UT/ORNL team uses Kraken to simulate nanoparticles in effort to control their movement
by Gregory Scott Jones
If one could peer into a crystal ball and indeed see the future of technology, the ball might well be empty. That is, unless you look very, very closely, on the order of billionths of a meter. The advent of nanotechnology promises to transform our lives in ways that are still being understood—everything from medicine to electronics to energy production stands to benefit in a disproportionately large way.
Nanomachines, or devices developed at the atomic or molecular scale, show enormous promise in their ability to revolutionize technology. Imagine molecule-size robots attacking viruses in your body or transistors designed from the ground up, atom by atom, enabling electronic devices far more powerful than today’s and with much lower power requirements. These advances are being developed the world over. However, there is still plenty of work to be done. For starters, if these technologies are to reach their full potential and actually enter the marketplace, researchers need to understand how to manage motion at the nanoscale, which means controlling the movements of nanoparticles, or pieces of nanomachines that behave as units and are roughly the size of a molecule.
This crucial threshold is still being explored across a wide swath of scientific arenas, but researchers at the University of Tennessee (UT), Oak Ridge National Laboratory (ORNL), and Los Alamos National Laboratory (LANL) just took a leap forward. Miguel Fuentes-Cabrera, Humberto Terrones , and Jason D. Fowlkes, all of ORNL, Bradley H. Rhodes from UT, Mike Baskes from LANL, and Phillip Rack and Mike Simpson of both ORNL and UT recently completed a suite of simulations on UT’s Kraken supercomputer to show that using lasers to melt metals on a substrate, or a surface upon which metals are deposited, induces a process known as dewetting, which in turn can be used to control the motion of metallic nanodroplets. Their findings were recently published in the September 27 issue of ACS Nano.
The outcome will set the stage for plenty of future research, said Fuentes-Cabrera, adding that “everything happens very fast . . . when lasers heat the metals, the liquid only lasts a few nanoseconds, and this is when these big computers become very helpful, when the phenomenon is reachable.”
Fuentes-Cabrera points to a European team [Habenicht, A.; Olapinski, M.; Burmeister, F.; Leiderer, P.; Boneberg, J. Science 2005, 309, 2043-2045] that heated gold triangles deposited on graphite with lasers, and noticed that occasionally the triangles would deform with so much energy as they dewet that they would literally detach and jump off their graphite base with velocities of the order of 20 meters per second. Fuentes Cabrera’s collaborators observed the same phenomenon when they simulated copper triangles on graphite, but sought to change the velocity with which their copper droplets leapt from the substrate.
Figure 1: From dewetting to jumping. Side-view of the evolution of a nanostructure (blue) deposited on graphite (pink) from its initial shape (circle, square, equilateral triangle, and isosceles triangle) to an ejected nanodroplet. The velocity of the ejected nanodroplet depends on the initial shape: the nanodroplet travels further the more symmetric its initial nanostructure.
“Dewetting and wetting is what happens when you place a liquid on top of a surface,” said Fuentes-Cabrera. “For example, if water was placed on a substrate, and it ‘liked’ the substrate, the water would spread out. This is wetting. Dewetting is the opposite, and in this case water ‘gets away’ from the substrate by collapsing in on itself and forming a droplet. Copper does not like graphite, so if you start from a pancake-like structure, such as a circle, a square, or a triangle, copper will dewet and collapse into a droplet. Sometimes the energy gained during dewetting is large enough to overcome the adhesion of the droplet to graphite, and then the droplet leaps.”
The team found they could change the velocity by changing the initial shape. For instance, a circle jumps faster than a square, which jumps faster than an equilateral triangle, which jumps faster than an isosceles triangle—and it all has to do with the temporal asymmetry of the mass coalescence.
“In other words,” added Fuentes-Cabrera, “collapsing occurs at the same time for a circle: all the atoms are moving inward and when they meet they do not have anywhere else to go but upwards. In an isosceles triangle, some atoms reach the center faster, where they collapse and just when they are ready to move up, there comes more atoms from another part of the triangle, bumping and slowing everybody enough to curtail jumping.”
Essentially, the simulations provide proof that a new tool, lasers, can be used to control the movements of nanoparticles. And not only can you make the particles jump, but you can specifically control “how” they jump, a key point in the design of novel materials built at the nanoscale.
The simulations that took place on Kraken, a Cray XT5 capable of 1.17 petaflops , will inevitably be used to refine experiments. But, said Fuentes-Cabrera, they have been extremely helpful in that while researchers might understand how the theory works, they only see the first and last stages of the phenomenon in experimental environments. With simulation, however, the entire process is observable, shedding a rare light on a complex, and practically invisible, course of events.
The team used a classical molecular dynamics simulation code known as LAMMPS to represent the phenomenon on Kraken. The application is renowned for its ability to accurately represent soft materials such as polymers, and solid-state materials, such as metals and semiconductors, all of which could possibly benefit from a better understanding of the behavior and control of nanoparticles.
“LAMMPS is working very well with these systems,” said Fuentes Cabrera, who added that the project consisted of a series of simulations that consumed from 300 to 1,000 cores for eight hours at a time. This is a good thing, he added, because it allows the team to run more and more simulations over time, a key metric in scientific computing because more runs in a given period of time mean a quicker time to solution for a given problem.
And while Kraken is one of the most powerful computers in the world, it is still not yet powerful enough to simulate the scale being performed in experiments. The phenomenon is already being observed at the scale being simulated, said Fuentes-Cabrera. However, in this case, as happens often in science, the answer has simply led to more questions, paving the way for Fuentes-Cabrera and collaborators and countless others around the world to explore this phenomenon in-depth and hopefully one day help bring nanomachinery to market.
Acknowledgements: This work was supported by the U.S. Department of Energy, Basic Energy Sciences, Materials Sciences and Engineering Division.