Feature - Supercomputing code helps develop new solar cells

Image courtesy of Patrick Moore.

If scientists could use simulations to zoom in on the atomic level of solar cells, the insight they gain could launch solar power into the next energy orbital.

Unfortunately, those simulations would require an exorbitant amount of computational power.

“Typically we need to simulate tens of thousands of atoms,” said Lin-Wang Wang, a scientist at Lawrence Berkeley National Laboratory. “For the conventional code, if the number of atoms increases by a factor of ten, the computational load increases by a factor of a thousand.”

In fact, the same problem arises with nano-scale simulations of a wide variety of materials. That’s why Wang and his research team came up with the LS3DF code.

“We were thinking about how to improve the algorithm and have linear scaling,” said Wang. When an algorithm scales linearly, the computational cost increases at the same rate as the number of atoms.

“There have been many linear scaling algorithms out there,” said Wang. But until LS3DF came along, none of them could function on a supercomputer – a pre-requisite for simulating an alloy.

The LS3DF code uses a simple concept: divide and conquer. The code cuts the work up into small pieces, calculates them in parallel, and then puts them back together again.

At the moment a number of different materials are being studied using LS3DF. Among them is zinc tellurium oxide, a semiconductor alloy which researchers hope will prove to be a highly efficient material for use in solar cells.

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.

Today’s typical solar panel converts between seven and 17 percent of sunlight into electricity – a far cry from the 28 percent efficiency of fossil fuel generators. To make matters worse, because the initial cost of solar panels is so high, generating electricity with solar panels is typically three to four times as expensive as buying it from the local power company.

Solar panels generate electricity when they absorb the sun’s heat. The heat causes electrons to jump from the valence level, where they are locked in place, to the conduction band, where they can move freely and carry a current. If the sunlight that’s hitting the panel has energy too low to boost electrons to the conduction band, the solar panel can’t convert that sunlight into electricity.

The researchers investigating the zinc tellurium oxide are hoping that the oxygen in the material will introduce an intermediate energy level – much like a landing on a staircase – that can absorb smaller amounts of energy. If they are right, adding the oxygen to zinc telluride solar cells will increase the efficiency from 30 percent to 60 percent.

To reach that goal, the researchers are using supercomputers at Oak Ridge National Laboratory, the National Energy Research Scientific Computing Center and Lawrence Berkeley National Laboratory.

The first step is to submit a simulation job to one of the machines to which they have access. Their job goes in a queue of waiting computations. It can take a day or two to get to the front of the line, but when they do, the job only takes a few hours because the machines are so powerful.

“We look at the result, check its scientific meaning, do some analysis and try to understand the result,” said Wang. “Then we resubmit another job which is related to the first job but modifying some aspects, maybe checking some other aspects of the problem.”

Even if the alloy turns out to be too expensive for mass manufacture, the simulations will have accomplished something important. “This is really the first of its kind,” said Wang. “The hope is that if it can be worked out and prove the concept, then we can try other systems with more banded elements.”

—Miriam Boon, iSGTW