36th European Photovoltaic Solar Energy Conference and Exhibition

RECENT PROGRESS OF SOLAR CELL DEVELOPMENT FOR CPV APPLICATIONS AT AZUR SPACE

R.H. van Leest1*, D. Fuhrmann1, A. Frey1, M. Meusel1, S.K. Reichmuth2, M. Steiner2 and G. Siefer2
1 AZUR SPACE Solar Power GmbH , Theresienstraße 2, 74072 Heilbronn, Germany
2 Fraunhofer Institute for Solar Energy Systems ISE, Heidenhofstraße 2, 79110 Freiburg, Germany

*Corresponding author: Rosalinda.vanLeest@azurspace.com

ABSTRACT: The next generation concentrator photovoltaic (CPV) product of AZUR SPACE will be a five-junction
upright metamorphic (5J-UMM) cell with a target efficiency of 46% under concentrator standard testing conditions
(CSTC: AM1.5d, 25°C). Additionally the cell design has been optimized for operating conditions (operation
temperatures up to 90°C). Over the course of four iterations of 5J-UMM cells the current performance could be
increased to 42.7% at 768 Suns. This cell is still significantly current mismatched. For a properly current matched cell
with further improvements efficiencies up to 46% are expected. Additionally cells were implemented in AZUR
SPACE’s Enhanced Fresnel Assemblies (EFAs) and C3PV Modules. The EFAs show similar results for both triple (3J)
and quintuple (5J) cells. The module, however, shows 11% enhanced output power when equipped with 5J cells.
Keywords: III-V Semiconductors, Multi-Junction Solar Cell, Concentrator Cells, System Performance

1 INTRODUCTION

Over the past decade, AZUR SPACE has continually

improved the performance of its triple-junction products
[1, 2] for space and concentrator photovoltaic (CPV)
applications, resulting in the market introduction of 3G30
for space applications in 2012 and 3C44 for CPV
applications in 2014. The triple-junction technology has
reached a high level of matureness, leaving little room for
further improvements. Still, there is a strong market
demand for cells with even higher performance, both for
space and terrestrial CPV applications.
In order to reach these higher efficiencies, new cell
concepts with four or more junctions need to be considered
[3,4,5]. Compared to other approaches, such as diluted
Nitride cells, inverted metamorphic (IMM) growth and
semiconductor bonding technology (SBT), upright
metamorphic (UMM) growth offers a cost-competitive
option as it requires only one growth run on a Ge substrate
and no specialized and therefore expensive processing
techniques (such as substrate removal, transfer to carrier,
semiconductor bonding). On the other hand, metamorphic
growth poses challenges to material quality in general and
more specifically for the Al-containing quaternary
materials AlInGaAs and AlInGaP, necessary for four- or
five-junction UMM devices. The recent market Figure 1: Schematic representations of the 4J-UMM cell
introduction of AZUR’s 4G32-Advanced [6], a four- design for space (a) and the 5J-UMM cell design for CPV
junction (4J) AlInGaP/AlInGaAs/InGaAs/Ge UMM cell (b).
for space applications (see Figure 1a), with an average
beginning-of-life (BOL) efficiency of 31.8% and the best For optimum performance under operation conditions
end-of-life (EOL) efficiency available (28.7% @ AM0, the subcell bandgaps (at 25°C) were optimized in such a
1367 W/m², 25°C, 1E15 1MeV e-/cm²), shows that way that none of the subcell bandgaps shifts into (water)
sufficient material quality can be obtained with UMM absorption bands in the terrestrial AM1.5d spectrum at
growth. operational temperatures up to 90°C, which enables a
The 4J-UMM cell for space [6] is the basis for a five- stable current-matching under operation conditions.
junction (5J) UMM cell for CPV applications (see Figure Ideally further development steps will be made to include
1b) [7]. The fifth InGaP junction is implemented between subcell current-matching that accounts for losses in optical
the AlInGaAs and AlInGaP subcells of the 4J structure. elements and for spectral conditions at typical CPV
This offers the benefit of a higher efficiency potential, locations.
while at the same time only minor technical challenges
arise. Additionally, adding a fifth junction increases Voc
and lowers Jsc, which is beneficial for a CPV cell where 2 APPROACH
electrical outputs are hampered by resistance losses, which
are dependent on current density. This new 5J-UMM CPV In a first iteration, a fifth InGaP junction was
cell design has a target efficiency of 46% under high implemented between the AlInGaP J1 and AlInGaAs J2 of
concentration (500-1000x), a target Voc > 5.5V and a target the 4G32(-Advanced) structure. In a second iteration, the
Jsc of 9.4 A/cm² both at 1000x, all at CSTC (concentrator distributed Bragg mirror typically used in space cells was
standard testing conditions: AM1.5d, 25°C). removed and the bandgaps were optimized for CPV.

586
 36th European Photovoltaic Solar Energy Conference and Exhibition

Subsequently, the subcell materials and tunnel diodes were

optimized for better performance. Finally, in a fourth
iteration, the FF was improved and the current-matching
was modified. On cells from each iteration, Quantum
Efficiency (QE) were performed with a grating
monochomator set-up and IV measurements under
concentrated light were performed with the Quadflash set-
up (four filtered flash lamps) at the CalLab of Fraunhofer
ISE.
With cells from the third iteration Enhanced Fresnel
Assemblies (EFAs, see Figure 2a) and CPV modules
(AZURs C3PV Design [8], see Figure 2b) with 5J cells
were built, measured and compared to EFAs and modules
with 3J cells. The EFAs were tested at a concentration of
837x at AZUR SPACE, while the module measurements
took place outdoor at Fraunhofer ISE (Freiburg, Figure 3: External quantum efficiencies (EQE) at 25ºC for
Germany). Between 1 and 10 August 2019 two modules iteration 1 (red), 2 (blue) and 3 (green) plotted together
(1 equipped with 3J cells, 1 equipped with 5J cells) were with the AM1.5d spectrum (grey) for clarity. The bandgap
mounted on the same tracker and IV Data were collected. shift between iteration 1 and 2 and the material quality
No corrections for irradiance, spectrum or temperature improvement in iteration 2 and 3 are clearly visible.
were made. During this time period the typical ambient Measurements performed at ISE CalLab.
temperature was 26.3°C and the typical DNI 801W/m². An
average at CSOC like (Concentrator Standard Operating Table 1: Photocurrent Density (J) for each of the five
Conditions) conditions was calculated over the data points subcells, total photocurrent Density (JJ1+JJ2+JJ3+JJ4+JJ5),
at DNIs between 700 and 1100 W/m². open circuit Voltage (Voc) at 1000 Suns, maximum
efficiency (ηmax) and concentration at which ηmax is
reached for iterations 1-3.

Iteration Iteration Iteration

1 2 3
JJ1 [mA/cm²] 9.4 9.4 10.4
JJ1 [mA/cm²] 8.2 8.8 9.1
JJ1 [mA/cm²] 9.4 8.9 8.6
JJ1 [mA/cm²] 8.8 8.9 9.1
JJ1 [mA/cm²] 10.5 12.9 13.0
JJ1+JJ2+JJ3+JJ4+JJ5 46.3 48.9 50.2
Figure 2: Images of an AZUR SPACE Enhanced Fresnel [mA/cm²]
Assembly (EFA) (a) and C3PV Module (b). Voc @1000 Suns 5.32 5.47 5.52
[V]
ηmax [%] 38.4 41.2 41.0
Concentration 390 486 589
3 RESULTS & DISCUSSION ηmax [Suns]

3.1 Cell Performance In Table 1 the subcell photocurrent densities, the total
The results of the first three iterations are summarized current density (JJ1+JJ2+JJ3+JJ4+JJ5), Voc at 1000 Suns, ηmax
in Figure 3 and Table 1. A detailed description and and the concentration at which ηmax occurs are displayed
analysis of the first three iterations can be found in [7]. In for iterations 1-3. The bandgap optimization between
Figure 3 the external quantum efficiencies (EQE) of iterations 1 and 2 is reflected by the increase in Voc at 1000
iteration 1 to 3 are plotted together with the AM1.5d Suns and the increase of the Ge (J5) subcell photocurrent
spectrum. The first notable feature in this graph is the density. The sum of the subcell photocurrent densities
bandgap shift between iterations 1 and 2. For iteration 1 (JJ1+JJ2+JJ3+JJ4+JJ5) increases over the three iterations,
the J4 bandgap lies within one of the water absorption indicating that more light is converted and thus that the
bands in the AM1.5d spectrum and the J3 bandgap is at an material quality was improved. The increase in material
absorption band edge. Since the QE shifts towards higher quality can also be deduced from the increase in V oc at
wavelengths for higher operation temperatures, this means 1000 Suns between iteration 2 and 3. Most notable for the
that the current-matching between these two subcells will third iteration is the substantial increase in J1 photocurrent
change dramatically with temperature. For the second density, indicating a significant improvement in material
iteration the band gaps were optimized in such a way that quality for this particular subcell. In combination with the
only at temperatures > 90°C the J3 and J4 bandgap start to slight reduction of the J3 photocurrent density, this leads
shift into the absorption bands. This allows for a relatively to a strong current-mismatch in the upper four subcells.
stable current-matching between J3 and J4 for This strong current-mismatch hampers the cell
temperatures up to 90°C. The second notable feature in performance, but with proper current-matching higher
Figure 1 is the increase in J3 QE for the second iteration efficiencies are to be expected. The average current
and the increase in J1 QE for the third iteration. This shows density of J1 to J4 is 9.3 mA/cm², which constitutes an
that the material quality and design of these subcells was increase of 8% compared to the 8.6 mA/cm² of the
improved in each of these iterations. currently limiting cell. Assuming an 8% increase in current

587
 36th European Photovoltaic Solar Energy Conference and Exhibition

density results in an 8% increase in efficiency, and thus outdoor has approximately 11% higher output power
efficiencies > 44% can be expected. compared to the standard 3J module. Since the EFA and
For the fourth iteration the Voc could be improved with Module design are not optimized for 5J cells yet, this is a
another 20 mV at 1000 Suns and an improvement in very promising result. Additionally it also confirms that
current-matching resulted in slight (+1%) increase in the for CPV cells it is important to adjust the design to CSOC
limiting subcell photocurrent density which is now at 8.7 rather than to CSTC alone.
mA/cm². Figure 4 shows the Voc, FF and Efficiency versus
Concentration for a cell from the fourth iteration (black Table 2: Average output power of enhanced Fresnel
squares), the results from a cell from the third iteration Assemblies (EFAs) equipped with triple and quintuple
with similar cell size (5.2mm*5.2mm) and grid design cells measured (uncalibrated) at 837x and average output
(200 µm grid spacing) are plotted for comparison (red power measured outdoor at CSOC like conditions of
squares). The most notable improvement in the fourth modules equipped with triple or quintuple cells and ratio
iteration is the FF which remains > 85% for concentrations between quintuple- and triple-junction output power of the
up to 1000 Suns. This FF improvement is realized by an EFAs and Modules.
improvement in J1 sheet resistance. This improvement has
the additional benefit that larger grid spacing can be used, Output Power EFA Output Power Module
resulting in higher current densities and therefore higher [W] [W]
efficiency. Combined altogether these improvements lead 3J 7.735 300
to efficiencies > 42% with a maximum of 42.7% at 768 5J 7.719 332
Suns, as can be seen in Figure 4. This efficiency is already 3J/5J 0.998 1.107
higher than the performance of typical triple-junction solar
cells with similar grid and cell dimensions at high
concentration. There is still significant improvement
potential (enhancement of total photocurrent density,
improved current-matching), which shows the huge
potential of the 5J CPV cell. Assuming proper current-
matching and further optimization of the total current
density, efficiencies up to 46% are expected. The target
current density of 9.4 mA/cm² presents an increase of 8%
compared to the currently limiting 8.7 mA/cm². Assuming
the target current density can be achieved and the increase
in current density results in the same increase in efficiency,
an efficiency > 46% can be expected. Even if only 9.25
mA/cm² can be achieved, efficiencies greater than 45.5%
seem realistic.

Figure 4: Open-circuit Voltage (Voc), Fill Factor (FF) and

Efficiency plotted versus concentration for a cell from
Iteration 3 (red squares) and a cell from Iteration 4 (black
squares). Measurements performed under the QuadFlash
set-up at Fraunhofer ISE.

Figure 5: Power at maximum power point (P mpp) (a) and

3.2 Module Performance Efficiency (b) plotted versus measurement time for a
Cells from the third iteration (41% efficiency @589 module equipped with 3J cells (green dots), and a module
Suns, see Table 1) were implemented into assemblies and equipped with 5J cells (blue dots). Measurements took
then mounted in a module. Table 2 shows the average place outdoor at Fraunhofer ISE in Freiburg, Germany.
output power of the triple- (3J) and quintuple-junction (5J) The typical ambient temperature was 26.3°C and the
EFAs, along with the average outdoor output power at typical DNI 801 W/m².
CSOC like conditions of modules equipped with the 3J and
5J EFAs. The output power of the EFAs is approximately In Figure 5 the power at the maximum power point
0.2% higher for the standard 3J EFAs, indicating that the Pmoo (Figure 5a) and the Efficiency (Figure 5b) are plotted
5J cells show similar performance compared to the 3J cells versus measurement time for two modules (one equipped
in the lab. On the other hand, the 5J module measured with 3J cells plotted in green, one equipped with 5J cells

588
 36th European Photovoltaic Solar Energy Conference and Exhibition

plotted in blue) mounted at the same tracker at Fraunhofer Conference on Concentrator Photovoltaics, Miyazaki,
ISE in Freiburg, Germany. Each cluster of data points Japan, 2013.
represents a measurement day, with the total measurement [2] V. Khorenko et al., “Roadmap for the next generation
time being 7 days over a time span of 8 days. Both in P mpp of European Space Solar Cells”, 29th EUPVSEC,
and Efficieny it can be clearly seen that most of the time Amsterdam, The Netherlands, 2014.
the 5J datapoints have significantly higher values than the [3] J. Geisz et al., “50% or Bust: Status of Six-Junction
3J data points, only at the beginning and the end of the day Concentrator Solar Cells”, CPV-15, Fes, Morocco,
they start to overlap. This indicates that the better system 2019.
performance of the 5J is achieved for a broad range of [4] F. Dimroth et al., “Four-junction wafer-bonded
conditions and not just at one specific set of operating concentrator solar cells”, IEEE Journal of
conditions, highlighting the huge potential of the 5J cells Photovoltaics 6 (2016).
under operating conditions. Further investigations (effect [5] M. Meusel et al, “III-V Multijunction Solar Cells –
of local spectrum, temperature, etc.) of the 5J module From Current Space and Terrestrial Products to
performance are ongoing. Modern Cell Architectures”, Proceedings of the 5th
World Conference on Photovoltaic Energy
Conversion, Valencia, Spain, 2010.
4 SUMMARY & OUTLOOK [6] T. Torunski et al., “Development and Qualification of
Upright Metamorphic 4J Space Solar Cells 4G32C-
On the basis of the 4G32-Advanced 4J-UMM cell for Advanced”, Space Power Workshop, Torrance, CA,
space applications [6] a 5J-UMM cell for CPV US, 2019.
applications has been developed [7] at AZUR SPACE. The [7] R. van Leest et al., „Recent Progress of Multi-Junction
target efficiency for this cell is 46% at 500-1000 Suns Solar Cell Development for CPV Applications at
concentration under concentrator standard testing AZUR SPACE”, Proceedings of the 15th International
conditions (CSTC: AM1.5d, 25°C). Additionally the cell Conference on Concentrator Photovoltaics, Fes,
will be optimized for system performance (primarily Morocco, 2019.
operation temperatures up to 90°C). The material quality [8] R. Löckenhoff et al., „C3PV: Mass Produced EFA
of the subcells and the FF of the top cell could be Receivers for a Franchised Module Technology”,
significantly improved. This has led to a maximum Proceedings of the 15th International Conference on
efficiency of 42.7% at 768 Suns, which is already better Concentrator Photovoltaics, Fes, Morocco, 2019.
than typical triple-junction cells with similar design. These
cells were still significantly current-mismatched, which
implies that there is a significant improvement potential.
Assuming the target current density of 9.4 mA/cm² can be
achieved and that the increase in current density results in
a similar increase in efficiency, efficiencies > 46% at
CSTC can be reached. First system tests with 5J cells (41%
efficiency), show that assemblies (EFAs) with these cells
show similar performance as 3J assemblies, but that the
average outdoor output power at CSOC like conditions of
the first 5J module is already 11% higher compared to the
output power of a 3J module. This confirms the huge
potential of the 5J cell and the importance of optimizing
for system performance, rather than for standard test
conditions alone.
The most important issue still to be addressed is
current-matching of the upper four subcells and the
increase of the total current generation. This will be the
focus for the next 5J cells, with the goal of achieving 46%
efficiency at CSTC. Further optimization towards
operation conditions (effects of optical elements, different
spectral conditions) will be addressed. Finally further
module tests and first stability tests are in preparation.

5 ACKNOWLEDGEMENTS

This work was supported in part by the German

Federal Ministry for Economic Affairs and Energy
(BMWi) under the project QuintUMM (Contract number
0324152).

6 REFERENCES

[1] D Fuhrmann et al., “About 42%-class CPV cells and

pathways beyond”, Proceedings of the 9th International

589