National Renewable Energy Laboratory, Golden, CO 80401, USA
The large-scale use of photovoltaics is becoming a reality. The total worldwide solar cell production in 2009 was >10 gigawatts (GW), mostly in the form of flat-plate Si solar cells. This is more than ten times the worldwide production in 2000, representing remarkable progress. Silicon modules have reached efficiencies of 20%, and the cost has been reduced to under $5/watt. However, in the context of world energy consumption, 3 GW is a small number. The phenomenal growth of the photovoltaics (PV) industry in the last seven years has been limited by the availability of purified silicon. The capital investment to ramp up production of purified silicon is daunting, especially since the use of Si by the PV industry has surpassed that by the integrated-circuit industry. One solution to this problem is to use “concentrator technology.” Concentrators are discussed in detail in Chapter 10, but the principle is simple: lenses or mirrors focus (concentrate) the sunlight from a larger area onto a smaller solar cell. Concentration ratios of 500 or more are often used in actual systems. At these concentration ratios, high efficiency is more important than the cost of the cell. High cell efficiency is also of value for space applications (Chapter 9), providing reductions in size and weight of the PV module.
For both concentrator and space applications, there is thus the need for the highest-efficiency cells possible. In the quest for high efficiencies, the fundamental limitation on the efficiency of a conventional single-junction cell is a significant obstacle. For such a cell, photons with energy greater than the cell’s bandgap have their excess energy lost as heat, while photons with energy below the bandgap are not absorbed and all their energy is lost. The multijunction approach provides an effective way of overcoming this obstacle, by dividing up the spectrum into several spectral regions and converting each region with a cell (junction) whose bandgap is tuned to that region. The concept is illustrated in Figure 8.1, which shows the solar spectrum divided up into two regions for conversion by a two-junction cell (specifically, a GaInP/GaAs cell, the basis of the industry-standard GaInP/GaAs/Ge three-junction design, and which will be discussed extensively in this chapter).
Handbook of Photovoltaic Science and Engineering, Second Edition Edited by Antonio Luque and Steven Hegedus © 2011 John Wiley & Sons, Ltd. ISBN: 978-0-470-72169-8
Figure 8.1 Schematic of GaInP/GaAs two-junction solar cell, showing the spectral regions converted by each junction. When the structure is grown on a Ge substrate, there is an option for introducing a third junction in the low bandgap (0.7 eV) Ge substrate, thus boosting the voltage and efficiency of the overall device. ARC = Antireflection Coating; TJ = tunnel junction. Dimensions are not to scale
While the multijunction concept is straightforward, the path to an actual high-efficiency commercially viable multijunction cell is quite complex. In 1984, researchers at the National Renewable Energy Laboratory (NREL) conceived and began work on the GaInP/GaAs two-junction solar cell, which would become the first commercially viable multijunction cell . A schematic of the cell is shown in Figure 8.2. The cell consists of a GaxIn1-xP top cell (with a bandgap of 1.8 to 1.9 eV) grown monolithically on a lattice-matched interconnecting tunnel junction and a GaAs bottom cell. As shown in Figure 8.2, for x « 0.5, GaxIn1-xP has the same lattice constant as GaAs with a bandgap energy between 1.8 and 1.9 eV. Prior to this, several groups were working on tandem device designs that theoretically should achieve efficiencies approaching 36-40%. These included mechanical stacks of a high-bandgap top cell on a Si bottom cell and monolithic combinations of AlGaAs, GaAs, and GaInAs or GaAsP on Si. However, the mechanical stacks were viewed as too costly and cumbersome. Minimizing the defects generated by the lattice mismatch between top and bottom cells in some of the monolithic structures (i. e. GaAs on GaInAs or GaAsP on Si) was a challenging problem. The AlGaAs/GaAs tandem cell is lattice-matched with a theoretical efficiency of 36% . However, the sensitivity of AlGaAs to trace levels of oxygen present in all growth systems and source materials made it difficult to produce this cell with high yield in a production environment and, thus, limited its use. The novel GaInP/GaAs cell design proposed by NREL trades manufacturability (i. e. lattice-matched top and bottom cells and oxygen-tolerant device materials) for a slightly lower theoretical efficiency of 34%.
By most standards, progress was rapid (see Figure 8.3). Despite initial problems with the growth of GaInP by metalorganic chemical vapor deposition (MOCVD) and complications associated with an anomalous red shift of the bandgap energy, by 1988 reasonably good GaInP top
Figure 8.3 Champion multijunction cell efficiencies, showing the historic development of these cells. The most recent champions are summarized in Table 8.1
cells could be fabricated [3-5]. In 1990, efficiencies greater than 27% one-sun air-mass 1.5 global (AM1.5G) were achieved by changing the top cell thickness to achieve current matching [6, 7]. This tuning of the top cell thickness can also be used to achieve current matching under different solar spectra for different applications, as NREL demonstrated over the next three years, setting records at AM1.5G with an efficiency n = 29.5% , at 160-suns AM1.5D with n = 30.2% , and at one-sun AM0 with n = 25.7%  (see Chapter 17 for discussion of AM0 and AM1.5 spectra.). In 1994, it was discovered that the GaInP/GaAs tandem cells had very good radiation tolerance for operating in space. Kurtz and coworkers published results for a GaInP/GaAs cell with П = 19.6% (AM0) after irradiation with 1 MeV electrons at a fluence of 1015 cm-2 , a standard radiation dose used to compare various solar cells. This efficiency was higher than the beginning – of-life efficiency for an unirradiated Si solar cell. These attributes soon attracted the attention of the commercial sector. Production of GaInP/GaAs solar cells (on Ge substrates) began around 1996, and the first GaInP/GaAs-powered satellite was launched in 1997.
Today, multijunction cells using GaInP are standard for both space and terrestrial concentrator applications, with more than a dozen companies capable of growing these cells. The cell
Table 8.1 Record solar-cell efficiencies. Unless otherwise specified, the cells were fabricated from single-crystal materials and the measurements were two-terminal [11, 12]
*One sun is defined as 1000 W/m2.
structures continue to evolve and other material combinations may become important, as described in Section 8.9. Current record solar-cell efficiencies are given in Table 8.1.
This chapter discusses the principles and operation of multijunction solar cells fabricated from III-V semiconductor compounds and alloys, with a particular emphasis on multijunction cells containing GalnP and Ga(In)As. III-V semiconductors have several characteristics that make them especially suitable for solar cells. A wide selection of these materials is available with direct bandgaps, and therefore, high absorption coefficients, in the ~0.7 to 2eV range of interest for solar cells; GaAs, with a bandgap of 1.42 eV, and Gao.5Ino.5P, with a bandgap of —1.85 eV, are especially
important examples. Both n – and p-type doping of these materials are generally straightforward, and complex structures made from these materials can be grown with extremely high crystalline and optoelectronic quality by high-volume growth techniques. As a result, Ш-V cells have achieved the highest single-junction efficiencies. Although these single-junction efficiencies are only slightly higher than the impressive efficiencies achieved by the best silicon cells, the ease of fabricating complex Ш-V structures (including layers with different bandgaps) makes possible the creation of Ш-V multijunction cells with efficiencies in excess of 40%, significantly exceeding the efficiencies of all single-junction devices. As shown in Figure 8.3, the champion efficiencies of these cells have been increasing at a rate of almost 1%/yr. Production versions of the cells have stayed within a few percent of the champion efficiencies. With further development, probably including the addition of a fourth junction, it is likely that the efficiencies will continue to increase, ultimately surpassing 45 or even 50%.