Freezing a Droplet to Stop the Ice Right Now

Advances in simulating water molecules in droplets reveal surfaces that may be resistant to ice formation.

Glass sphere with a dark gray sponge like mass filling the right side
Image courtesy of GE and Oak Ridge National Laboratory
Million molecule simulation of ice formation in a single water droplet. The location of the first water molecules to freeze within the droplet depends on properties of the supporting surface.

The Science

Researchers from GE Global Research improved computer programs that simulate water’s behavior using an approach that is both more accurate and time saving. The improvements enabled scientists to simulate a million molecule water droplet – the most comprehensive simulation of water freezing on a surface ever performed.

The Impact

In cold climates, ice build-up on surfaces of critical infrastructure such as offshore wind turbines can significantly reduce efficiency and require further cost and energy investments to de-ice. The new simulation approach has enabled sufficiently accurate models of the freezing process to screen for surface properties that resist ice build-up, paving the way for large cost and energy saving non-icing material surfaces.


Ice build-up on critical infrastructure, such as offshore oil and gas  drilling and production rigs and wind turbines, operating in extremely cold environments significantly reduces performance, efficiency, and requires a further investment of money and energy to de-ice. To speed development of non-icing surfaces for cold climate wind turbines, researchers at GE turned to modeling and simulation to better understand the fundamental properties of ice formation. The first step was to develop a physical model accurate enough to capture the necessary physics yet computationally fast enough to be useful. To balance speed and accuracy, researchers developed a three-body molecular dynamics code able to take advantage of the graphical processing unit (GPU) accelerator speed-up on Titan, the world’s fastest open-access supercomputer located at Oak Ridge National Laboratory. The code included more accurate physics than the previous GPU-enabled two-body simulation code while still saving computation time through the GPU accelerators. For further time savings, the researchers used a computationally faster but physically simpler model of water molecules, the “mW.” The researchers tested their approach in a detailed simulation of a million molecule water droplet freezing on a surface. The new code is 5.1 times faster, was experimentally validated, and is sufficiently accurate for use in a host of studies. The code was published in 2013; now, the GE research team is using the code to investigate the detailed physics of droplet freezing and screen for properties that will lead to non-icing surfaces for wind turbines as well as oil and gas drilling rigs.


Masako Yamada
General Electric (GE)


GE Global Research (Niskayuna, NY)

Computational resources provided at Oak Ridge Leadership Computing Facility (Oak Ridge National Laboratory, Oak Ridge TN) through the Advanced Scientific Computing Research (ASCR) Leadership Computing Challenge (ALCC).


W. M. Brown, and M. Yamada, “Implementing Molecular Dynamics on Hybrid High Performance Computers – Three-Body Potentials.” Computer Physics Communications 184, 12 (2013). [DOI: 10.1016/j.bbr.2011.03.031]

Related Links

Windpower Engineering and Development, Jan. 2014

Oak Ridge National Laboratory, Oct 25, 3013

Highlight Categories

Program: ASCR

Performer: SC User Facilities , ASCR User Facilities , OLCF

Additional: Technology Impact