Adrianus (Adri) van Duin
Force Fields Fill the Gap in Materials Simulations
Adri van Duin, a chemist with a theoretical bent, is happy to find himself surrounded by practical engineering types intent on solving real world problems. It is a good sign, he says, when theory guys are employable in departments that have the word engineering in their title. Only five or ten years ago such a thing would have been unthinkable, van Duin believes. It shows that the gap between science and engineering is getting smaller.
Adri van Duin’s theoretical research fits nicely into that gap. The inventor of a computational method called reactive force fields (ReaxFF), van Duin’s simulations lie between the highly accurate but computationally expensive quantum mechanical-based simulations, and the larger scale simulation tools that describe physical interactions in molecules and materials but cannot describe chemical reactions in which bonds break and reform.
A recently appointed faculty in the Department of Mechanical and Nuclear Engineering, van Duin began to develop his force field method during his first postdoctoral appointment at the University of Newcastle upon Tyne in northeast England. In that university’s Department of Fossil Fuels and Environmental Geochemistry, he worked on simulations of how water and oil interact with mineral surfaces, an important consideration for oil producers trying to determine which sort of underground reservoirs are likely retain oil.
At Newcastle, he studied the processes that turn a “big, messy polymer called kerogen” into oil and gas over millions of years of applied heat and pressure in the subsurface layers of the earth. These processes are extremely difficult to replicate in the laboratory, and because kerogen doesn’t dissolve in any known solvent, it is almost impossible to analyze except by pyrolysis, or burning it and looking at the leftover fragments. “I thought it would be attractive if we could do this from a simulation perspective so that we could have a random polymer and heat it up and see where bonds break and where new bonds are being formed and basically follow the process,” van Duin explains.
The problem was that quantum simulations, while capable of working accurately with small molecules of 20 atoms or so, were too computationally demanding to handle polymers like kerogen with unusual structures made up of a couple of thousand atoms. Larger models, called force fields, were available in the literature for a few materials, but they needed to be further developed so that they could simulate chemical interactions. Through support from a Royal Society fellowship, van Duin was given time to improve the force field models to the point where they could predict the correct reaction products from experiments and predict them in the right time frame. The result was a new simulation method called ReaxFF that is now widely used in molecular dynamics simulations.
As his research began to take off, van Duin was invited to join a large computational chemistry group at the California Institute of Technology under the leadership of William Goddard. Whereas at Newcastle he had been primarily involved with the chemistry of oil interactions, at Caltech he was surrounded by people with interests in many different materials. Together they managed to make the force field simulations work across a large part of the periodic system, testing them on metals, metal oxides, and condensed phase/gas phase interactions.
“These force fields are not in any way a replacement for quantum simulations,” van Duin is quick to point out. Reactive force fields only exist because they are developed using quantum data, and certain material properties, especially those directly dependent on electronic interactions like conductivity, cannot be understood with these simulations. However, once a structure is predicted by force field simulations, conductivity can be calculated by going back and using quantum methods.
Force fields fill a substantial void between quantum simulations and continuum simulations commonly used by engineers, which don’t take individual molecules into account, van Duin suggests. Communicating between researchers working at the atomistic and picosecond scale and those who deal in millimeters and seconds is not easy. But force field simulations, which work on a time and length scale in between, can help both sides, he believes. Working with both groups means that he can validate and improve his simulations from below and above, from atomistic simulations and from experiment. Once validated, he can use his force fields to suggest further experiments, helping researchers to focus their resources on those experiments most likely to gain valuable results. “Nowadays people are doing very small scale experiments,” van Duin remarks. “They can do nanometer scale and nanosecond resolution, and that is something I can directly provide them simulations for. Doing experiments in that size and time range is expensive. Force fields run on very modest computational resources, on pretty much any laptop you have around.” In principal, he says, this is a method that works with any kind of material or interface. You can put any combination of materials together in the simulation box.
With his newly formed research group, van Duin has settled in at Penn State and is eager to work with people in both the quantum chemistry field and with those who use continuum methods, such as combustion chemists and aerospace engineers. “I would love to see what kind of materials people are thinking of so I can make my force fields and start doing simulations.” One of his principle aims is to find materials that can sustain higher temperatures in a combustion environment. Higher temperatures make for cleaner combustion and more efficient energy output. But for that a new class of materials would need to be developed, and high temperature experiments in the range of 1500 to 1700K are not trivial, he says. For that reason, simulations to rule out unsuitable materials would be valuable.
He is currently teaching a mechanical engineering undergraduate course and hopes to start teaching a graduate course in the fall of 2009. His goal is to let students in engineering know that there is something below continuum they can use that will take atomistic scale resolution into account and still be relevant for engineering applications. “That kind of communication between quantum and force field and force field and continuum, that is something I would very much like to introduce in courses in this department,” van Duin concludes.
