Sunteți pe pagina 1din 6

ARTICLE IN PRESS

Journal of Crystal Growth 298 (2007) 748753 www.elsevier.com/locate/jcrysgro

Using MOVPE growth to generate tomorrows solar electricity


Sarah Kurtz, Daniel Friedman, John Geisz, William McMahon
National Renewable Energy Laboratory (NREL), Golden, CO 80401, USA Available online 4 December 2006

Abstract Growing interest in renewable energy combined with the excellent performance of compound-semiconductor solar cells inspires a vision of large-scale deployment of solar systems using these cells. Three-junction, GaInP/Ga(In)As/Ge solar cells have demonstrated an efciency of 39%. Many companies are developing solar concentrator systems based on these commercially available cells. This paper reviews compound-semiconductor solar-cell technology from the viewpoint of researchers experienced with metal organic vapor-phase epitaxy (MOVPE) growth of other compound-semiconductor devices. We describe the generic structure of todays multijunction solar cells and some of the challenges of implementing these structures. Several new device structures are currently being investigated toward achieving efciencies above 40%. The potential for growth of this technology is huge, because high-efciency (multijunction) concentrator systems have a chance to capture a signicant part of the large-scale electricity photovoltaic market. r 2006 Elsevier B.V. All rights reserved.
PACS: 81.15.Kk; 73.50.Pz; 73.40.Kp; 84.60.Jt; 42.79.Ek Keywords: A3. Metal organic vapor-phase epitaxy; B1. GaInP; B2. Semiconducting IIIV materials; B3. Multijunction solar cells

1. Introduction 1.1. Growth of the photovoltaic (PV) industry Worldwide, the demand for energy is growing, and the end is nowhere in sight. At the same time, it is becoming more difcult to nd new reserves of crude oil, and global warming has raised concerns about the increased use of fossil fuels. In response, there is growing interest in alternative sources of energy. This includes solar, the worlds must abundant energy resource. Solar energy will ultimately have to become one of the worlds primary energy sources, but it currently supplies less than 0.1% of the worlds electricity. This situation could change in the next few years. In 2004 and 2005, the photovoltaic (PV) industry doubled (see Fig. 1) with annual sales now surpassing $10 billion. This growth is increasing public awareness and stimulating substantial capital investment in new manufacturing capabilities.

For years, PV systems have been too expensive to compete with conventional electricity production methods. Much of todays PV market is subsidized by government funds or by green tag programs. However, as the production volume has grown, costs have decreased, and PV systems are beginning to be competitive with conventional electricity in locations with high electricity prices. In Japan, an incentive program has been successful in creating a stable market even when the incentive has disappeared. 1.2. Opportunity for concentrator solar cells More than 95% of PV systems today are based on crystalline silicon technologies, but the growth of the PV industry in 2005 was limited by a shortage of silicon feedstock. Although the supply of silicon feedstock (puried silicon) is increasing, the investment in silicon purication is currently not keeping pace with the demand. This situation has opened the door for competing PV technologies that use less semiconductor material per watt of electricity generated. One strategy for reducing the amount of semiconductor material is to use mirrors or lenses to focus the sunlight onto tiny solar cells. In this

Corresponding author. Tel.: +1 303 384 6475; fax: +1 303 384 6531.

E-mail address: sarah_kurtz@nrel.gov (S. Kurtz). 0022-0248/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2006.10.176

ARTICLE IN PRESS
S. Kurtz et al. / Journal of Crystal Growth 298 (2007) 748753
40

749

1600 1400 MW of shipments 1200 1000 800 600 400 200 0 1998

Maximum efficiency (%)

Rest of world US Europe Japan

30

20
GaAs AlGaAs/GaAs GaInP/GaAs GaInP/GaAs/Ge GaInP/GaAs/GaInAs

10

0 1985 1990 1995 Year 2000 2005

1999

2000

2001

2002

2003

2004

2005

Year

Fig. 1. Growth of the photovoltaic industry in recent years. Data from the Prometheus Institute.

Fig. 2. History of monolithic, multijunction cell development. Twoterminal, series-connected data are included. The efciencies were measured under a variety of conditions. All of the most recent measurements are under concentration. Most of these data were taken from the Proceedings of the IEEE Photovoltaic Specialists Conference.

case, the major cost is in the large-area optics and other balance-of-system costs such as tracking. Such a system will generate more electricity if a very high-efciency cell, such as a multijunction compound-semiconductor cell, is placed at the focal point, leading to the possibility of highefciency systems at a comparable or lower cost. Once this technology is developed, the capital investment for scale up of production is expected to be only a fraction of that for conventional silicon, providing a means for higher industry growth rates. Such systems, based on the high performance of compound-semiconductor materials, have the potential to revolutionize the industry and generate a signicant portion of tomorrows solar electricity. 1.3. Opportunity for IIIV solar cells Although it has long been recognized that multijunction solar cells have the potential to outperform single-junction cells, it was not until about 1990 that higher efciencies were actually demonstrated [1,2]. A history of efciencies measured for some IIIV monolithic solar cells is shown in Fig. 2. In 1990, an efciency over 27% was reported for two different two-junction solar cells: AlGaAs/GaAs and GaInP/GaAs. Until 1990, AlGaAs/GaAs had been considered the preferred approach and was explored by a number of laboratories, whereas the GaInP/GaAs cell was explored only at NREL. The primary advantage of the GaInP/GaAs cell was high efciency in a lattice-matched conguration, like the AlGaAs/GaAs cell, but without the purity requirements necessary for Al-containing active layers. In the 1990s, the Applied Solar Energy Corporation, Spectrolab, and EMCORE began manufacturing the GaInP/GaAs cells for space applications. More recently, GaInP/Ga(In)As/Ge three-junction concentrator cells became available off the shelf for terrestrial applications from Spectrolab and EMCORE. Several other companies

or research labs are also supplying concentrator cells in limited quantities, and, stimulated by a growing demand, about a dozen companies around the world are developing a capability to manufacture these or related cells. If these new companies are as successful as the current ones in establishing manufacturing capabilities, enough cells for 41 GW of concentrator systems could be produced each year. In contrast, concentrator companies (mostly Amonix and Solar Systems) installed less than 1 MW of concentrator systems (using silicon cells) in 2005. Although both companies are currently working to incorporate the GaInP/Ga(In)As/Ge cells into their systems, they have not yet begun to use these cells in volume. The potential for growth of this industry is huge because high-efciency (multijunction) concentrator systems have a chance to capture a signicant part of the large-scale electricity PV market. If electric utilities embrace concentrator systems, the growth potential is staggering. However, there is a huge uncertainty as to when the concentrator industry will develop multijunction-based product and capture enough of the market to provide business for the new cell companies. Nevertheless, the wider availability of multijunction concentrator cells will encourage investors and reduce prices, supporting faster growth of the industry, overall. This paper reviews compound-semiconductor solar-cell technology from the viewpoint of researchers experienced with metal organic vapor-phase epitaxy (MOVPE) growth of other compound-semiconductor devices. We describe the generic structure of todays multijunction solar cells and some of the challenges of implementing these structures. We highlight the importance of nearly ideal material quality and the challenge of controlling dopant diffusion while achieving high-conductance tunnel junctions. Looking forward, we discuss research opportunities for future high-efciency solar cells.

ARTICLE IN PRESS
750 S. Kurtz et al. / Journal of Crystal Growth 298 (2007) 748753

2. High-performance solar cells MOVPE is the technique used currently for the manufacture of IIIV solar cells. MOVPE provides precise control of layer compositions, thicknesses, and doping while providing nearly perfect crystalline quality. Although the cost of MOVPE growth is signicant, it is acceptable for space applications and is projected to be adequate for terrestrial PV applications that use lenses or mirrors to concentrate the light by about 1000 times. 2.1. Fundamentals of operation of solar cells Sunlight is absorbed by a solar cell, exciting valence band electrons into the conduction band, as shown in Fig. 3(a). These excess carriers are minority carriers: holes in n-type material and electrons in p-type material. Solar cells separate the excess minority carriers from the majority carriers at a pn junction and deliver them to an outside circuit, as shown by the white arrows in Fig. 3(a). Some of the excess carriers recombine before reaching the pn junction. Any excess carriers that recombine non-radiatively are lost from the process. Thus, the biggest challenge of creating a high-efciency solar cell is to remove all defects that can lead to non-radiative recombination. Radiative recombination can also decrease the efciency of a solar cell, but cannot be entirely eliminated because it represents the reverse of the absorption process. For high-quality epitaxial layers, the biggest loss of excess carriers is at the surfaces. The recombination of excess carriers at the front and the back of the solar cells can be reduced by continuing the crystal lattice with a different material with a larger bandgap to reduce the concentration of defects and build in a eld that deects the minority carriers. In general, a potential barrier with a height of about ve times the thermal energy, kT, is sufcient to contain the excess carriers in the active region.

The surfaces around the perimeter of the solar cell cannot be passivated as effectively as the front and back surfaces, but do not usually dominate the performance of the solar cell. A multijunction cell combines two or more of the structural units shown in Fig. 3(a) with highly doped shorting (tunnel) junctions or some other means of interconnection. Todays commercial high-efciency, multijunction cells are grown on germanium with three active junctions, as shown in Fig. 3(b). The spectrum is naturally sorted as each subcell passes light of sub-bandgap energies on to the lower subcells. Each subcell includes a pn junction with passivation (cladding) layers on the front and back, as shown in Fig. 3(a). 2.2. Effect of high diode quality factor on performance In many ways the structure of a solar cell is similar to that of a light-emitting diode (LED). A solar cell can be run backwards, showing a large-area light emission with a color indicative of the bandgap. However, the optimal design of an LED differs from that of a solar cell, which must be thick enough to absorb light near the band edge while efciently utilizing the strongly absorbed, highenergy light. In contrast, LEDs usually use quantum wells and are optically engineered for a narrow wavelength range. Here we choose one of the differences between the function of solar cells and LEDs to guide the material requirements for solar cells. Because LEDs are operated further in forward bias than solar cells, an LED is less affected by non-ideal diode behavior. The diode equation is given by J V J 0 expqV =nkT , (1)

where J(V) is the current density of the diode as a function of the voltage across the diode, J0 is a prefactor that

Fig. 3. (a) (Left): schematic describing function of solar cell, not to scale. The light enters through the passivating window layer on the left and is absorbed by the n-type and p-type layers. The arrows show how the minority carriers are repelled from the passivating (cladding) layers and separated in the eld region. (b) (Right): structure of three-junction GaInP/Ga(In)As/Ge cell.

ARTICLE IN PRESS
S. Kurtz et al. / Journal of Crystal Growth 298 (2007) 748753 751

depends on the design of the diode, q is the electronic charge, n is the ideality factor, k is Boltzmanns constant, and T is the temperature. For an ideal diode, n is unity. In practice, n may take any value or multiple values, such as J V J 01 expqV =kT J 02 expqV =2kT , (2)

where J01 and J02 are prefactors dependent on the diode. Fig. 4(a) shows an example of a GaAs diode that follows Eq. (2). It shows dark current with an n 2 slope under low forward bias, but returns to more ideal behavior as the forward bias is increased. 2.3. Diode quality factor and material quality The n 2 component of the dark current is a result of non-radiative recombination, e.g., perimeter recombination or recombination at dislocations that thread through the junction. In an LED, the n 2 current does not result in light emission, whereas the n 1 current causes emission of light, some fraction of which is emitted from the LED. Although some of the n 1 current causes non-radiative recombination, and should be minimized, neglecting it in this analysis does not affect our qualitative conclusions. The effect of an increased n 2 current on an LED can be estimated from the relative magnitudes of the n 1 and n 2 currents. The relative decrease in performance of an LED with increased n 2 current is shown in Fig. 4(b) for a range of bias voltages. The data show that higher n 2 current becomes unimportant if the LED is operated with adequate forward bias. Most GaAs LEDs are operated with bias voltage 41.5 V. In contrast, a GaAs solar cell operates at about 0.9 V under one-sun illumination. At this

lower bias voltage, an increased n 2 current can cause substantial degradation of performance, as shown in Fig. 4(b). Thus, a key challenge in the design and growth of solar cells is achieving nearly ideal diode quality for bias voltages near the open-circuit voltage. To achieve this in GaAs cells, it is necessary to limit the threading dislocation density to o106/cm2. Other materials show slightly different requirements on the dislocation densities. Notably, GaN LEDs have achieved good efciencies despite high threading dislocation densities, but many GaN LEDs show non-ideal diode behavior and the threading dislocations requirement for GaN solar cells has not yet been determined. Crystallographic defects in solar cell materials are minimized when clean, highly polished, single-crystal wafers are used for epitaxial growth. Careful matching of the lattice constants of the wafer and the epitaxial layer allow nearly perfect crystal quality. Mismatched layers with high quality can also be grown if the strain and relaxation properties can be controlled in order to limit the dislocation density. An extension of the analysis shown in Fig. 4 implies that solar cells made from non-ideal mismatched material will benet from operation under concentrated sunlight, which increases the operating voltage. 2.4. Dopant diffusion Mechanical stacks or other congurations can be used to interconnect multijunction solar cells, but, so far, tunnel junctions (as shown in Fig. 3(b)), have been the preferred connection scheme. MOVPE growth of tunnel junctions is

10
Optimal operating bias for photocurrent of 30 mA/cm2

b
Relative performance

1.00
GaAs LED at 1.3 V 1.4 V 1.5 V

0.95
n=2

1 Current (mA/cm2)
GaAs Data

0.90

0.1

GaAs solar cell operated at optimal bias (~0.9V)

0.85

0.01
n=1

0.80

0.001 0.7 0.8 0.9 1.0 Voltage (V) 1.1


4 5 67 2 3 4 5 67 2 3 4 5 67

10-8

10-7

10-6

Prefactor for n=2 current (mA/cm2)

Fig. 4. (a) Dark currentvoltage data for a GaAs solar cell and indication of how the slope indicates a diode quality factor between 1 and 2, depending on the forward bias. The n 1 and n 2 lines were calculated as 9 1017 exp(V/0.026) and 3.6 109 exp(V/0.052) mA/cm2, respectively. (b) Effect of increased dark current with diode quality factor of 2 on the performance of GaAs diodes. In each case, the dark current was assumed to be 9 1017 exp(V/0.026)+Prefactor exp(V/0.052) mA/cm2. The relative efciencies for the LEDs were calculated from the ratio of the n 2 current to the total current at the specied bias. Most GaAs LEDs are recommended for operation at greater than 1.5 V. The solar cell efciencies were calculated assuming superposition and 30 mA/cm2 of photocurrent.

ARTICLE IN PRESS
752 S. Kurtz et al. / Journal of Crystal Growth 298 (2007) 748753

challenging because of the need to very abruptly change the dopant type, while maintaining very high carrier concentrations. Fortunately, commercial MOVPE systems have minimal memory effects because the surface-area-tovolume ratio is relatively low. Nevertheless, the performance of tunnel junctions can be compromised if dopants diffuse during subsequent growth. The complexity of the diffusion mechanism complicates its control. Ironically, the dopant diffusion may be most strongly affected by growth conditions (such as the group V pressure) in layers far from where the diffusion takes place. Fig. 5 shows an example of how zinc diffusion can be affected by subsequent growth of an n-type layer [3]. Deppe [4] described how Fermi-level pinning during growth can lead to non-equilibrium concentrations of point defects that then induce zinc diffusion in nearby layers. The solution to the zinc diffusion for GaAs HBTs (that Deppe studied) was to replace the zinc with carbon, but carbon doping does not work well in phosphide materials. Thus, in GaInP/GaAs/Ge solar cells, dopant diffusion is controlled by manipulating point-defect concentrations, adding diffusion barriers, and minimizing annealing of the tunnel junctions. 3. Research opportunities for multijunction, compoundsemiconductor solar cells In theory, the solar cell with the highest possible efciency is one based on a collection of semiconductor materials with bandgaps chosen to match the energies in the solar spectrum. Numerous publications have detailed the precise bandgap combinations that yield the highest theoretical efciencies [58]. In practice, it is easy to identify combinations of materials that match the solar spectrum, but difcult to identify ways to achieve nearideal performance with each of those materials in a convenient geometry. For example, GaInN materials with bandgaps spanning the solar spectrum have been proposed as a pathway to 460% efciency, but, practically, it is not clear how the low-bandgap GaInN alloys can be used in solar cells. As of May 2006, the structure shown in

Fig. 3(b) held the maximum efciency record39% measured at a concentration of 240 times the solar intensity [9]. As described above, this structure has been a signicant commercial success for space applications, with perhaps the best-known application being the power source for the Mars rovers. Fig. 6 shows other structures that are being explored to reach even higher efciencies, detailing the efciencies that have been achieved so far. Whereas the Fig. 3(b) structure has achieved the highest efciency to date, the structures shown in Figs. 6(a) and (d) have achieved almost the same efciency and have the potential to achieve much higher efciencies. As more junctions are added, the efciency gains are smaller, and the electrical energy produced under a variable spectrum increases even less. One study suggests that there will be no advantage to using more than six series-connected junctions [10]. Future research opportunities include study of mismatched alloy systems [1113], new alloys such as the dilute nitride alloys [1418], and methods for integrating highquality materials. Successful wafer bonding and methods for substrate removal/reuse could open up many new pathways. Ideally, the top-cell bandgap for cells with more than four junctions should approach 2.4 eV, but none of the arsenides or the phosphides has direct-gap materials above 2.2 eV. Thus, GaInN may provide the only approach to a 2.4 eV cell. For high-bandgap cells, there is need to develop non-absorbing passivating (cladding) layers to retain excellent blue response. The modularity and the exibility of the precise choices of bandgaps imply that the next generation of multijunction cells is limited only by our imaginations and our ability to implement our ideas with high-quality materials. 4. Summary The history of IIIV cell development was reviewed, showing that most of the recent development has been focused on GaInP/Ga(In)As/Ge cells. These cells have now achieved an efciency of 39% and are being enthusiastically investigated for use in high-efciency solar

a
Zn, Si, Se (cm-3)

1022 1020 1018 1016 1014 0.6 0.8 1.0 1.2 1.4 1.6 Depth (m)

b
Zn, Si, Se (cm-3)

1022 1020 1018 1016 1014 0.6 0.8 1.0 1.2 1.4 1.6 Depth (m)

Fig. 5. Secondary ion mass spectroscopy data showing dopant proles in GaAs solar cells [3]: (a) This sample was grown with Se doping in the emitter and shows very little movement of the zinc. (b) This sample used different conditions for the growth of the emitter, including Si as the dopant. This sample showed movement of the zinc into the back of the layer in addition to an accumulation of zinc in the junction region.

ARTICLE IN PRESS
S. Kurtz et al. / Journal of Crystal Growth 298 (2007) 748753 753

Fig. 6. Multijunction designs currently being explored for higher efciency. The exact bandgaps and compositions vary. Efciencies that have been achieved so far include: (a) 38.8% [9] for a structure with bandgaps 1.8 eV/1.3 eV/0.7 eV; (b) work has focused on GaInNAs subcell [1418]; (c) 21%, 5.3 V [19], 45 V [13]; (d) 37.9% [12].

concentrator systems to provide large-scale electricity generation. The need for excellent material quality and low dopant diffusion were described as two of the special MOVPE growth considerations for multijunction solar cells. New structures that can reach 440% efciency were shown, with related research including study of new alloy systems, mismatched materials, and new methods for integrating materials. Acknowledgments We thank S. Moon for assistance with the manuscript. The writing of this paper was funded under DOE Contract no. DE-AC36-99GO10337. References
[1] B.-C. Chung, G.F. Virshup, J.C. Schultz, in: Proceedings of the 21st IEEE PVSC, 1990, p. 179. [2] J.M. Olson, S.R. Kurtz, A.E. Kibbler, P. Faine, Appl. Phys. Lett. 56 (1990) 623. [3] S.R. Kurtz, J.M. Olson, D.J. Friedman, R. Reedy, in: Proceedings of the 26th PVSC IEEE, New York, 1997, p. 819. [4] D.G. Deppe, Appl. Phys. Lett. 56 (1990) 370. [5] J.C.C. Fan, B.-Y. Tsaur, B.J. Palm, in: Proceedings of the 16th IEEE PVSC, 1982, p. 692. [6] A. Marti, G.L. Araujo, Sol. Energy Mater. Sol. Cells 43 (1996) 203. [7] M.W. Wanlass, T.J. Coutts, J.S. Ward, K.A. Emery, T.A. Gessert, C.R. Osterwald, in: Proceedings of the 22nd PVSC IEEE, 1991, p. 38.

[8] S.R. Kurtz, D. Myers, J.M. Olson, in: Proceedings of the 26th PVSV IEEE, New York, 1997, p. 875. [9] R.R. King, D.C. Law, C.M. Fetzer, R.A. Sherif, K.M. Edmondson, S. Kurtz, G.S. Kinsey, H.L. Cotal, D.D. Krut, J.H. Ermer, N.H. Karam, in: Proceedings of the 20th European Photovoltaic Solar Energy Conference, 2005, p. 118. [10] K. Araki, M. Yamaguchi, M. Kondo, H. Uozumi, in: Proceedings of the Third World Conference on PV Energy Conversion, 2003, p. 307. [11] R.R. King, D.C. Law, K. Edmondson, C.M. Fetzer, R.A. Sherif, G. Kinsey, D.D. Krut, H.L. Cotal, N.H. Karam, in: Proceedings of the Fourth World Conference on Photovoltaic Energy Conversion IEEE, 2006, p. 760. [12] M.W. Wanlass, P. Ahrenkiel, D.S. Albin, J.J. Carapella, A. Duda, K. Emery, D.J. Friedman, J. Geisz, K. Jones, A. Kibbler, J. Kiehl, S. Kurtz, W.E. McMahon, T. Moriarty, J.M. Olson, A.J. Ptak, in: Proceedings of the Fourth World Conference on Photovoltaic Energy Conversion IEEE, 2006, p. 729. [13] F. Dimroth, R. Beckert, M. Meusel, U. Schubert, A.W. Bett, Prog. Photovolt. 9 (2001) 165. [14] A.J. Ptak, D.J. Friedman, S. Kurtz, R.C. Reedy, J. Appl. Phys. 98 (2005) 094501. [15] S. Kurtz, S.W. Johnston, J.F. Geisz, D.J. Friedman, A.J. Ptak, in: Proceedings of the 31st PVSC IEEE, 2005, p. 595. [16] D.J. Friedman, J.F. Geisz, S.R. Kurtz, J.M. Olson, J. Crystal Growth 195 (1998) 409. [17] J.F. Geisz, J.M. Olson, D.J. Friedman, M.J. Romero, R.C. Reedy, K.M. Jones, in: Proceedings of the 31st PVSC IEEE, 2005, p. 695. [18] J.F. Geisz, D.J. Friedman, Semicond. Sci. Technol. 17 (2002) 769. [19] R.R. King, C.M. Fetzer, D.C. Law, K. Edmondson, H. Yoon, G. Kinsey, D.D. Krut, J.H. Ermer, P. Hebert, B.T. Cavicchi, N.H. Karam, in: Proceedings of the Fourth World Conference on Photovoltaic Energy Conversion IEEE, 2006, p. 757.

S-ar putea să vă placă și