For decades, polycrystalline CdTe thin films for solar applications have been restricted to grain sizes of microns or less whereas other semiconductors such as silicon and perovskites have produced devices with grains ranging from less than a micron to more than 1 mm. Because the lifetimes in as-deposited polycrystalline CdTe films are typically limited to less than a few hundred picoseconds, a CdCl2 treatment is generally used to improve the lifetime; but this treatment may limit the achievable hole density by compensation. Here, we establish methods to produce CdTe films with grain sizes ranging from hundreds of nanometers to several hundred microns by close-spaced sublimation at industrial manufacturing growth rates. Two-photon excitation photoluminescence spectroscopy shows a positive correlation of lifetime with grain size. Large-grain, as-deposited CdTe exhibits lifetimes exceeding 10 ns without Cl, S, O, or Cu. This uncompensated material allows dopants such as P to achieve a hole density of 1016 cm−3, which is an order of magnitude higher than standard CdCl2-treated devices, without compromising the lifetime.
Silicon technologies employ a variety of grain sizes ranging from the nanometer to millimeter scales, as well as single-crystal and amorphous Si films. For solar applications, multi-crystalline silicon (mc-Si) with grains on the order of millimeters currently dominates the market as a compromise between the material quality afforded by large grains and the cost savings afforded by not growing single crystals.1 Thin-film technologies—including CdTe, Cu2ZnSn(S,Se)4, and CuInGaSe2—have predominantly focused on the natural grain size resulting from growing films a few microns thick, starting with non-semiconductor substrates such as glass and metal.2 Just as Si performance gradually improves with grain size,3,4 it is possible that thin film technologies can do so as well. CdTe PV technology has recently matched mc-Si technology on both cost and efficiency; an achievement made primarily by maximizing the photocurrent.5 Further absorber layer improvements can increase the open-circuit voltage (Voc) from about 850 mV to >1 V, the efficiency towards 25%, and allow a levelized cost of electricity lower than fossil fuels.6
CdTe grain-size studies have focused on varying parameters within the standard process sequence such as oxygen content and substrate deposition temperature; however, the materials have mostly been studied with current density-voltage (J-V) device measurements, and lifetimes have not been frequently reported.7,8 Evaluating the effect of grain size on film or device electronic properties has been difficult because the size is usually limited to a few microns, and its influence is convoluted with other variables such as structural defects, defect chemistry, voids, and inhomogeneity. Additionally, it has been difficult experimentally to separate bulk recombination from surface and grain-boundary recombination.
Industry, academia, and national laboratories have deposited polycrystalline CdTe by various methods including close-spaced sublimation (CSS), vapor transport deposition, electrodeposition, metal-organic chemical vapor deposition (MOCVD), and sputtering, and have universally observed the lifetime to be very short—on the order of tens to hundreds of picoseconds in the as-deposited material.9,10 Consequently, the entire community relies on a CdCl2 treatment to enhance lifetime and device efficiency. This treatment has been shown to reduce grain-boundary recombination11,12 and to influence transport.12–15 However, the inability to increase the hole density above 1015 cm−3 in standard CdCl2-treated polycrystalline CdTe by multiple institutions indicates that the CdCl2 treatment creates a compensated material with a stubborn defect chemistry. This is arguably one of the main reasons for the stagnation of the Voc between 800 and 900 mV in standard polycrystalline cells for the past two decades. Higher hole densities may be attainable via defect chemistries without CdCl2 treatment, but this will require producing the as-deposited material with carrier lifetimes exceeding 1 ns at deposition rates relevant for manufacturing. Presently, it is not clear if lifetimes in as-deposited films are limited by intrinsic defects introduced during fast deposition, source impurities, or structural defects including the surface, grain boundaries, or stacking faults within the grain interior, which could be subsequently removed by recrystallization during the CdCl2 process.16
In this work, we vary the CdTe grain size from several hundred nm to several hundred μm in test structures, and characterize the minority-carrier lifetimes with time-resolved photoluminescence (TRPL) spectroscopy. TRPL lifetimes have been shown to correlate well with the Voc and in turn the conversion efficiency of CdTe solar cells.17 We photoexcite carriers with sub-bandgap photons which excite carriers at the focus of an excitation beam by nonlinear two-photon optical absorption. Thereby we can avoid recombination effects at the top and bottom surfaces and primarily probe the grain interior and the boundaries between grains.18,19 This is a distinct advantage over most TRPL measurements, which are performed with photon energies above the bandgap and limited to generation near the surface. For instance, 630-nm (1.97-eV) excitation yields a 200-nm absorption depth in CdTe.20
Sub-bandgap photons of 1.11-eV energy were generated by an amplified Yb:KGW laser delivering femtosecond pulses with a repetition rate of 1.1 MHz, powering an optical parametric amplifier (OPA). Both the laser excitation light and sample photoluminescence were guided through a multimode fiber, and TRPL signals were acquired using time-correlated single-photon counting.21 A bandpass filter of center wavelength 819 nm and bandwidth 44 nm was used before the photon counter to isolate the CdTe PL emission. The beam diameter at the focal point is estimated to be 20 μm, and the axial extent is about 60 μm. The focal point was positioned so that the vast majority of excitation occurs in the bulk and scanned across different sample locations.
Two-dimensional numerical solutions of the coupled drift-diffusion equations for charge carriers were performed to simulate the TRPL measurements.22 In the simulations, the bulk of the sample is uniformly illuminated with a short light pulse, and the PL signal, which is proportional to the radiative recombination rate, is calculated as a function of time and position during the decay for various grain sizes, predicting the measured lifetime τ. Carriers were assumed to have hole and electron mobilities of 80 and 320 cm2/(V s), respectively.23,24
Optical microscopy images were taken with a Nomarski microscope (Zeiss) at 100× magnification. The methods for estimating grain size from optical microscope images are subject to significant error when the surface topography or etching features are not well aligned with the grain structure. Therefore, the average grain size for each sample was determined by analyzing electron backscatter diffraction (EBSD) data. Grain-grouping and grain-size calculations in EBSD are based on the crystallographic orientation of grains, which are independent of topography; hence, EBSD yields a more robust quantification of grain-size.25 In EBSD, the diffracted electrons from the beam of a scanning electron microscope (SEM) are collected, forming Kikuchi patterns on the surface of a phosphor screen.26 These patterns are characteristic of the crystalline structure of the material, including the orientation. EBSD measurements were performed with an EDAX EDS-EBSD Pegasus system and Hikari camera. To prevent shading effects, the samples were ion milled in a JEOL cross-section polisher with an Ar beam prior to EBSD measurements.
Cathodoluminescence (CL) measurements were performed in a JEOL JSM-7600 FESEM operating at 5 kV accelerating voltage, beam current of ∼1 nA, and temperature of 6 K. Spectra were collected with a Horiba CLUE system equipped with a Peltier-cooled Sincerity Si CCD.
We varied the grain size from several hundred nanometers to several hundred microns in CSS-grown films by (i) varying the substrate temperature with a constant H2 ambient, and (ii) varying the ambient from reducing (pure H2), partially reducing (H2 + He), inert (pure He), to oxidizing (O2 + He) with different levels of O2. Additional experiments adjusted the total pressure; however, in this case, the grain-size variation was small compared to the changes achieved by varying ambient composition or substrate temperature. The source material was 5N CdTe (Alfa Aesar) and was deposited by CSS with a source temperature of 700 °C.27,28 Films were grown 50–100 μm thick to largely remove the impact of interface recombination. The substrates were Corning 7059 borosilicate glass, either bare or coated with Mo; the latter aids film adhesion. Post-deposition ex-situ P diffusion with Cd overpressure at 500–700 °C was performed on some samples in order to achieve p-type doping; this process does not alter the grain size.29,30 Secondary-ion mass spectrometry measured ∼1017–1018 cm−3 P incorporation, and the capacitance-voltage (CV) measurements on large-grain films indicate ∼1016 cm−3 hole density.30
The grain size varied concomitantly over the range of deposition conditions such that the largest grains were obtained with a H2 background and the smallest grains were obtained with O2 in the growth ambient. Nelson et al. have observed that polycrystalline CdTe films produced in- or subjected to an oxidizing environment resisted grain growth, whereas those which did not have oxygen or oxides at the grain boundaries coalesced to larger grain sizes under certain treatments.31 These results are consistent with the fact that oxygen encourages grain nucleation and may lead to small grain sizes.32
Fig. 1(a) shows that for films grown in H2 ambient, the grain size steadily increases with substrate temperature from 475 °C to 600 °C. Higher growth temperatures supply the necessary energy for grain coalescence into larger grains. These data also indicate that the average grain size is not greatly affected by deposition onto an amorphous substrate (glass) as opposed to a fine-grain film (Mo/glass). Similarly, Fig. 1(b) shows that the grain size increases as the total growth-chamber pressure increases from 1 to 200 Torr for both He and O2 + He (1:16) growth ambient.
(a) Grain size vs. growth temperature for CdTe films grown in various atmospheres on glass or Mo/glass substrates. (b) The effect of total ambient pressure on growth in He or He + O2 on pure glass substrates.
(a) Grain size vs. growth temperature for CdTe films grown in various atmospheres on glass or Mo/glass substrates. (b) The effect of total ambient pressure on growth in He or He + O2 on pure glass substrates.
Fig. 2 shows the optical microscopy images and EBSD grain color maps taken on the polished back surfaces of four representative samples. Different colors in the EBSD data are used for distinguishing separate grains but do not correspond to specific crystal orientations. These samples were grown in He at 600 °C (a) and (e) and in H2 at varying substrate temperature (b), (c), (d), (f), (g), and (h), they display average grain sizes ranging from 1.5 μm to 88 μm.
Optical microscopy images of four samples grown in He at 600 °C (a) and in H2 at 475 °C (b), 550 °C (c), and 600 °C (d) with average grain sizes of 1.5 μm, 8.4 μm, 38.3 μm, and 88.1 μm, respectively. (e)–(h) EBSD grain color maps of the same films, respectively, but not on the same locations. Colors in the EBSD maps are used to denote individual grains, but are not associated with specific crystal orientations.
Optical microscopy images of four samples grown in He at 600 °C (a) and in H2 at 475 °C (b), 550 °C (c), and 600 °C (d) with average grain sizes of 1.5 μm, 8.4 μm, 38.3 μm, and 88.1 μm, respectively. (e)–(h) EBSD grain color maps of the same films, respectively, but not on the same locations. Colors in the EBSD maps are used to denote individual grains, but are not associated with specific crystal orientations.
CL intensity images of CdTe films with varying grain size are displayed in Figs. 3(a)–3(c). The grain boundaries are dark relative to the grain interiors, which indicate that the non-radiative recombination rate is significantly higher at the boundaries than within the grains in these non-CdCl2-treated films. As the grain size increases, some intragrain defects are observed that may limit bulk interior lifetimes, while the relative contribution of grain boundary (GB) recombination decreases significantly.
CL intensity images for three non-CdCl2-treated samples with average grain size 1.5 μm (a), 8.4 μm (b), and 38.3 μm (c). Notice the different scale bars.
CL intensity images for three non-CdCl2-treated samples with average grain size 1.5 μm (a), 8.4 μm (b), and 38.3 μm (c). Notice the different scale bars.
Figs. 4(a) and 4(b) show a representative PL emission spectrum and two-photon excitation (2PE) TRPL decay curve, respectively. The tail of the TRPL curve is fit to an exponential decay to extract the bulk carrier lifetime (τ),18 which is affected by recombination within grains and at grain boundaries, but not at the outer surfaces of the CdTe film. Fig. 4(c) shows that the lifetime increases with grain size, reaching values as high as 30 ns for the largest grains. P doping did not reduce the measured lifetimes as seen in Fig. 4(c) except for the smallest grain sizes, so dopants such as P are compatible with long lifetimes in large-grain polycrystalline CdTe. Grain-boundary and grain-interior recombination can vary with growth conditions and contribute to the scatter observed in the lifetime data in Fig. 4(c). The ex-situ P diffusion process with Cd overpressure preferentially incorporates P along GBs and makes the bulk stoichiometry more Cd-rich.30 These effects can affect GB recombination and decrease intra-grain recombination, respectively,6,33,34 and may explain the difference in lifetime between the as-deposited and P-doped samples. For instance, the films with the smallest grain sizes are most sensitive to GB recombination and less sensitive to intra-grain recombination, which might explain why they appear to exhibit slightly lower lifetimes when P-doped. Additionally, Fig. 2(b) indicates that samples with intermediate average grain sizes can have a mix of small and large grains, which might add to the spread in the lifetime data. Still, amongst these variations, there appears to be an overriding grain size effect. A first order fit shown in Fig. 4(c) was generated by numerical drift-diffusion simulations of the TRPL data as described above, assuming a constant intra-grain lifetime of τi = 20 ns and a constant grain-boundary recombination velocity of S = 5 × 104 cm s−1. According to this model, the measured lifetime is dominated by grain boundary recombination for small grains, and progressively increases with grain size to approach the intra-grain lifetime, τi, due to the increased average time for carriers to reach the grain boundaries. Thus, intrinsic defects and source impurities do not limit the lifetime to values below 20 ns in CdTe films not treated by CdCl2. Furthermore, the data indicate that dopants such as P can produce hole concentrations higher than standard CdCl2-treated devices while being compatible with long lifetimes in large-grain polycrystalline CdTe.
Representative PL spectrum (a) and 2PE TRPL decay curve (b) for large grain (85 μm) films; the black line in (b) shows the exponential decay fit to extract the bulk lifetime. (c) Bulk lifetime (τ) versus average grain diameter. Red symbols are for samples subjected to P doping treatment, black are as-deposited. The average lifetimes were determined by taking multiple measurements across the sample, and error bars show the standard deviation. The latter is affected by sample spatial inhomogeneties including the grain-size distribution shown in Fig. 2. Blue symbols and lines show lifetimes from diffusion, drift, and recombination simulations assuming intra-grain lifetime τi = 20 ns and grain boundary recombination velocity S = 5 × 104 cm s−1.
Representative PL spectrum (a) and 2PE TRPL decay curve (b) for large grain (85 μm) films; the black line in (b) shows the exponential decay fit to extract the bulk lifetime. (c) Bulk lifetime (τ) versus average grain diameter. Red symbols are for samples subjected to P doping treatment, black are as-deposited. The average lifetimes were determined by taking multiple measurements across the sample, and error bars show the standard deviation. The latter is affected by sample spatial inhomogeneties including the grain-size distribution shown in Fig. 2. Blue symbols and lines show lifetimes from diffusion, drift, and recombination simulations assuming intra-grain lifetime τi = 20 ns and grain boundary recombination velocity S = 5 × 104 cm s−1.
In summary, we have showed that carrier lifetimes increase significantly with grain size in polycrystalline CdTe deposited by close-space sublimation in the absence of CdCl2. This indicates that fast deposition, source impurities, and intrinsic defects do not limit the lifetimes in present PV-grade as-deposited CdTe, and suggests that the CdCl2 treatment in thin small-grain films increases lifetimes primarily by reducing grain boundary and surface recombination. Increasing the grain size in the absence of CdCl2 offers a path to achieve carrier concentrations beyond that of standard CdCl2-treated devices while maintaining lifetimes sufficient to overcome historical barriers.
We thank Dr. Darius Kuciauskas for useful discussion and assistance with the measurements. This research was supported by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, under Contract No. DE-AC36-08GO28308. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes.
