Presented at the 24th European Photovoltaic Solar Energy Conference and Exhibition, Hamburg, Germany, 21-25 Sep. 2009
BAND-GAP-ENGINEERED ARCHITECTURES FOR HIGH-EFFICIENCY
MULTIJUNCTION CONCENTRATOR SOLAR CELLS
R. R. King, A. Boca, W. Hong, X.-Q. Liu, D. Bhusari, D. Larrabee, K. M. Edmondson,
D. C. Law, C. M. Fetzer, S. Mesropian, and N. H. Karam
Spectrolab, Inc., 12500 Gladstone Ave., Sylmar, CA 91342 USA
ABSTRACT: Beginning with maximum theoretical efficiencies from detailed balance calculations, we evaluate the
real-world energy loss mechanisms in a variety of high-efficiency multijunction cell architectures such as inverted
metamorphic 3- and 4-junction cells, as a step toward closing the gap between theory and experiment. Experimental
results are given on band-gap-engineered lattice-matched and metamorphic 3-junction cells, and on 4-junction
terrestrial concentrator cells. A new world record 41.6%-efficient solar cell is presented, the highest efficiency yet
demonstrated for any type of solar cell.
Keywords: III-V Semiconductors, Concentrator Cells, High-Efficiency, Multijunction Solar Cell,
Gallium Arsenide Based Cells, Lattice-Mismatched, Metamorphic
INTRODUCTION
.
Current Density / Incident Intensity (A/W )
0.3
1
contact
AR
2
HIGH-EFFICIENCY CELL ARCHITECTURES
The main candidates for new high-efficiency
multijunction cell designs draw on the principles of
finding semiconductor band gaps that are more suited for
conversion of the solar spectrum, e.g., metamorphic
materials and Al- or N-containing compounds; dividing
the solar spectrum more finely through the use of
multiple subcells; and driving the cell farther out of
equilibrium through the use of high incident light
intensity (higher concentration ratios).
Figures 1-4 show cell schematics, concepts, and
corresponding modeled subcell and multijunction cell
light I-V characteristics, for four of these candidate
architectures for high-efficiency multijunction (MJ)
terrestrial concentrator cells. The upright metamorphic
p-AlGaInP BSF
p ++-TJ
n++-TJ
W
id
n-GaInP window
n-GaInAs emitter
p-GaInAs base
ell
C
p
To
p-GaInP base
M
e-E
g
el
nn
Tu
dle
id
ll
Ce
p-GaInP BSF
p-GaInAs
step-graded
buffer
el
nn
Tu
p++-TJ
n++-TJ
nucleation
tio
nc
Ju
m
tto
Bo
n+-Ge emitter
p-Ge base
and substrate
n
ll
Ce
MJ cell
0.25
subcell 1
subcell 2
0.2
subcell 3
0.15
0.1
0.05
contact
Lattice-Mismatched
or Metamorphic (MM)
0
0
0.5
1
1.5
2
2.5
3
3.5
Voltage (V)
2.1 3-junction Eg1/ Eg2/ 0.67 eV cell efficiency
Eg1 = Subcell 1 (Top) Bandgap (eV) .
Multijunction
solar
cells
for
concentrator
photovoltaic (CPV) systems can have over 70%
theoretical efficiency. The key question for terrestrial
electricity generation, however, is "how much of this
efficiency potential can be realized in practical solar cells
and converted to kilowatt-hours at the electric meter?"
Experimental III-V multijunction solar cells have
demonstrated efficiencies over 40% since 2006 [1], and
are the highest efficiency PV technology known
presently. But various mechanisms analyzed below still
result in a chasm between the efficiencies promised from
theoretical calculations, and those reached in practice.
Beginning with maximum theoretical efficiencies
from detailed balance calculations, we evaluate the realworld energy loss mechanisms in a variety of highefficiency multijunction cell architectures such as
inverted metamorphic 3- and 4-junction cells, as a step
toward closing the gap between theory and experiment.
Experimental results are given on band-gap-engineered
lattice-matched and metamorphic 3-junction cells, and on
4-junction terrestrial concentrator cells. A new world
record 41.6%-efficient solar cell is presented, the highest
efficiency yet demonstrated for any type of solar cell.
This experimental Spectrolab concentrator cell, with a
III-V lattice-matched 3-junction design and reduced
metal coverage fraction, has been independently verified
at NREL under standard test conditions (25°C, AM1.5D,
ASTM G173-03 spectrum), at 364 suns (36.4 W/cm2).
n+-GaInAs
n-AlInP window
n-GaInP emitter
240 suns (24.0 W/cm2), AM1.5D (ASTM G173-03), 25oC
2 Ideal efficiency -- radiative recombination limit
MM
40.7%
1.9
LM
40.1%
1.8
54%
1.7
52%
1.6
50%
48%
1.5
46%
44%
1.4
42%
40%
1.3
1.0
1.1
1.2
1.3
1.4
38%
1.5
1.6
Eg2 = Subcell 2 Bandgap (eV)
Disordered GaInP top subcell
Ordered GaInP top subcell
Figure 1. Cross-sectional diagram, modeled I-V curves
and iso-efficiency contour plot for upright metamorphic
(MM) 3-junction cell architecture.
(MM) 3-junction design [2-5] in Fig. 1, e.g., the 1.78/
1.30/ 0.67 eV cell modeled in the I-V chart, uses
metamorphic GaInP and GaInAs to lower the band gaps
of the upper two subcells to a more advantageous
theoretical combination, as shown by the arrow in the
contour plot of cell efficiency.
The inverted
metamorphic (IMM) 3-junction [6-8] design in Fig. 2,
e.g., the 1.90/ 1.40/ 1.00 eV cell of the I-V chart, instead
raises the band gap of the third and bottommost subcell,
through the use of MM materials. The arrow in the
contour plot of theoretical efficiency vs. subcell 2 and
subcell 3 band gaps again indicates the shift toward more
optimum band gaps for solar conversion. The general 4junction cell architecture [9] shown in Fig. 3 may be
.
Current Density / Incident Intensity (A/W)
0.16
Ge or GaAs substrate
Growth
Direction
Ge or GaAs substrate
cap
1.9 eV (Al)GaInP subcell 1
1.4 eV GaInAs subcell 2
graded MM buffer layers
0.14
0.12
0.1
MJ cell
subcell 1
0.08
subcell 2
0.06
subcell 3
0.04
0.02
1.0 eV GaInAs subcell 3
0
Growth on Ge or GaAs substrate,
followed by substrate removal from
sunward surface
0
0.5
1
1.5
2
2.5
3
3.5
4
Voltage (V)
3-junction 1.9 eV/ Eg2/ Eg3 cell efficiency
1.6
2
o
500 suns (50 W/cm ), AM1.5D (ASTM G173-03), 25 C
X
Eg2 = Subcell 2 Bandgap (eV) .
Ideal efficiency -- radiative recombination limit
1.5
1.4
53%
52%
51%
1.3
of a lattice-matched (LM), upright metamorphic, inverted
metamorphic, or wafer-bonded design. This type of cell
trades off current for increased voltage, as shown in the
modeled I-V chart for a metamorphic 1.90/ 1.55/ 1.20/
0.67 eV 4-junction cell, giving lower I2R series resistance
losses, and the efficiency can be substantially raised by
adjusting the subcell 2 and 3 band gaps, as shown in the
contour plot. An upright metamorphic 4-junction cell
with this subcell band gap combination is shown in Fig.
4. The 4-junction design can be extended to 5, 6 or more
junctions, as in the 5-junction inverted metamorphic cell
shown in schematic in Fig. 5a, and the 5-junction cell in
Fig. 5b in which the desired band gaps are achieved by
semiconductor-to-semiconductor bonding, or wafer
bonding, subcells grown on different substrates. Finally,
Fig. 6 shows a cell schematic, modeled I-V curves, and
spectrum partition for a more radical 6-junction
terrestrial concentrator cell architecture [9], with band
gaps of 2.00/ 1.78/ 1.50/ 1.22/ 0.98/ 0.67 eV.
50%
metal
gridline
48%
1.2
46%
1.1
1.8-eV (Al)GaInP cell 1
44%
1.55-eV AlGaInAs cell 2
1
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.2-eV GaInAs cell 3
Eg3 = Subcell 3 Bandgap (eV)
transparent buffer
Figure 2. Schematic diagram, modeled I-V curves and
iso-efficiency contour plot for inverted metamorphic
(IMM) 3-junction cell architecture.
0.67-eV Ge cell 4
and substrate
Figure 4. Schematic diagram of a 4-junction upright
metamorphic solar cell.
0.25
Current Density / Incident Intensity (A/W)
contact
AR
AR
cap
(Al)GaInP Cell 1
1.9 eV
wide-Eg tunnel junction
AlGa(In)As Cell 2
1.6 eV
wide-Eg tunnel junction
Ga(In)As Cell 3
1.4 eV
tunnel junction
Ga(In)As buffer
nucleation
Ge Cell 4
and substrate
0.67 eV
MJ cell
subcell 1
0.2
subcell 2
subcell 4
metal
gridline
0.1
2.0-eV AlGaInP cell 1
0.05
1.7-eV AlGaInAs cell 2
1.4-eV GaInAs cell 3
0
0
1
2
3
4
5
Voltage (V)
back contact
1.7
gr o Ge o
wt r G
h s aA
ub s
str
ate
subcell 3
0.15
4-junction 1.9 eV/ Eg2/ Eg3/ 0.67 eV cell efficiency
2
o
500 suns (50 W/cm ), AM1.5D (ASTM G173-03), 25 C
Eg2 = Subcell 2 Bandgap (eV) .
transparent buffer
1.1-eV GaInAs cell 4
metal
gridline
2.0-eV AlGaInP cell 1
semiconductor
bonded
interface
1.7-eV AlGaInAs cell 2
1.4-eV GaInAs cell 3
transparent buffer
1.1-eV GaInPAs cell 4
0.75-eV GaInAs cell 5
0.75-eV GaInAs cell 5
X
1.6
Ideal efficiency -- radiative recombination limit
(a)
(b)
Figure 5.
Schematic diagrams of high-efficiency
terrestrial concentrator solar cells with (a) a 5-junction
inverted metamorphic cell structure, and (b) a 5-junction
wafer-bonded cell structure in which the desired band
gaps are achieved by semiconductor-to-semiconductor
bonding subcells grown on different substrates.
1.5
58%
1.4
56%
1.3
54%
1.2
50%
38%
46%
1.1
34%
42%
1
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
Eg3 = Subcell 3 Bandgap (eV)
Figure 3. Schematic diagram, modeled I-V curves and
iso-efficiency contour plot for 4-junction terrestrial in the
concentrator cell architectures, which may have a latticematched (LM), upright metamorphic, inverted
metamorphic, or wafer-bonded design.
Cell designs with high theoretical efficiency often
call for band gaps that are a challenge to reach in high
quality semiconductor materials on conventional
substrates, e.g., semiconductors with >2.0 eV and those
~1 eV at the Ge lattice constant. The main nextgeneration terrestrial CPV cell candidates – such as
upright and inverted metamorphic cells, cells with 4, 5,
and 6 junctions, and multijunction cells formed by waferbonding subcells with different lattice constants grown
on two or more substrates – each represent a different
contact
2.0 eV
wide-Eg tunnel junction
GaInP Cell 2 (low Eg)
1.78 eV
wide-Eg tunnel junction
AlGa(In)As Cell 3
1.50 eV
wide-Eg tunnel junction
Ga(In)As Cell 4
1.22 eV
tunnel junction
GaInNAs Cell 5
0.98 eV
tunnel junction
Ga(In)As buffer
nucleation
0.12
0.08
MJ cell
0.06
subcell 3
0.04
subcell 4
subcell 5
0.02
4J
50%
4J
3J
45%
4J
subcell 6
0
0
back contact
1
2
3
4
5
6
7
40%
3J
Voltage (V)
3J & 4J MM solar cells
AM1.5D, ASTM G173-03, 1000 W/m2
1.4
Utilization efficiency of photon energy
1-junction cell
3-junction cell
1.2
6-junction cell
600
500
1
400
0.8
300
0.6
200
0.4
100
0.2
0
35%
1
0
0
0.5
1
1.5
2
2.5
Photon Energy (eV)
3
3.5
10
100
Incident Intensity (suns)
500
1000
10000
(1 sun = 0.100 W/cm2)
.
700
Intensity per Unit Photon Energy
(W/m 2 . eV)
subcell 1
subcell 2
Ge Cell 6
and substrate
0.67 eV
4
Figure 6. Schematic diagram, modeled I-V curves and
partition of the solar spectrum for a 6-junction terrestrial
concentrator cell design with low series resistance losses
and very high efficiency.
approach toward the same goal of reaching the most
advantageous subcell band gaps for energy conversion in
a multijunction cell.
Adding more subcells in a
multijunction stack increases efficiency by partitioning
the solar spectrum into smaller slices, but also increases
sensitivity to the non-ideal properties of tunnel junctions,
and current matching among subcells.
Higher
concentration brings the available voltage at the cell
terminals closer to the maximum allowed by the band
gap, but also increases series resistance power loss, and
adds to the cost of the concentrating optics. Further realworld issues include the cost and manufacturing
robustness of some new high-efficiency cell designs, and
the actual energy harvested by solar cells with 3, 4, 5,
and 6 junctions, considering the current match among
subcells with varying solar spectrum over the course of
the day and year [9].
3
3J
55%
0.1
Efficiency (%)
Current Density / Incident Intensity (A/W)
(Al)GaInP Cell 1
Detailed balance limit efficiency
Radiative recombination only
Series res. and shadowing, optimized grid spacing
Normalized to experimental efficiency
60%
0.14
AR
cap
Photon utilization efficiency
.
AR
MULTIJUNCTION CELL MODELING
Differences between solar cell maximum theoretical
performance limits and experimental efficiencies are
investigated below, in order to guide research to identify
and minimize the gap between the two. The maximum
efficiency allowed by the second law of thermodynamics
for a multijunction cell can be determined by finding the
photogenerated current density Jph in each subcell, and
applying the principle of detailed balance [10] to
calculate the diode saturation current density Jo of each
subcell. This upper limit on efficiency is calculated for
3- and 4-junction metamorphic solar cells as a function
of incident intensity in Fig. 7, using the expression for Jo
in Eq. 1, calculated from detailed balance [10-13].
Figure 7.
Calculated efficiency of 3J and 4J
metamorphic solar cells using successively more realistic
models: 1) detailed balance; 2) radiative recombination
only; 3) series resistance and grid shadowing; and 4)
normalization to experimental 3J cell efficiency.
Jo =
2πkTq
(E g + kT )2 e − E g / kT
3 2
hc
(1)
In this physical model, not only are non-ideal
recombination mechanisms like Shockley-Read-Hall
(SRH) recombination assumed to be zero, but even the
fundamental recombination mechanism of radiative
recombination is suppressed, and the only emission
escaping the solar cell is that due to thermal radiation,
treating the solar cell as a non-ideal blackbody at the cell
temperature, that emits only in the wavelength range
where the solar cell also absorbs light, i.e., for photon
energies greater than the solar cell band gap [10]. For
solar cells with direct band gap semiconductors, as in IIIV multijunction cells, the actual emission spectrum is
quite different, consisting mainly of photon emission
from radiative recombination at the band gap energy of
each subcell. One interpretation for this difference is that
if photon recycling, in which photons emitted by the
radiative recombination process are reabsorbed to create
an electron-hole pair elsewhere in the semiconductor,
were to capture the photon from nearly every electronhole recombination event, the radiative recombination
emission spectrum would be suppressed leaving only the
non-ideal blackbody spectrum of the detailed balance
model [14].
In a sequence of increasingly realistic physical
models, the case in which radiative recombination – the
inverse of the essential process of photogeneration of
electron-hole pairs in a solar cell – is considered to be the
only recombination mechanism is treated next in Fig. 7.
At 500 suns for example, the calculated efficiency of the
4J MM cell considered here goes from 55% in the
detailed balance model, to 51% in the radiative
recombination model, in which none of the emission
from radiative recombination is assumed to be recaptured
by photon recycling.
Next, the losses associated with the series resistance
and shadowing of the metal grid on the multijunction cell
are calculated in Fig. 7, assuming a technologically
relevant metal grid width of 6 μm, and with the grid
EXPERIMENTAL RESULTS
A new record efficiency for a photovoltaic cell has
been achieved, with 41.6% efficiency measured under
standard test conditions for concentrator cells (25°C,
AM1.5D, ASTM G173-03 spectrum) at 364 suns (36.4
W/cm2). This record efficiency cell, built at Spectrolab
with a lattice-matched 3-junction structure and reduced
grid coverage fraction, has the highest efficiency yet
demonstrated for any type of solar cell. The 41.6%
efficiency result was independently verified by
measurements at the National Renewable Energy
Laboratory (NREL). The light I-V characteristic of this
Figure 8. Light I-V characteristic for record efficiency
41.6%-efficiency concentrator solar cell, built at
Spectrolab with a lattice-matched 3-junction cell
structure, and independently verified at NREL.
In Fig. 9, the efficiency, Voc, and FF of this record
cell are plotted with incident intensity varying from 0.9
to 940 suns. The efficiency peaks at 41.6% at 364 suns
and drops for higher concentrations due to series
resistance, but is still over 40% at 820 suns, and is 39.8%
at 940 suns. The fill factor shows a broad plateau over
~2 decades in incident intensity, rising rapidly between 1
and ~4 suns due to increasing excess carrier
concentration, and decreasing above ~400 suns due to
series resistance. The Voc increases at an average rate of
~80 mV per decade in incident intensity, per junction, in
the intensity range from 1 to 10 suns. From 10 to 100
suns this rate is 75 mV per decade, while from 100 to
1000 suns it drops to 61 mV per decade. Thus over the
100 to 1000 sun range of interest, the average Voc
increase with concentration is fit fairly well by using
diode ideality factor n = 1 for each of the 3 subcells.
44
Efficiency
0.98
41.6%
Voc x 10
42
0.96
Voc fit, 100 to 1000 suns
40
0.94
FF
38
0.92
36
0.90
34
0.88
32
0.86
30
0.84
28
0.82
26
0.80
24
0.1
1.0
10.0
0.78
1000.0
100.0
2
Incident Intensity (suns) (1 sun = 0.100 W/cm )
Figure 9. Measured light I-V parameters of the 41.6%efficient solar cell as a function of incident intensity.
The cell is still over 40% efficient at 820 suns, and has
39.8% efficiency at 940 suns.
Fill Factor (unitless)
3
cell is plotted in Fig. 8, showing measured parameters of
Voc = 3.192 V, Jsc/intensity = 0.1468 A/W, FF = 0.8874 ,
and efficiency = 41.6 % at 364 suns.
Efficiency (%) and Voc x 10 (V)
spacing reoptimized at each concentration level.
Although the 3J cell has had similar or even higher
calculated efficiency up to this point, the advantage of
lower current density in the 4J cell begins to show in the
series resistance and shadowing calculations in Fig. 7; at
500 suns the 4J cell design is ~48.5%, with the 3J cells
nearly 1% lower in absolute efficiency. The difference
between 4J and 3J cells becomes more pronounced for
even higher concentrations.
Finally, the lowest set of curves in Fig. 7 show the
predicted efficiency when normalization factors for Voc,
Jsc, and FF observed on experimental 3J cells are applied
to the calculated efficiency. These curves normalized to
experimental efficiency should give the closest
correspondence to actual cell efficiency. At 500 suns for
the specific band gap combinations considered in Fig. 7
(1.90/ 1.64/ 1.30/ 0.67 eV for 4J and 1.78/ 1.30/ 0.67 for
3J), the calculated efficiency normalized to experimental
cell efficiency is ~43% for 4J cells, compared to ~41%
for 3J cells. Note that although 3J cells are calculated to
have similar or higher performance compared to 4J cells
for the ideal efficiency cases of detailed balance and
radiative recombination only, the high-voltage, lowcurrent design of 4J cells gives them a significant
efficiency advantage at 500 suns and above for the more
realistic cases where series resistance, grid shadowing,
and normalization to experimental cell values are taken
into account. An important question, requiring both
modeling and field testing to answer, is how solar cells
with 4 (or more) junctions will fare compared to 3J cells
under the real solar spectrum which changes as a
function of sun angle and meteorological conditions [9].
The difference between each of the physical models
treated in Fig. 7 presents an opportunity for bringing the
performance of actual multijunction cells closer to their
theoretical potential. The difference between detailedbalance and radiative-recombination-only models might
be bridged if photon recycling were made very efficient,
such that nearly every photon from a radiative
recombination event were reabsorbed to create an
electron-hole pair elsewhere in the semiconductor. The
cell performance change associated with grid series
resistance and shadowing can be addressed through
improved grid technology and through the use of highvoltage, low-current cell designs such as those with 4 or
more junctions. Finally, the further difference with
respect to the actual Voc, Jsc, and FF of the best
experimental cells can be addressed by device
improvements such as semiconductor layers with higher
bulk lifetime and better interface passivation to reduce
SRH recombination, and more transparent, highly
conductive tunnel junctions.
the target average production efficiencies of these cell
generations: 37.0%, 37.5%, and 38.5% efficiency at 500
suns (50.0 W/cm2), AM1.5D, ASTM G173-03 spectrum,
25°C, respectively, for the C1MJ, C2MJ and C3MJ
products. Additionally, although it is for a smaller
sample size, the distribution of the C3MJ cell,
Spectrolab's next concentrator cell product, is markedly
narrower and sharper than earlier cell generations.
40%
% of Population
The light I-V characteristics of the 41.6%-efficient,
GaInP/GaInAs/Ge 3-junction cell are shown in Fig. 10,
for a range of incident intensities from 2.6 to 940 suns.
The rise in open-circuit voltage with increased light
intensity is evident. Beyond about 400 suns, the Voc
continues to increase, but the fill factor and Vmp are
clearly degraded, due to series resistance.
Figure 11 is a chart of the best research cell
efficiencies confirmed by the National Renewable
Energy Laboratory, showing the most recent 41.6%efficient 3-junction GaInP/GaInAs/Ge solar cell. It is
interesting to note that not only do multijunction
concentrator cells have by far the highest efficiency of
any photovoltaic cell technology, they are currently
enjoying a more rapid rate of increase than other
technologies such as crystalline silicon, thin-film
compound semiconductor cells such as CIGS and CdTe,
and amorphous Si cells.
ηAVG = 38.2%
C1MJ
35%
C2MJ
30%
C3MJ
ηAVG = 37.5%
25%
ηAVG = 36.9%
20%
15%
10%
5%
41.6%
39.5%
38.5%
39.0%
38.0%
37.5%
37.0%
36.5%
Figure 12. Cell efficiency distributions for successive
Spectrolab terrestrial concentrator cell products, C1MJ,
C2MJ, and C3MJ, with 37.0%, 37.5%, and 38.5%
efficiency targets, respectively, at 500 suns.
6.6
0.08
36.0%
Efficiency η at Max. Power
2.6
0.1
35.5%
Inc. Intensity (suns)
2
1 sun = 0.100 W/cm
0.12
35.0%
0.14
34.5%
0%
34.0%
Current Density / Incident Intensity (A/W)
0.16
17.6
59.8
0.06
127.3
364.2
0.04
604.8
0.02
940.9
0
0
0.5
1
1.5
2
2.5
3
3.5
Voltage (V)
Figure 10. Series of light I-V curves measured over a
range of incident intensities.
Figure 11. Chart of best research cell efficiencies for a
variety of solar cell technologies, as confirmed by the
National Renewable Energy Laboratory, showing the
most recent 41.6%-efficient multijunction cell. Chart
courtesy of Larry Kazmerski, NREL.
Many of the efficiency improvements of past
experimental cells are beginning to impact production
terrestrial concentrator cell efficiencies as well. The
efficiency distributions of successive Spectrolab
terrestrial concentrator solar cell product generations, the
C1MJ, C2MJ, and most recent C3MJ cell products, are
shown in Fig. 12. The cell distributions match well with
Looking beyond the C3MJ production cell, the C4MJ
cell has a 40% target average efficiency in production,
and later concentrator cell products, C5MJ, C6MJ, and so
on, are slated to have still higher production efficiencies.
Fig. 13 shows preliminary data for a small number of
experimental full-process 1-cm2 cells with an upright
metamorphic (MM) 3-junction cell structure, a candidate
for the C4MJ cell. The light I-V data at 500 suns (50
W/cm2), AM1.5D, ASTM G173-03, 25°C in Fig. 13 are
corrected to correspond to standard calibration methods
in production light I-V testing. These prototype cells
with 5%-In GaInAs in the middle cell base include a
variety of experiments in cell structure, not a frozen
design, so the distribution width is expected to tighten
markedly when the same cell structure is grown and
processed day in and day out. Even so, the preliminary
average efficiency measured for this experimental batch
is 39.6%, and there are a substantial number of cells with
efficiency above the 40% target.
To reach the ever higher efficiencies of future cell
generations, the advanced cell architectures described in
the first sections of the paper are being investigated. The
performance of 4-junction terrestrial concentrator solar
cells, one family of these advanced designs, is plotted in
Figs. 14 and 15. Figure 14 shows the measured external
quantum efficiency as a function of wavelength for the
individual subcells in a lattice-matched 4-junction cell.
The photogenerated current density Jph in each subcell
can be found by integrating with the standard AM1.5D
(ASTM G173-03) solar spectrum, also plotted in Fig. 14.
The current densities at one sun (0.100 W/cm2) for this
particular 4J cell are 9.0 mA/cm2 for AlGaInP subcell 1,
8.8 for AlGaInAs subcell 2, 9.5 for GaInAs subcell 3,
and 21.7 mA/cm2 for Ge subcell 4. As seen in Fig. 14,
opportunities for improving these concentrator 4J cells
include increasing the quantum efficiency in the AlGaInP
and AlGaInAs subcells 1 and 2 up to the ~90% level of
0.16
Corrected LIV Data for
Prototype 3J Metamorphic (MM) 5%-In Cells
10 runs, 22 wafers, 205 cells
Current Density / Inc. Intensity (A/W) .
.
5% 3J-MM
0.14
0.12
0.10
3J Conc.
Cell
0.08
3J Conc.
Cell
Metamorphic
V oc
Jsc /inten.
V mp
FF
conc.
area
0.06
0.04
2.911
0.1596
2.589
0.875
240
0.267
41.6%
AM1.5D,
low -AOD spectrum
4J Conc.
Cell
LM, 822 suns
3.192 V
0.1467 A/W
2.851 V
0.887
364
suns
0.317 cm2
Eff. 40.7%
0.02
3J Conc.
Cell
Lattice-matched
3.251
0.1467
2.781
0.841
822
0.317
40.1%
AM1.5D,
ASTM G173-03
4J Cell
4.398 V
0.0980 A/W
3.950 V
0.856
500 suns
0.208 cm2
36.9%
AM1.5D,
ASTM G173-03
Independently confirmed meas.
25°C
AM1.5D,
ASTM G173-03
Prelim. meas.
25°C
40.8%
40.5%
40.2%
39.9%
39.6%
39.3%
39.0%
38.7%
38.4%
38.1%
37.8%
37.5%
37.2%
36.9%
36.6%
36.3%
36.0%
35.7%
0.00
0
Average
Std. dev.
Voc (V)
3.090
0.022
FF
0.845
0.009
5
100
1600
90
1400
80
1200
60
50
AlGaInAs subcell 2
1.66 eV
Ge subcell 4
0.72 eV
AM1.5D
ASTM G173-03
40
1000
800
600
30
400
20
200
10
0
300
500
700
900
1100
1300
Wavelength (nm)
1500
1700
0
1900
Figure 14. Measured external quantum efficiency for
each of the 4 subcells in a lattice-matched 4-junction
AlGaInP/ AlGaInAs/ GaInAs/ Ge terrestrial concentrator
cell, along with the irradiance of the standard AM1.5D
(ASTM G173-03) solar spectrum for comparison.
In Fig. 15, the illuminated I-V curve of a prototype 4junction terrestrial concentrator cell is plotted, showing
its high-voltage, low-current characteristics which lead to
lower resistive power loss compared with 3-junction
cells. These early prototype 4-junction concentrator cells
have reached up to 36.9% efficiency at 500 suns (50
W/cm2) at Spectrolab in unconfirmed measurements.
The independently confirmed light I-V curves of the
earlier record 40.7%-efficient metamorphic 3-junction
cell [1], the present record 41.6%-efficient latticematched 3-junction cell at 364 suns, and the same cell
measured to have 40.1% efficiency at 822 suns, are also
plotted for comparison.
Intensity Per Unit Wavelength
(W/(m2μ m))
External Quantum Efficiency (%)
the GaInAs subcell 3, improved current balance among
the top 3 subcells, and use of metamorphic materials in
4J cells to achieve a better band-gap-engineered
multijunction device structure as discussed above.
AlGaInP subcell 1
1.95 eV
GaInAs subcell 3
1.39 eV
All subcells
1.5
2
2.5
3
3.5
4
Figure 15. Light I-V measurements for a 4-junction
terrestrial concentrator cell with 4.398 V open-circuit
voltage and preliminary measured efficiency of 36.9% at
500 suns, compared with independently confirmed
measurements on an earlier record 40.7% MM 3J cell [1],
and the present record 41.6% LM 3J cell at 364 suns,
measured to have 40.1% efficiency at 822 suns.
Eff
39.6%
0.8%
Figure 13. Efficiency distribution and average light I-V
parameters for prototype full-process 1-cm2-area
metamorphic 3-junction cells, as described in the text.
70
1
Voltage (V)
3J MM Cell Efficiency Bin
Isc (A)
7.601
0.135
0.5
SUMMARY
The structures of several of the main candidate
designs for next-generation 3- to 6-junction solar cells
are described, and optimization of the subcell band gap
combinations is discussed. Theoretical efficiency is
calculated for terrestrial 3- and 4-junction concentrator
solar cells as a function of incident intensity, beginning
with the most idealized case of detailed balance
efficiency, and progressing through a series of
successively more realistic models, in order to identify
theoretically allowed areas for further efficiency
improvement. Experimental results are presented for
present and future generations of production concentrator
cells, for experimental 4-junction terrestrial concentrator
cells, and for a new record 41.6%-efficient latticematched 3-junction cell, the highest efficiency yet
demonstrated for any type of solar cell.
ACKNOWLEDGMENTS
The authors would like to thank Carl Osterwald, Keith
Emery, Larry Kazmerski, Martha Symko-Davies, Fannie
Posey-Eddy, Holly Thomas, Russ Jones, Jim Ermer,
Peichen Pien, Dimitri Krut, Hector Cotal, Mark Osowski,
Joe Boisvert, Geoff Kinsey, Kent Barbour, Mark
Takahashi, and the entire multijunction solar cell team at
Spectrolab. This work was supported in part by the U.S.
Dept. of Energy through the NREL High-Performance
Photovoltaics (HiPerf PV) program, the DOE
Technology Pathways Partnership (TPP), and by
Spectrolab.
REFERENCES
[1] R. R. King et al., Appl. Phys. Lett., 90, 183516,
4 May 2007.
[2] R. R. King et al., Proc. 28th IEEE Photovoltaic
Specialists Conf. (2000), Anchorage, Alaska, p. 982.
[3] F. Dimroth, U. Schubert, and A.W. Bett, IEEE
Electron Device Lett., 21, 209 (2000).
4.5
[4] T. Takamoto et al., Proc. 3rd World Conf. on
Photovoltaic Energy Conversion (2003), Osaka, Japan,
p. 581.
[5] F. Dimroth et al., Proc. 34th IEEE Photovoltaic
Specialists Conf. (2009), Philadelphia, Pennsylvania.
[6] M. W. Wanlass et al., Proc. 31st IEEE Photovoltaic
Specialists Conf. (2005), p. 530.
[7] R. R. King et al., Proc. 20th European Photovoltaic
Solar Energy Conf. (2005), Barcelona, Spain, pp. 118123.
[8] J. F. Geisz et al., Appl. Phys. Lett., 93, 123505
(2008).
[9] R. R. King et al., Proc. 21st European Photovoltaic
Solar Energy Conf. (2006), Dresden, Germany, pp.
124-128.
[10] W. Shockley and H. J. Queisser, J. Appl. Phys., 32,
510 (1961).
[11] M. A. Green, Third Generation Photovoltaics –
Advanced Solar Energy Conversion, Springer-Verlag,
Berlin-Heidelberg (2003).
[12] J. Nelson, The Physics of Solar Cells, Imperial
College Press, London (2003).
[13] S. Kurtz et al., Prog. Photovolt.: Res. Appl., 16,
537 (2008).
[14] Personal communication, Richard M. Swanson,
Sarah Kurtz.
[15] R. R. King et al., Proc. 23rd European
Photovoltaic Solar Energy Conf., Valencia, Spain,
Sep. 1-5, 2008.