REGULAR PAPER

Evolution of Multijunction Solar Cell Technology for Concentrating Photovoltaics

Russell K. Jones, James H. Ermer, Christopher M. Fetzer, and Richard R. King
Jpn. J. Appl. Phys. 51 (2012) 10ND01
Reprinted from
#2012 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
Evolution of Multijunction Solar Cell Technology for Concentrating Photovoltaics
Russell K. Jones, James H. Ermer, Christopher M. Fetzer, and Richard R. King
Spectrolab, Inc., 12500 Gladstone Avenue, Sylmar, CA 91342, U.S.A.
Received December 13, 2011; accepted February 10, 2012; published online October 22, 2012
Multijunction solar cells have evolved from their original development for space missions to displace silicon cells in high concentrating photovoltaic
(CPV) systems. Todays three-junction lattice-matched production cells have efficiency of 3939.5% under high concentration, and there appears
to be little opportunity for further efficiency gain with this three-junction technology. Future generations of CPV cells will exploit more than three
junctions, with metamorphic subcells, or both technical approaches to achieve efficiencies >45%. As new designs seek closer current matching
and further spectral splitting, atmospheric variability will necessitate careful modeling to optimize energy output. These new cells will also be
higher cost, which will favor higher CPV system concentration. # 2012 The Japan Society of Applied Physics
1. Historical Perspective
Optical concentration has been explored since the earliest
days of the photovoltaic (PV) industry, as a means to both
boost cell eciency and leverage the high cost of solar cells.
Early concentrating photovoltaic (CPV) systems used the
technology of the day, i.e., almost exclusively silicon,
although the potential of multijunction cells, including the
structure that was ultimately successfully commercialized in
todays technology, was recognized by the late 1970s.
1)
Strong interest in GaAs from the space community for its
radiation hardness motivated the development of GaAs
devices, on lightweight Ge substrates. Bedair et al. demon-
strated the rst multijunction solar cell with epitaxially-
grown subcells monolithically interconnected with an
epitaxially grown tunnel junction, the basic approach used
today, with an AlGaAs/GaAs two-junction cell in 1979.
2)
Several years later, Olsen et al.
3)
demonstrated the dual
junction GaInP/GaAs cell, and this technology was
deployed on space missions in the mid-1990s. An active
Ge bottom subcell was developed, using the Ge growth
substrate, resulting in three-junction (3J) GaInP/GaInAs/Ge
cells which were rst launched on spacecraft in 2000.
These successes sparked immediate interest in the
terrestrial CPV community. Numerous experiments were
tried using the new multijunction cells,
4,5)
and to develop-
ment of cells optimized for terrestrial use, leading to a series
of record terrestrial concentrator cell eciencies,
69)
includ-
ing the rst solar cell of any type to reach over 40%
eciency
8)
in 2006. These achievements led to incorpora-
tion of multijunction cells in place of silicon cells in the
leading CPV systems,
10,11)
and to the development of
numerous new systems. Commercialization of the cell
technology has steadily progressed. Figure 1 shows the
progression of production eciency distributions for
commercial CPV products made by Spectrolab to date, with
steadily increasing performance.
12,13)
Until the introduction of the C4MJ technology in 2011,
all production cells had been lattice-matched. The C4MJ
technology, by contrast, has an upright metamorphic
structure (see Fig. 2).
2. Multijunction Technology Roadmap
The record cell eciency for conventional lattice-
matched 3J cells is 41.6%.
7)
This record has been matched
by 3J upright metamorphic cells,
14)
and has been surpassed
by two dierent approaches:
. A 42.3% eciency 3J cell with metamorphic bottom
cell achieved with bifacial growth;
15)
and
. A 3J lattice-matched cell with an undisclosed lattice-
matched technology has achieved 43.5% eciency.
16)
C4MJ
=39.9%
C3MJ+
=39.2%
C3MJ
=38.7%
C2MJ
=37.7%
C1MJ
=36.9%
0%
5%
10%
15%
20%
25%
30%
35%
40%
3
4
.
0
%
3
4
.
5
%
3
5
.
0
%
3
5
.
5
%
3
6
.
0
%
3
6
.
5
%
3
7
.
0
%
3
7
.
5
%
3
8
.
0
%
3
8
.
5
%
3
9
.
0
%
3
9
.
5
%
4
0
.
0
%
4
0
.
5
%
4
1
.
0
%
4
1
.
5
%
%

o
f

C
e
l
l
s
C1MJ
C2MJ
C3MJ
C3MJ+
C4MJ
Fig. 1. Spectrolab production eciency history.
(a) Lattice-Matched (b) Metamorphic
Wide-Eg
tunnel
junction
GaInP top
cell
Ge bottom
cell
Ga(In)As
middle cell
Tunnel
junction
Buffer
region
Back Surface Field
p-GaInP base
n-Ga(In)As emitter
n
+
-Ge emitter
Back Surface Field
n-GaInP emitter
n-AlInP window
cap
contact
AR
p-Ge base
and substrate
contact
n-Ga(In)As buffer
p
++
-TJ
n
++
-TJ
p-Ga(In)As base
nucleation
n-GaInP window
p
++
-TJ
n
++
TJ
Coat
p-GaInP base
n-GaInAs emitter
n
+
-Ge emitter
Back Surface Field
n-GaInP emitter
n-AlInP window
cap
contact
AR
p-Ge base
and substrate
contact
p
++
-TJ
n
++
-TJ
p-GaInAs base
nucleation
n-GaInP window
p
++
-TJ
n
++
-TJ
Back Surface Field
Coat
p-GaInAs
step-graded buffer
Fig. 2. Lattice matched and metamorphic epitaxial structure.
Japanese Journal of Applied Physics 51 (2012) 10ND01
10ND01-1 # 2012 The Japan Society of Applied Physics
REGULAR PAPER
http://dx.doi.org/10.1143/JJAP.51.10ND01
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
In addition, recent record AM0 eciency results for a 4J
inverted metamorphic (IMM) have surpassed the best
eciencies of any 3J cells,
17)
demonstrating for the rst
time a device with more than three junctions surpassing the
best 3J devices. Taken together, these record devices
demonstrate that promising avenues for improvement exist
with both lattice-matched and metamorphic structures.
Spectrolab has a broad research program pursuing a
variety of development paths toward improved eciency
(see Fig. 3). This program includes the above structures, as
well as semiconductor bonding.
A common strategy for all the proposed improved device
structures is to more closely match the currents between
subcells. In both of the record cells cited above, the bottom
Ge cell has been replaced by a wider bandgap material
which is current matched (or nearly so) to the top two cells
and provides a higher voltage than Ge. Cells with more
junctions seek to further split the solar spectrum and
inherently rely on current matching over a broader range
of sub-bands.
3. Implications of Advanced Designs
The trend toward more junctions, and more closely matched
currents between multiple subcells, has potential perfor-
mance implications as discussed below. In addition, all of
the technology options on the roadmap can be expected to
increase production cost.
3.1 Performance
The current balance for a given cell varies with both
atmospheric variations and the temperature of the cell.
13,18)
Indeed, the eect of cell temperature has been shown to
degrade performance in bifacial growth cells with a closely
matched bottom cell current.
15)
On the other hand, King
et al. have shown by analysis that, at least considering air
mass variation only, a 6J cell outperforms 3J, 4J, and 5J cells
throughout the day except within 30 min of sunrise or sunset,
when the available energy is low.
5,19)
This nding is useful
in that it shows the added junctions deliver higher energy
over a broad range of spectral content from changing air
mass throughout the day.
Figures 4 and 5 provide a simplied look at the variation
in current balance due to major atmospheric variables. The
SMARTS model
20)
was used to model the solar spectrum at
Daggett, California at 3 pm on the autumnal equinox (as a
representative CPV location and spectral condition) based on
TMY3 site data.
21)
The TMY3 data were examined to nd
the minimum and maximum broadband AOD and precipi-
table water, and SMARTS spectra were generated with
the minimum and maximum values to represent extreme
bounds of atmospheric conditions at a typical CPV site. The
broadband aerosol optical density (AOD) was converted to a
spectral AOD value as described by Myers.
22)
Figure 4 illustrates that AOD has a broadband eect that
is most pronounced in the shorter wavelength region,
whereas precipitable water (W) aects the longer wave-
length portions of the spectrum with very little eect on the
short wavelengths. Thus both inuence the current balance
but in dierent ways.
Figure 5 shows that this is indeed the case. The rela-
tive currents are shown for C3MJ+ (1.88 eV/1.4 eV/
0.67 eV) and a 5J lattice-matched cell with GaInNAs
C5MJ
TJ/ITJ
Lattice-
Matched
(LM)
2001
2005 34%
C1MJ
20052007
LM
36.5%
39.3%
C2MJ
2008
LM
37.5%
40.1%
Metamorphic materials
development
1.6 and 1.7 eV subcell development
high- wafer
process
development
Epitaxial Ge, SiGe, and dilute nitride subcell development
Lattice Matched (LM)
Metamorphic (MM)
Semiconductor Bonding
Epitaxial Lift-Off
(ELO) Technology
4-5J
LM
43% proj.
4-5J
LM
45% proj.
6J
LM
45% proj.
Epi optimized
for CPV
improved fab.
process
C3MJ
2009
LM
38.5%
41.0%
improved epi
Meta-
morphic
(MM)
2006
40.7% Inverted
meta-
morphic
(IMM)
3J
44% proj.
Inverted
meta-
morphic
(IMM)
4J
47% proj.
4J SBT
47% proj.
5J SBT
52% proj.
Legend:
Average production efficiency
Champion efficiency
Production technology
Experimental/future technology
4J Upright
MM
43% proj.
5J Upright
MM
45% proj.
3J Bifacial
MM
43% proj.
4J Bifacial
MM
45% proj.
C3MJ+
2010
LM
39.2%
41.6%
improved fab.
process
C4MJ
2011
Upright
MM
40%
41.6%
2001 2003 2005 2007 2009 2011 2013 2015
Fig. 3. Multijunction cell technology roadmap.
R. K. Jones et al. Jpn. J. Appl. Phys. 51 (2012) 10ND01
10ND01-2 # 2012 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
for the fourth subcell (2.1 eV/1.71 eV/1.4 eV/1.12 eV/
0.67 eV). The C3MJ+ cell is top-cell-limited in all
conditions except low AOD, high W. The 5J cell, with the
top two cells closely current-matched under the standard
(ASTM-173G) direct spectrum, fares rather well with these
atmospheric extremes, with subcell 1, 2, or 4 limiting
performance depending on the condition.
It should be noted that this analysis neglects the band edge
shifts due to temperature. The fact that cell temperature is
somewhat correlated with atmospheric conditions (e.g., high
air mass occurs near twilight when cell temperature is lower)
highlights the need for full annual modeling of cell and
system performance, taking into account system optics
transmittance, cell temperature, and spectral variability in
order to arrive at optimal bandgap tailoring.
3.2 Cost
In the space industry, satellites and associated launch costs
run to hundreds of millions of dollars, and for commercial
satellites, the revenue potential is directly proportional to the
power generated by the solar panels. High eciency is
therefore extremely leveraging in space applications. In CPV
systems the cell eciency similarly leverages the balance of
systems cost. If the complete plant costs are known, one can
easily calculate the impact on the overall project economics
for dierent cell price and eciency combinations.
Figure 6 illustrates this calculation using hypothetical
cost and eciency numbers for concentrating silicon, and
low-cost and high-cost multijunction cells. Cell e-
ciencies are as noted, and are assumed to be a function of
concentration based on empirical t to 3J cell data. The
balance of system cost is assumed to increase superlinearly
with increasing concentration (secondary axis).
Assuming the balance of system cost is independent of
concentration, the higher eciency MJ cells will always
0%
50%
100%
150%
200%
250%
ASTM-G173 Min AOD
min W
Min AOD
max W
Max AOD
min W
Max AOD
max W
Top
Middle
Bottom
C3MJ+
S
u
b
c
e
l
l

C
u
r
r
e
n
t

(
a
s

%

o
f

m
i
n
i
m
u
m
)
0%
50%
100%
150%
200%
250%
ASTM-G173 Min AOD
min W
Min AOD
max W
Max AOD
min W
Max AOD
max W
Subcell 1
Subcell 2
Subcell 3
Subcell 4
Subcell 5
5JLM
(dilute
nitride)
S
u
b
c
e
l
l

C
u
r
r
e
n
t

(
a
s

%

o
f

m
i
n
i
m
u
m
)
Fig. 5. (Color) Current balance under varying simulated atmospheric
conditions for C3MJ+ and a ve-junction lattice-matched cell with a dilute
nitride fourth junction.
$-
$100
$200
$300
$400
$500
$600
$1.50
$2.00
$2.50
$3.00
0 200 400 600 800 1000 1200 1400
Concentration
S
y
s
t
e
m

C
o
s
t

(
$
/
W
)
B
a
l
a
n
c
e

o
f

S
y
s
t
e
m

C
o
s
t

(
$
/
m

)
MJ $3$4/cm, 36.8%@500X
MJ $5$8/cm, 40.1%@500X
Si $0.01/cm,
26.7%@500X
Fig. 6. (Color) Impact of cell eciency and cost on system economics
(assuming increasing system cost with concentration). Numbers for
illustrative purposes only.
0
0.2
0.4
0.6
0.8
1
1.2
800 1000 1200 1400 1600 1800
I
r
r
a
d
i
a
n
c
e

(
w
/
m
n
m
)
AOD=0.0082, W=0.3cm
AOD=0.0082, W=3.4cm
AOD=0.3195, W=0.3cm
AOD=0.3195, W=3.4cm
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 500 1000 1500 2000 2500 3000 3500 4000
I
r
r
a
d
i
a
n
c
e

(
w
/
m
n
m
)
AOD=0.0082, W=0.3cm
AOD=0.0082,W=3.4cm
AOD=0.3195, W=0.3cm
AOD=0.3195, W=3.4cm
C3MJ+ spectrum partition
5JLM (dilute nitride) partition
(nm)
(nm)

Fig. 4. (Color) SMARTS simulated spectra for AOD and precipitable

water extremes. The model parameters are based on TMY3 data for Daggett,
CA at 3 pm on the autumnal equinox, except the AOD and precipitable
water inputs are varied by the extremes for the site.
R. K. Jones et al. Jpn. J. Appl. Phys. 51 (2012) 10ND01
10ND01-3 # 2012 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
result in a lower system cost ($/W) at suciently high
concentration. By the same token, however, the roadmap to
higher eciency will most likely drive CPV systems to ever-
increasing concentration.
4. Conclusions
Multiple avenues exist for further growth of multijunction
solar cell eciency. These avenues all necessitate much
more careful and comprehensive modeling than has been the
practice to date to ensure optimized performance in the eld.
To the extent the higher eciency cells increase manufac-
turing cost, the trend to higher concentration systems will
continue.
1) L. M. Fraas and R. C. Knechtli: Proc. 13th IEEE Photovoltaic Specialists
Conf., 1978, p. 886.
2) S. M. Bedair, M. F. Lamorte, and J. R. Hauser: Appl. Phys. Lett. 34 (1979)
38.
3) J. M. Olson, S. R. Kurtz, and A. E. Kibbler: Proc. 20th IEEE Photovoltaic
Specialists Conf., 1988, p. 777.
4) R. A. Sherif, R. R. King, H. L. Cotal, D. C. Law, C. M. Fetzer, K.
Edmondson, G. S. Glenn, G. Kinsey, D. Krut, and N. H. Karam: Proc. 21st
European Photovoltaic Solar Energy Conf., 2006.
5) R. A. Sherif, R. R. King, N. H. Karam, and D. R. Lillington: Proc. 31st
IEEE Photovoltaic Specialists Conf., 2005, p. 17.
6) R. R. King, C. M. Fetzer, K. M. Edmondson, D. C. Law, P. C. Colter, H. L.
Cotal, R. A. Sherif, H. Yoon, T. Isshiki, D. D. Krut, G. S. Kinsey, J. H.
Ermer, Sarah Kurtz, T. Moriarty, J. Kiehl, K. Emery, W. K. Metzger, R. K.
Ahrenkiel, and N. H. Karam: Proc. 19th European Photovoltaic Solar
Energy Conf., 2004, p. 3587.
7) R. R. King, R. A. Sherif, D. C. Law, J. T. Yen, M. Haddad, C. M. Fetzer,
K. M. Edmondson, G. S. Kinsey, H. Yoon, M. Joshi, S. Mesropian, H. L.
Cotal, D. D. Krut, J. H. Ermer, and N. H. Karam: Proc. 21st European
Photovoltaic Solar Energy Conf., 2006, p. 124.
8) R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H.
Yoon, R. A. Sherif, and N. H. Karam: Appl. Phys. Lett. 90 (2007) 183516.
9) 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:
Proc. 24th European Photovoltaic Solar Energy Conf., 2009, p. 55.
10) G. S. Kinsey, R. A. Sherif, H. L. Cotal, P. Pien, R. R. King, R. J. Brandt,
W. G. Wise, E. L. Labios, K. F. Wan, M. Haddad, J. M. Lacey, C. M.
Fetzer, P. Verlinden, J. Lasich, and N. H. Karam: IEEE 4th World Conf.
Photovoltaic Energy Conversion, 2006, p. 625.
11) R. Gordon, K. W. Stone, D. Dutra, A. Gray, R. Hurt, R. Boehm, M. J. Hale,
and F. Posey-Eddy: Proc. 4th Int. Conf. Solar Concentrators, 2007, p. 221.
12) R. K. Jones, P. Hebert, P. Pien, R. R. King, D. Bhusari, R. Brandt, O. Al
Taher, C. Fetzer, J. Ermer, A. Boca, D. Larrabee, X. Q. Liu, and N. Karam:
Proc. 35th IEEE Photovoltaic Specialists Conf., 2010, p. 189.
13) J. H. Ermer, R. K. Jones, P. Hebert, P. Pien, R. R. King, D. Bhusari, R.
Brandt, O. Al-Taher, C. Fetzer, G. S. Kinsey, and N. Karam: IEEE J.
Photovoltaics 2 (2012) 209.
14) R. R. King, D. Bhusari, D. Larrabee, X.-Q. Liu, E. Rehder, K. Edmondson,
H. Cotal, R. K. Jones, J. H. Ermer, C. M. Fetzer, D. C. Law, and N. H.
Karam: Prog. Photovoltaics (in press) [DOI: 10.1002/pip.1255].
15) P. T. Chiu, S. J. Wojtczuk, X. Zhang, C. Harris, D. Pulver, and M.
Timmons: Proc. 37th IEEE Photovoltaic Specialists Conf., 2011, p. 771.
16) NREL conrms world-record 43.5% eciency on Solar Junctions CPV
cell [http://www.pv-tech.org/news] (April 11, 2011).
17) P. Patel, D. Aiken, A. Boca, B. Cho, D. Chumney, B. Clevenger, A.
Cornfeld, N. Fatemi, Y. Lin, J. Mccarty, F. Newman, P. Sharps, J. Spann,
M. Stan, J. Steinfeldt, and T. Varghese: IEEE J. Photovoltaics 2 (2012)
377.
18) G. Kinsey and K. Edmondson: Prog. Photovoltaics 17 (2009) 279.
19) R. R. King, D. Bhusari, A. Boca, D. Larrabee, X.-Q. Liu, W. Hong, C. M.
Fetzer, D. C. Law, and N. H. Karam: Prog. Photovoltaics 19 (2011) 797.
20) C. Gueymard: Sol. Energy 71 (2001) 325.
21) S. Wilcox and W. Marion: Users Manual for TMY3 Data Sets, Tech. Rep.
NREL/TP-581-43156, Revised May 2008.
22) D. R. Myers, K. Emery, and C. Gueymard: Proc. 29th IEEE Photovoltaic
Specialists Conf., 2002, p. 923.
R. K. Jones et al. Jpn. J. Appl. Phys. 51 (2012) 10ND01
10ND01-4 # 2012 The Japan Society of Applied Physics