or in far away planets, or down to the "nano" world of devices that are just the size of one billionth of a meter. Best of all, these computational tools increase in speed and accuracy at the same speed as computers (i.e. doubling in performance every 14 months – try to do that with your car, or your microscope!), and are shared across the world under a democratic model where every scientist can download them for free, and contribute to the effort.
To "understand", "predict" and "design" the properties and performance of new materials and devices, using quantum-mechanical simulations: this is the mission statement for the Laboratory for Theory and Simulation of Materials, headed by Professor Nicola Marzari. The scientist starts from the laws of quantum mechanics and electromagnetism that govern the behaviour of electrons and atoms, transforms those into computer algorithms, and unleashes massive computational simulations to understand or discover new materials. Since no physical laboratory or any other major infrastructure are required, the capabilities are only limited by human ingenuity and computer performance. "We are on the edge of a revolution – we have now started to obtain very precise descriptions of the molecular architecture and the performance of a given material well before setting foot into a real laboratory." says Professor Marzari. "I find it magical", he adds.
From nanotechnology to energy storage
The list of applications seems endless. Take nanotechnology – we need to understand how electrons race within miniature components such as the billion-more transistors of a current computer chip, without burning it down, each of them obediently performing its assigned task. Sometimes the devices are so small that the electrons go ballistic (a technical term, we are told…), and move effortlessly without any resistance. Graphene (that prompted a Nobel prize in 2010) could unleash such potentialities, but there is a race to engineer its properties for information processing, and to make it compatible with well-established silicon technology. Another hurdle: dissipating all that heat. "The smaller a component becomes, the more heat it gives off, and the more it can be damaged," says Professor Marzari. "We therefore must have a precise control over all the heat-generating and heat-dissipating mechanisms, in order to get the fastest, most reliable device." The group started working on graphene well before it was first described in real experiments, as part of their effort on novel carbon electronics, and predicted some of its most striking properties, such as the capability to contract, rather than expand with temperature, and to carry extremely large amounts of heat.
The team is also working on improving performance and discovering new materials for energy applications. The goal is to find novel ways to harvest energy (e.g. from the sun), and to either store it or covert it into fuels, for later use. The list of projects includes new electrolytes for lithium-ion batteries, new catalysts for fuel cells, and thermoelectric materials (i.e., materials that can transform heat into electricity), all of which could potentially be improved via Professor Marzari’s approach.
To improve fuel cells, the group is looking for a way to replace the platinum in the electrodes with another material that would be just as effective but earth-abundant, and thus less expensive. Another line of inquiry involves determining how electrodes could be made more efficient by controlling the shape of the catalyst (the sprinkle that increases the speed of chemical reactions) at the nanoscale. Metallic nanoparticles can be used to make very active catalysts, but they don’t last very long. "We need to strike a balance between having a catalyst that is active, but not so active that it disappears – it’s always a balancing act" comments Professor Marzari.
The potential applications of the scientist’s work are not limited to nanotechnology and energy. The group recently developed a novel way to describe the response of a material to a magnetic field – this translated into a new algorithm to calculate chemical shifts measured in nuclear magnetic resonance experiments, and Prof Marzari is using this capability to try and decode the microscopic structure of cement (in collaboration with another faculty member in Materials, Prof Scrivener), but also to study the shape and structure of gallstones that form in the bile tract. Precise characterization of these crystalline bodies would help doctors identify the most efficient means of dissolving them. "This is an example of a new theoretical approach becoming right away a new computational algorithm, and giving us new capabilities to interact with our experimental colleagues", he states.
A global "group-sourced" database
Since simulations require only a software infrastructure to be run, it becomes natural to use methods and approaches of computer science to organize the research. These include running calculations in database-driven modes, using machine-learning techniques to discover hidden laws in the data, and storing all the calculations run by the group in a permanent repository. In particular, Professor Marzari wants to leverage the fact that a single computer is all one needs to undertake state-of-the-art digital simulations — without a lab. "Thanks to the continuous increase in our computing capabilities, we can make more, and ever-more-precise, predictions, without running up against the constraints imposed by physical experiments," says Professor Marzari. With this in mind, he and his team have developed a "materials’ informatics" platform capable of investigating thousands of systems at a time, rather than processing them one after another. Together with the open-source environment of the computational tools used (www.wannier.org and www.quantum-espresso.org) the plan is to create a global database of materials’ properties, and to store the results of every calculation run worldwide. "Even in the most far-away places, any researcher with Internet access will be able to contribute to this effort and then add the results to a global database. This is a way of democratizing science and building global group-sourced knowledge," says Professor Marzari – "I think we have a revolution in the making".
Article par Laure-Anne Pessina
Photos par Alban Kakulya