Microsc. Microanal. 21, 927–935, 2015 doi:10.1017/S143192761500077X
© MICROSCOPY SOCIETY OF AMERICA 2015
Characterization of Sputtered CdTe Thin Films with Electron Backscatter Diffraction and Correlation with Device Performance Matthew M. Nowell,1 Michael A. Scarpulla,2 Naba R. Paudel,3,* Kristopher A. Wieland,3,† Alvin D. Compaan,3 and Xiangxin Liu3,‡ 1
EDAX-TSL, 392 East 12300 South Suite H, Draper, UT 84020, USA Departments of Materials Science and Engineering and Electrical and Computer Engineering, University of Utah, Salt Lake City, UT 84112, USA 3 Department of Physics and Astronomy, University of Toledo, Toledo, OH 43606, USA 2
Abstract: The performance of polycrystalline CdTe photovoltaic thin films is expected to depend on the grain boundary density and corresponding grain size of the film microstructure. However, the electrical performance of grain boundaries within these films is not well understood, and can be beneficial, harmful, or neutral in terms of film performance. Electron backscatter diffraction has been used to characterize the grain size, grain boundary structure, and crystallographic texture of sputtered CdTe at varying deposition pressures before and after CdCl2 treatment in order to correlate performance with microstructure. Weak fiber textures were observed in the as-deposited films, with (111) textures present at lower deposition pressures and (110) textures observed at higher deposition pressures. The CdCl2-treated samples exhibited significant grain recrystallization with a high fraction of twin boundaries. Good correlation of solar cell efficiency was observed with twin-corrected grain size while poor correlation was found if the twin boundaries were considered as grain boundaries in the grain size determination. This implies that the twin boundaries are neutral with respect to recombination and carrier transport. Key words: CdTe, sputtering, EBSD, thin film
I NTRODUCTION The effects of grain boundaries in polycrystalline films on solar cell performance are not well understood. Grain boundaries can act as recombination sites, reducing performance (Visoly-Fisher et al., 2004). However, grain boundaries can also form a localized electric field, either through structural or impurity effects (Yan et al., 2007), which can improve majority and minority carrier separation, and collection and thus improve performance (Li et al., 2014). Enhanced photocurrent has been observed at grain boundaries (Visoly-Fisher et al., 2003, 2006; Li et al., 2014). Grain boundaries could also act as localized collection sites for defects, improving the carrier collection within the interior of the neighboring grains (Li et al., 2014). Grain boundaries can also act as diffusion pathways during the CdCl2 annealing treatment and facilitate high-efficiency devices. Grain boundary density would be expected to influence device performance and therefore average grain size (GS) could be a useful metric for correlation with device performance (Yan et al., 2000). Major et al. (2010) found a linear correlation Received November 14, 2014; accepted May 8, 2015 *Corresponding author. [email protected] †Current address: Corning Incorporated, One Riverfront Plaza, Corning, NY 14831 USA ‡Current address: Institute of Electrical Engineering, Chinese Academy of Sciences, Beijing 100190, P.R. China
between efficiency and GS in CdTe when the close spaced sublimation (CSS) growth rate was controlled by pressure. Eisenbarth et al. (2009) found limited correlation for Copper– Indium–Gallium–Selenide (CIGS) with GS dependent on the Ga/In ratio. However, these studies have not attempted to classify the electrical performance of different grain boundary types. The GS of magnetron-sputtered CdTe films can be controlled with the sputter gas pressure as well as with growth temperature. In this work we have used the sputter gas pressure to control the as-deposited film characteristics and a standard CdCl2 treatment that produces considerable grain recrystallization. Whereas the 8-µm-thick CSS films with average GSs below ~2 µm (Major et al., 2010) showed poor cell performance, the 2.3-µm-thick sputtered films in this study with average GSs of 500 nm perform similar to the CSS films with larger ~4 µm grains. This implies that the electrical properties of the grain boundaries depend on the growth method. Before CdCl2 treatment cell performance is poor. Therefore, this study focuses on the relationship between the cell performance and film properties after the “activation” treatment. After CdCl2 activation films exhibit a large number of twin boundaries, and we are particularly interested in how these twin boundaries exhibit different electrical and diffusional properties relative to random high-angle grain boundaries (Abou-Ras et al., 2009). To characterize the grain boundary structure and differentiate twin boundaries from random high-angle grain
928
Nowell, Scarpulla & Paudel et al.
boundaries, electron backscatter diffraction (EBSD) was used to characterize the CdTe films to correlate measured microstructural features with device performance. EBSD allows for the concurrent measurement of GS, crystallographic texture, and grain boundary character distribution (GBCD) of a material (Schwarzer et al., 2009). These results also aid in understanding the active mechanisms of film deposition and growth as well as recrystallization and grain growth accompanying the CdCl2 treatment.
M ATERIALS AND M ETHODS Three sets of CdTe films were grown using radio frequency (RF) magnetron sputtering. Two sets were used for EBSD characterization and another set was used for cell performance characterization. The EBSD-characterized films were grown on a commercial SnO2:F (Pilkington TEC7, Pilkington, North America) glass substrate while the performance-characterized films were grown on a TEC15 + HRT (high resistivity transparent) substrate. The different substrate choice was incidental and based on prior experience we expect the different substrates not to affect grain growth. Substrates were ~76 × 76 × 3 mm. For all sets, a 0.11-µm-thick CdS layer was sputter deposited at 18 mTorr Ar pressure and an RF power of 35 W at a rate of 9 nm/min. The Ar gas pressure during CdTe deposition was selected as the primary variable. Films were deposited at constant RF power of 20 W, substrate temperature of 250°C, and with pressures (deposition rates) of 5 mTorr (21 nm/min), 18 mTorr (16.5 nm/min), and 30 mTorr (13 nm/min). The small variation in final CdTe film thickness is not expected to influence device performance (Plotnikov et al., 2009). One set of the EBSD-characterized films was kept in the as-grown state. The other EBSD set and the set used for cell fabrication received a 30 min 387°C CdCl2 treatment in air. Films received a methanol rinse but no chemical etching. Cells were made by evaporating back contacts through a mask using 3 nm of Cu and 20 nm of Au and heat treating at 150°C for 45 min. Before both EBSD and device characterization, samples were cut into ~5 × 10 mm pieces, and samples from the center and edge of the deposition area were analyzed to investigate any effects of lateral inhomogeneity. Further details on the sputtering procedure are given elsewhere (Shao et al., 1996; Gupta & Compaan, 2004; Compaan, 2006). Initially both the as-grown and CdCl2-treated films were examined with EBSD in the as-deposited condition without any EBSD-specific surface preparation. While some EBSD patterns were observed in this state, surface roughness caused signal shadowing problems and the overall EBSD pattern indexing rate (as described below) was below 50%, which was considered too low for accurate characterization. A number of different EBSD preparation techniques for CdTe films have been described previously (Moutinho et al., 2008): in this work, lowincidence surface-milling (LISM) was used (Zaefferer et al., 2008). An FEI Quanta 3D FEG Dual Beam instrument (EDAXTSL, Ametak (Material Analysis Division)) equipped with
both electron and focused Ga+ ion beams [focused ion beam (FIB)] was used to cut a flat analysis plane at 1.5° to the surface before EBSD. This angle is smaller than the 10–20° typically reported for LISM, but was required due to the thickness of the CdTe thin films and resulted in better resolution through the depth of the film. A mill box of 30 × 1.5 × 2.0 µm was positioned on the film surfaces and milled at 30 kV beam energy and 1 nA of beam current with an approximate milling time of 10 min. In this geometry, the Z-depth of the milling box essentially controls the length of the cut on the sample surface. While the 30 kV ion beam energy cuts produced high-quality EBSD patterns from these (and other) CdTe films, it has been previously reported that high-energy FIB milling can affect EBSD pattern quality in semiconducting materials (Matteson et al., 2002; Michael & Giannuzzi, 2007), and other authors have observed a strong dependence of EBSD pattern quality with FIB milling energy on some CIGS photovoltaic (PV) thin films (Bhatia et al., 2010; Hlaing et al., 2011). While the samples could be analyzed with EBSD using the dual beam instrument, in this work the samples were transferred to an FEI XL-30 FEG SEM (EDAX-TSL, Ametak) to maximize sample throughput. This scanning electron microscope (SEM) was equipped with an EDAX-TSL Hikari EBSD detector (EDAX-TSL, Ametak) running OIM Data Collection version 5.32. The samples were tilted so that the milled analysis plane was at 15° relative to the vertical electron beam. Note that the 1.5° incident angle of the FIB cut results in a slight offset of the measured orientation relative to the film surface normal orientation, which was corrected during postprocessing data analysis. The EBSD patterns were collected at 20 kV and an approximate beam current of 3 nA at a 14 mm working distance. The EBSD detector was set at 100 EBSD patterns per second acquisition at a resolution of 96 × 96 pixels and the camera gain set to a value of ~5 dB to optimize the signal intensity. For mapping, EBSD patterns were automatically collected and analyzed at a rate of 100 points/s from an 18.5 × 18.5 µm area using a hexagonal sampling grid with a 25 nm spacing resulting in ~6.4 × 105 measurements in 107 min. Before map collection, the region of interest was imaged for ~30 min under the same beam conditions to allow the instrument and sample to stabilize and avoid any drifting issues. The mapping location was positioned so that the bottom of the EBSD images was the FIB-etched area nearest the CdTe surface that was completely smooth after milling. Thus, we avoided the transition region between the as-processed surface and the prepared surface. The collected EBSD mapping data was analyzed using EDAX-TSL OIM Analysis version 5.31. The EBSD indexing rate, which is the percentage of measurement points that are analyzed correctly, was determined by the fraction of points with a confidence index (CI) >0.2 after a grain CI standardization procedure (Field, 1997; Nowell & Wright, 2005) relative to the total number of points collected. A systematic noise-reduction routine was also applied. Low-confidence points remaining after these steps were excluded from subsequent analysis.
Characterization of Sputtered CdTe Thin Films
RESULTS AND D ISCUSSION As-Grown CdTe Films Figure 1 shows an EBSD Image Quality (IQ) map (Wright & Nowell, 2006) from the center region of the as-grown film deposited at 30 mTorr, which is generally representative of the microstructure observed in all the as-grown films. Figure 1a shows the entire collection area, where clearly many grains have been observed, while Figure 1b shows a magnified region within the sampled area highlighting that the 25 nm step size used is adequate to resolve the observed microstructure. The total numbers of grains measured, excluding edge grains for more accurate GS measurements, are 7,868 for 5 mTorr pressure, 6,377 for 18 mTorr pressure, and 3,392 for 30 mTorr pressure for the center regions. While these numbers of grains are lower than the >10,000 typically desired for statistically reliable texture analysis,
Figure 1. EBSD image quality map from center region of as-deposited 30 mTorr film showing (a) full region of interest and (b) magnified region of interest. EBSD, electron backscatter diffraction.
929
comparisons of the measured textures from samples prepared by FIB with larger area analyses performed on broad ion-beam-prepared samples (Nowell, 2002) of the same films with >100,000 grains measured show good correlation and are deemed acceptable. In this work, grains are defined by boundaries with an angular misorientation >5° and must contain at least 2 pixels of similar orientation. The 2 pixel minimum gain threshold was selected to maximize spatial resolution performance with the 25 nm sampling step size and, when used in conjunction with only high CI data, was deemed acceptable. Further details are given elsewhere on quantitative texture measurements (Wright et al., 2007) and on the determination of grains (Wright et al., 2000) using EBSD. The indexing rate for these films was ~80–90%. The lower confidence regions were primarily located between well-defined and indexed grains. This may be due to a disordered region arising between adjacent grains during coalescence (Nalwa, 2002). A disordered crystallographic structure would inhibit EBSD pattern formation and result in low-confidence data points. However, based on experience with Cu(In,Ga)Se2 films, it is possible that a lower energy FIB mill would improve pattern quality between the defined grains (Bhatia et al., 2010). Figure 2a shows average GS (diameter) and texture index (TI) of the films as a function of deposition pressure and sampling location (center or edge). TI, as defined by Bunge (1993), allows numerical characterization of the degree of preferred orientation without detailing the specific orientation distribution. This approach was used because the films exhibited different preferred orientations. The TI values vary between 1 for completely random orientation distributions to ∞ for an ideal single crystal. The TIs for these as-sputtered films are distributed approximately between values of 1.15–1.6 indicating generally weak textures. For comparison, we have measured a physical vapor deposited (PVD) CdTe film deposited at 325°C and PVD aluminum film that both showed the canonical strong (111) fiber textures with TI≈9. Lateral heterogeneity in film microstructures was examined by collecting data from the center and edge of the films. The TI of the edge of samples was lower than the TI for the center of samples for all deposition pressures. The difference in GS center to edge depended on pressure; however, average GS at 5 mTorr was ~175 nm with little variation across the film. At 18 and 30 mTorr, the GS at the edge remained ~175 nm, but GS at the center increased with pressure to 300 nm at 30 mTorr. Since film thickness is about 15% less at the edges, the growth rate is lower; this would predict larger grains at the edges. This indicates that the higher ion and atom bombardment in the center is playing a role in nucleation and/or grain growth although the details are not understood (Gupta & Compaan, 2004). Figure 2b presents the data for GS and TI as functions of depth through the films. The shallow 1.5o angle of the LISM preparation allows the EBSD data to be partitioned according to depth with reasonable statistics. Since the bottom of the images corresponds to 75–100 nm below the CdTe
930
Nowell, Scarpulla & Paudel et al.
Figure 2. Average GS and TI versus deposition pressure for the as-deposited films at (a) different lateral locations and (b) as functions of depth in the film. GS, grain size; TI, texture index.
surface, the top represents a depth of about 500 nm. Both GS and TI values decrease with depth below the surface. GSs decrease by 10–15% in 400 nm. This is consistent with a selected oriented growth in which the grains with faster growth gradually dominate. Longer FIB milling times could be used to study the full film thickness but this was not the focus of this study. The observed weak texture in the films of this study is also observed using the maximum values of the (111) and (110) pole figures, as shown in Figures 3a and 3b for three deposition pressures. The pole figures are presented normalized to the statistically expected texture values for random grain orientations. Films deposited at 5 mTorr have a weak (111) fiber texture while the films deposited at 18 and 30 mTorr have weak (110) fiber textures. These were confirmed with the initial measurements made on the as-received samples without LISM. The presence of two primary textures, while low in strength, suggests two different microstructural evolution processes. While the textures of the underlying CdS layers were not measured, all CdS
Figure 3. CdTe (111) and (110) pole figures for deposition pressures of 5, 18, and 30 mTorr for: (a) as-sputtered films and (b) CdCl2-treated films.
layers were deposited at 18 mTorr, which should eliminate CdS orientation effects as a cause of this difference. X-ray diffraction measurements show evidence of in-plane compressive stress producing strain in the growth direction that is 0.5% at 5 mTorr diminishing to 0.1% at 30 mTorr. The (111) texture typically develops due to surface and interface energy minimization effects (Thompson & Carel, 1995, 1996). The (111) texture present in the 5 mTorr film is stronger than the (110) texture in the other films. While deposition temperature and RF power were held constant for all films, both the rate and kinetic energy of atoms impinging on the growth interface at 5 mTorr will be higher due to the lower gas pressure and fewer gas collisions. For CSS CdTe films (Major et al., 2010) a similar decrease in (111) texture was observed as growth rate was reduced at higher pressures. This was interpreted as due to greater competition from the faster-growing (111) grains because of the higher nucleation
Characterization of Sputtered CdTe Thin Films
density at faster growth rates. The larger impinging atom energy produces more strain and may also play a role in development of the (111) texture, but the strain energy will affect recrystallization during CdCl2 treatment discussed later. Consonni et al. (2008) reported evolution of (110) textures driven by strain energy minimization in 1,000-µmthick CdTe films, but it is not clear that strain effects are great enough to affect our 2 µm films. EBSD can also be used to characterize small angle misorientations within grains (Brewer et al., 2009). Here we used the metric of local orientation spread (LOS), where a kernel (third nearest neighbors in this case) is centered on each point (Wright et al., 2011). The misorientation between each point and every other point in the kernel is then calculated (excluding values >5° corresponding to high-angle grain boundaries). The average value for the kernel is then calculated and assigned to that point. The average LOS values for the central region were 0.90° for 5 mTorr, 0.71° for 18 mTorr, and 0.65° for 30 mTorr. These results again suggest that the increased ion bombardment and kinetic energy of the sputtered material associated with lower pressures cause a measureable change in the crystalline microstructure.
931
These misorientation defects could influence device performance by inducing charge carrier traps.
CdCl2-Treated CdTe Films Figure 4 shows EBSD IQ and secondary electron (SE) SEM signal maps from the CdCl2-treated center region samples. These structures are vastly different than the as-deposited microstructures, and show that recrystallization and grain growth has occurred as observed for other low-temperature grown CdTe films (Romeo et al., 2000; Zoppi et al., 2006). The near-vertical lines in the SE images are due to curtaining effects during the FIB preparation. EBSD indexing rates are significantly higher for the treated films, with >98% indexing for the 18 and 30 mTorr deposition pressures and >92% for the 5 mTorr films. The interfaces between adjacent grains are much sharper than the as-deposited films. Voids are visible for these CdCl2-treated films in both types of images, although they are more clearly seen in the SE image. Atomic force microscopy measurements were used to confirm that these were voids rather than, for example, secondary phases of differing conductivity. The observation
Figure 4. EBSD image quality (a–c) and SEM secondary electron (d–f) images for CdCl2-treated films at 5, 18, and 30 mTorr deposition pressures, respectively. Voids are more easily visible in the secondary electron images. (Note: sampling was taken from the center piece of the sputtered film.) EBSD, electron backscatter diffraction; SEM, scanning electron microscope.
932
Nowell, Scarpulla & Paudel et al.
of voids via SEM examination of the treated films away from the FIB-prepared area for EBSD analysis confirmed that these defects were not artifacts of FIB preparation. These voids were also not observed in the as-deposited state before CdCl2 treatment when imaging the FIB-milled surface. The area fractions of voids calculated using SE gray-scale image thresholding were 14.5, 1.6, and 3.9% for the 5, 18, and 30 mTorr deposition pressures, respectively. The larger fraction of voids within the 5 mTorr films explains the lower indexing rate relative to the other samples. The fact that the EBSD indexing rates and void fractions do not total 100% is explained by the fact that indexable EBSD patterns are still obtained from some portions of the surface identified as voids by SE image intensity value. Figure 5 shows pseudo-grain maps for the 18 mTorr center region sample. Typically, grain maps have randomly colored grains to highlight GS and morphology. In this work, the random color maps have been transformed into a 12-bit gray-scale image using ImageJ (Abramoff et al., 2004). To visually delineate adjacent grains that may have similar gray-scale values, black lines have been drawn around each determined grain. It is apparent from this image that CdCl2-treated microstructures also contain high fractions of twin boundaries, indicating that twinning occurs during the recrystallization process. Using EBSD mapping to characterize the GBCD allows categorization of grain boundaries as low-angle grain boundaries (5–15° misorientations), twin boundaries [Σ3 or Σ9 coincident site lattice (CSL) boundaries in this case], or random high-angle grain boundaries (misorientations >15° not meeting CSL criteria). Misorientations 5° misorientation (i.e., all grains comprising the GBCD) have been included in the grain calculations. In Figure 5b, the twin boundaries have been selectively ignored during the grain calculations to produce what is termed here a twin-corrected (TC) grain structure. The average GS distributions for both the standard and TC grain structures are shown in Figure 6 as a function of deposition pressure and sample location (center or edge). Note in all cases the GSs, defined with or without twins, are significantly larger in the centers of the samples than at the edge locations. The fiber textures initially present have disappeared in the CdCl2-treated films, and the textures have generally weakened and become less recognizable as was shown in the pole figures (Fig. 3). Although fewer grains were examined than with the as-deposited films (typically between 1,500 and 3,000 grains), the measured textures were again compared against broad ion-beam-prepared areas with >150,000 grains and good correlation found. The texture weakening is not unexpected with the amount of twinning observed, but it was not possible to systematically retrace the observed
Figure 5. EBSD grain maps (a) including (standard grain size) and (b) excluding twin boundaries as grain boundaries (twin corrected) from the center of the 18 mTorr CdCl2-treated sample. Grain boundaries are drawn as solid black lines. (Note: sampling was taken from the center piece of the sputtered film.) EBSD, electron backscatter diffraction.
orientations back to the original deposited textures using twinning relationships, as has been done previously in CIGS films (Abou-Ras & Pantleon, 2007). However, the textures were generally weak in both the as-deposited and treated samples; because of this, it is difficult to determine if selected orientation grain growth (and twinning) is occurring or if
Characterization of Sputtered CdTe Thin Films
933
Figure 6. Average grain size and TC grain size versus deposition pressure for CdCl2-treated samples. TC, twin-corrected.
new recrystallized nuclei are being generated and then proceeding to grow and twin. For CdTe, new recrystallized nuclei appear to explain regrowth for low-temperature grown films (Moutinho et al., 1998, 2008), but not if the film growth was done at high temperature (≥550°C) (Romeo et al., 2000; Zoppi et al., 2006; Manivannan et al., 2008). The textures at the edge of the treated samples are again weaker than those in the center region, but there is little correlation between the textures unlike the as-deposited films. The LOSs of the treated films were significantly lower than the as-deposited films, with values of 0.49°, 0.41°, and 0.39° for the 5, 18, and 30 mTorr deposition pressures, respectively, for the center regions. This result is consistent with the deposition pressure-dependent strain providing an additional driving force for recrystallization resulting in different retained orientation spreads after recrystallization and grain growth.
Correlation of Device Performance with Measured Microstructures The second set of CdCl2-treated samples received 3 mm diameter back contacts to complete an array of solar cells and their performance was characterized. To correlate performance with the EBSD-characterized microstructure, the four dot cells nearest the center and nearest an edge of each film were chosen to obtain averages. In rare cases for which such a dot cell appeared anomalous compared with the other 140 cells on the film, an adjacent cell was chosen to average. The averaged power-conversion efficiency and fill factor (FF) as functions of either standard or TC average GS are shown in Figures 7a and 7b, respectively. Six independent grain-size data points are created with three deposition pressures and center and edge locations for each. Note that this assumes that the major variable is GS and that, for example, sputter pressure is of secondary importance.
Figure 7. a: Average efficiency and FF versus nontwin-corrected average grain size for CdCl2-treated samples. b: Average efficiency and FF versus twin-corrected grain size for CdCl2-treated samples. The performance parameters shown in (a) and (b) are corresponding data points of three pressures and two locations shown in Figure 6. FF, fill factor.
The correlation of performance with standard GS (uncorrected for twinning) is poor; however, a general trend of increasing device efficiency and FF with GS emerges once the twin boundaries are ignored in the grain boundary angular distributions. The general trend is analogous to an inverse Hall–Petch-type relationship with efficiency and FF increasing as TC GS increases. This is consistent with twin boundaries exhibiting different recombination properties than random high-angle grain boundaries—in fact it suggests that the twin boundaries do not contribute to nonradiative recombination at low carrier generation levels. Cathodoluminescence measurements (corresponding to high injection conditions) on CIGS films have shown similar correlation with EBSD grain boundary type (Abou-Ras et al., 2009). Twin boundaries typically have lower grain boundary energies than random high-angle grain boundaries and are expected to have different diffusional properties that may affect doping and passivation mechanisms. In addition, twin boundaries are likely to have lower densities of recombination site defects. In 8-µm-thick films grown by
934
Nowell, Scarpulla & Paudel et al.
CSS, a plateauing of efficiency versus GS was reported at the point at which series resistance is no longer dominated by grain boundaries but rather by the contacts (Major et al., 2010). Our data for these 2 µm sputtered films show a similar plateau only with TC GS. The observed 100–200 nm voids, as well as electronic defects likely to occur at points having large LOS, could also be limiting the cell performance. While these results show a general correlation of cell performance with microstructure, this work represents a statistical approach characterizing a large number of grain boundaries. In addition, random high-angle grain boundaries are expected to have varying properties leading to varying cell performance. Enhanced photocurrents have been observed by two groups at some grain boundaries (Visoly-Fisher et al., 2003, 2006; Smith et al., 2004; Li et al., 2014). While these enhanced boundaries have not been characterized in terms of misorientation relationships, the present work suggests using a FIB-prepared sample for EBSD crystallographic characterization to develop a structure versus performance matrix beyond the general grain boundary characterization methodology used in this work. It would be desirable to characterize grain boundaries according to those that are advantageous, disadvantageous, and neutral to cell performance. This would of course lead to the ability to engineer specific grain boundaries to improve performance (Fraas, 1978; Randle, 2010), which has been discussed for thin films previously (Field et al., 2005; Park et al., 2005; Park & Field, 2006).
S UMMARY
AND
CONCLUSIONS
CdTe films sputtered at three different deposition pressures were characterized before and after CdCl2 treatment with EBSD in order to correlate microstructural measurements with PV cell performance. A LISM FIB sample-preparation technique allowed for measurements of microstructure as a function of film depth during EBSD. This revealed an increase of GS and texture strength during film growth suggesting an oriented growth process. The CdCl2-treated samples had a recrystallized microstructure with larger GSs and weakened fiber texture relative to the as-deposited films. A high fraction of twin boundaries was observed in all of the CdCl2-treated films. Weak fiber textures were observed in all of the as-deposited films with a (111) texture observed for lower deposition pressures and a (110) texture observed for higher deposition pressures. The (111) texture arises at the higher growth rates that occur at lower pressure as also seen with other low-temperature growth of CdTe. However, the texture development is likely to also be influenced by the substantial strain in the low-pressure-grown films. An important aspect of this work was the establishment of correlations observed between increasing TC GS and improved efficiency and FF, which were absent if the standard GS was considered. This is attributed to the twin boundaries being much less active in carrier recombination and presenting lower barriers to carrier transport than do random high-angle grain boundaries.
ACKNOWLEDGMENTS The authors would like to thank Stuart Wright and John Carpenter at EDAX-TSL for assistance and discussions on low-incidence surface-milling focused ion beam preparation and CdTe microstructures, Elizabeth Lund at the University of Utah for performing the atomic force microscopy measurements, David Field at Washington State University for valuable discussions on thin film textures, and Dohyoung Kwon at the University of Toledo for some of the X-ray diffraction measurements. M.A.S. acknowledges support from the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering under Award DE-SC0001630.
REFERENCES ABOU-RAS, D., KOCH, C.T., KÜSTNER, V., VAN AKEN, P.A., JAHN, U., CONTRERAS, M.A., CABALLERO, R., KAUFMANN, C.A., SCHEER, R., UNOLD, T. & SCHOCK, H.W. (2009). Grain-boundary types in chalcopyrite-type thin films and their correlations with film texture and electrical properties. Thin Solid Films 517, 2545–2549. ABOU-RAS, D. & PANTLEON, K. (2007). The impact of twinning on the local texture of chalcopyrite-type thin films. Phys Stat Sol (RRL) 1, 187–189. ABRAMOFF, M.D., MAGALHAES, P.J. & RAM, S.J. (2004). Image processing with imageJ. Biophoto Inter 11, 36–43. BHATIA, A., DALE, P.J., NOWELL, M.M. & SCARPULLA, M.A. (2010). Pulsed laser processing of electrodeposited CuInSe2 photovoltaic absorber thin films. Proc Mater Res Soc 1268, 103. BREWER, L.N., FIELD, D.P. & MERRIMAN, C.C. (2009). Mapping and assessing plastic deformation using EBSD. In Electron Backscatter Diffraction in Materials Science, 2nd ed. Schwartz, A.J., Kumar, M., Field, D.P. & Adams, B.L. (Eds.), pp. 251–262. Springer. BUNGE, H.J. (1993). Texture Analysis in Materials Science: Mathematical Methods. Gottingen, Germany: Cuvillier Verlag. COMPAAN, A.D. (2006). Photovoltaics: Clean power for the 21st century. Sol Energy Mater Sol Cells 90, 2170–2180. CONSONNI, V., FEUILLET, G. & GERGAUD, P. (2008). Plasticity induced texture development in thick polycrystalline CdTe: Experiments and modeling. J Appl Phys 103, 063529. EISENBARTH, T., UNOLD, T., CABALLERO, R., KAUFMAN, C.A., ABOU-RAS, D. & SCHOCK, H.W. (2009). Origin of defects in CuIn1-xGaxSe2 solar cell with varied Ga content. Thin Solid Films 517, 2244–2247. FIELD, D.P. (1997). Recent advances in the application of orientation imaging. Ultramicroscopy 67, 1–9. FIELD, D.P., SANCHEZ, J.E., PARK, N.J. & BESSER, P.R. (2005). Texture evolution in thin Cu films and lines, materials science forum. In Textures of Materials—ICOTOM 14, Van Houtte, P. & Kestens L. (Eds.), Vols 495–497, pp. 1323–1332. Leuven, Belgium: Trans Tech Publications. FRAAS, L.M. (1978). Basic grain‐boundary effects in polycrystalline heterostructure solar cells. J Appl Phys 49, 871. GUPTA, A. & COMPAAN, A.D. (2004). All-sputtered 14% CdS/CdTe thin-film solar cell with ZnO: Al transparent conducting oxide. Appl Phys Lett 85, 684–686. HLAING, O.O.W.M., JOHNSON, J.L., BHATIA, A., LUND, E.A., NOWELL, M.M. & SCARPULLA, M.A. (2011). Grain size and texture of Cu2ZnSnS4 thin films synthesized by co-sputtering binary sulfides and annealing: Effects of processing conditions and sodium. J Electron Mater 40, 2214–2221.
Characterization of Sputtered CdTe Thin Films LI, C., WU, Y., POPLAWSKY, J., PENNYCOOK, R.J., PAUDEL, N.R., YIN, W., HAIGH, S.J., OXLEY, M.P., LUPINI, A.R., AL-JASSM, M., PENNYCOOK, S.J. & YAN, Y. (2014). Grain-boundary-enhanced carrier collection in CdTe solar cells. Phys Rev Let 112, 156103. MAJOR, J.D., PROSKURYAKOV, Y.Y., DUROSE, K., ZOPPI, G. & FORBES, I. (2010). Control of grain size in sublimation-grown CdTe, and the improvement in performance of devices with systematically increased grain size. Sol Energy Mater Sol Cells 94, 1107–1112. MANIVANNAN, V., ENZENROTH, R.A., BARTH, K.L., KOHLI, S., MCCURDY, P.R. & SAMPATH, W.S. (2008). Microstructural features of CdTe PV thin film devices. Thin Solid Films 516, 1209–1213. MATTESON, T.L., SCHWARZ, S.W., HOUGE, E.C., KEMPSHALL, B.W. & GIANNUZZI, L.A. (2002). Electron backscattering diffraction investigation of focused ion beam surfaces. J Electron Mater 31, 33–39. MICHAEL, J.R. & GIANNUZZI, L.A. (2007). Improved EBSD sample preparation via low energy Ga + FIB ion milling. Microsc Microanal 13, 926–927. MOUTINHO, H.R., AL-JASSIM, M.M., LEVI, D.H., DIPPO, P.C. & KAZMERSKI, L.L. (1998). Papers from the 44th National Symposium of the AVS. San Jose, CA, USA: AVS. 1251pp. MOUTINHO, H.R., DHERE, R.G., ROMERO, M.J., JIANG, C.S., TO, B. & AL-JASSIM, M.M. (2008). Electron backscatter diffraction of CdTe thin films: Effects of CdCl2 treatment. J Vac Sci Tech A 26, 1068–1073. NALWA, H.G. (2002). Handbook of Thin Film Materials: Semiconductor and Superconductor Thin Films. San Diego, CA: Academic Press. NOWELL, M.M. (2002). Ion beam preparation of passivated copper integrated circuit structures for electron backscatter diffraction/ orientation imaging microscopy analysis. J Electron Mater 31, 23–32. NOWELL, M.M. & WRIGHT, S.I. (2005). Orientation effects on indexing of electron backscatter diffraction patterns. Ultramicroscopy 103, 41–58. PARK, N.-J. & FIELD, D.P. (2006). Predicting thickness dependent twin boundary formation in sputtered Cu films. Scripta Mater 54, 999–1003. PARK, N.-J., FIELD, D.P., NOWELL, M.M. & BESSER, P.R. (2005). Effect of film thickness on the evolution of annealing texture in sputtered copper films. J Electron Mater 34, 1500–1508. PLOTNIKOV, V.V., VASKO, A.C., COMPAAN, A.D., LIU, X., WIELAND, K.A., ZELLER, R.M., LI, J. & COLLINS, R.W. (2009). Magnetron sputtering for II-VI solar cells: Thinning the CdTe. Proc Mater Res Soc 1165, M09. RANDLE, V. (2010). Grain boundary engineering: An overview after 25 years. Mater Sci Tech 26, 253–261. ROMEO, A., BATZNER, D.L., ZOGG, H. & TIWARI, A.N. (2000). Recrystallization in CdTe/CdS. Thin Solid Films 361–362, 420–425. SCHWARZER, R.A., FIELD, D.P., ADAMS, B.L., KUMAR, M. & SCHWARTZ, A.J. (2009). Present state of electron backscatter diffraction and prospective developments. In Electron Backscatter Diffraction in Materials Science, 2nd ed. Schwartz, A.J., Kumar, M., Field, D.P. & Adams B.L. (Eds.), pp. 1–20. Springer.
935
SHAO, M., FISCHER, A., GRECU, D., JAYAMAHA, U., BYKOV, E., CONTRERAS-PUENTE, G., BOHN, R.G. & COMPAAN, A.D. (1996). Radio-frequency-magnetron-sputtered CdS/CdTe solar cells on soda-lime glass. Appl Phys Lett 69, 3045–3048. SMITH, S., ZHANG, P., GESSERT, T. & MASCARENHAS, A. (2004). Near-field optical beam-induced currents in CdTe/CdS solar cells: Direct measurement of enhanced photoresponse at grain boundaries. Appl Phys Lett 85, 3854. THOMPSON, C.V. & CAREL, R. (1995). Texture development in polycrystalline thin films. Mater Sci Eng B 32, 211–219. THOMPSON, C.V. & CAREL, R. (1996). Stress and grain growth in thin films. J Mech Phys Solids 44, 657–673. VISOLY-FISHER, I., COHEN, S.R. & CAHEN, D. (2003). Direct evidence for grain boundary depletion in polycrystalline CdTe from nanoscale-resolved measurement. Appl Phys Lett 82, 556–558. VISOLY-FISHER, I., COHEN, S.R., GARTSMAN, K., RUZIN, A. & CAHEN, D. (2006). Understanding the beneficial role of grain boundaries in polycrystalline solar cells from single grain boundary scanning probe microscopy. Adv Funct Mater 16, 649–660. VISOLY-FISHER, I., COHEN, S.R., RUZIN, A. & CAHEN, D. (2004). How polycrystalline devices can outperform single-crystal ones: Thin film CdTe/CdS solar cells. Adv Mater 16, 879–883. WRIGHT, S.I. (2002). Investigation of coincident site lattice boundary criteria in Cu thin films. J Electron Mater 31, 50–54. WRIGHT, S.I., FIELD, D.P. & DINGLEY, D.J. (2000). Advanced software capabilities for automated EBSD. In Electron Backscatter Diffraction in Materials Science, Schwartz, A.J., Kumar, M. & Adams B.L. (Eds.), pp. 141–152. New York, NY: Kluwer Academic/Plenum Publishers. WRIGHT, S.I. & LARSEN, R.J. (2002). Extracting twins from orientation imaging microscopy scan data. J Microsc 205, 245–252. WRIGHT, S.I. & NOWELL, M.M. (2006). EBSD image quality mapping. Microsc Microanal 12, 72–84. WRIGHT, S.I., NOWELL, M.M. & BINGERT, J.F. (2007). A comparison of textures measured using X-ray and electron backscatter diffraction. Metall Mater Trans A 38, 1845–1855. WRIGHT, S.I., NOWELL, M.M. & FIELD, D.P. (2011). A review of strain analysis using EBSD. Microsc Microanal 17, 316–329. YAN, Y., AL-JASSIM, M.M., JONES, K., WEI, S.-H. & ZHANG, S.B. (2000). Observation and first-principles calculations of buried wurtzite phases in zinc-blende CdTe thin films. Appl Phys Lett 77, 1461–1463. YAN, Y., JONES, K.M., JIANG, C.S., WU, X.Z., NOUFI, R. & AL-JASSIM, M.M. (2007). Understanding the defect physics in polycrystalline photovoltaic materials. Physica B 401–402, 25–32. ZAEFFERER, S., WRIGHT, S.I. & RAABE, D. (2008). Three dimensional orientation microscopy in a focused ion beam-scanning electron microscope: A new dimension of microstructure characterization. Metall Mater Trans A 39, 374–389. ZOPPI, G., DUROSE, K., IRVINE, S.J.C. & BARRIOZ, V. (2006). Grain and crystal texture properties of absorber layers in MOCVD-grown CdTe/CdS solar cells. Semicond Sci Technol 21, 763–770.
© MICROSCOPY SOCIETY OF AMERICA 2015
Characterization of Sputtered CdTe Thin Films with Electron Backscatter Diffraction and Correlation with Device Performance Matthew M. Nowell,1 Michael A. Scarpulla,2 Naba R. Paudel,3,* Kristopher A. Wieland,3,† Alvin D. Compaan,3 and Xiangxin Liu3,‡ 1
EDAX-TSL, 392 East 12300 South Suite H, Draper, UT 84020, USA Departments of Materials Science and Engineering and Electrical and Computer Engineering, University of Utah, Salt Lake City, UT 84112, USA 3 Department of Physics and Astronomy, University of Toledo, Toledo, OH 43606, USA 2
Abstract: The performance of polycrystalline CdTe photovoltaic thin films is expected to depend on the grain boundary density and corresponding grain size of the film microstructure. However, the electrical performance of grain boundaries within these films is not well understood, and can be beneficial, harmful, or neutral in terms of film performance. Electron backscatter diffraction has been used to characterize the grain size, grain boundary structure, and crystallographic texture of sputtered CdTe at varying deposition pressures before and after CdCl2 treatment in order to correlate performance with microstructure. Weak fiber textures were observed in the as-deposited films, with (111) textures present at lower deposition pressures and (110) textures observed at higher deposition pressures. The CdCl2-treated samples exhibited significant grain recrystallization with a high fraction of twin boundaries. Good correlation of solar cell efficiency was observed with twin-corrected grain size while poor correlation was found if the twin boundaries were considered as grain boundaries in the grain size determination. This implies that the twin boundaries are neutral with respect to recombination and carrier transport. Key words: CdTe, sputtering, EBSD, thin film
I NTRODUCTION The effects of grain boundaries in polycrystalline films on solar cell performance are not well understood. Grain boundaries can act as recombination sites, reducing performance (Visoly-Fisher et al., 2004). However, grain boundaries can also form a localized electric field, either through structural or impurity effects (Yan et al., 2007), which can improve majority and minority carrier separation, and collection and thus improve performance (Li et al., 2014). Enhanced photocurrent has been observed at grain boundaries (Visoly-Fisher et al., 2003, 2006; Li et al., 2014). Grain boundaries could also act as localized collection sites for defects, improving the carrier collection within the interior of the neighboring grains (Li et al., 2014). Grain boundaries can also act as diffusion pathways during the CdCl2 annealing treatment and facilitate high-efficiency devices. Grain boundary density would be expected to influence device performance and therefore average grain size (GS) could be a useful metric for correlation with device performance (Yan et al., 2000). Major et al. (2010) found a linear correlation Received November 14, 2014; accepted May 8, 2015 *Corresponding author. [email protected] †Current address: Corning Incorporated, One Riverfront Plaza, Corning, NY 14831 USA ‡Current address: Institute of Electrical Engineering, Chinese Academy of Sciences, Beijing 100190, P.R. China
between efficiency and GS in CdTe when the close spaced sublimation (CSS) growth rate was controlled by pressure. Eisenbarth et al. (2009) found limited correlation for Copper– Indium–Gallium–Selenide (CIGS) with GS dependent on the Ga/In ratio. However, these studies have not attempted to classify the electrical performance of different grain boundary types. The GS of magnetron-sputtered CdTe films can be controlled with the sputter gas pressure as well as with growth temperature. In this work we have used the sputter gas pressure to control the as-deposited film characteristics and a standard CdCl2 treatment that produces considerable grain recrystallization. Whereas the 8-µm-thick CSS films with average GSs below ~2 µm (Major et al., 2010) showed poor cell performance, the 2.3-µm-thick sputtered films in this study with average GSs of 500 nm perform similar to the CSS films with larger ~4 µm grains. This implies that the electrical properties of the grain boundaries depend on the growth method. Before CdCl2 treatment cell performance is poor. Therefore, this study focuses on the relationship between the cell performance and film properties after the “activation” treatment. After CdCl2 activation films exhibit a large number of twin boundaries, and we are particularly interested in how these twin boundaries exhibit different electrical and diffusional properties relative to random high-angle grain boundaries (Abou-Ras et al., 2009). To characterize the grain boundary structure and differentiate twin boundaries from random high-angle grain
928
Nowell, Scarpulla & Paudel et al.
boundaries, electron backscatter diffraction (EBSD) was used to characterize the CdTe films to correlate measured microstructural features with device performance. EBSD allows for the concurrent measurement of GS, crystallographic texture, and grain boundary character distribution (GBCD) of a material (Schwarzer et al., 2009). These results also aid in understanding the active mechanisms of film deposition and growth as well as recrystallization and grain growth accompanying the CdCl2 treatment.
M ATERIALS AND M ETHODS Three sets of CdTe films were grown using radio frequency (RF) magnetron sputtering. Two sets were used for EBSD characterization and another set was used for cell performance characterization. The EBSD-characterized films were grown on a commercial SnO2:F (Pilkington TEC7, Pilkington, North America) glass substrate while the performance-characterized films were grown on a TEC15 + HRT (high resistivity transparent) substrate. The different substrate choice was incidental and based on prior experience we expect the different substrates not to affect grain growth. Substrates were ~76 × 76 × 3 mm. For all sets, a 0.11-µm-thick CdS layer was sputter deposited at 18 mTorr Ar pressure and an RF power of 35 W at a rate of 9 nm/min. The Ar gas pressure during CdTe deposition was selected as the primary variable. Films were deposited at constant RF power of 20 W, substrate temperature of 250°C, and with pressures (deposition rates) of 5 mTorr (21 nm/min), 18 mTorr (16.5 nm/min), and 30 mTorr (13 nm/min). The small variation in final CdTe film thickness is not expected to influence device performance (Plotnikov et al., 2009). One set of the EBSD-characterized films was kept in the as-grown state. The other EBSD set and the set used for cell fabrication received a 30 min 387°C CdCl2 treatment in air. Films received a methanol rinse but no chemical etching. Cells were made by evaporating back contacts through a mask using 3 nm of Cu and 20 nm of Au and heat treating at 150°C for 45 min. Before both EBSD and device characterization, samples were cut into ~5 × 10 mm pieces, and samples from the center and edge of the deposition area were analyzed to investigate any effects of lateral inhomogeneity. Further details on the sputtering procedure are given elsewhere (Shao et al., 1996; Gupta & Compaan, 2004; Compaan, 2006). Initially both the as-grown and CdCl2-treated films were examined with EBSD in the as-deposited condition without any EBSD-specific surface preparation. While some EBSD patterns were observed in this state, surface roughness caused signal shadowing problems and the overall EBSD pattern indexing rate (as described below) was below 50%, which was considered too low for accurate characterization. A number of different EBSD preparation techniques for CdTe films have been described previously (Moutinho et al., 2008): in this work, lowincidence surface-milling (LISM) was used (Zaefferer et al., 2008). An FEI Quanta 3D FEG Dual Beam instrument (EDAXTSL, Ametak (Material Analysis Division)) equipped with
both electron and focused Ga+ ion beams [focused ion beam (FIB)] was used to cut a flat analysis plane at 1.5° to the surface before EBSD. This angle is smaller than the 10–20° typically reported for LISM, but was required due to the thickness of the CdTe thin films and resulted in better resolution through the depth of the film. A mill box of 30 × 1.5 × 2.0 µm was positioned on the film surfaces and milled at 30 kV beam energy and 1 nA of beam current with an approximate milling time of 10 min. In this geometry, the Z-depth of the milling box essentially controls the length of the cut on the sample surface. While the 30 kV ion beam energy cuts produced high-quality EBSD patterns from these (and other) CdTe films, it has been previously reported that high-energy FIB milling can affect EBSD pattern quality in semiconducting materials (Matteson et al., 2002; Michael & Giannuzzi, 2007), and other authors have observed a strong dependence of EBSD pattern quality with FIB milling energy on some CIGS photovoltaic (PV) thin films (Bhatia et al., 2010; Hlaing et al., 2011). While the samples could be analyzed with EBSD using the dual beam instrument, in this work the samples were transferred to an FEI XL-30 FEG SEM (EDAX-TSL, Ametak) to maximize sample throughput. This scanning electron microscope (SEM) was equipped with an EDAX-TSL Hikari EBSD detector (EDAX-TSL, Ametak) running OIM Data Collection version 5.32. The samples were tilted so that the milled analysis plane was at 15° relative to the vertical electron beam. Note that the 1.5° incident angle of the FIB cut results in a slight offset of the measured orientation relative to the film surface normal orientation, which was corrected during postprocessing data analysis. The EBSD patterns were collected at 20 kV and an approximate beam current of 3 nA at a 14 mm working distance. The EBSD detector was set at 100 EBSD patterns per second acquisition at a resolution of 96 × 96 pixels and the camera gain set to a value of ~5 dB to optimize the signal intensity. For mapping, EBSD patterns were automatically collected and analyzed at a rate of 100 points/s from an 18.5 × 18.5 µm area using a hexagonal sampling grid with a 25 nm spacing resulting in ~6.4 × 105 measurements in 107 min. Before map collection, the region of interest was imaged for ~30 min under the same beam conditions to allow the instrument and sample to stabilize and avoid any drifting issues. The mapping location was positioned so that the bottom of the EBSD images was the FIB-etched area nearest the CdTe surface that was completely smooth after milling. Thus, we avoided the transition region between the as-processed surface and the prepared surface. The collected EBSD mapping data was analyzed using EDAX-TSL OIM Analysis version 5.31. The EBSD indexing rate, which is the percentage of measurement points that are analyzed correctly, was determined by the fraction of points with a confidence index (CI) >0.2 after a grain CI standardization procedure (Field, 1997; Nowell & Wright, 2005) relative to the total number of points collected. A systematic noise-reduction routine was also applied. Low-confidence points remaining after these steps were excluded from subsequent analysis.
Characterization of Sputtered CdTe Thin Films
RESULTS AND D ISCUSSION As-Grown CdTe Films Figure 1 shows an EBSD Image Quality (IQ) map (Wright & Nowell, 2006) from the center region of the as-grown film deposited at 30 mTorr, which is generally representative of the microstructure observed in all the as-grown films. Figure 1a shows the entire collection area, where clearly many grains have been observed, while Figure 1b shows a magnified region within the sampled area highlighting that the 25 nm step size used is adequate to resolve the observed microstructure. The total numbers of grains measured, excluding edge grains for more accurate GS measurements, are 7,868 for 5 mTorr pressure, 6,377 for 18 mTorr pressure, and 3,392 for 30 mTorr pressure for the center regions. While these numbers of grains are lower than the >10,000 typically desired for statistically reliable texture analysis,
Figure 1. EBSD image quality map from center region of as-deposited 30 mTorr film showing (a) full region of interest and (b) magnified region of interest. EBSD, electron backscatter diffraction.
929
comparisons of the measured textures from samples prepared by FIB with larger area analyses performed on broad ion-beam-prepared samples (Nowell, 2002) of the same films with >100,000 grains measured show good correlation and are deemed acceptable. In this work, grains are defined by boundaries with an angular misorientation >5° and must contain at least 2 pixels of similar orientation. The 2 pixel minimum gain threshold was selected to maximize spatial resolution performance with the 25 nm sampling step size and, when used in conjunction with only high CI data, was deemed acceptable. Further details are given elsewhere on quantitative texture measurements (Wright et al., 2007) and on the determination of grains (Wright et al., 2000) using EBSD. The indexing rate for these films was ~80–90%. The lower confidence regions were primarily located between well-defined and indexed grains. This may be due to a disordered region arising between adjacent grains during coalescence (Nalwa, 2002). A disordered crystallographic structure would inhibit EBSD pattern formation and result in low-confidence data points. However, based on experience with Cu(In,Ga)Se2 films, it is possible that a lower energy FIB mill would improve pattern quality between the defined grains (Bhatia et al., 2010). Figure 2a shows average GS (diameter) and texture index (TI) of the films as a function of deposition pressure and sampling location (center or edge). TI, as defined by Bunge (1993), allows numerical characterization of the degree of preferred orientation without detailing the specific orientation distribution. This approach was used because the films exhibited different preferred orientations. The TI values vary between 1 for completely random orientation distributions to ∞ for an ideal single crystal. The TIs for these as-sputtered films are distributed approximately between values of 1.15–1.6 indicating generally weak textures. For comparison, we have measured a physical vapor deposited (PVD) CdTe film deposited at 325°C and PVD aluminum film that both showed the canonical strong (111) fiber textures with TI≈9. Lateral heterogeneity in film microstructures was examined by collecting data from the center and edge of the films. The TI of the edge of samples was lower than the TI for the center of samples for all deposition pressures. The difference in GS center to edge depended on pressure; however, average GS at 5 mTorr was ~175 nm with little variation across the film. At 18 and 30 mTorr, the GS at the edge remained ~175 nm, but GS at the center increased with pressure to 300 nm at 30 mTorr. Since film thickness is about 15% less at the edges, the growth rate is lower; this would predict larger grains at the edges. This indicates that the higher ion and atom bombardment in the center is playing a role in nucleation and/or grain growth although the details are not understood (Gupta & Compaan, 2004). Figure 2b presents the data for GS and TI as functions of depth through the films. The shallow 1.5o angle of the LISM preparation allows the EBSD data to be partitioned according to depth with reasonable statistics. Since the bottom of the images corresponds to 75–100 nm below the CdTe
930
Nowell, Scarpulla & Paudel et al.
Figure 2. Average GS and TI versus deposition pressure for the as-deposited films at (a) different lateral locations and (b) as functions of depth in the film. GS, grain size; TI, texture index.
surface, the top represents a depth of about 500 nm. Both GS and TI values decrease with depth below the surface. GSs decrease by 10–15% in 400 nm. This is consistent with a selected oriented growth in which the grains with faster growth gradually dominate. Longer FIB milling times could be used to study the full film thickness but this was not the focus of this study. The observed weak texture in the films of this study is also observed using the maximum values of the (111) and (110) pole figures, as shown in Figures 3a and 3b for three deposition pressures. The pole figures are presented normalized to the statistically expected texture values for random grain orientations. Films deposited at 5 mTorr have a weak (111) fiber texture while the films deposited at 18 and 30 mTorr have weak (110) fiber textures. These were confirmed with the initial measurements made on the as-received samples without LISM. The presence of two primary textures, while low in strength, suggests two different microstructural evolution processes. While the textures of the underlying CdS layers were not measured, all CdS
Figure 3. CdTe (111) and (110) pole figures for deposition pressures of 5, 18, and 30 mTorr for: (a) as-sputtered films and (b) CdCl2-treated films.
layers were deposited at 18 mTorr, which should eliminate CdS orientation effects as a cause of this difference. X-ray diffraction measurements show evidence of in-plane compressive stress producing strain in the growth direction that is 0.5% at 5 mTorr diminishing to 0.1% at 30 mTorr. The (111) texture typically develops due to surface and interface energy minimization effects (Thompson & Carel, 1995, 1996). The (111) texture present in the 5 mTorr film is stronger than the (110) texture in the other films. While deposition temperature and RF power were held constant for all films, both the rate and kinetic energy of atoms impinging on the growth interface at 5 mTorr will be higher due to the lower gas pressure and fewer gas collisions. For CSS CdTe films (Major et al., 2010) a similar decrease in (111) texture was observed as growth rate was reduced at higher pressures. This was interpreted as due to greater competition from the faster-growing (111) grains because of the higher nucleation
Characterization of Sputtered CdTe Thin Films
density at faster growth rates. The larger impinging atom energy produces more strain and may also play a role in development of the (111) texture, but the strain energy will affect recrystallization during CdCl2 treatment discussed later. Consonni et al. (2008) reported evolution of (110) textures driven by strain energy minimization in 1,000-µmthick CdTe films, but it is not clear that strain effects are great enough to affect our 2 µm films. EBSD can also be used to characterize small angle misorientations within grains (Brewer et al., 2009). Here we used the metric of local orientation spread (LOS), where a kernel (third nearest neighbors in this case) is centered on each point (Wright et al., 2011). The misorientation between each point and every other point in the kernel is then calculated (excluding values >5° corresponding to high-angle grain boundaries). The average value for the kernel is then calculated and assigned to that point. The average LOS values for the central region were 0.90° for 5 mTorr, 0.71° for 18 mTorr, and 0.65° for 30 mTorr. These results again suggest that the increased ion bombardment and kinetic energy of the sputtered material associated with lower pressures cause a measureable change in the crystalline microstructure.
931
These misorientation defects could influence device performance by inducing charge carrier traps.
CdCl2-Treated CdTe Films Figure 4 shows EBSD IQ and secondary electron (SE) SEM signal maps from the CdCl2-treated center region samples. These structures are vastly different than the as-deposited microstructures, and show that recrystallization and grain growth has occurred as observed for other low-temperature grown CdTe films (Romeo et al., 2000; Zoppi et al., 2006). The near-vertical lines in the SE images are due to curtaining effects during the FIB preparation. EBSD indexing rates are significantly higher for the treated films, with >98% indexing for the 18 and 30 mTorr deposition pressures and >92% for the 5 mTorr films. The interfaces between adjacent grains are much sharper than the as-deposited films. Voids are visible for these CdCl2-treated films in both types of images, although they are more clearly seen in the SE image. Atomic force microscopy measurements were used to confirm that these were voids rather than, for example, secondary phases of differing conductivity. The observation
Figure 4. EBSD image quality (a–c) and SEM secondary electron (d–f) images for CdCl2-treated films at 5, 18, and 30 mTorr deposition pressures, respectively. Voids are more easily visible in the secondary electron images. (Note: sampling was taken from the center piece of the sputtered film.) EBSD, electron backscatter diffraction; SEM, scanning electron microscope.
932
Nowell, Scarpulla & Paudel et al.
of voids via SEM examination of the treated films away from the FIB-prepared area for EBSD analysis confirmed that these defects were not artifacts of FIB preparation. These voids were also not observed in the as-deposited state before CdCl2 treatment when imaging the FIB-milled surface. The area fractions of voids calculated using SE gray-scale image thresholding were 14.5, 1.6, and 3.9% for the 5, 18, and 30 mTorr deposition pressures, respectively. The larger fraction of voids within the 5 mTorr films explains the lower indexing rate relative to the other samples. The fact that the EBSD indexing rates and void fractions do not total 100% is explained by the fact that indexable EBSD patterns are still obtained from some portions of the surface identified as voids by SE image intensity value. Figure 5 shows pseudo-grain maps for the 18 mTorr center region sample. Typically, grain maps have randomly colored grains to highlight GS and morphology. In this work, the random color maps have been transformed into a 12-bit gray-scale image using ImageJ (Abramoff et al., 2004). To visually delineate adjacent grains that may have similar gray-scale values, black lines have been drawn around each determined grain. It is apparent from this image that CdCl2-treated microstructures also contain high fractions of twin boundaries, indicating that twinning occurs during the recrystallization process. Using EBSD mapping to characterize the GBCD allows categorization of grain boundaries as low-angle grain boundaries (5–15° misorientations), twin boundaries [Σ3 or Σ9 coincident site lattice (CSL) boundaries in this case], or random high-angle grain boundaries (misorientations >15° not meeting CSL criteria). Misorientations 5° misorientation (i.e., all grains comprising the GBCD) have been included in the grain calculations. In Figure 5b, the twin boundaries have been selectively ignored during the grain calculations to produce what is termed here a twin-corrected (TC) grain structure. The average GS distributions for both the standard and TC grain structures are shown in Figure 6 as a function of deposition pressure and sample location (center or edge). Note in all cases the GSs, defined with or without twins, are significantly larger in the centers of the samples than at the edge locations. The fiber textures initially present have disappeared in the CdCl2-treated films, and the textures have generally weakened and become less recognizable as was shown in the pole figures (Fig. 3). Although fewer grains were examined than with the as-deposited films (typically between 1,500 and 3,000 grains), the measured textures were again compared against broad ion-beam-prepared areas with >150,000 grains and good correlation found. The texture weakening is not unexpected with the amount of twinning observed, but it was not possible to systematically retrace the observed
Figure 5. EBSD grain maps (a) including (standard grain size) and (b) excluding twin boundaries as grain boundaries (twin corrected) from the center of the 18 mTorr CdCl2-treated sample. Grain boundaries are drawn as solid black lines. (Note: sampling was taken from the center piece of the sputtered film.) EBSD, electron backscatter diffraction.
orientations back to the original deposited textures using twinning relationships, as has been done previously in CIGS films (Abou-Ras & Pantleon, 2007). However, the textures were generally weak in both the as-deposited and treated samples; because of this, it is difficult to determine if selected orientation grain growth (and twinning) is occurring or if
Characterization of Sputtered CdTe Thin Films
933
Figure 6. Average grain size and TC grain size versus deposition pressure for CdCl2-treated samples. TC, twin-corrected.
new recrystallized nuclei are being generated and then proceeding to grow and twin. For CdTe, new recrystallized nuclei appear to explain regrowth for low-temperature grown films (Moutinho et al., 1998, 2008), but not if the film growth was done at high temperature (≥550°C) (Romeo et al., 2000; Zoppi et al., 2006; Manivannan et al., 2008). The textures at the edge of the treated samples are again weaker than those in the center region, but there is little correlation between the textures unlike the as-deposited films. The LOSs of the treated films were significantly lower than the as-deposited films, with values of 0.49°, 0.41°, and 0.39° for the 5, 18, and 30 mTorr deposition pressures, respectively, for the center regions. This result is consistent with the deposition pressure-dependent strain providing an additional driving force for recrystallization resulting in different retained orientation spreads after recrystallization and grain growth.
Correlation of Device Performance with Measured Microstructures The second set of CdCl2-treated samples received 3 mm diameter back contacts to complete an array of solar cells and their performance was characterized. To correlate performance with the EBSD-characterized microstructure, the four dot cells nearest the center and nearest an edge of each film were chosen to obtain averages. In rare cases for which such a dot cell appeared anomalous compared with the other 140 cells on the film, an adjacent cell was chosen to average. The averaged power-conversion efficiency and fill factor (FF) as functions of either standard or TC average GS are shown in Figures 7a and 7b, respectively. Six independent grain-size data points are created with three deposition pressures and center and edge locations for each. Note that this assumes that the major variable is GS and that, for example, sputter pressure is of secondary importance.
Figure 7. a: Average efficiency and FF versus nontwin-corrected average grain size for CdCl2-treated samples. b: Average efficiency and FF versus twin-corrected grain size for CdCl2-treated samples. The performance parameters shown in (a) and (b) are corresponding data points of three pressures and two locations shown in Figure 6. FF, fill factor.
The correlation of performance with standard GS (uncorrected for twinning) is poor; however, a general trend of increasing device efficiency and FF with GS emerges once the twin boundaries are ignored in the grain boundary angular distributions. The general trend is analogous to an inverse Hall–Petch-type relationship with efficiency and FF increasing as TC GS increases. This is consistent with twin boundaries exhibiting different recombination properties than random high-angle grain boundaries—in fact it suggests that the twin boundaries do not contribute to nonradiative recombination at low carrier generation levels. Cathodoluminescence measurements (corresponding to high injection conditions) on CIGS films have shown similar correlation with EBSD grain boundary type (Abou-Ras et al., 2009). Twin boundaries typically have lower grain boundary energies than random high-angle grain boundaries and are expected to have different diffusional properties that may affect doping and passivation mechanisms. In addition, twin boundaries are likely to have lower densities of recombination site defects. In 8-µm-thick films grown by
934
Nowell, Scarpulla & Paudel et al.
CSS, a plateauing of efficiency versus GS was reported at the point at which series resistance is no longer dominated by grain boundaries but rather by the contacts (Major et al., 2010). Our data for these 2 µm sputtered films show a similar plateau only with TC GS. The observed 100–200 nm voids, as well as electronic defects likely to occur at points having large LOS, could also be limiting the cell performance. While these results show a general correlation of cell performance with microstructure, this work represents a statistical approach characterizing a large number of grain boundaries. In addition, random high-angle grain boundaries are expected to have varying properties leading to varying cell performance. Enhanced photocurrents have been observed by two groups at some grain boundaries (Visoly-Fisher et al., 2003, 2006; Smith et al., 2004; Li et al., 2014). While these enhanced boundaries have not been characterized in terms of misorientation relationships, the present work suggests using a FIB-prepared sample for EBSD crystallographic characterization to develop a structure versus performance matrix beyond the general grain boundary characterization methodology used in this work. It would be desirable to characterize grain boundaries according to those that are advantageous, disadvantageous, and neutral to cell performance. This would of course lead to the ability to engineer specific grain boundaries to improve performance (Fraas, 1978; Randle, 2010), which has been discussed for thin films previously (Field et al., 2005; Park et al., 2005; Park & Field, 2006).
S UMMARY
AND
CONCLUSIONS
CdTe films sputtered at three different deposition pressures were characterized before and after CdCl2 treatment with EBSD in order to correlate microstructural measurements with PV cell performance. A LISM FIB sample-preparation technique allowed for measurements of microstructure as a function of film depth during EBSD. This revealed an increase of GS and texture strength during film growth suggesting an oriented growth process. The CdCl2-treated samples had a recrystallized microstructure with larger GSs and weakened fiber texture relative to the as-deposited films. A high fraction of twin boundaries was observed in all of the CdCl2-treated films. Weak fiber textures were observed in all of the as-deposited films with a (111) texture observed for lower deposition pressures and a (110) texture observed for higher deposition pressures. The (111) texture arises at the higher growth rates that occur at lower pressure as also seen with other low-temperature growth of CdTe. However, the texture development is likely to also be influenced by the substantial strain in the low-pressure-grown films. An important aspect of this work was the establishment of correlations observed between increasing TC GS and improved efficiency and FF, which were absent if the standard GS was considered. This is attributed to the twin boundaries being much less active in carrier recombination and presenting lower barriers to carrier transport than do random high-angle grain boundaries.
ACKNOWLEDGMENTS The authors would like to thank Stuart Wright and John Carpenter at EDAX-TSL for assistance and discussions on low-incidence surface-milling focused ion beam preparation and CdTe microstructures, Elizabeth Lund at the University of Utah for performing the atomic force microscopy measurements, David Field at Washington State University for valuable discussions on thin film textures, and Dohyoung Kwon at the University of Toledo for some of the X-ray diffraction measurements. M.A.S. acknowledges support from the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering under Award DE-SC0001630.
REFERENCES ABOU-RAS, D., KOCH, C.T., KÜSTNER, V., VAN AKEN, P.A., JAHN, U., CONTRERAS, M.A., CABALLERO, R., KAUFMANN, C.A., SCHEER, R., UNOLD, T. & SCHOCK, H.W. (2009). Grain-boundary types in chalcopyrite-type thin films and their correlations with film texture and electrical properties. Thin Solid Films 517, 2545–2549. ABOU-RAS, D. & PANTLEON, K. (2007). The impact of twinning on the local texture of chalcopyrite-type thin films. Phys Stat Sol (RRL) 1, 187–189. ABRAMOFF, M.D., MAGALHAES, P.J. & RAM, S.J. (2004). Image processing with imageJ. Biophoto Inter 11, 36–43. BHATIA, A., DALE, P.J., NOWELL, M.M. & SCARPULLA, M.A. (2010). Pulsed laser processing of electrodeposited CuInSe2 photovoltaic absorber thin films. Proc Mater Res Soc 1268, 103. BREWER, L.N., FIELD, D.P. & MERRIMAN, C.C. (2009). Mapping and assessing plastic deformation using EBSD. In Electron Backscatter Diffraction in Materials Science, 2nd ed. Schwartz, A.J., Kumar, M., Field, D.P. & Adams, B.L. (Eds.), pp. 251–262. Springer. BUNGE, H.J. (1993). Texture Analysis in Materials Science: Mathematical Methods. Gottingen, Germany: Cuvillier Verlag. COMPAAN, A.D. (2006). Photovoltaics: Clean power for the 21st century. Sol Energy Mater Sol Cells 90, 2170–2180. CONSONNI, V., FEUILLET, G. & GERGAUD, P. (2008). Plasticity induced texture development in thick polycrystalline CdTe: Experiments and modeling. J Appl Phys 103, 063529. EISENBARTH, T., UNOLD, T., CABALLERO, R., KAUFMAN, C.A., ABOU-RAS, D. & SCHOCK, H.W. (2009). Origin of defects in CuIn1-xGaxSe2 solar cell with varied Ga content. Thin Solid Films 517, 2244–2247. FIELD, D.P. (1997). Recent advances in the application of orientation imaging. Ultramicroscopy 67, 1–9. FIELD, D.P., SANCHEZ, J.E., PARK, N.J. & BESSER, P.R. (2005). Texture evolution in thin Cu films and lines, materials science forum. In Textures of Materials—ICOTOM 14, Van Houtte, P. & Kestens L. (Eds.), Vols 495–497, pp. 1323–1332. Leuven, Belgium: Trans Tech Publications. FRAAS, L.M. (1978). Basic grain‐boundary effects in polycrystalline heterostructure solar cells. J Appl Phys 49, 871. GUPTA, A. & COMPAAN, A.D. (2004). All-sputtered 14% CdS/CdTe thin-film solar cell with ZnO: Al transparent conducting oxide. Appl Phys Lett 85, 684–686. HLAING, O.O.W.M., JOHNSON, J.L., BHATIA, A., LUND, E.A., NOWELL, M.M. & SCARPULLA, M.A. (2011). Grain size and texture of Cu2ZnSnS4 thin films synthesized by co-sputtering binary sulfides and annealing: Effects of processing conditions and sodium. J Electron Mater 40, 2214–2221.
Characterization of Sputtered CdTe Thin Films LI, C., WU, Y., POPLAWSKY, J., PENNYCOOK, R.J., PAUDEL, N.R., YIN, W., HAIGH, S.J., OXLEY, M.P., LUPINI, A.R., AL-JASSM, M., PENNYCOOK, S.J. & YAN, Y. (2014). Grain-boundary-enhanced carrier collection in CdTe solar cells. Phys Rev Let 112, 156103. MAJOR, J.D., PROSKURYAKOV, Y.Y., DUROSE, K., ZOPPI, G. & FORBES, I. (2010). Control of grain size in sublimation-grown CdTe, and the improvement in performance of devices with systematically increased grain size. Sol Energy Mater Sol Cells 94, 1107–1112. MANIVANNAN, V., ENZENROTH, R.A., BARTH, K.L., KOHLI, S., MCCURDY, P.R. & SAMPATH, W.S. (2008). Microstructural features of CdTe PV thin film devices. Thin Solid Films 516, 1209–1213. MATTESON, T.L., SCHWARZ, S.W., HOUGE, E.C., KEMPSHALL, B.W. & GIANNUZZI, L.A. (2002). Electron backscattering diffraction investigation of focused ion beam surfaces. J Electron Mater 31, 33–39. MICHAEL, J.R. & GIANNUZZI, L.A. (2007). Improved EBSD sample preparation via low energy Ga + FIB ion milling. Microsc Microanal 13, 926–927. MOUTINHO, H.R., AL-JASSIM, M.M., LEVI, D.H., DIPPO, P.C. & KAZMERSKI, L.L. (1998). Papers from the 44th National Symposium of the AVS. San Jose, CA, USA: AVS. 1251pp. MOUTINHO, H.R., DHERE, R.G., ROMERO, M.J., JIANG, C.S., TO, B. & AL-JASSIM, M.M. (2008). Electron backscatter diffraction of CdTe thin films: Effects of CdCl2 treatment. J Vac Sci Tech A 26, 1068–1073. NALWA, H.G. (2002). Handbook of Thin Film Materials: Semiconductor and Superconductor Thin Films. San Diego, CA: Academic Press. NOWELL, M.M. (2002). Ion beam preparation of passivated copper integrated circuit structures for electron backscatter diffraction/ orientation imaging microscopy analysis. J Electron Mater 31, 23–32. NOWELL, M.M. & WRIGHT, S.I. (2005). Orientation effects on indexing of electron backscatter diffraction patterns. Ultramicroscopy 103, 41–58. PARK, N.-J. & FIELD, D.P. (2006). Predicting thickness dependent twin boundary formation in sputtered Cu films. Scripta Mater 54, 999–1003. PARK, N.-J., FIELD, D.P., NOWELL, M.M. & BESSER, P.R. (2005). Effect of film thickness on the evolution of annealing texture in sputtered copper films. J Electron Mater 34, 1500–1508. PLOTNIKOV, V.V., VASKO, A.C., COMPAAN, A.D., LIU, X., WIELAND, K.A., ZELLER, R.M., LI, J. & COLLINS, R.W. (2009). Magnetron sputtering for II-VI solar cells: Thinning the CdTe. Proc Mater Res Soc 1165, M09. RANDLE, V. (2010). Grain boundary engineering: An overview after 25 years. Mater Sci Tech 26, 253–261. ROMEO, A., BATZNER, D.L., ZOGG, H. & TIWARI, A.N. (2000). Recrystallization in CdTe/CdS. Thin Solid Films 361–362, 420–425. SCHWARZER, R.A., FIELD, D.P., ADAMS, B.L., KUMAR, M. & SCHWARTZ, A.J. (2009). Present state of electron backscatter diffraction and prospective developments. In Electron Backscatter Diffraction in Materials Science, 2nd ed. Schwartz, A.J., Kumar, M., Field, D.P. & Adams B.L. (Eds.), pp. 1–20. Springer.
935
SHAO, M., FISCHER, A., GRECU, D., JAYAMAHA, U., BYKOV, E., CONTRERAS-PUENTE, G., BOHN, R.G. & COMPAAN, A.D. (1996). Radio-frequency-magnetron-sputtered CdS/CdTe solar cells on soda-lime glass. Appl Phys Lett 69, 3045–3048. SMITH, S., ZHANG, P., GESSERT, T. & MASCARENHAS, A. (2004). Near-field optical beam-induced currents in CdTe/CdS solar cells: Direct measurement of enhanced photoresponse at grain boundaries. Appl Phys Lett 85, 3854. THOMPSON, C.V. & CAREL, R. (1995). Texture development in polycrystalline thin films. Mater Sci Eng B 32, 211–219. THOMPSON, C.V. & CAREL, R. (1996). Stress and grain growth in thin films. J Mech Phys Solids 44, 657–673. VISOLY-FISHER, I., COHEN, S.R. & CAHEN, D. (2003). Direct evidence for grain boundary depletion in polycrystalline CdTe from nanoscale-resolved measurement. Appl Phys Lett 82, 556–558. VISOLY-FISHER, I., COHEN, S.R., GARTSMAN, K., RUZIN, A. & CAHEN, D. (2006). Understanding the beneficial role of grain boundaries in polycrystalline solar cells from single grain boundary scanning probe microscopy. Adv Funct Mater 16, 649–660. VISOLY-FISHER, I., COHEN, S.R., RUZIN, A. & CAHEN, D. (2004). How polycrystalline devices can outperform single-crystal ones: Thin film CdTe/CdS solar cells. Adv Mater 16, 879–883. WRIGHT, S.I. (2002). Investigation of coincident site lattice boundary criteria in Cu thin films. J Electron Mater 31, 50–54. WRIGHT, S.I., FIELD, D.P. & DINGLEY, D.J. (2000). Advanced software capabilities for automated EBSD. In Electron Backscatter Diffraction in Materials Science, Schwartz, A.J., Kumar, M. & Adams B.L. (Eds.), pp. 141–152. New York, NY: Kluwer Academic/Plenum Publishers. WRIGHT, S.I. & LARSEN, R.J. (2002). Extracting twins from orientation imaging microscopy scan data. J Microsc 205, 245–252. WRIGHT, S.I. & NOWELL, M.M. (2006). EBSD image quality mapping. Microsc Microanal 12, 72–84. WRIGHT, S.I., NOWELL, M.M. & BINGERT, J.F. (2007). A comparison of textures measured using X-ray and electron backscatter diffraction. Metall Mater Trans A 38, 1845–1855. WRIGHT, S.I., NOWELL, M.M. & FIELD, D.P. (2011). A review of strain analysis using EBSD. Microsc Microanal 17, 316–329. YAN, Y., AL-JASSIM, M.M., JONES, K., WEI, S.-H. & ZHANG, S.B. (2000). Observation and first-principles calculations of buried wurtzite phases in zinc-blende CdTe thin films. Appl Phys Lett 77, 1461–1463. YAN, Y., JONES, K.M., JIANG, C.S., WU, X.Z., NOUFI, R. & AL-JASSIM, M.M. (2007). Understanding the defect physics in polycrystalline photovoltaic materials. Physica B 401–402, 25–32. ZAEFFERER, S., WRIGHT, S.I. & RAABE, D. (2008). Three dimensional orientation microscopy in a focused ion beam-scanning electron microscope: A new dimension of microstructure characterization. Metall Mater Trans A 39, 374–389. ZOPPI, G., DUROSE, K., IRVINE, S.J.C. & BARRIOZ, V. (2006). Grain and crystal texture properties of absorber layers in MOCVD-grown CdTe/CdS solar cells. Semicond Sci Technol 21, 763–770.
