Ultra-lightweight and ﬂ exible inverted metamorphic four junction solar cells for space applications

. Inthisworkaninvertedmetamorphicfourjunction(IMM4J)solarcellwith30.9%conversionef ﬁ ciency in beginning of life conditions under the AM0 (1367W/m 2 ) spectrum is presented. Additionally, our newest improvedIMM3Jcell,consistingofGa 0.51 In 0.49 P/GaAs/Ga 0.73 In 0.27 Assubcells,with30.6%ef ﬁ ciencyisalsoshown. The IMM4J solar cells consist of Al 0.05 Ga 0.46 In 0.49 P/Al 0.14 Ga 0.86 As/Ga 0.89 In 0.11 As/Ga 0.73 In 0.27 As subcells and are epitaxially grown by metal organic vapor phase epitaxy (MOVPE) on a GaAs substrate. These IMM solar cells achieve power-to-mass ratios of 3W/g or more, which is more than three times higher than standard germanium basedtripleorfourjunctionspacesolarcells.Thelossesincomparisontothesimulatednear-termpotentialef ﬁ ciency of 33.8% for the IMM4J are analyzed in detail. Furthermore, the irradiation behavior for 1 MeV electron ﬂ uences of 1 (cid:1) 10 14 e − /cm 2 and 2.5 (cid:1) 10 14 e − /cm 2 for the IMM4J cells was investigated. A roadmap to further develop this concepttowards anIMM5J witharealisticbeginof life (BOL) ef ﬁ ciency potentialof35.9%underAM0 ispresented.


Introduction
Solar cells made of III-V semiconductor materials are typically used in space applications because, in addition to a high radiation tolerance [1], they also show the highest possible efficiencies. A direct wafer bonded five junction solar cell from Spectrolab demonstrated already a conversion efficiency of 36.0% [2]. A wafer bonded based four junction solar cell from Fraunhofer ISE showed an efficiency of 35.2% under AM0 conditions [3]. Recently, NREL presented an inverted metamorphic triple-junction cell with an middle cell absorber that includes a GaInAs/GaAsP superlattice with an efficiency of 34.2% under the AM0 reference spectrum [4]. Typical commercial available space solar cells from Spectrolab, Solaero and Azur Space [5][6][7] show efficiencies above 31% under the AM0 spectrum at begin of life.
Increasing number of missions with satellites employing electric propulsion systems and ever-increasing data volumes to be processed on satellites leading to a steady increase of the energy demand for e.g. communication satellites in geo-stationary orbits. At the same time, a reduction in mission launch cost is an important design goal of satellite manufacturers and operators today. Thin, flexible solar modules are an interesting option to fulfill these requirements, by maximizing the power-to-mass ratio and minimizing the stowage volumes [8]. Suitable thin, flexible solar cells with a high power to mass ratio are often realized by an inverted metamorphic multijunction solar cell structure and have already been demonstrated [9][10][11][12]. Such cells are grown in an inverted fashion: the highest bandgap subcell first, and the lowest bandgap subcell last. After the growth, the substrate is removed in a non-destructive manner by applying the Epitaxial Lift-Off (ELO) process that was used to manufacture the IMM3J solar cells [11]. The ELO process separates the IMM layer structure from the substrate by selectively etching away a thin intermediate AlAs release layer, without damaging the IMM layer structure nor the substrate [12,13]. This enables multiple reuses of the substrate which results in considerable savings in costs and precious raw materials. After the substrate removal, thin film processing is used to produce ultra-thin and thus also ultra-light flexible solar cells with a high power-to-mass ratio (>3 W/g), while maintaining high efficiencies. Sharp demonstrated IMM3J solar cells with a design optimized for 1 MeV electron irradiation fluences with 1 Â 10 15 e − /cm 2 which showed end of life efficiencies of 26.5% [10].
In general, increasing the number of subcells and spreading the range of bandgap energies is a key towards increasing solar cell conversion efficiencies. Progressing from an IMM triple to a quadruple junction can be done in two ways, either by adding a lower bandgap energy (<1.0 eV) bottom GaInAs cell together with an extended metamorphic buffer due to the higher lattice mismatch or by adding a higher bandgap energy (>1.9 eV) top cell by increasing the aluminum content in AlGaInP. The BOL efficiency potential of the lower bottom bandgap energy concept is slightly higher compared to the higher top bandgap energy IMM4J design. However, the lower bottom bandgap concept has several disadvantages: It is less irradiation hard due to the additional GaInAs subcell [14], it operates at higher temperatures due to the higher absorption in the wavelength range from 1200 nm to 1600 nm and this concept significantly increases the epitaxial layer thickness to overcome the lattice mismatch, which leads to longer and more expensive epitaxy processes. In order to minimize the epitaxy thickness and to avoid the other disadvantages, in this work the maximum lattice mismatch to the GaAs substrate and thus the composition of the bottom subcell was kept at Ga 0.73 In 0.27 As. In other words, the bottom cell bandgap was kept at about 1 eV while the top subcell bandgap was raised by 50 meV. The potential near term efficiency of this IMM4J design is 33.8% according to realistic simulations, which is already 3% abs higher than the realistic potential of the IMM3J. It will be shown, how this IMM concept can be further developed by increasing the top subcell bandgap to 2.10 eV while keeping the bottom subcell bandgap constant at 1.0 eV. This finally leads to an IMM5J solar cell with a realistic efficiency potential of 35.9% (AM0, BOL). We consider this the most cost-effective approach to increase the efficiency of IMM solar cells, since it minimizes the increase in epitaxial layer thickness. The suggested solar cell structures are schematically displayed in Figure 1.

Methods
The inverted grown metamorphic four junction cells (IMM4J) presented in this paper consist of Al 0.05 Ga 0.46 In 0.49 P/ Al 0.14 Ga 0.86 As/Ga 0.89 In 0.11 As/Ga 0.73 In 0.27 As absorber materials whereby the two top cells are lattice matched to GaAs and for each of the metamorphic GaInAs sub-cells a step graded GaInP buffer is used to adjust the lattice constant (see Fig. 2 for a schematic sketch of the layer structure). The absorber materials and their thicknesses were chosen to attempt current matching at begin of life (BOL) conditions under the AM0 (1367 W/m 2 ) reference spectrum. Due to the early development stage, as it will be shown below, the target of current matching under BOL conditions was not fully reached yet. Close to fully transparent tunnel diodes are located between the subcells to allow a series connection of all junctions. In Figure 2 each junction and both metamorphic buffers are shown on the lattice constant versus bandgap plot for III-V semiconductors. An etchstop layer stack is located prior to the first junction to allow subsequent removal of the substrate. The complete structure has a total thickness of 13 mm and was realized by metal organic vapor phase epitaxy (MOVPE) in an AIXTRON 2800G4-TM planetary reactor in an 8 Â 4 00 configuration. The details to the growth of previous generations of IMM3J solar cells and to the full two dimensional opto-electrical model simulations can be found in Schön et al. [11].
A thin metal layer stack is applied on top of the epitaxial layers directly after the growth, which serves as mirror for the bottom cell (J4) and as a backside contact for the entire multijunction solar cell. After the substrate removal, the metal layer stack further serves as a mechanical carrier for the ultra-thin epitaxial layers. The substrate removal was realized in this case by etching away the GaAs substrate. The resulting thin-films were subsequently processed into thin-film IMM solar cells using standard lithography, metallization and wet chemical etching processes. In addition, an anti-reflection coating of ZnS/MgF 2 was applied by thermal evaporation. The fully processed thin-film IMM4J solar cells were weighted with a laboratory weight scale and reached very low areal mass densities of 14 mg/cm 2 . All wafer removal and thin-film processing steps were carried out by tf2 devices in Nijmegen, the Netherlands.
For the analyses and optimization of the solar cells, numerical simulations were carried out. Two-dimensional opto-electrical models of the different 4J and 5J cell designs were elaborated in Sentaurus TCAD [15] based on the model for the IMM3J end of life (EOL) solar cell [11]. The optical model is based on the Transfer-Matrix Method, while the charge carrier transport, tunnel processes and recombination are considered in the electrical part. Material properties and models for GaInP, GaInAs and AlGaAs were taken from previous studies [11] and the optical and electrical parameters for AlGaInP were extracted from measurements at single junction solar cells. For ternary materials with varying compositions the optical material parameters were generated by using a morphing algorithm presented in reference [16]. For an estimate of realistic efficiency potentials that are reachable with the available material quality, simulations were carried out with Shockley-Read-Hall lifetimes extracted from former measurements on inverted grown samples. We use smaller electron lifetimes for the AlGaInP base, namely 1 ns (1.95 eV) down to 0.5 ns (2.10 eV) compared to the Ga 0.51 In 0.49 P base (1.4 ns [11]).
Calibrated characterization under an AM0 (1367 W/m 2 ) reference equivalent spectrum and external quantum efficiency measurements of every subcell of the IMM cells were conducted at the CalLab Fraunhofer ISE with adjustable multi-source sun simulators [17,18]. The IMM4J solar cells were sent to TU Delft for 1 MeV electron irradiation. Half of the cells received a fluence of 1 Â 10 14 e -/cm 2 and the second half received a fluence of 2.5 Â 10 14 e -/cm 2 to simulate typical doses expected for low earth orbit missions. The irradiated cells were treated with a subsequent annealing according to the ECSS standard [19] by being exposed to the AM0 spectrum for 48 h and then being kept at an elevated temperatures of 60°C for 24 h. IV curves, EQEs and spectral electroluminescence were again measured on the irradiated cells after the annealing. Subcell open circuit voltages were determined using the reciprocity relation described in [20] with calibrated external quantum efficiency and spectral electroluminescence measurement data of the IMM4J solar cells. Handling during characterization, shipping, and irradiation was done as carefully as possible using standard tweezers and equipment, but still the many steps resulted in cracks and defects (see Fig. 3) in the thin film solar cells. To obtain more reliable data, the cells must be mounted on temporary substrates immediately after processing and weight measurements.

Begin of life
The newest generation (Gen IV) of IMM3J solar cells reached an efficiency of 30.6% (AM0, BOL) with a fill factor of 83%, a J sc of 16.5 mA/cm 2 and a V oc of 3.06 V. The efficiency gain compared to the former solar cells presented in reference [11] is mainly caused by an improved current matching and thus higher J sc . In Figure 4 the light IV curve of the Gen IV of IMM3J is shown together with the IV curves for the first and second generations of IMM4J solar cells. For comparison purpose the simulation targets of the IMM3J and IMM4J are shown. For comparison, the simulated near-term efficiency potential and the already achieved efficiencies for the IMM3J and IMM4J are shown. Keep in mind that the IMM3J is optimized for EOL conditions, whereas the 4J and 5J structures for BOL. The sketch of the layer structure shows schematically that increasing the number of junctions is not increasing the total epitaxy thickness heavily. Note that for an IMM4J cell with top subcell bandgap up to 2.1 eV, all other absorber material bandgap energies stay the same.
The first generation of the IMM4J reached already 30% efficiency under the AM0 reference spectrum and the second generation of IMM4J exceeds the newest generation of IMM3J with an efficiency of 30.9% a fill factor of 82.8%, a J sc of 12.55 mA/cm 2 and a V oc of 4.07 V. Gen I was current limited by subcell 4 and suffered additionally from a relatively low fill factor of 81.1% (see Fig. 4). The current limiting subcell (J4) of this cell has likely a low parallel resistance which explains the significantly lower than expected fill factor. In Gen II the current matching and thus the J sc (12.55 mA/cm 2 ) was improved. The current is now limited by subcell 2 (see Fig. 5) which led to an increased FF (see Fig. 4). However, the current matching can be improved further in following generations. Just by adjusting the absorber thickness of the subcells, the next generation could achieve already a J sc of 12.9 mA/cm 2 , which is very close to the near-term potential predicted by simulation of 13.0 mA/cm 2 . Further, the EQE plot (Fig. 5) reveals a current loss of approximately 0.3 mA/cm 2 in the Gen II compared to the simulation (gray, dashed line). To determine whether this is due to parasitic absorption in the tunnel diode or recombination in the AlGaAs emitter is the target of current investigations.
Additionally, the V oc , which is correlated to the material quality, is already close to the simulated values (see Fig. 4) and therefore the main improvement potential for the IMM4J is in the fill factor. The fill factor could be improved by avoiding handling caused cracks and shunts (see EL images in Fig. 3) in the thin film devices [11] and by lowering the series resistance (R s ). The difference in slopes around V oc between the simulation and the Gen II device suggests the FF loss to be dominated by the series resistance. Measurements of Gen I IMM3J solar cells at different temperatures (not shown) reveal that this resistor has probably a lower temperature sensitivity as expected from an ohmic type resistor. The root cause of this unexpected high series resistance is still under investigation.
The subcell open circuit voltages were determined from spectral electroluminescence data. The resulting values for the Gen II IMM4J in BOL and after electron irradiation (see Sect. 3.2) are shown together with the results from the latest generation of IMM3J for comparison in Figure 6. The detailed balance limit for the V oc as a function of bandgap is shown together with a 123 mV downshifted curve, which serves as guide for the eye for equal absorber material expectations. In BOL condition J2 shows by far the highest material quality and reaches a similar voltage difference to the detailed balance limit as the GaAs middle cell in the IMM3J. The Al 0.05 Ga 0.46 In 0.49 P absorber shows a slightly lower quality as the Ga 0.51 In 0.49 P cell in the IMM3J but nevertheless the higher bandgap allows a gain in V oc by about 40 mV. The difference of the shifted detailed balance limit line to the V oc values of the metamorphic GaInAs subcells suggest, that the largest improvement of the BOL V oc can be achieved by improving the growth and design of especially J3. W OC (E g /e -V oc ) values of 410 mV have been already demonstrated in literature [21] for such metamorphic GaInAs cells. Thus, a V oc increase of about 30 mV in J3 and of about 20 mV in J4 seems possible in future developed devices. Improvements of the current matching, fill factor and W OC will lower the difference to the simulation target in following generations. Compared to the IMM3J device these improvements would lead not only to an efficiency gain but also to a higher power-to-mass ratio, since the increase in weight and thickness from IMM3J to IMM4J is negligible.
In Figure 1 our roadmap towards an IMM5J with a simulated BOL efficiency of 35.9% assuming current material qualities is shown together with the measured and simulated near-term potential efficiencies of the different development stages of our IMM3J and IMM4J solar cells. The bottom subcell is in all cases kept at a bandgap of 1.00 eV, keeping the maximum lattice mismatch to the GaAs substrate and the absorber material composition (Ga 0.73 In 0.27 As) constant for all the presented IMM designs. This allows the usage of the same metamorphic step graded buffers to reach the required lattice constants for the metamorphic bottom subcells. Therefore, the total epitaxial growth thickness and growth time is only minimally increased with increasing number of junctions. This allows similar areal mass densities of 11.27 mg/cm 2 as already demonstrated by the Gen IV of IMM3J for all the IMM solar cells designs presented in this work. The simulated near-term potential efficiency (34.2%) of the IMM4J solar cell design with a 2.10 eV top subcell is plotted in Figure 7 with the areal mass density of the Gen IV IMM3J as current target for comparison.

After electron irradiation
The IMM4J solar cells from the second generation have been irradiated with 1 MeV electrons with a fluence of 1 Â 10 14 e − /cm 2 and 2.5 Â 10 14 e − /cm 2 to simulate typical missions in low earth orbits. The open circuit voltages after irradiation and subsequent annealing (as described in Sect. 2) are shown in Figure 8 in absolute comparison to the BOL values. The solar cells V oc degrade after 1 Â 10 14 e − /cm 2 on average by 220 mV and 270 mV after 2.5 Â 10 14 e − /cm 2 . In comparison, the IMM3J reported by Sharp (3100 mV, BOL) degrades in V oc after a fluence of 2.5 Â 10 14 e − /cm 2 by about 215 mV [10]. The higher loss in V oc of the IMM4J after nominally the same fluences is caused due to the fact, that the IMM4J starts in BOL conditions already with very high V oc values (4071 mV, BOL). In Figure 4 the light IV curves under AM0 illumination for the two best irradiated cells are shown. The degradation results in an efficiency of 28.4% after 1 Â 10 14 e − /cm 2 and 27.5% after 2.5 Â 10 14 e -/cm 2 fluences. The second generation of IMM4J solar cells was designed for high BOL efficiencies and thus suffer in comparison to references (see Fig. 7) stronger under the same irradiation fluences. To achieve higher efficiency potentials after irradiation the solar cell must be designed in a way that the radiation hardest subcell (i.e. the top cell) is the current limiting junction in non-irradiated conditions. Subcell voltages for the irradiated IMM4J cells have been calculated as described already in Section 2 and are also plotted in Figure 6 in comparison to the corresponding BOL values. The arsenic based subcells all degraded in V oc to approximately the same offset to the radiative limit, which means that already at this low irradiation dose, the minority carrier lifetime is strongly limited by the irradiation induced damage in those subcells. The phosphorus based top cell is the radiation hardest material in the IMM4J and is therefore only slightly degraded in V oc . The Al 0.05 Ga 0.46 In 0.49 P top cell only loses 12 mV in V oc after 2.5 Â 10 14 e -/cm 2 fluence irradiation. More than half of the loss in V oc (145 mV of 280 mV) in the IMM4J is related to the degeneration in the Al 0.14 Ga 0.86 As subcell (J2) after the irradiation. The average subcell W OC of the IMM4J degraded from 435 mV in BOL condition to 510 mV after 2.5 Â 10 14 e -/cm 2 .
The main approach of inverse grown metamorphic multijunction solar cells is to maximize the power-to-mass ratio without lowering the efficiency potential. In Figure 7 the power-to-mass ratios versus efficiencies of commercially available multijunction solar cells are compared to the presented IMM cells before and after irradiation. The solar cells based on direct growth on Germanium and subsequent substrate thinning show power-to-mass ratios slightly below 1 W/g. The IMM3J and IMM4J junction solar cells presented in this work and literature data from Sharp [10] and MicroLink [22] show 3.5 to 4.5 times higher power-tomass ratios while showing comparable efficiencies in BOL and soft irradiation conditions. For the latest generation of our IMM3J solar cell unfortunately no irradiation data is available yet, but from simulations and irradiation data from previous generations [11], it can be assumed that the cell will behave comparable to the IMM space solar cell  The parallel curve (shifted downwards by 123 mV) to the radiative limit [23] serves as guide to the eye and can be interpreted as guide for equal material expectation.
from Sharp [10]. Additionally, the spatial electroluminescence images (see Fig. 3) of the solar cells after irradiation indicate that more defects are introduced during the many intermediate handling steps with standard tweezers of the thin film devices. For more reliable irradiation data, the cells must be mounted on temporary substrates directly after processing and weight measurements.

Conclusion
We further improved the efficiency and power-to-mass ratio of our IMM3J solar cell compared to our previous publication [11] and reach an efficiency of 30.6%. The latest generation of our IMM3J optimized for EOL performance has a power-tomass ratio of 3.7 W/g and shows the lightweight potential of these devices after full processing. The second generation of IMM4J solar cells already outperforms the latest generation of the IMM3J space solar cells with an efficiency of 30.9% under the AM0 spectrum. The simulated near-term target of the IMM4J solar cell with a bandgap spread from 1.96 eV to 1.00 eV is 33.8% and can realistically be achieved in following generations. This result is a steppingstone towards the realization of an IMM5J based on the same platform. The proposed IMM5J features a 2.10 eV top subcell and the same 1.00 eV bottom subcell, with a realistic efficiency potential of 35.9%. The IMM solar cells show a very high power-to-mass ratio over 3 W/g, which can be achieved by all designs presented in this work and outperform in this category the commercial available space solar cells based on thinned Germanium substrates by more than a factor of three. After electron irradiation, to simulate low earth orbit operations, the IMM4J solar cells optimized for BOL conditions showed 28.4% after a 1 MeV electron fluence of 1 Â 10 14 e -/cm 2 and 27.5% after 2.5 Â 10 14 e -/cm 2 in efficiency. The power-tomass ratio of the current IMM4J generation in EOL condition is 2.6 W/g. This high power-to-mass ratio, the compatibility with flexible solar generators with very low stowage volumes and the possibility of growth substrate reuse by applying the ELO process can significantly reduce launch costs for satellites equipped with IMM based solar panels. Especially for earth low orbit missions this approach is a very attractive option in future missions.  results available yet. The 4G32C from Azur [5] and the XTE-SF from Spectrolab [6] are based on thinned Germanium substrates whereas the cells from Sharp [10] and MicroLink [22] are IMM3J based solar cells. The IMM based solar cells have a more than three times higher power-to-mass ratio in comparison to the thinned Germanium based devices. The IMM-a from Solaero [7] is an IMM based solar cell but permanently bonded to a robust substrate and therefore not as lightweight as the other IMM based cells. The ISE/tf2 simulation target assumes an areal mass density of 11.27 mg/cm 2 (as for the Gen IV IMM3J) and the realistic efficiency potential of 35.9% of the IMM4J design with a bandgap spread from 2.10 eV to 1.00 eV.