DEVICE CHARACTERIZATION OF (AgCu)(InGa)Se 2 SOLAR CELLS William Shafarman 1, Christopher Thompson 1, Jonathan Boyle 1, Gregory Hanket 1, Peter Erslev 2, J. David Cohen 2 1 Institute of Energy Conversion, University of Delaware, Newark, DE 19716 2 Physics Department, University of Oregon, Eugene, OR, 97403 ABSTRACT Ag-alloying of Cu(InGa)Se 2 thin films presents the possibility to increase the bandgap with improved structural properties as a result of a lower melting temperature. (AgCu)(InGa)Se 2 films were deposited by elemental co-evaporation and the resulting solar cell behavior was characterized. While the bandgap in the highest efficiency Cu(InGa)Se 2 cells is ~1.15 ev, Ag alloying allows the bandgap to be increased to 1.3 ev with an increase in V OC, no loss in device efficiency, and fill factors up to 80%. With high Ga content to increase bandgap > 1.5 ev, Ag alloying improves solar cell efficiency. Analysis of the device behavior shows that the basic mechanisms controlling (AgCu)(InGa)Se 2 solar cells and limiting performance with wide bandgap are comparable to those with Cu(InGa)Se 2. Finally the effect of Na in (AgCu)(InGa)Se 2 devices is shown to be comparable to that with Cu(InGa)Se 2 including a decrease in V OC attributed to interface recombination with insufficient Na. INTRODUCTION High efficiency Cu(InGa)Se 2 based thin film solar cells typically have absorber layers with [Ga]/[In+Ga] 0.25 and a bandgap (E g) of ~1.15 ev. Wider bandgap absorber layers are desirable because their higher operating voltage with lower current density can be used to improve module performance, especially at elevated temperatures encountered in most applications. The chalcopyrite CuInSe 2 material system provides several options for alloying the film to increase the bandgap. Alloying with silver to form (AgCu)(InGa)Se 2 has the potential advantage of increasing the bandgap while also decreasing the melting temperature of the material. This might enable the growth of films with reduced structural disorder and defects and, consequently, improved device performance. We have previously shown promising solar cell results with (AgCu)(InGa)Se 2 absorber layers including 17.6% efficiency and V OC = 710 mv with E g = 1.3 ev, and 13.0% efficiency and V OC = 890 mv with E g = 1.6 ev [1]. Further, Nakada et al. reported a cell with V OC = 950 mv and 9.3 % efficiency using a Ag(InGa)Se 2 absorber layer with E g = 1.6 ev [2]. We have also shown that Ag-alloying gives improved sub-bandgap transmission [ 3 ] and sharper bandtails [ 4 ], suggesting improved structural quality. Optical bandgaps of (AgCu)(InGa)Se 2 films have been measured. E g increases by ~ 0.2 ev from Cu(InGa)Se 2 to Ag(InGa)Se 2 with a bowing parameter b = 0.2 for all values of fixed Ga content. In contrast, E g increases from 1.04 ev for CuInSe 2 to 1.67 ev for CuGaSe 2 [5], so Ga alloying has a greater impact for increasing bandgap. In this paper (AgCu)(InGa)Se 2 films have been deposited over a wide compositional range by elemental coevaporation. We characterize the solar cell behavior with these absorber layer films and show improvements in cell efficiency with Ag addition. In addition, the effect of Na incorporation in (AgCu)(InGa)Se 2 devices is addressed. EXPERIMENTAL PROCEDURES (AgCu)(InGa)Se 2 films have been deposited by elemental co-evaporation onto Mo-coated soda lime glass (SLG) substrates at 550 C. In one case, (AgCu)(InGa)Se 2 was also deposited on a substrate with an Al 2O 3 barrier layer between the glass and Mo to reduce Na diffusion from the glass by about 2 orders of magnitude. All films were deposited with time-invariant fluxes to produce homogeneous through-film composition. The films, spanning a composition range with x [Ga]/[In+Ga] 0.3, 0.5, and 0.8 and w [Ag]/[Ag+Cu] from 0 to 0.8, are listed in Table 1 along with their bandgaps calculated from the composition [3]. All films are group I deficient, as indicated by the ratio [Ag+Cu]/[In+Ga], and 2 µm thick. Table 1. Composition ratios of (AgCu)(InGa)Se 2 films, determined from energy dispersive x-ray spectroscopy and optical bandgaps. x w Ag+Cu E g In+Ga (ev) 0.28 0.00 0.87 1.18 0.29 0.16 0.88 1.20 0.27 0.28 0.82 1.20 0.31 0.46 0.79 1.26 0.31 0.77 0.75 1.35 0.48 0.00 0.91 1.32 0.45 0.16 0.86 1.30 0.46 0.22 0.81 1.32 0.44 0.46 0.94 1.34 0.49 0.77 0.77 1.46 0.80 0.00 0.87 1.55 0.80 0.16 0.85 1.55 0.80 0.21 0.76 1.55 0.81 0.49 0.82 1.59 0.81 0.76 0.86 1.66 Solar cells were completed with a standard structure of SLG/Mo/(AgCu)(InGa)Se 2/CdS/ZnO/ITO/grids and 0.5 cm 2 978-1-4244-5892-9/10/$26.00 2010 IEEE 000325
total area cells were delineated by mechanical scribing. All results are reported without the benefit of an antireflection layer. Device characterization includes J-V measurements under AM1.5 illumination, temperature dependent J-V measurements, voltage-bias dependent quantum efficiency, and transient photocapacitance (TPC) spectroscopy. Device Performance RESULTS and DISCUSSION Solar cells were fabricated in sets with fixed Ga fraction and increasing Ag fraction. V OC, FF and efficiency (η) are shown in Fig. 1 for the best cell at each composition. In each set, V OC increases with Ag addition as expected for increasing E g. The V OC increase is sub-linear with increasing Ga fraction or bandgap, as with Cu(InGa)Se 2 and other alloys [6,7]. However, greater improvement in V OC with increasing Ag addition is demonstrated by the set with x = 0.8 and E g from 1.55 1.66 ev. The highest values of FF are achieved with Ag alloying and x = 0.3 or 0.5. Values of FF = 80% are particularly noteworthy because there is no bandgap gradient to assist current collection as is created in Cu(InGa)Se 2 deposited with a 3-step evaporation process. This high FF cannot be attributed to changes in parasitic losses such as shunt or series resistances. Instead, it may be an indication of improved minority carrier collection length. The efficiencies with Ag alloying are, within the experimental spread, the same for the samples with x = 0.3 and 0.5 despite the increased E g and V OC with higher Ga. In devices without Ag, η decreased from 16 to 15 %. Diode Analysis The current voltage results shown in Fig. 1 were analyzed using the procedure described previously [8] to determine diode parameters including the forward current (J O), the diode quality factor (A), the series resistance (R S) and the shunt conductance (G) as defined by the diode equation: J = J o exp q(v R s J) J o J L +GV. (1) AkT The diode analysis is partly shown in Fig. 2 where an (AgCu)(InGa)Se 2 sample with x = 0.3 and w = 0.3 is compared to one with x = 0.8 and w = 0.2. The light and dark J-V curves are shown at the top. In the middle, the derivative dv/dj is plotted vs. (J+J SC-GV) -1, where G was determined from the minimum slope dj/dv in reverse bias. The slope on this plot determines A and the intercept determines R S. Finally, the bottom plot is a semilogarithmic plot of (J+J SC-GV) vs. (V-R SJ) where a fit to the linear region at high voltage gives J O and A. A distinct difference between the samples is seen in the second and third plots. The low Ga cell has similar behavior between the light and dark curves over nearly 2 orders of Figure 1. V OC, FF, and efficiency vs. w or devices with x = 0.3, 0.45 and 0.8. magnitude in current while with the high Ga device the light data deviates from the dark data and is not exponential. This is indicative of a voltage dependent collection of photo-generated current J L(V) that the above diode analysis does not take into account [8]. As a result. the light J-V data on the high Ga cell cannot be reliably fit to determine the diode parameters under illumination. QE measurements of the same devices are shown in Fig. 3 at two different voltage biases, 0V and -1V. There is negligible difference in the QE curves of the low Ga cell but a significant increase in QE with reverse voltage bias 978-1-4244-5892-9/10/$26.00 2010 IEEE 000326
Figure 2. Diode analysis of (AgCu)(InGa)Se 2 devices with x = 0.3, w = 0.3, and E g = 1.20 ev (left) and E g = 1.55 ev, x = 0.8 and w = 0.2 (right). Data in blue is measured in the dark and data in red under AM1.5 illumination. Figure 3. QE measurements under white light bias at 0V (red) and -1V (blue) for the devices in Fig. 2. 978-1-4244-5892-9/10/$26.00 2010 IEEE 000327
for the high Ga cell. This bias dependence confirms the interpretation of the voltage dependent current collection loss in the J-V data with high Ga. The J L(V) effect can be quantitatively characterized by calculating the current J QE(V) from the integral over wavelength of the product of the QE and the AM1.5 spectrum and then determining the ratio at different voltage bias. For the devices in this work, this ratio J QE(-1V) / J QE(0V) was determined. All devices with x 0.3 and 0.5, except the device with x = 0.49 and w = 0.77, had values of this ratio in the range 1.00 1.02. This indicates that there is little increase in current with the additional field created by the reverse voltage bias. For the devices with x 0.8 and the excepted cell with x = 0.49 this ratio increases to 1.07 1.10 so the added field has a significant effect on improving collection of photogenerated current. Thus, the voltage-dependent collection, which has been previously noted for wide bandgap Cu(InGa)Se 2 [6], is observed in all the wide bandgap cells with E g > 1.4 ev, with no apparent difference due to Ag alloying. This leads to a loss in FF and, to a lesser extent, J SC [9]. One result of the diode analysis is the determination of the diode quality factor which can be an indicator of the recombination that limits device performance. These values, determined by analysis as in Fig. 2, are shown in Fig. 4. With x = 0.3 or 0.5 comparable values of A were obtained from both the dark and light J-V data. These values of A in the range 1.4 ± 0.2 are typical of wellbehaved Cu(InGa)Se 2 devices. However, with x = 0.8, the dark J-V data gave A 2.0 while A could not be determined for the light case due to the J L(V) shown in Fig. 2. These J-V characteristics are consistent with V OC controlled by Shockley-Read-Hall (SRH) recombination in the space charge region of the absorber layer [10]. A value of A 2 is consistent with the trap states for recombination situated nearer to mid-gap in energy [11]. We have previously characterized the sub-bandgap optical spectrum in Cu(InGa)Se 2 and (AgCu)(InGa)Se 2 using transient photocapacitance measurements on completed devices [4]. In addition to sharper bandtails which suggest better structural quality of the (AgCu)(InGa)Se 2, the TPC measurements reveal a deep defect band which remains located at 0.8 ev above the valence band in all of the alloy samples. This defect is close to midgap for the samples with x = 0.8 so that the SRH recombination is most favorable. Thus this may explain the higher diode factors and relatively lower V OC with the wider bandgap (AgCu)(InGa)Se 2 devices. Sodium Effect The beneficial role of Na in Cu(InGa)Se 2 has been extensively studied though there is still no conclusive understanding of its role in improving device performance. (AgCu)(InGa)Se 2 films were deposited in a single run with x = 0.46, w = 0.22, and E g = 1.32 ev on soda lime glass Figure 4. Diode quality factor A for devices in Fig. 1. and glass with a Na diffusion barrier. The J-V curves are shown in Fig. 5 for the resulting devices. With the reduced Na diffusion V OC decreased by 120 mv and FF decreased from 79 % to 62 %. The decrease in V OC and roll-over behavior are typical of Cu(InGa)Se 2 devices with insufficient Na. To characterize the mechanism controlling V OC and the decrease with low Na, the temperature dependence of V OC was measured [8,12] as shown in Fig. 6. Extrapolation to T = 0K gives the activation energy (E A) for recombination according to V oc = E A q AkT q ln J oo, (2) J L which follows from Eq. 1 with an Arrhenius behavior for the forward current J O. The cell on a SLG substrate has E A = E g which indicates that the dominant recombination occurs in the absorber layer and is again consistent with SRH recombination in the (AgCu)(InGa)Se 2 space charge region. Two differences are seen with the Na barrier substrate. First is a lower E A which has been attributed to interface recombination limiting V OC [12]. Second is a saturation of V OC as temperature decreases which indicates a freezing out of the recombination. CONCLUSIONS (AgCu)(InGa)Se 2 solar cells with a wide range of absorber layer compositions have been characterized and it has been shown that the Ag alloying can produce improved device performance. In particular, while the bandgap in the highest efficiency Cu(InGa)Se 2 cells is ~ 1.15 ev, Ag alloying allows the bandgap to be increased to 1.3 ev with an increase of up to 100 mv in V OC and no loss in device efficiency. With E g > 1.5 ev, V OC, FF, and efficiency all increase with increasing Ag content. The best (AgCu)(InGa)Se 2 cells have FF = 80 % which may be an indication of improved minority carrier collection length, 978-1-4244-5892-9/10/$26.00 2010 IEEE 000328
although additional characterization would be needed to confirm this. Such an improvement would be consistent with previously reported indications of improved structural quality of the (AgCu)(InGa)Se 2 films based on optical properties. comparable to that with Cu(InGa)Se 2. The main effect is a decrease in V OC attributed to interface recombination. ACKNOWLEDGEMENTS The authors acknowledge the assistance of Josh Cadoret, Dan Ryan, and Kevin Hart for deposition and device fabrication. This work was supported in part by the DOE Next Generation PV Program. REFERENCES Figure 5. J-V curves in the dark and under AM1.5 illumination for devices on soda lime glass and on glass with a Na diffusion barrier. [1] G. Hanket, J. Boyle and W. Shafarman, Proc. 34 th IEEE PVSC, (2009). [2] T. Nakada, et al., Mater. Res. Soc. Symp. Proc. 865, 2005, p. F11.1.1. [3] J. Boyle, G. Hanket and W. Shafarman, Proc. 34 th IEEE PVSC, Philadelphia, PA, (2009). [4] P. Erslev, et al., Mater. Res. Soc. Symp. Proc. 1165, 1165-M01-07 (2009). [5] P. Paulson, R. Birkmire and W. Shafarman, J. Appl. Phys. 94, 879 (2003). [6] W. Shafarman, R. Klenk and B. McCandless, J. Appl. Phys. 79, 7324 (1996). [7] R. Herberholz, et al., Solar En. Mat. and Solar Cells, 49, 227 (1997). [8] S. Hegedus and W. Shafarman, Prog. in Photovoltaics 12, 155 (2004). [9] M. Eron and A. Rothwarf, Appl. Phys. Lett. 44, 131 (1984). [10] J. Phillips, et al., Phys. Status Solidi B 194, 31 39 (1996). [11] C. Sah, R. Noyce and W. Shockley, Proc. Inst. Radio Engrs. 45, 1228 1243 (1957). [12] C. Thompson, et al. Proc. 33 rd IEEE PVSC (2008). Figure 6. The temperature dependence of V OC for the devices with E g = 1.32 ev in Fig. 5 and extrapolation to T=0 to determine E A. The basic device behavior that controls the (AgCu)(InGa)Se 2 solar cells is comparable to that with Cu(InGa)Se 2. Analysis of J-V and V OC-T measurements are consistent with SRH recombination in the absorber layer as the limiting mechanism for V OC. With E g > 1.5 ev, the increase in V OC is not proportional to the bandgap which may be due to more efficient recombination through trap states close to midgap. The wide bandgap cells are also limited by poor minority carrier collection which primarily reduces FF. Finally the effect of Na in (AgCu)(InGa)Se 2 devices has been shown to be 978-1-4244-5892-9/10/$26.00 2010 IEEE 000329