National Center for Computational Sciences

Nanostructure simulations show problems and solutions

The figure above illustrates intermediate electron states within the zinc tellurium band gap introduced by the addition of oxygen atoms to the alloy. The figure on the right shows electrons in the material that are available to conduct electricity.

Image credit: Lin-Wang Wang

Scientists from Lawrence Berkeley National Laboratory (LBNL) will use the Jaguar supercomputer at Oak Ridge National Laboratory (ORNL) to help give life to solar power as a substantial energy source. Led by Lin-Wang Wang, the research team will use 2 million processor hours on ORNL’s Cray XT Jaguar to simulate solar panel materials at the atomic level, where both the inefficiencies and possible improvements begin.

“With the Jaguar machine, we can run a job in one hour which will otherwise take days on other machines,” Wang said. “This will enable us to check the results quickly, and submit new runs. Such interaction between humans and the computer is important for material science research.”

Solar technology has the potential to power everything from lawn mowers and cars to family homes and towering office buildings, and to do it with almost zero pollution. A few places in the country already use solar energy for some of their power, mainly in areas that receive a lot of sunlight such as California and Hawaii. Combined, these sunny spots provide less than 1 percent of the nation’s power. The goal in the United States is to have solar cells that can provide 7 percent of the nation’s energy needs by 2012.

The most advanced solar cells operate at about 40 percent efficiency. However, the type that an average person could afford runs between 15 percent and 25 percent efficiency. That is, the solar cells only convert up to 25 percent of the sunlight that hits them into electricity. Before solar energy can become widely used, it must become less expensive and more efficient.

Finding better materials from which to make solar panels is a large part of the battle. Solar panels typically use silicon and thin films made of varying combinations of elements. However, silicon, which is also used in computers and electronics, is very expensive to produce. Researchers can now simulate the atomic properties of photovoltaic materials, which will help them discover new materials.

Wang and his team will use Jaguar, the world’s most powerful supercomputer for open scientific research, to model the nanostructures of potential materials. Executing up to 1.64 quadrillion calculations per second (petaflops), Jaguar will simulate these materials at the nanoscale in unprecedented detail.

Solar cells are also called photovoltaic cells, as they rely on the photovoltaic effect to function. The photovoltaic effect occurs when an electron is transmitted from a material after the material absorbs sunlight.

It begins when the electrons become excited as the energy increases. They leave their stable, lowest energy state and begin jumping to higher energy levels that are unoccupied by other electrons, much like hopping up stairs on both feet. Before you jump to the next stair, you have to build up enough force in your legs to get there.

The space between each electron energy level is known as the band gap. Materials with smaller band gaps—known as semiconductors—require less energy to excite and move electrons. They make good solar panels because less sunlight is needed to move the electrons to higher energy states. The more easily electrons can move up stairs, the more electric current you can generate.

Wang’s team will investigate the electronic properties of zinc tellurium oxide, a semiconductor alloy. They will explore the role of oxygen atoms in the zinc tellurium, specifically whether the oxygen atom will introduce an intermediate electron state in the middle of the zinc tellurium band gap. Such an intermediate state is like a middle step in a stair, so the electron will not need to jump all the way from the bottom level to the top level. Instead it can first jump to the middle level, then to the top level. Theoretically, such a system can increase the solar cell efficiency from 30 percent to 60 percent. The team will also investigate how bending a zinc oxide wire changes its internal electric field, which can change the electron conductivity and other properties of the nanowire.

Wang is also exploring nanorods with dimensions ranging from 5 to 50 nanometers (billionth of a meter). One such nanorod has a core of cadmium selenide and a shell of cadmium sulfide. Both compounds are semiconductors and are already used in thin film solar panels. The research team seeks to find out more about these materials, such as whether the electron will stay in the same place.

Determining the electronic structure of a material requires mapping the movement of every electron. In this project, each nanostructure consists of tens of thousands of atoms, and each atom contains 6 to 16 electrons. To deal with this enormous amount of information, Wang and his team invented the LS3DF computer code, which they will use on Jaguar. This code divides the nanostructure into pieces. Each segment then goes to a group of processors on Jaguar, which will execute the necessary calculations for that piece of the nanostructure. LS3DF is unprecedented in that it can take the separate bits of information from each group of processors and put it back together to obtain accurate information about the nanostructure as a whole.

Running LS3DF on Jaguar will be a milestone for nanoscience. It will be the first electronic structure code run on a petascale computer and will open doors for future investigation. For inventing LS3DF, Wang and his team from LBNL won the 2008 Gordon Bell Prize for advancements in algorithms. Before LS3DF, every time the number of atoms increased by a factor of 2, the computational cost would increase by a factor of 8. Now the number of atoms and the computational cost increase linearly.

Learning about the microscopic properties such as band gap and electron movement in materials leads to a more complete understanding of atomic scale processes. Aided by supercomputing, researchers will solve these tiny mysteries and bring us another step closer to cheap and effective solar energy.

“Combining new algorithms with big computers like Jaguar, we will eventually be able to simulate how the electron moves in a nanosystem, from its excitation after the sunlight absorption, to its transport to the surface and interface, and for some of them being trapped in defect states, and some of them being collected by the electrodes to generate electricity,” Wang said. “Such detailed understanding is essential for designing new solar cells. It has taken 30 years for people to understand fully the simple thin film silicon solar cell. Hopefully, with the large scale simulation, it will take less time to understand the more complicated nanocells”.

–Elizabeth Storey

Elizabeth Storey is a science-writing intern at the National Center for Computational Sciences.