Wide spectral coverage (0.7–2.2 eV) lattice‐matched multijunction solar cells based on AlGaInP, AlGaAs and GaInNAsSb materials

We report on the progress in developing lattice‐matched GaAs‐based solar cells with focus on developing AlGaInP, AlGaAs, and GaInNAsSb materials, aiming at achieving a wide spectral coverage, that is, 0.7–2.2 eV. To this end, we first benchmark the performance of an upright four‐junction GaInP/GaAs/GaInNAsSb/GaInNAsSb solar cells grown by molecular beam epitaxy on p‐GaAs substrates with bandgaps of 1.88, 1.42, 1.17, and 0.93 eV, respectively. The four‐junction cell exhibited an efficiency of ~39% at 560‐sun illumination while showing good electrical performance even up to 1000 suns. As a first step to further improve the efficiency toward 50% level, we demonstrate AlGaInP (>2 eV) and GaInNAsSb (<0.8 eV) subcells. We prove that AlGaInP cells with 0.1 Al composition would exhibit current‐matching condition when being incorporated in a five‐junction architecture together with two GaInNAsSb bottom and AlGaAs top junctions. Furthermore, current matching required for a six‐junction tandem architecture is achieved for an Al composition of 0.26. Overall, the results open a practical path toward fabrication of lattice‐matched solar cells with more than four junctions.


| INTRODUCTION
III-V semiconductor materials are commonly used for fabricating the most efficient multijunction solar cells, with record efficiency set to 47.1% for an architecture with six monolithic junctions fabricated with an inverted metamorphic process. 1 Metamorphic approaches usually require growth of thick buffer layers between junctions that have different lattice constants. Ideally, the entire multijunction solar cell structure would be monolithically grown and would have the same lattice constant for all the junctions. In theory, such lattice-matched (LM) III-V solar cell with more than four junctions (4J) can achieve efficiencies of over 50% under concentrated sunlight, that is, concentrated photovoltaics (CPV) operation, and over 35% for space operation. It would also be beneficial to fabricate the structure in an upright architecture, which simplifies the epi-wafer processing, increases the yield, and requires handling of only one wafer. Upright III-V triple-junction (3J) solar cells have been fabricated for more than a decade, [2][3][4] and recently, also 4J solar cells have been reported with metamorphic 2 and LM approaches. [5][6][7][8] The upright LM approach is particularly interesting, because besides the fact that it avoids the use of metamorphic buffer layers, it can make use of the same tunnel junctions that have been long time optimized for the LM 3J solar cells.
Moreover, the LM approach is the most efficient strategy in terms of material utilization given their reduced thickness, a feature that becomes more evident for architectures with four or more subcells.
One of the essential steps in developing the LM multijunction approach has been the progress in epitaxy of GaInNAsSb junction materials with bandgap covering the 0.8-1 eV spectral range. To this end, an upright LM 4J GaInP/GaAs/GaInNAsSb/GaInNAsSb solar cell monolithically grown on GaAs by molecular beam epitaxy (MBE) was reported recently. 8 This first demonstration of 4J design including two dilute nitride subcells, which is schematically shown in Figure 1, exhibited an efficiency of 25% at one-sun illumination, whereas an efficiency of 37% was already demonstrated under 100-sun illumination. It was also estimated that efficiency of 47% seems feasible under higher concentration for an optimized design.
In this paper, we present the progress in developing 4J LM solar cell outlining the need to further extend the spectral coverage toward 0.7 eV at the long wavelengths and beyond 2 eV for the short wavelengths. To this end, we present the development of narrow bandgap (0.7 eV) GaInNAsSb junctions and of AlGaInP high-bandgap (>2 eV) junctions, enabling a more efficient conversion in a wider spectral range. Additional developments toward achieving the 50% efficiency target are also discussed.

| EXPERIMENTS
Epitaxial solar cell structures were grown upright on p-GaAs substrates using Veeco GEN20 and VG Semicon V90 MBE systems.
Detailed description for the MBE processes related to GaInNAsSb fabrication are given elsewhere. [9][10][11]  The current-voltage (IV) characteristics of all cells were measured at 1 sun. In addition, the LM 4J solar cell was measured with concentration up to 1000 suns. These measurements were performed using a commercial steady-state OAI 7-kW TriSOL solar simulator. External quantum efficiency (EQE) measurements were performed using an inhouse built system for measuring single-and multijunction solar cells.
In addition, the internal quantum efficiencies (IQE) for AlGaInP solar cells were estimated using EQE and reflectance data. The short-circuit current densities (J sc ) of the AlGaInP single-junction solar cells and LM 4J subjunctions were estimated from EQE and IQE data. ASTM G173-3 AM1.5D (1000 W/m 2 ) spectrum was used as the reference spectrum. The performance of the LM 4J solar cell was modeled using simple diode equations that have been earlier validated for dilute nitride-based 3J and 4J solar cells. 8

| RESULTS AND DISCUSSION
The EQE results of the best LM 4J is presented in Figure 2. As it can be seen, this structure covers efficiently the spectral range of 350-1310 nm, in which intra-band efficiency can theoretically reach over 52% efficiency under 1000-sun illumination. 8  To this end, we will later discuss the results obtained in developing high-bandgap >1.9-eV AlGaInP and low-bandgap <0.9-eV GaInNAsSb materials.
The IV characteristics of the LM 4J solar cell are shown in The CPV performance of the cell is summarized in Table 1 and shown in Figure 4 using a CPV solar simulator. The results show that the improved LM 4J of this work has significantly lower resistive losses than the LM 4J demonstrated in ref. 8 , which in turn helps to achieve higher FF values and an efficiency of about 39%. In addition, the V oc is over 4.1 V, and the cell maintains high efficiency even close to 1000 suns. These improvements are attributed to better charge carrier transport properties of the top GaInP solar cell compared to the top AlGaAs solar cell used in ref. 8 . The front grid pattern was also modified, and the size of solar cell device was increased from 1 × 1 to 2 × 2 mm. We want to note here that the simulator is target spectrum is AM1.5D, but when the measured subjunction J sc values are compared to the corresponding EQE-calibrated J sc values, we find the following bias nonidealities (misfits). The top cell current is well calibrated and matches the EQE data, but the second, third, and fourth subjunctions are overbiased to 1.093, 1.095, and 1.269 times, respectively, when compared with the photocurrent generation determined from EQE. These numbers show that the 4J solar cell could not be measured using exact subcell biasing, which resulted in top SC-limited measurement, even though when matched to the AM1.5D standard the third cell should limit. For more accurate efficiency determination, the solar cell should be measured under certified CPV conditions. Based on a separate spectrally dynamic analysis (not included here), we estimate that FF might be underestimated by 2-3 percentage points due to the bias mismatch and spectral impurity. We note that when the intensity of CPV simulator is reduced to close to 1 sun while maintaining the spectrum, the FF is only 81%, which is actually lower than what we get when the 1-sun setup is used for the OAI simulator. This points that under high intensities, the FF values are slightly underestimated when the CPV setup is used. We estimate that the reason for the difference in FF at 1 sun originate from difference in the blue end intensity, which is lower for CPV measurement mode. Further investigations will be conducted to confirm the exact FF values at high intensities. We estimate that the current overgeneration in the fourth cell does not play a significant role for assessing the 4J performance, because it was not limiting or even close to be the limiting subjunction. The values in the table are temperature corrected using separately determined reference data. We want to emphasize that even though the spectrum used for the 4J performance evaluation is not pure enough for efficiency verification, the results show that the FF is not severely limited by resistive losses up to 946 suns. This shows that after fine-tuning, these SCs could be used realistically in high CPV systems.
The LM 4J performance could be improved by further reducing the reflectance of the antireflection coating that has an average value of 3%. Moreover, the collection losses of the junctions are estimated to amount for about 8% and are caused mainly by the transmission losses of the bottom junction. The J sc distribution of the LM 4J solar cell is graphically illustrated in Figure 5. We note that the estimated absorption losses in the top tunnel junction are only 1%.
Next, we focus on presenting the main results obtained for the high-bandgap and low-bandgap heterostructures required in architectures with more than 4J. It is generally accepted that growing highperformance AlGaInP solar cells, LEDs, and lasers is challenging independent of the growth method. Typically, the more there is Al in AlGaInP, the more severe the challenges are. Nonoptimal material characteristics have been caused, for instance, by oxygen and boron impurities, [14][15][16] partly originating from the reactive nature of Al and strong bonds of Al compounds. In addition, issues with doping activation and low charge carrier mobilities can increase the resistive losses and cause low charge collection efficiencies. These limitations have been also reflected in our study that reported the EQE response of the AlGaInP solar cells and shown in Figure 6; it can be seen that for Al compositions ≥0.5, the EQE response is significantly reduced. The   Finally, Figure 9 illustrates also the 1-sun V oc values for single-