19 Dec 2003
For III-V solar cell manufacturers, it is Earth rather than space that represents the final frontier. However, recent advances in the performance of cells and high-concentration optical modules suggest that multi-junction photovoltaics can infiltrate the silicon-dominated terrestrial energy market. Michael Hatcher reports on the latest developments.
While multi-junction III-V cells now dominate the solar energy market for space applications, their cheaper silicon rivals have maintained a stranglehold on the terrestrial photovoltaics industry. Large-scale terrestrial solar power plants that are currently being built in Arizona and Australia rely on silicon cells, despite recent advances that have seen multi-junction cell efficiencies reach almost double that of their silicon counterparts.
Late last year the US company Amonix installed the first 35 kW unit of the 1 MW solar energy plant that it is building in Arizona. By June this year, the plant's installed capacity had reached almost 600 kW. Meanwhile, the Japanese solar systems builder Kyocera is stepping up its manufacturing output to 120 MW per year to meet increased demand for residential solar energy.
Given this dominance of silicon technology in Earth-bound photovoltaics, the opportunities for III-V technology might seem limited. However, the latest III-V technology from Emcore, Spectrolab and Sharp looks likely to change this scenario in the medium term.
This year, all three companies have pushed the efficiency of their multi-junction cells towards the 40% mark. In May, Japan-based electronics giant Sharp announced the development of its 36.5% efficiency InGaP/InGaAs/Ge cell at the 3rd World Conference on Photovoltaic Energy Conversion in Osaka. At the 11th Biannual Workshop on OMVPE in July, Boeing subsidiary Spectrolab of Sylmar, CA, announced that it had developed a 36.9% terrestrial concentrator cell (figure 1). Emcore Photovoltaics has produced its own terrestrial triple-junction cells with an efficiency of 30% at 200 suns concentration, and recently won a contract to supply the solar panels for the two STEREO satellites being built by Johns Hopkins University.
In terrestrial applications, direct sunlight must be concentrated onto the multi-junction cells using focusing optics. A significant advantage of these concentrator systems is that fewer solar cells are required to achieve a specific power output, thus replacing large areas of semiconductor materials with relatively inexpensive optics that provide optical concentration. The higher cost of multi-junction cells is offset by the use of fewer cells. Their higher efficiency means that only a small fraction of the total cell area is required to generate the equivalent power output, compared with crystalline silicon or thin-film flat-plate modules.
Nasser Karam, Spectrolab's vice-president for Advanced Technology, told Compound Semiconductor that recent improvements in space cells should result in better efficiency for terrestrial cells. "During the last few years, multi-junction solar cells have doubled the power output of large commercial satellites, and substantially improved their revenue-generating capability. We believe that further optimization of terrestrial concentrator cells will yield the potential to surpass 40% conversion efficiency," he said.
One of the major stumbling blocks in adapting multi-junction cells for terrestrial use has been the tunnel junction used to combine the currents produced from each of the three sub-cells. Compared with space applications, terrestrial solar modules with a high sunlight concentration ratio generate a much higher current density than the tunnel junctions had originally been designed to handle. Further complications arise from the need to extract current and heat efficiently from the cells without causing hot spots that can degrade cells.
Spectrolab has been tackling the tunnel junction issue on two fronts, by developing both lattice-matched and metamorphic solar cells. In the first approach, the idea is to match the lattice spacing in all three junctions to address the terrestrial spectrum. However, in both dual- and triple-junction cells, the bandgap combination of 1.81.9 eV for the GaInP top cell and 1.424 eV for the GaAs junction is not ideal for either the space or terrestrial spectrum. According to Spectrolab, the conversion efficiency can be improved by grading the composition of a GaInAs buffer on a germanium substrate. In this lattice-mismatched or "metamorphic" approach (figure 2), the top two sub-cells can then be grown with a higher indium content and larger lattice constant than the lattice-matched cells. This lowers the bandgap of both the GaInP and GaInAs cells, providing a better match to the terrestrial solar spectrum.
Karam said that the metamorphic and lattice-matched structures were "neck and neck" in terms of the progress made so far. The lattice-matched structure has shown a 32.5% conversion efficiency, with the metamorphic cell reaching 31.3%. "We believe that both can be improved further to reach the 40% efficiency goal under concentration in the future," he said.
Crucially, Spectrolab's new cell and receiver designs are said to be capable of supporting concentrator modules that increase the sun's irradiation by up to a factor of 1000. In Spectrolab's tests on cells optimized to operate at 300 suns concentration, the cells successfully withstood prolonged illumination under 997 suns concentration, although conversion efficiency dropped to 27.9% from 34.2% at 161 suns.
Mark O'Neill is president of Entech Solar, a company based in Keller, TX, that builds concentrator solar energy systems. Entech made the 720 Fresnel lenses that were used to concentrate sunlight onto 3600 dual-junction cells used to power NASA's Deep Space 1 probe launched in 1998, which included the first solar array to use dual-junction cells as a spacecraft's major power source.
One of Entech's latest developments for the space market is a prototype wing (figure 3) that houses four complete 50 cm long photovoltaic receivers consisting of 14 triple-junction cells each. Spectrolab and Emcore each supplied two of the receiver modules. Entech has come up with a new prismatic cell cover that eliminates the normal gridline shadowing loss on the solar cells, and O'Neill says that the best of the receivers achieved 31% average cell efficiency and 27.5% net lens/ receiver efficiency.
O'Neill and his company have been fielding silicon-based terrestrial systems for the past 15 years, and developing terrestrial concentrators based on multi-junction cells for the past four years. The aim is to double the operational solar-electric conversion efficiency of silicon-based systems.
In order to challenge silicon-based cells on the terrestrial front, O'Neill has calculated that III-V concentrator cells must operate at a few hundred suns concentration. "Our approach to using such cells in the next generation of our terrestrial equipment is relatively simple operate at a high enough concentration ratio (about 400 suns) so that the cells are so small that they become cost-effective," he explained. The calculation is based on the assumption that multi-junction cells are around two orders of magnitude more expensive per cm2 than silicon cells. By focusing sunlight in two planes in a 21:1 ratio, the concentration increases to 440 suns.
O'Neill is bullish on the prospects for multi-junction cells in terrestrial applications: "The multi-junction cells developed for spacecraft in the 1990s provide such a huge performance advantage over silicon that the terrestrial concentrator market is certain to explode when such cells are incorporated into cost-effective systems," he said. "At the end of the day, I believe that high performance will win on the ground just like it has already won in space."
Karam claims that Spectrolab is the only company to have deployed multi-junction cells in both terrestrial and space-based photovoltaic systems. "We have had our multi-junction cells operating inside several high-concentration modules. The longest test data we have is on a terrestrial system that has been in the field for six months now. I am confident that the technology is viable," he said. While all of the terrestrial systems built so far are relatively small, and capable of generating kilowatt output, Karam predicts that megawatt-scale concentrator systems will be deployed in the next decade. He told Compound Semiconductor that for III-V-based concentrator modules to infiltrate terrestrial photovoltaics the key requirement now is financial support: "We are looking for governments and utilities to embrace the technology. If individual companies are expected to put up all the investment, entering the terrestrial market will take a very long time."
Emcore Photovoltaics' director of sales and marketing Patrick Park says that his customers are looking to introduce commercial terrestrial systems in 2005. "Governments and utilities are warming up to it," he said.
The concentrator systems need direct sunlight to work. As a result, the likely deployment of III-V concentrator photovoltaics will be restricted to regions of reliable sunshine. Spectrolab has drawn up a map of the places where concentrator photovoltaic systems are most likely to succeed.
According to Karam, there is plenty more scope for further improving the cell efficiencies by adding more junctions. Four-, five- and six-junction cells are all being pursued. In the Spectrolab technology roadmap (see table), triple-junction cell efficiency based on the existing structure is predicted to reach 40% in 2005, with five-junction devices expected to reach 45% in 2007.
It is not just Spectrolab that is looking at increasing the number of junctions per cell. Researchers at the Fraunhofer Institute for Solar Energy Research in Freiburg, Germany, have fabricated five-junction cells that are based on layers of AlGaInP, GaInP, AlGaInAs, GaInAs and Ge grown by MOVPE (figure 4). These particular cells have been designed with extra radiation hardness in mind for space applications, but could also be adapted for terrestrial use.
Although further improvements in cell efficiency would undoubtedly be welcome, the critical issue now appears to be one of financial support for the developers of the III-V solar cells and modules to enable the technology to get a foothold in terrestrial energy. Sarah Kurtz, a senior scientist at the National Renewable Energy Laboratory in Golden, CO, is urging the US funding bodies to pool some resources to fund research into III-V photovoltaics and solid-state lighting based on HB-LEDs. Currently, the two areas are dealt with by different funding bureaus.
Kurtz believes that there are technological issues common to both LEDs and concentrator photovoltaics that could be addressed in joint studies. For instance, concentrator cells in the field will go through cycles of getting very hot during the day and cold at night. After thousands of such heating cycles, delamination of bonded parts within the cells could become a problem. Kurtz says that the experience gained by LED manufacturers could come in useful here. Protecting the cells from weathering effects such as ultraviolet exposure will also need to be dealt with and, again, the experience that LED manufacturers have in encapsulating devices could be invaluable.
A completely different breed of compound semiconductor materials is also making headway in photovoltaics. First Solar of Perrysburg, OH, is taking the low-efficiency, low-cost approach with its CdTe and CdS solar modules. The modules are based on CdTe and CdS films deposited by vapor transport deposition. This process involves vaporizing CdTe or CdS powders, using a carrier gas to transport the vapor and allowing the material to settle on a glass superstrate.
First Solar began pilot-line production in 2002 and is planning to increase output from the current level of 2 MW per year to 25 MW per year by November 2004. The average module conversion efficiency is only 8%, but First Solar is aiming to increase this to 13% over the next five years by improving the deposition technique to reduce variations in grain size, crystallographic orientation and structural defect density.
Michael Hatcher is associate editor of Compound Semiconductor magazine.