Inverted metamorphic GaInAs solar cell grown by dynamic hydride vapor phase epitaxy

Metamorphic III–V epitaxy enables devices with tunable bandgaps, which have the potential to satisfy the demands of applications such as photovoltaics,^1,2 thermophotovoltaics,^3,4 solid-state lighting,⁵ photodetectors,⁶ high electron mobility transistors,⁷ and lasing.⁸ Metamorphic materials and devices are costly due to the significant growth thickness/time necessary to make low defect density compositionally graded buffers (CGBs) because of the low throughput of present growth methods. Dynamic hydride vapor phase epitaxy (D-HVPE) is an epitaxial growth technique with the potential to dramatically reduce device costs through the use of low-cost inputs (elemental metals, HCl) and potentially higher throughput via in-line operation.^9,10 Recently, we developed CGBs grown by D-HVPE at rates up to 1 μm/min with low defect density.¹¹ The development of high-performance metamorphic devices on these grades is the next step to validate D-HVPE as a capable metamorphic device growth method, because the performance of these devices is sensitive to the material quality of the CGB.

Recent interest in D-HVPE to reduce the cost of III–V photovoltaics has led to the development of HVPE-grown GaInP/GaAs tandem solar cells with up to ∼27% efficiency.^12,13 The addition of lower bandgap metamorphic GaInAs junctions is a way to increase the efficiency of these tandem devices^14,15 with ∼1 eV being near-optimal for a three junction device. Here, we present an inverted metamorphic Ga_0.71In_0.29As solar cell device as one example of a metamorphic device grown by D-HVPE. We evaluate the defect structure and the photovoltaic performance of the device, finding no insurmountable barriers to achieving equivalent performance to incumbent growth methods. These results show that D-HVPE is a viable method to deposit high-performance metamorphic devices for any application with potentially high throughput and potentially low cost.

All materials were grown in our atmospheric pressure D-HVPE reactor described previously,^11,16 using GaCl, InCl, AsH₃, and PH₃. H₂Se and diethylzinc were used as n- and p-type dopants, respectively. The substrate was Zn-doped (001) GaAs offcut 4° toward the (111)B plane. The growth rates for the GaAs, GaInP, and GaInAs layers were 1.0, 0.8, and 0.6 μm/min, respectively. Figure 1 shows the growth sequence (left) and nominal device structure (right). The device used a rear heterojunction architecture with the layers grown in an inverted sequence relative to their top-to-bottom order in the final processed device. The p–n junction was formed between the n-type GaInAs emitter and the p⁺-GaInP layer, which also served as the back contact. The net donor density of the emitter was measured as 5 × 10¹⁷ cm⁻³ by the capacitance–voltage technique. We note that this was the lowest doping density we could obtain with our present H₂Se source. The substrate started in the heat up chamber (HUC) and was heated to the deposition temperature of 650 °C under AsH₃ while the flows for the GaAs buffer, etch stop, and GaAs contact were equilibrated. Next, the substrate was inserted into growth chamber (GC) 1, GC2, and then GC1, in order, to deposit those three layers. Then the substrate was parked in the HUC under AsH₃ while the flows for the Ga_1−xIn_xP CGB were equilibrated in GC1. The substrate was then transferred to GC1 to grow the static (grown all in one chamber)¹¹ Ga_1−xIn_xP CGB. The CGB consisted of a series of nominally 0.8 μm steps formed by varying the ratio of GaCl and InCl in the chamber with a constant input group V to group III precursor ratio of ∼4. The final capping layer of the grade, which also served as the front passivating window layer for the solar cell, was grown thicker than the individual steps to maximize relaxation in that layer. Toward the end of the capping layer deposition, the GaInAs flows were turned on in GC2, and then the substrate was transferred to that chamber to begin deposition of the GaInAs emitter. Finally, the substrate was retracted to GC1 to deposit the GaInP:Zn layer, and then to the HUC to cool under PH₃, completing the growth. The time required to grow the CGB and device layers, with a total thickness of ∼8 μm, was just under 10 min using these relatively modest, for HVPE,^10,17 growth rates.

We used high resolution x-ray diffraction reciprocal space mapping to measure composition and the strain state of the cap and emitter layers from the symmetric (004) and asymmetric glancing-exit (115) reflections. We used high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) to characterize the defect structure of the device. Energy dispersive x-ray spectroscopy (EDS) was conducted in the STEM to study the compositional profile of the device. The device was produced using inverted processing techniques¹⁸ with a total device area of 0.11 cm² and a 0.10 cm² active area. The device was bonded to a Si handle, and the substrate was removed. A bilayer MgF₂/ZnS anti-reflection coating was evaporated onto the device after etching of the GaAs contact layer. We note that the graded buffer remained within the processed device and filtered the light reaching the active layers of the solar cell. External quantum efficiency (EQE) was measured on a system using chopped white light and a monochromator, and light current–density voltage (J–V) curves were measured under the AM1.5G spectrum to characterize the solar photovoltaic performance of the device. We performed dark J–V and electroluminescence (EL) measurements as a way to further characterize the J–V behavior, calculating an implied junction voltage from the measured external radiative efficiency as a function of injection current.¹⁹ The threading dislocation density (TDD) in the GaInAs emitter was measured by cathodoluminescence (CL) in a scanning electron microscope operating with an accelerating voltage of 15 kV and a beam current of 1.3 nA. The GaInP graded buffer and cap were etched away before the CL measurement using concentrated hydrochloric acid, revealing the GaInAs emitter. The sample was cooled to 80 K, and 4.0 × 10⁻⁴ cm² of area was analyzed over multiple images. Secondary ion mass spectrometry (SIMS) was used to measure the Se and Zn dopant concentrations ([Se] and [Zn], respectively) near the nominal p–n junction position.

First, we analyzed the structural characteristics of the device. Figure 2(a) shows a reciprocal space map of the (004) reflection taken with a (110) x-ray plane of incidence. The graded buffer exhibits a slight negative tilt in ω, implying the absence of atomic ordering in the graded buffer.^11,20 The GaInAs reciprocal lattice point sits at a shorter ω–2θ angle than the GaInP cap/window reciprocal lattice point, because the cap is not completely relaxed. Table I shows the results of the XRD analysis of the GaInP cap/window and GaInAs emitter layers. The GaInP cap has a composition of Ga_0.19In_0.81P and exhibits compressive strains of −0.25% and −0.42% in the respective [110] and [−110] in-plane directions. The GaInAs layer has a composition of Ga_0.71In_0.29As and is well-matched to the in-plane lattice constants of the Ga_0.19In_0.81P cap, as shown in Table I, implying that nearly strain-free growth of this layer occurred without relaxation. Figure 2(b) shows a [110]-oriented HAADF STEM image of a different device grown with the same recipe taken after inversion and processing. The misfit dislocations are confined to the graded buffer region, and no misfits appear at the interfaces between the GaInAs emitter and GaInP in the imaged area. We note that the TEM foil is roughly 0.10–0.15 μm thick. Figure 2(c) shows the compositional profile of the device near the active region obtained from a STEM EDS linescan. Note that the growth direction is from left to right. The interfaces between the GaInAs and GaInP window and p⁺-GaInP appear abrupt. However, there does appear to be some slight grading of In in the GaInAs emitter with the In decreasing (and Ga increasing) with depth in the structure. Figure 2(d) shows a CL image of the GaInAs emitter taken after the GaInP grade and the cap layer were selectively etched away. The TDD determined from an analysis of multiple images is 2.3 × 10⁶ cm⁻². This density is consistent with TDD values that we demonstrated previously in GaInP graded buffers with this lattice constant.¹¹ We observe a low density of dark lines, presumed to be misfit dislocations, in some of the images. Figure 2(e) shows SIMS profiles of [Se] and [Zn] in the device near the p–n junction region. [Se] is ∼3–4 × 10¹⁷ cm⁻³ in the GaInAs emitter, and the profile is relatively flat. This density is roughly equal to the capacitance–voltage derived net donor density, within the expected error of the SIMS measurement. [Zn] in the p-GaInP layer is ∼3 × 10¹⁹ cm⁻³. There is apparent diffusion of Zn roughly 0.3 μm into the GaInAs emitter, which likely alters the placement of the p–n junction from the nominal position.

Next, we measured the photovoltaic characteristics of the device to analyze its quality. Figure 3(a) shows the EQE of the device. The sharp cutoff at shorter wavelengths is due to absorption in the GaInP CGB, which filters the device. The longer wavelength band edge is somewhat sloped, likely due to the slightly graded concentration in the GaInAs emitter. The effective device bandgap, E_G, determined by the detailed-balance equivalent method²¹ is 0.996 eV. The EQE peaks at 84% and decreases from shorter to longer wavelengths corresponding to an increase in the reflectance, because the absorber layer is not optically thick. There is some loss of collection due to reflection of the front surface and grids (∼7%), but this peak EQE value suggests incomplete minority carrier collection. Figure 3(b) shows the light J–V curve for the device acquired under an AM1.5G illumination condition. The open-circuit voltage (V_OC) is 0.589 V, and the bandgap-voltage offset, W_OC = E_G/q–V_OC, is 0.407 V. W_OC is a function of the recombination current in the device and is commonly used to characterize solar cell peformance.²² France et al.²³ demonstrated a ∼1 eV inverted metamorphic GaInAs device grown by organometallic vapor phase epitaxy (OMVPE) that exhibited a W_OC of 0.38 V. This metamorphic device performance represents the state-of-the-art and implies similar material quality to our D-HVPE-grown device. We note that the OMVPE-grown device used a front homojunction design with a 0.1 μm-thick emitter and 2.9 μm base, whereas this device uses a rear heterojunction design. The Ga_1−xIn_xP CGB used in the OMVPE device exhibited a threading dislocation density (TDD) of ∼1.0 × 10⁶ cm⁻²,^23,24 which is lower than the TDD measured in the emitter of our device and likely explains some of the difference in W_OC. The fill factor (FF) of the D-HVPE-grown device is 73.9%, which is lower than the 81.2% demonstrated in the OMVPE device. We analyzed the dark J–V and EL-derived J–V, shown in Fig. 3(c), to better understand the limits on the FF. The close agreement between the dark J–V and EL-derived J–V near the one-sun short-circuit current (J_SC) rules out any series resistance limitation on the FF of the light J–V, because the EL-derived J–V is not sensitive to the series resistance. The slope of the dark J–V near J_SC indicates a diode ideality factor of n = 2, suggesting that the dominant source of dark current is recombination in the depletion region.²⁵ This explains the reduced FF, because the J–V curve of an n = 2 limited diode is more rounded near the maximum power point (the point on the curve where the product JV is maximized) than an n = 1 limited diode. Ignoring series resistance effects, a higher fill factor is expected at higher current density, because the EL dark current approaches an n = 1 slope there, but the key to improving the one-sun FF and V_OC is to reduce the n = 2 dark current contribution. We expect that minor optimization of the CGB design will reduce TDD and, by extension, the dark current. Elimination of the misfit dislocations at the emitter/window interface will also reduce the dark current. Their source is presently unknown, but we suspect that elimination of the observed compositional grading in the emitter could help eliminate the observed misfits. Reducing diffusion of Zn into the emitter could also improve the dark current by reducing compensation of the emitter. This can be accomplished by reducing the doping of the p-GaInP layer.²⁶ Finally, the emitter doping density of 5 × 10¹⁷ cm⁻³ is higher than typical, likely resulting in reduced minority carrier lifetime and increased dark current. We suspect that reduction of the doping density to a more standard doping of ∼1 × 10¹⁷ cm⁻³, combined with these other improvements, will increase the EQE, V_OC, and FF, enabling equal performance to state-of-the-art metamorphic solar cells.

In conclusion, we showed that high growth rate D-HVPE compositional grading enables high-performance metamorphic devices as evidenced by the characteristics of our inverted metamorphic GaInAs solar cell. This result indicates that D-HVPE is a viable technique for the deposition of metamorphic devices with tunable properties, meaning that there is potential to dramatically reduce the cost of metamorphic devices thanks to potentially higher throughput and lower input costs than present technology. In addition to photovoltaics, these results are impactful for all other metamorphic device applications, including solid state lighting, photodetection, transistors, and lasers.

The authors would like to thank David Guiling for materials growth and Evan Wong for device processing. This work was authored by Alliance for Sustainable Energy, LLC, the manager and operator of the National Renewable Energy Laboratory for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. The information, data, or work presented herein was funded by the Advanced Research Projects Agency-Energy (ARPA-E), U.S. Department of Energy, Award No. 15/CJ000/07/05. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. 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.