CROSS-SECTIONAL MEASUREMENTS OF
CRYSTALLINE GaAs SOLAR CELLS
M. K. Herndon,a W. C. Bradford, and R. T. Collins,b Physics Department, Colorado School of Mines, Golden, CO 80401
B. E. Hawkins and T. F. Kuech, Department of Chemical Engineering, University of Wisconsin-Madison, Madison, WI 53706
D. J. Friedman and S. R. Kurtz, National Renewable Energy Laboratory, 1617 Cole Blvd, Golden, CO 80401
Near Field Scanning Optical Microscopy (NSOM) was used to study cleaved edges of GaAs solar cell devices. Using visible light for excitation, the NSOM acquired spatially resolved traces of the photocurrent response across the various layers in the device. For excitation energies well above the bandgap, carrier recombination at the cleaved surface had a strong influence on the photocurrent signal. Decreasing the excitation energy, which increased the optical penetration depth, allowed the effects of surface recombination to be separated from collection by the pn junction. Using this approach, the NSOM measurements directly observed the effects of a buried minority carrier reflector/passivation layer.
With a direct band gap of ~1.43eV at room temperature, GaAs is suitably matched to the solar spectrum for use in photovoltaics. Reported efficiencies for crystalline GaAs solar cells are in excess of 25%. A number of different devices with very similar layer structures and growth conditions were used in this study. The devices were grown by atmospheric pressure metal-organic chemical vapor deposition (MOCVD) on (100) GaAs cut 2° toward (110). The typical structure began with a 0.2mm p+ layer of Zn doped GaAs deposited on a p+ GaAs substrate. A ~0.1mm thick layer of In0.48Ga0.52P (InGaP), doped p-type with Zn to a nominal density of ~1017cm-3, was grown next. This layer was added to prevent minority photocarriers from diffusing into the substrate as discussed below. The p-type absorber, ~2.5mm of GaAs with Zn doping of ~3 x 1016cm-3, was deposited next. The n-type junction layer was grown next with a thickness of ~0.1mm and Se doping of ~2 x 1017cm-3. An n-type layer of InGaP, doped with Se to 8 x 1017cm-3, was grown on the n-GaAs layer to function as an etch stop and a minority carrier reflector. A 0.5mm layer of n+ GaAs was then grown to allow low resistance ohmic contacts to be made. A gold contact grid was deposited on top of this final GaAs layer. The device was then placed in an etch that removed the final layer, except where it was protected by the gold contacts. Layer thicknesses and doping concentrations reported are nominal values estimated from calibration runs, except for the doping of the p-type absorber, which was determined from C-V measurements. To explore the effects of the InGaP layers, an additional device structure was grown. It had nominally the same configuration as the standard devices discussed above, but with GaAs layers of comparable doping substituted for both InGaP layers. For this particular device, the dopants in the n and p-type layers were Si and C, respectively, and the doping concentration of the p-type absorber layer was higher (~6 x 1017cm-3). Unless otherwise noted, the measurements being discussed were taken on the samples containing InGaP.
Two important parameters, the diffusion length of minority carriers, Le,h, and depletion width of the junction, W, are helpful in understanding the photocurrent response of the device as a function of position. The total width of the depletion region, W, was found from zero bias C-V measurements to be 0.28mm. An estimate of the minority carrier diffusion length for electrons in the p-type GaAs absorber layer can be made using reported values for the mobility, me~4000 cm2V-1s-1, and minority carrier lifetime, te~80ns (see for example, Ref. 3, 4 and 5). This yields a diffusion length, Le~ 30mm, which is much longer than the p-type absorber layer thickness. Therefore, when scanning the device edge, as light from the NSOM probe creates electron-hole pairs in the p-type GaAs layer, a significant number of the minority carriers created should be able to diffuse to the p-n junction to be collected, even if they are created outside of the depletion region. A similar estimate of the minority carrier diffusion length of holes in the n-type layer using mh~80 cm2V-1s-1 , and tp~20ns for the hole mobility and lifetime yields roughly Lp~2mm, which is again much larger than the layer thickness.
The sample was cleaved, then mounted with the newly cleaved edge perpendicular to the NSOM tip. The sample holder also provided electrical contacts to the front and back of the device so that the NSOM-induced photocurrent could be measured. NSOM probes having 100nm apertures were made by tapering and metal coating single mode optical fibers. Laser light was mechanically chopped and then coupled into the fiber probe, which was used as an excitation source. Chopping the light allowed the NSOM-induced photocurrent to be collected by a lock-in amplifier referenced to the chopping frequency. Shear force feedback was used to maintain a constant probe-sample separation. This allowed surface topography to be obtained simultaneously with photocurrent. Three lasers, a 632.8nm HeNe laser, a tunable Ti:Sapphire laser and an Argon-Ion laser, were used as excitation sources.
The NSOM probe was raster scanned across the cleaved edge, with the fast raster axis parallel to the device layers. Scanning started at the substrate and progressed toward the top face of the device. By scanning this way, the probe only scanned off the front edge of the device one time during the acquisition of an image, minimizing tip damage. The sample position was adjusted so that the probe fell off the edge near the end of each measurement. This identified the edge and helped determine where the various layers in the sample were located.
An example of a spatially resolved photocurrent line scan obtained from the cleaved edge of a GaAs device is shown in Figure 1. Excitation energy was 1.96eV. The NSOM measurement began by acquiring a two-dimensional photocurrent image of the surface. Because of the high uniformity of these crystalline devices, the variation in photocurrent signal in the direction parallel to the wafer surface was negligible. Therefore, the data along this direction was averaged to obtain a one-dimensional trace of photocurrent as a function of position in the direction perpendicular to the layers in the device as shown in Figure 1. The locations of major interfaces and regions are indicated on the plot as a guide. Collection is seen when the probe scans off the front face of the device because light is being coupled in through the top of the device, as it would under normal operating conditions. A peak in collection is observed when the probe is located above the metallurgical p-n junction. The collection decreases as the probe moves away from the p-n junction and further into the p-type absorber. When the probe passes over the interface between the p-type GaAs and the buried InGaP layer, the slope changes and collection decreases more rapidly. This is presumably because the InGaP layer is reflecting most of the carriers back toward the p-n junction, and blocking current carriers created in the substrate from being collected by the junction.
A program used to simulate solar cell operation, SimWindows, was used to model the cross sectional measurements. This one-dimensional model can calculate the light generated current passing through the device, and allows the size and location of the light source to be specified. The program allows the user to input a device file that simulates the film's composition, including layer thicknesses, alloy compositions, and doping densities. Our GaAs device structure was input into the simulation and the minority carrier recombination lifetime was varied.
The results of this simulation are also shown in Figure 1, assuming a recombination lifetime of 80ns. A relatively uniform collection is seen across the p-type layer of the device. When carriers are created within a diffusion length of the junction, they can diffuse toward or away from the junction. The purpose of the InGaP layer is to reflect those diffusing away back toward the junction where they can be collected, hence increasing device efficiency. The uniform collection across the p-type layer seen in the model is a result of this effect. Also, there is no collection from the substrate because the InGaP layer blocks the carriers from crossing. Properties of the InGaP such as the presence of surface recombination at the interface and the band offsets between GaAs and InGaP influence the performance of this blocking layer. Olson et al. have reported much lower interface recombination velocities for InGaP/GaAs interfaces than for similar Al1-xGaxAs/GaAs interfaces. Al1-xGaxAs is an alternative minority carrier blocking layer used in GaAs solar cells.
It is obvious that the measured photocurrent from the side of the p-type layer away from the p-n junction is much smaller relative to the photocurrent at the junction than the model predicts. By reducing the minority carrier lifetime in the simulation to 0.3ns, a closer match to the actual data is obtained. It is unlikely, however, that such a lifetime is correct. Instead, it seems likely that recombination at the cleaved surface,, an effect not considered by the one-dimensional simulation, is causing reduced collection in the p-type region. Several tests were performed to examine this possibility, as discussed below.
In order to verify this, a variety of surfaces, expected to have varying densities of surface states, were studied. For example, collection in the p-type layer dropped by 30% over a period of 4 days when a fresh cleave, such as the one shown in Fig. 1, was exposed to atmosphere. This was presumably due to degradation of the surface, and therefore introduction of more surface states, with exposure to atmosphere.
Additionally, a number of different excitation wavelengths were used to explore the influence of the cleaved surface. Lower energy (longer wavelength) light will propagate further into the material before creating carriers than higher energy (shorter wavelength) light. Therefore, by increasing the wavelength of the light coupled into the fiber probe, the NSOM measurements can be made less sensitive to the surface. The results of the wavelength dependent measurements are shown in Figure 2. The orientation of the device is the same as in Figure 1. In these experiments, we found that as the excitation energy decreased, the amount of photocurrent collected from the p-type absorber became comparable to that collected at the junction, and the shape of the photocurrent trace more closely resembled that predicted by the simulation. This is presumably the result of the lower energy light being absorbed further from the surface, and therefore seeing less of an effect from surface states.
There are several interesting aspects to these measurements. First, we observe photocurrent for energies below the nominal GaAs bandgap. Presumably this is due to the doping of the layers which tends to broaden the absorption edge and enhance below bandgap absorption. The spatial resolution at energies near the band edge is also intriguing. We would expect to have very limited spatial resolution at these energies because the absorption length here is many microns long. In particular, the drop in photocurrent at the InGaP layer should be significantly broadened. However, measurements taken in this region consistently show this drop occurring over a distance of less than 1.5mm (as seen in Figure 2), indicating a spatial resolution of at least this value. Since the index of refraction of the sample is greater than that of air, the cone angle of light from the probe will decrease significantly as it enters the device, enhancing resolution. This effect alone, however, would not produce the high resolution we observe. We can speculate that once the absorption length is sufficiently large, the spatial resolution is not set by the absorption length but instead by the evanescent modes emanating from the tip. However, the reduction in absorption coefficient as the energy falls below bandgap should lead to a more rapid drop in signal intensity than observed. We do not, at present, fully understand what determines the resolution for energies near bandgap, and are developing models of field emission from the tip and of interaction of the field with the semiconductor to gain better insight into the resolution we observe.
As a final test of the performance of the InGaP blocking layer, the device fabricated without the InGaP layers was studied. A photocurrent trace obtained from this device is shown in the inset of Figure 2. Also shown on the plot is a trace for the standard device. Both measurements were made with an excitation energy of 1.42eV, which is very close to the GaAs bandgap. Collection in the p-type material for the device containing no InGaP drops much more rapidly with distance from the junction and extends into the p+ substrate verifying the minority carrier reflector/passivation effect of the InGaP layer. We note that the increased doping in the p-type region of this device discussed above reduces the minority carrier diffusion length somewhat. However, the diffusion length is still sufficiently long that simulations of a device with this doping but also including an InGaP layer gave results similar to those shown in Fig. 1. In addition, simulations of the device without InGaP showed qualitative agreement with the measurements.
In conclusion, we have shown that surface recombination has a strong influence on cross sectional NSOM photocurrent measurements of GaAs diodes. Changing the excitation energy allows surface recombination to be separated from collection in the pn junction. Additionally, we have directly observed the effects of minority carrier reflector layers which are included in solar cells to increase efficiency.
The authors gratefully acknowledge the technical support of J. Kintner and O. Wolf. This material is based upon work supported by the National Science Foundation under Grant No. DMR-9704780. The University of Wisconsin group would like to acknowledge funding from the Army Research Office, the National Science Foundation and facilities support from the UW-Madison MRSEC, a NSF-funded center. The NREL group would like to acknowledge the support from the U.S. Department of Energy under Contract No. DE-AC36-98-GO10337.
2. S. M. Vernon, S. P. Tobin, V. E. Haven, L. M. Geoffroy, and M. M. Sanfacon, in 22nd IEEE Photovoltaics Specialists Conference, Las Vegas, 1991, p. 353-357.
3. J. R. Lowney and H. S. Bennett, J. Appl. Phys. 69, 7102 (1991).
4. S. Tiwari and S. L. Wright, Appl. Phys. Lett. 56, 563 (1990).
5. R. K. Ahrenkiel and M. S. Lundstrum, Minority Carriers in III-V Semiconductors: Physics and Applications (Academic Press, San Diego, 1993).
6. G. A. Valaskovic, M. Holton, and G. H. Morrison, Appl. Optics 34, 1215 (1995).
7. D. W. Winston, Ph.D. Thesis, University of CO, 1996.
8. S. R. Kurtz, J. M. Olson, and A. Kibbler, in 21st IEEE Photovoltaics Specialists Conference, 1990, p. 138.
9. J. M. Olson, R. K. Ahrenkiel, D. J. Dunlavy, B. Keyes, and A. E. Kibbler, Appl. Phys. Lett. 55, 1208 (1989).
10. M. Liehr and H. Luth, J. Vac. Sci. Technol. 16, 1200 (1979).
11. W. E. Spicer, I. Lindau, P. Skeath, and C. Y. Su, J. Vac. Sci. Technol. 17, 1019 (1980).
12. H. C. Casey Jr., D. D. Sell, and K. W. Wecht, J. Appl. Phys. 46, 250 (1975).
Figure 1. Photocurrent collected from the cleaved edge of a GaAs
solar cell (solid line) using 1.96eV light. Simulations of this device
are also shown using the expected electron recombination lifetime of ~80ns
and a much shorter recombination lifetime of 0.3ns.
Figure 2. Photocurrent collected from the cleaved edge of a GaAs solar cell as a function of excitation energy. Each curve has been normalized to the power coupled into the fiber tip. Traces were aligned laterally using topographic measurements, which allow the position of the front edge of the device to be identified. The inset shows data taken at 1.42eV on the device that does not contain the buried InGaP minority carrier reflector/blocking layer, compared to that taken from the standard device.