Solar Energy Materials & Solar Cells 143 (2015) 190–197

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells

journal homepage: www.elsevier.com/locate/solmat

The Passivated Emitter and Rear Cell (PERC): From conception to mass
production
Martin A. Green n
Australian Centre for Advanced Photovoltaics, School of Photovoltaic and Renewable Energy Engineering, University of New South Wales (UNSW), Sydney
2052, Australia

art ic l e i nf o a b s t r a c t

Article history: Improved solar cell efficiency is the key to ongoing photovoltaic cost reduction, particularly as economies
Received 31 March 2015 of scale propel module-manufacturing costs towards largely immutable basic material costs and as
Received in revised form installation costs become an increasingly large contributor to total system costs. To enable manufacturers
28 June 2015
to move past the 20% cell energy conversion efficiency figure in production, high-efficiency PERC (Pas-
Accepted 29 June 2015
sivated Emitter and Rear Cell) sequences are being increasingly brought online. Most new photovoltaic
manufacturing capacity added in the second half of 2014 was PERC-based, making PERC now the cell
Keywords: technology with second-highest production capacity, with the latest industry roadmap anticipating PERC
Silicon solar cells will become the dominant commercial cell technology by 2020. The first paper describing the PERC cell
PERC
appeared in 1989, although the structure was conceived several years earlier. The attractive technical
High efficiency solar cells
features were the reduction of rear surface recombination by a combination of dielectric surface passi-
vation and reduced metal/semiconductor contact area while simultaneously increasing rear surface
reflection by use of a dielectrically displaced rear metal reflector. The key issues in the development of
this technology and its commercial implementation are described, including a review of recent adoption
rates and the way these are likely to evolve in the future.
& 2015 Elsevier B.V. All rights reserved.

1. Background technology efficiency being its high open-circuit voltage (Voc). A contributor to
this high voltage (658.4 mV at 25 °C) was the reduced area of the
The first paper describing the PERC cell appeared in 1989 [1], top surface contact, the first demonstration of improved conver-
although this device was first described in 1983 in a UNSW sion efficiency by this approach, and surface passivation by
(University of New South Wales) final grant report [2] and as a dielectric oxide, both features that the PERC cell sought to emulate
deliverable in a subsequent grant proposal [3], accompanied in on the rear surface. The PESC cell retained the Al alloyed BSF/BSR
both cases by the drawing shown in Fig. 1. The attractive feature (back surface field/ back surface reflector) developed in the late
was the elegant way in which the PERC cell incorporated three 1960s [6] that had featured in earlier generations of high efficiency
attributes into the rear contacting scheme that earlier work at cells and has been used in most commercial cells to date.
Any solar cell can be associated with an effective diode
UNSW and elsewhere had shown were important to obtaining
saturation current density, Jo, calculated from its measured Voc and
high efficiency. These were the reduction of rear surface recom-
short-circuit current density, Jsc, as:
bination by a combination of dielectric surface passivation and
reduced metal/semiconductor contact area, with simultaneously ⎛ qVoc ⎞
Jo = Jsc /[exp ⎜ ⎟ − 1]
increased rear surface reflection by use of a dielectrically displaced ⎝ kT ⎠ (1)
rear metal reflector.
where T is the absolute temperature and kT/q is the thermal vol-
Around the time the PERC cell was proposed, the highest
tage (25.693 mV at 298.15 K or 25 °C). This cell had a creditable Jo
confirmed efficiency for a Si cell was 19.1% [4], estimated as 18.4%
of 270 fA/cm2 at 25 °C, almost evenly divided between contribu-
efficient by present standards [5]. The cell structure was a rela-
tions from the combination of top surface and contact recombi-
tively simple UNSW planar PESC cell (Passivated Emitter Solar
nation and from bulk and rear contact recombination [7].
Cell) of Fig. 2 with the main feature responsible for its high Apparently the first published suggestion of reduced contact
area as a way of reducing contact recombination and its con-
n
Tel.: þ 61 2 9385 4018. tribution to Jo was made at UNSW almost a decade earlier [8]. In
E-mail address: [email protected] the Crowell-Sze thermionic-emission/diffusion theory of metal/

http://dx.doi.org/10.1016/j.solmat.2015.06.055
0927-0248/& 2015 Elsevier B.V. All rights reserved.
 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197 191

Fig. 4. Oxide passivation as a way of reducing surface recombination in MIS solar

cells as well as related MINP cells, with the latter differentiated by an n-type dif-
fused layer under the oxide.

Fig. 1. First diagram of PERC cell [2,3] (PERT configuration: Passivated Emitter, Rear
Totally-diffused). More information on different PERC configurations is given sub-
sequently in Fig. 8, with redefined acronyms to better reflect present usage.

Fig. 5. Early history of Voc improvement in Si cells, demonstrating the effectiveness

Fig. 2. The PESC solar cell, the most efficient Si cell at the time of PERC cell proposal of UNSW oxide passivation.
[4].
researchers suggested the same approach as in Fig. 3 as a way of
achieving this low contact, implemented by COMSAT Labs nearly a
decade later, slightly earlier than in PESC cells, but with less
resounding outcomes [12].
For similar reasons, dielectric passivation of the top surface also
had not been important in earlier generations of silicon cells. The
good interfacial properties of thermally grown oxides were well
known from their use in Metal/Oxide/Semiconductor (MOS)
transistors. UNSW led efforts to improve open circuit voltage in
the late 1970s, using the inversion layer MIS (Metal/Insulator/
Semiconductor) cell structure of Fig. 4, relying on oxide passiva-
tion in non-contacted areas as well as charge in the overlying
dielectric. These MIS devices were first reported as giving Voc
above conventional cells in 1976 [13].
Fig. 3. Reduced contact area as a way of reducing contact recombination [8]. MIS and related MINP (Metal/Insulator/N–P junction) devices
subsequently became the first silicon cells to display Voc above
650 mV in 1978 and 678 mV in 1981 (Fig. 5). The latter voltage
semiconductor contacts [9], currents associated with these con-
tacts are described in terms of effective recombination velocities corresponds to J0 values of 100 pA/cm2 with 26 pA/cm2 assigned to
for both electrons and holes [10]. It was suggested reducing con- recombination at the top oxide-passivated surface [14]. This oxide
tact area as in Fig. 3 could reduce such effective velocities [8], passivation has been maintained through successive generations
improving the open-circuit voltage of Schottky diode solar cells. of UNSW high voltage, high efficiency devices.
The specific approach shown in Fig. 3 to reduce contact area was The first paper applying the benefits of such thin oxide passi-
employed in the first experimental PESC cells, prior to delivery of vation to more conventional p–n junction device structures
the photolithography masks used to fabricate the 19.1% cell of appeared in 1978 [15]. This was also an important paper in that it
Fig. 2, with 687 mV Voc confirmed for a device with this structure also seems to be the first to use PECVD silicon nitride, now the
by NASA-Lewis in Sept. 1983 [2], the highest ever independently industry standard, as an antireflection (AR) coating, although
confirmed value at this stage. Motorola had been using non-PECVD nitrides in production since
Almost contemporaneously with the UNSW paper [8], Lind- 1977 [16].
mayer and Allison of COMSAT Laboratories, inventors of the violet The UNSW interest in dielectrically displaced rear reflectors
cell that led to substantial efficiency improvements in the early arose from the results shown in Fig. 6, published by Solarex in
1970s, also suggested use of reduced contact area in a subse- 1976 [17], showing the measured reflectance nominally at the rear
quently published patent application [11]. The shallower emitters Al/Si BSF/BSR interface as a function of alloying conditions. Trying
(top diffused layers) in these devices had directed attention to to understand these data (later realising drop at high temperature
emitter surface recombination, negligible in earlier generations of was due to surface roughening and consequent light-trapping),
cells because of their deep emitter diffusions. Independently, these expected reflection values were calculated theoretically.
192 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197

This led to the realisation that the high refractive index of almost ideal 98%. Knowledge of these large optical benefits pro-
silicon considerably supressed rear metal reflection and that vided additional incentive for implementing the PERC structure.
inserting a lower index dielectric layer between the metal and Successful implementation of the PERC cell was not immediate.
silicon would significantly increase this reflection. Experiments Introducing top-surface texturing of the PESC cell of Fig. 2, also
subsequently conducted at UNSW as part of an undergraduate suggested in the previous grant documents [2,3], gave more rapid
thesis project produced the results shown in Fig. 7, showing progress with 20% cell efficiency demonstrated in 1985 [5].
Attention was then diverted to applying these developments to
reflection from as-deposited Al could be increased from 89% to an
silicon concentrator cells to meet contractual requirements, with
this work resulting in the first 20% efficient photovoltaic module
[18]. For the PERC cell, boron diffusion capability needed to be
established and perfected, with this proving more challenging
than for the phosphorus diffusions already established. Some tips
from the solar cell group at Stanford University regarding the
advantages of chlorine-based furnace processing proved most
helpful here, allowing the whole family of PERC cells (Fig. 8) to be
experimentally investigated. The first high efficiency PERC cells
were fabricated in 1988, with 21.8% efficiency confirmed at Sandia
in October 1988 (20.9% by present standards).

2. Efficiency and processing optimisation

These initial results fuelled the ongoing improvements in sili-

con cell efficiency on p-type monocrystalline substrates to 25%
(Fig. 9). Applying the approach to multicrystalline substrates led to
the first multicrystalline cell of efficiency above 20%, by present
Fig. 6. Measured reflectance nominally at the rear alloyed Al/Si interface as a
function of alloying conditions [17]. standards, in 1998 [5]. Initial UNSW application to n-type sub-
strates using reversed doping polarities gave lower efficiency than
on p-type substrates, with an efficiency of 21.9% (22.1% by present
standards [5]) demonstrated in 1991, with the lower efficiency
arising from the increased challenges involved in performing
large-area B diffusions. Much later in 2005, an inverted rear
emitter structure increased this efficiency to 22.7% (22.9% by
present standards), equalling the efficiency of the best ever n-type
solar cell at that time. A significant subsequent independent
development was the recognition of the excellent surface passi-
vation properties of Al2O3. This material had long been used as the
low-index layer in double-layer AR coatings for space cells. An
early report [19] of its excellent passivation properties for p-type
surfaces went largely unnoticed until a new efficiency mark of
Fig. 7. Early UNSW measurements of effects of intervening dielectric layer on rear
23.2% (23.4% by present standards) on n-type substrates [20] was
Al reflection (evaporated SiO was used as a dielectric since the ability to grow thick established in 2006. This dielectric has proved important for
thermal oxides had not been established at UNSW at the time). subsequent PERC commercialisation.

Fig. 8. The Passivated Emitter and Rear Cell (PERC) family. (a) Simple PERD cell (Passivated Emitter, Rear Directly-contacted); (b) PERL cell (Passivated Emitter, Rear Locally-
doped); (c) PERT cell (Passivated Emitter, Rear Totally-diffused); (d) PERF cell (Passivated Emitter, Rear Floating-junction). The PERC configurations now most widely
implemented are the PERL and PERT.
 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197 193

Fig. 11. Metallisation pattern used for 68 mm  68 mm buried contact cells.

Fig. 9. Evolution of Si cell efficiency [5].

oxide and driving it through the oxide by laser heating at localised
points. We found this approach to be more problematic, with
difficulties attributed to the layer of phosphorus that wraps
around to the rear during the top surface diffusion. Although it
was possible to control the thermal environment during normal
buried contact cell processing to prevent detrimental con-
sequences from this layer [26], this was not possible in the tran-
sient laser-processing environment. Subsequently, this problem
was resolved and the approached perfected at the Fraunhofer
Institute [27], by the simple expedient of masking the rear during
front surface diffusion.
The lasers available both to us and to our licensees at that time
were better suited to forming continuous grooves than dot or dash
patterns when large cell quantities were involved. Buried contact
cell work consequently took the direction of exploring bifacial rear
grooved structures. PERC integration was achieved by a hybrid
sequence, using laser processing for the top cell surface but che-
Fig. 10. Buried contact solar cell. mical etching to open holes in the rear oxide, with hole positions
defined by photolithography. This approach took hybrid, large area
3. Large area cells and modules cell performance well above 20%. Sixteen of these early PERC
hybrid cells were encapsulated into a nominally 800-cm2 module
UNSW had earlier pioneered the use of lasers in solar cell in 1992 resulting in the first confirmed 20% flat-plate photovoltaic
processing [21,22] with this resulting in the buried contact solar module efficiency (19.9% certified in 1992, equivalent to 20.1%
cell of Fig. 10 [23], developed as a commercial version of the PESC under present standards). These results were improved upon in
cell. These cells relied upon plated Ni/Cu metallisation, with 1993 [28] by implementing an all-photolithographic PERC process
excellent field performance from early modules deployed by that retained the metallisation pattern of Fig. 10, to allow the cells
licensee BP Solar recently reported after 20 years field exposure to be used interchangeably with previous pilot line output. About
5 kW of UNSW pilot line cells were supplied to 3 teams for the
[24].
1993 Solar World Challenge, the solar car race across Australia,
A pilot line was established at UNSW in the late 1980s to assist
including about 1 kW of PERC cells representing the first com-
in transferring this technology to licensees. In 1990, this pilot line
mercial PERC cell sales. For the 1996 race, a further 8 kW of PERC/
was diverted to the fabrication of large areas of cells for solar car
PERL cells were supplied to 3 teams. Forty cells made during this
racing, at that time attracting the interest of major automobile
period were encapsulated into a nominally 800-cm2 module giv-
companies as a promotional vehicle. The 68 mm  68 mm cells
ing confirmed 22.7% module efficiency [29] (22.9% by present
fabricated on this line used the “finger intensive” cell metallisation
standards, still the highest for a silicon module at the time of
pattern of Fig. 11, made possible by the high conductivity and writing).
reduced width of the buried contacts.
Two laser-based processing approaches were investigated in
the early 1990s as a way of integrating the new PERC rear structure 4. Commercialisation
into this laser-based sequence. One was to use the laser to ablate
small openings in a rear oxide to provide the desired small area Since the mid-1980s, most manufacturers have used a common
contact points, with Al subsequently deposited and sintered. This manufacturing process for fabricating silicon solar cells based on
was the most successful in our work, producing cells with high Voc the use of phosphorus diffused, boron doped silicon wafers with
and good fill factors [25]. The equivalent of this contacting screen-printed silver paste top contacts and rear contacts based on
approach is also the most widely used in present commercial PERC screen-printed Al pastes [30]. The latter are alloyed to form an “Al-
sequences. The second involved deposition of Al on top of the BSF” under the rear contact. Multicrystalline wafers and plasma
194 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197

deposited silicon nitride antireflection coatings have been used 100%

Other n- type
commercially from about the same time. Over recent years, the

New Si Capacity, %
80% HJT/Rear J
standard Al-BSF process accounted for over 90% of silicon solar cell
production, itself accounting for over 90% of total photovoltaic PERC
60%
module production. The remainder of the latter production is Al-BSF
made up of thin-film technologies based on CdTe, CIGS (Copper– 40%
Indium–Gallium–Selenide), and amorphous silicon (a-Si), with Source:
slowly declining combined market share. By 2013, the perfor- 20% Solarbuzz
mance of this dominant silicon Al-BSF technology had steadily
0%
improved to the stage where standard commercial cells were 2013 2014 2015 2016 2017 2018
approaching the efficiency level reached by PESC cells in the
Fig. 13. Share of new silicon-based manufacturing capacity of different cell
laboratory in the 1980s using a similar rear Al-BSF approach. To approaches [38].
progress further, a similar rear contact improvement was required.
Nothing is as powerful as an idea whose time has come. As the
100%
push to efficiencies above 20% in production intensified, many

Total Si Capacity, %
companies reported on progress with implementing PERC tech- 80%
nology. Suntech Power was one of the first, reporting on such Si-Tandem
progress in production implementation in 2009 [31] as the com- 60%
Rear J
pany's “Pluto” product series. Generation 1 product incorporated a
40% HJT
laser-based selective emitter process co-developed with UNSW
with standard Al-BSF, reaching production volumes of 0.5 GW/year 20% PERC
[32]. Generation 2 product incorporated PERC rear contacts,
0% Al-BSF
demonstrating 20.3% efficiency as independently confirmed in
2011 [32]. Schott Solar reported 18.7% efficiency using multi-
crystalline wafers and the PERC approach in 2010, resulting in a
Fig. 14. Expected market share of different Si cell technologies [39].
confirmed world-record 17.6% efficient module efficiency [33]. In
2012, Schott reported 21% monocrystalline silicon efficiency, the
imminent release of a high-efficiency PERC cell module and share of total new capacity additions in 2013 and accounted for the
licensing the company's version of PERC technology to equipment majority of new additions by the second half of 2014 (Fig. 13). This
supplier Schmid. Q-Cells reported 19.5% multicrystalline PERC cell displacement of the standard Al-BSF approach is expected to
efficiency in 2011, increasing confirmed module efficiency to 17.8% continue, with no new Al-BSF capacity expected to be added after
and then 18.5% [34]. Trina Solar recently increased the multi- 2017.
crystalline cell efficiency to a world record 20.8% and corre- This view is reiterated in the independent April 2015 photo-
sponding module efficiency to 19.2%, again using PERC cells. Tai- voltaic industry roadmap (ITRPV) [39]. As indicated in Fig. 14, the
wanese company, Global Sunrise, founded by former UNSW rapid growth of PERC production capacity in 2014 made PERC,
researchers, was one of the first companies to report on actual after the standard Al-BSF technology, the cell technology with the
PERC production experience. Working with equipment supplier, second highest established production capacity by year-end
Roth and Rau, a PERC production process compatible with screen- (ahead of rear junction and HIT/HJT a-Si heterojunction silicon
printed pastes went into production in 2012, with average cell cells, as well as CdTe, CIGS and a-Si thin-film cells). Over the next
efficiency above 20% reported [35]. few years, the dominance of new production capacity by PERC
GTM Research documents subsequent activity in 2013 and 2014 (Fig. 13) will steadily increase its share of total capacity with the
[36]. Equipment suppliers Schmid, Centrotherm, RENA, Solaytec, most recent ITRPV roadmap [39] showing PERC likely to become
Levitech, SINGULUS and Meyer Burger are mentioned as well as the dominant cell production technology by 2020 (Fig. 14).
manufacturers Neo Solar Power, SunEdison, JA Solar and WINAICO. This trend is being accelerated by the consolidation policies of
NDP Solarbuzz reported on the rapid rate of adoption of PERC into the Chinese government whereby manufacturers without access
manufacturing [37], with 2.5 GW PERC capacity estimated world- to good technology are being closed down by imposing cell effi-
wide in August, 2014, with this growing quickly (Fig. 12). ciency standards [40].
According to a separate NDP Solarbuzz report [38], the new
PERC capacity additions shown in Fig. 12, accounted for a growing
5. Production sequences and costings
2.5
Although PERC manufacturing capacity is growing quickly, the
technology is still early in its developmental cycle (Fig. 12) with far
PERC capacity, GW

2
Source: NDP Solarbuzz more differences between sequences in production than with the
much better established Al-BSF approach. Nonetheless, the sim-
1.5 plified process flow diagram for PERC cell fabrication, shown
compared to the standard Al-BSF approach in Fig. 15, broadly
1 captures the key features. As apparent from this figure, one
attraction of the PERC sequence is that it is largely compatible with
0.5 the standard Al-BSF sequence.
The first step in both cases is a wafer saw damage removal etch
0 and texturing, followed by emitter diffusion and etch. Often, a rear
2012 Q1
Q2
Q3
Q4

Q2
Q3
Q4

Q2
Q3
Q4
2013 Q1

2014 Q1

side polish etch is included at this stage for the PERC sequence, to
increase rear reflection and to reduce rear recombination [35,41].
This is a low-cost step since usually done in conjunction with the
Fig. 12. Rapid increase in PERC manufacturing capacity (2012–2014) [37]. required edge-junction isolation step. Usually a few microns of
 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197 195

Fig. 15. Simplified process flow for Al-BSF (upper) and PERC (lower) sequences.

silicon are removed in this polishing step, with consequent energy as the higher voltage capability of the PERC approach is fully
conversion efficiency gains reported as in the 0.4% to 1.5% (abso- exploited. In the laboratory, the performance of PESC devices,
lute) range [42,43], depending on the details of the process. representative of a fully developed Al-BSF approach, peaked at just
Instead of a simple silicon nitride AR coating deposition as in below 21% by present standards, while PERC/PERL sequences
the standard Al-BSF sequence, both AR and rear dielectric coating reached 25%, suggesting a 20% margin could eventually open up.
of the cell is required in the PERC process, with the rear coating
generally taking the form of either an aluminium oxide/silicon
nitride stack [35,44] or a silicon oxide/silicon nitride stack [45]. In 6. Performance potential
some processes, both the front and the rear depositions are done
in the same item of equipment [35]. The contact holes through the In mid-2015, the best performing near-commercial PERC mul-
rear dielectric are generally formed by laser ablation, although ticrystalline cells have efficiencies in the 20–21% range, while the
chemical etching is feasible. Contact screening and printing, cell best PERC monocrystalline cells have slightly higher values in the
testing and sorting occurs similarly in both sequences, although 21–22% range. There were confident expectations for both values
different paste compositions would be used for the rear contact for soon to surpass the top of these respective ranges, but how far can
PERC cells, to avoid damage to the rear dielectrics and void for- the technology go?
mation at the locally alloyed regions [43]. Voids arise from Several recent studies throw some light on this issue [52–55].
migration of some silicon through the rear Al layer away from the The most recent [55] shares the author's view that, now the
openings during alloying, never to return, although addition of industry has invested in a transition to PERC technology, this
silicon to the pastes can prevent this from becoming an issue. provides a path for manufacturers for incremental improvements
To allow soldering of the rear, Ag solder pads also need to be to values of cell efficiency to close to the 25% value demonstrated
screened to the rear side of the cell, as in the standard Al-BSF in the laboratory. Commercial PERC cells still perform well short of
approach. This is reported as being less harmful to cell perfor- the best laboratory devices in many ways, particularly in terms of
mance for PERC cells [46], since the area of high recombination Voc as determined by total cell Jo (Eq. (1)) and resistive losses [56].
velocity contact between the Ag layer and substrate is much The 25% efficient PERL devices have a total Jo of only 50 fA/cm2
smaller in this case, occurring only where there are openings in while a typical 20–21% efficient commercial PERC cell may have a
the rear dielectric. Contact between the pads and the Al rear value over 300 fA/cm2 with over half of this coming from the
metallisation areas occur through carefully controlled overlap emitter region [53,55,56].
regions [47]. This makes attention to emitter design important for realizing
Since at least three extra processing steps are involved, extra the highest efficiencies in the near term. The 25% devices have a Jo
cell processing costs are involved for PERC cells. One early study contribution from this region of only 15 fA/cm2 [56], showing
[48] estimates the cell processing cost as 26% higher/cell, reducing what is ultimately feasible. Values below 80 fA/cm2 are suggested
to 20% higher/W when the rather low assumed 5.6% performance as feasible for commercial devices in the near term using a
advantage is taken into account. At the cell level, with wafer costs selective emitter approach [53], as in the PERC cells of Fig. 8,
included but similarly leveraged by the higher efficiency, the although advanced homogeneous emitters are also regarded as an
additional manufacturing cost is estimated at 6%/W. At the module option [55]. Reducing emitter contributions to Jo will then bring
level, the same leveraging reduces the additional cost to 2%/W. At the rear contact of the PERC cell into focus as the performance-
the system level, PERC costs become marginally lower, in this limiting feature.
study. A manufacturer of both PERC and Al-BSF devices considers a Attention to the surface recombination velocity along the non-
much higher 11% performance advantage as a reasonable estimate, contacted regions of the rear surface [52] as well as to the for-
with 90% of this advantage able to be captured as lower transport mation of the doped region in the contact areas [53,55] is expected
costs, 70% as lower installation costs and over 50% as lower bal- to bring the rear surface contribution to Jo below that of the
ance of systems costs [49]. emitter, then elevating recombination in the bulk of the device to
PERC cells are now available in sufficient quantities that at least being the major contributor to Jo, determined by the combination
one market research company tracks spot prices for PERC cells as a of minority carrier lifetime and doping level in these regions.
separate category [50]. In mid-March 2015, the company reported B-doped, p-type wafers presently dominate commercial produc-
average spot prices for multicrystalline Al-BSF cells as 29.5–31.5 c/ tion. A key advantage is the near-unity liquid to solid segregation
W, depending on efficiency with 17.8% efficiency providing a coefficient of B that means its concentration remains reasonably
demarcation point, and 38.5 c/W for monocrystalline Al-BSF constant along the manufactured ingots, whether mono- or multi-
devices. PERC monocrystalline spot prices were reported as 46– crystalline. A disadvantage is that B forms a complex with O that is
48 c/W for cells of efficiency in the 20.4–20.6% range [50], corre- activated under illumination, restricting the bulk minority carrier
sponding to a 19–25% market premium that would seem to more lifetime in fielded devices [57–59].
than offset the increased processing costs. For this reason, there have been suggestions that the industry
With ongoing streamlining and standardisation of PERC needs to move as a whole to P-doped, n-type monocrystalline
sequences, the processing cost premium/cell relative to Al-BSF wafers, which is also a feasible substrate for high-performance
devices is expected to reduce [51]. At the same time, the perfor- PERC cells. Since P does not form such detrimental defects with O,
mance advantage over Al-BSF cells might be expected to increase this allows higher lifetimes in standard Czochralski (CZ) grown
196 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197

ingots. Companies such as SunPower and Panasonic use n-type on-going incremental improvements. Recent simulations confirm
wafers to produce modules attracting premium prices for the the feasibility of such efficiencies through the continuous devel-
high-efficiency end of the present market (double to triple those of opment of existing technologies [55].
standard product [60]). However, the lower P segregation coeffi-
cient means that there is a greater resistivity variation along the
CZ ingot than with B. Acknowledgement
To maintain tight production spreads and performance levels
close to champion devices, some manufacturers are said to use as The Australian Centre for Advanced Photovoltaics is supported
little as 35% of each P-doped ingot [61]. Using a greater portion of by the Australian Government through the Australian Renewable
the ingot may be one reason for the large differences noted with Energy Agency (ARENA) (SRI-001). Responsibility for the views,
other technologies, such as heterojunction devices, between information or advice expressed herein is not accepted by the
champion and production devices. Going to continuous CZ growth Australian Government. The author thanks the many colleagues
provides an opportunity to reduce resistivity variations, but this who contributed to PERC cell development within his group,
may be self-defeating, since the longer growth times are likely not notably Aihua Wang, Andrew Blakers, Jianhua Zhao and Stuart
consistent with obtaining the desired highest possible carrier Wenham.
lifetimes [61]. In any case, P-doping does not appear suitable for
the multicrystalline ingots that form the backbone of the present
industry, due to this dopant segregation issue.
Given these issues and the present industry momentum, it References
seems likely that p-type wafers will remain the industry norm
well into the future. If recently reported progress [57–59] with [1] A.W. Blakers, A. Wang, A.M. Milne, J. Zhao, M.A. Green, 22.8% Efficient Silicon
Solar Cell, Appl. Phys. Lett. 55 (1989) 1363–1365.
rapid and permanent deactivation of the B–O defect can be [2] M.A. Green, A.W. Blakers, J. Kurianski, S. Narayanan, J. Shi, T. Szpitalak, M.
brought into production, millisecond bulk lifetimes might be a Taouk, S.R. Wenham and M.R. Willison, Ultimate Performance Silicon Solar
standard option using p-type CZ wafers. This corresponds to a bulk Cells, Final Report, NERDDP Project 81/1264, Jan. 82-Dec. 83 (dated Feb., 1984).
[3] M.A. Green, High Efficiency Silicon Solar Cells, Proposal in response to RFP RB-
contribution to Jo of less than 40 fA/cm2[53]. Alternatively, other
4-04033, SERI (now NREL), March, 1984.
p-type dopants such as Ga or In that do not form harmful defects [4] M.A. Green, A.W. Blakers, J. Shi, E.M. Keller, S.R. Wenham, 19.1% efficient sili-
with O could be used for monocrystalline material, while main- con solar cell, Appl. Phys. Lett. 44 (1984) 1163–1165.
taining processing compatibility with B-doped multcrystalline [5] M.A. Green, The path to 25% silicon solar cell efficiency: History of silicon cell
evolution, Prog. Photovolt. 17 (2009) 183–189.
material. Lifetimes in the latter material are rapidly improving, [6] J. T. Burrill, D. Smith, K. Stirrup and W. J. King. The application of implantation
with recent reports of lifetimes of 0.4 ms averaged over a grown to thin silicon cells, in: Proceedings of the 6th IEEE Photovoltaic Specialists
ingot [62]. Conference, Cacoa Beach, 1967, pp. 81–97.
With such improved bulk lifetime, the emitter region may then [7] A.W. Blakers, M.A. Green, Oxidation condition dependence of surface passi-
vation in high efficiency silicon solar cells, Appl. Phys. Lett. 47 (1985) 818–820.
again require further refinement, putting another cycle of incre- [8] M.A. Green, Enhancement of Schottky solar cell efficiency above its semi-
mental improvements into effect. Resistive losses can be lowered empirical limit, Appl. Phys. Lett. 27 (1975) 287–288.
towards those of laboratory devices by increasing the number of [9] C.R. Crowell, S.M. Sze, Current transport in metal-semiconductor barriers,
Solid-State Electron. 9 (1966) 1035.
busbars, with a multi-wire approach suggested as the ultimate [10] M.A. Green, J. Shewchun, Minority carrier effects upon the small signal and
solution in this regard [55]. This allows contact by narrow plated steady-state properties of Schottky diodes, Solid-State Electron. (1973)
top contacts, with 24.4% efficiency for a 156 mm by 156 mm cell 1141–1150.
predicted by detailed modelling by such continuous development [11] J. Lindmayer and J. F. Allison, Dotted Contact Fine Geometry Solar Cell, U.S.
Patent 3,982,964, September 28, 1976. (Assigned to COMSAT).
of existing fabrication technology, leading to the conclusion that [12] A. Meulenberg and R.A. Arndt, Surface effects in high voltage silicon solar cells,
PERC cells will set a moving target and dominate the market for in: Proceedings of the 16th IEEE PV Specialists Conference, 1982, pp. 348–353.
some time to come [55]. [13] M.A. Green, R.B. Godfrey, MIS solar cell-general theory and new experimental
results for silicon, Appl. Phys. Lett. 9 (1976) 610–612.
[14] A.W. Blakers, M.A. Green, 678 mV open circuit voltage silicon solar cell, Appl.
Phys. Lett. 39 (1981) 483–485.
7. Conclusion [15] J.G. Fossum, E.L. Burgess, High efficiency p þ -n-n þ back-surface-field silicon
solar cells, Appl. Phys. Lett. 33 (1978) 238.
[16] T.G Sparks and R.A. Pryor, in: Proceedings of the 14th IEEE Photovoltaic Spe-
As manufacturers move past the 20% efficiency mark in pro- cialists Conference, San Diego, January 1980, pp. 783–786.
duction, the standard Al-BSF approach that has been the dominant [17] C.Y. Wrigley, High efficiency solar cells, in: Proceedings of the 12th IEEE
commercial approach for the last 30 years is in the process of Photovoltaic Specialists Conference, Baton Rouge, 1976, pp. 343–346.
[18] C.J. Chiang and E.H. Richards, A 20% efficient photovoltaic concentrator
being surpassed by a fundamentally higher efficiency approach.
module, conference record, 21st IEEE Photovoltaic Specialists Conference,
Although high silicon cell efficiency has been obtained in pro- Kissimmee, May 1990, pp. 861–863.
duction with both rear junction and heterojunction approaches, [19] R. Hezel, K. Jaeger, Low-Temperature surface passivation of silicon for solar
these require specialized high-quality, n-type monocrystalline cells, J. Electrochem. Soc. 136 (2) (1989) 518–523.
[20] J. Benick, B. Hoex, M.C.M. Van de Sanden, W.M.M. Kessels, O. Schultz, S.
wafers. The PERC cell approach has the advantage of being able to W. Glunz, High efficiency n-type Si solar cells on Al2O3-passivated boron
tolerate both monocrystalline and multicrystalline substrates of emitters, Appl. Phys. Lett. 92 (25) (2008) (253504-253504-3).
either polarity, while demonstrating similar efficiencies to the [21] S.R. Wenham and M.A. Green, Laser Grooved Solar Cell, Australian Patent
565,214; U.S. Patent 4,626,613, Japanese Patent No. 69933/95.
above approaches on good quality substrates.
[22] S.R. Wenham and M.A. Green, Buried Contact Solar Cell, Australian Patent
This robustness and compatibility with existing product pro- 570,309; U.S. Patents 4,726,850 and 4,748,130; Taiwanese Patent N1024881;
duction lines is considered likely to see the approach surpass the Indian Patent 164492, German Patent P3585697.1-08, South Korean Patent No.
standard approach by 2020 in terms of capacity share [39]. PERC 70109.
[23] Chee Mun Chong, S.R. Wenham, M.A. Green, High efficiency laser grooved,
uptake is likely to lead to an era of accelerated silicon solar cell buried contact solar cell, Appl. Phys. Lett. 52 (1988) 407–409.
performance increase as the full capabilities of this technology are [24] M. Alonso-Abella, F. Chenlo, A. Alonso and D. González, TOLEDO PV Plant
exploited. Just as commercial solar cell efficiencies have approa- 1 MWp – 20 years of operation, 30th EUPVSEC, Amsterdam, September 2014.
[25] Fuzu Zhang, Buried Contact Silicon Concentrator Solar Cells, The University of
ched the performance of the best laboratory Al-BSF solar cells, it is
New South Wales, Sydney, 1995 (Ph.D thesis).
anticipated that the PERC technology will also approach that of the [26] Chee Mun Chong, Buried Contact Solar Cells, University of New South Wales,
best laboratory cells with efficiency of 25% [5] through a process of Sydney, 1989 (Ph.D thesis).
 M.A. Green / Solar Energy Materials & Solar Cells 143 (2015) 190–197 197

[27] E. Schneiderlöchner, R. Preu, R. Lüdemann, S.W. Glunz, Laser-fired rear con- oxidation layer, Proceedings of the 42nd IEEE Photovoltaic Specialists Con-
tacts for crystalline silicon solar cells, Prog. Photovolt.: Res. Appl. 10 (1) (2002) ference, New Orleans, June 14–19, 2015.
29–34. [46] F. Kiefer, T. Brendemü hl, M. Berger, A. Lohse, S. Kirstein, N. Braun, M. Lehr,
[28] J. Zhao, A. Wang, M. Taouk, S.R. Wenham, M.A. Green, 20% efficient photo- F. Heinemeyer, V. Jung, A. Morlier, S. Blankemeyer, I. Kunze, R. Winter, N.-
voltaic module, Electron Device Lett. EDL-14 (1993) 539–541. P. Harder, T. Dullweber, M. Kö ntges, R. Brendel, Influence of solder pads to
[29] J. Zhao, A. Wang, E. Abbaspour-Sani, F. Yun, M.A. Green, Improved efficiency PERC solar cells for module integration, Energy Procedia 38 (2013) 368–374.
silicon solar cell module, IEEE Electron Device Lett. 18 (1997) 48–50. [47] C. Kohn, M. Hug, R. Kuebler, M. Krappitz and G. Kleer, Increase of the strength
[30] M.A. Green, Silicon photovoltaic modules: a brief history over the first 50 of screen printed silicon solar cells by post treatments, 25th European Pho-
years, Prog. Photovolt. 13 (2005) 447–455. tovoltaic Solar Energy Conference and Exhibition/5th World Conference on
[31] Z. Shi, S.R. Wenham, S.; J. Ji, Mass production of the innovative Pluto solar cell Photovoltaic Energy Conversion, 6–10 September 2010, Valencia, pp. 2062–
technology, in: Proceedings of the 34th IEEE Photovoltaic Specialists Con- 2065.
ference, Philadelphia, 2009. [48] S. Nold, N. Voigt, L. Friedrich, D. Weber, I. Hädrich, M. Mittag, H. Wirth, B.
[32] Z. Wang, P. Han, H. Lu, H. Qian, L. Chen, Q. Meng, N. Tang, F. Gao, Y. Jiang, J. Wu, Thaidigsmann, I. Brucker, M. Hofmann, J. Rentschand R. Preu, Cost Modeling of
W. Wu, H. Zhu, J. Ji, Z. Shi, A. Sugianto, L. Mai, B. Hallam, S.R. Wenham, silicon solar cell production innovation along the PV value chain, in: Pro-
Advanced PERC and PERL production cells with 20.3% record efficiency for ceedings of the 27th European Photovoltaic Solar Energy Conference and
standard commercial p-type silicon wafers, Prog. Photovolt.: Res. Appl. 20 (3) Exhibition, 2012.
(2012) 260–268. [49] JA Solar, Percium Mono 60/275-295 Data Sheet, September 2014.
[33] C. Arndt, D. Stichtenoth, N. Mü ller, M. Reinicke, D. Grote, U. Schröder, M. [50] 〈http://pv.energytrend.com/pricequotes.html〉.
Scherr, M. Weser, F. Lange, P. Voigt, P. Olwal, L. Schubert, F. Satian, M. Döring, [51] J. Rodriguez, PERC solar cells steadily gaining steam in PV manufacturing, Sol.
M. Neitzert, T. Geiler, K. Meyer, O. Gybin, F. Dreckschmidt, J. Neusel, S. Alt- Choice News (2015) 17 February (http://www.solarchoice.net.au/blog/news/
mannshofer, M. Mach, C. Markscheffel, A. Froitzheim, M. Mü ller, C. Koch, G. perc-solar-cells-steadily-gaining-steam-in-pv-160215).
Fischer and E. Schneiderloechner, Manufacturing of 300Wp Modules by [52] M. Mü ller, P.P. Altermatt, H. Wagner, G. Fischer, Sensitivity analysis of indus-
Industrialization of PERC Solar Cell Technology Including Crystal, Cell and trial multicrystalline PERC silicon solar cells by means of 3-D device simula-
Module Development, in: Proceedings of the 29th European Photovoltaic tion and metamodeling, IEEE J. Photovolt. 4 (2014) 107–113.
Solar Energy Conference and Exhibition, 2014. [53] P.P. Altermatt, K.R. McIntosh, A roadmap for PERC cell efficiency towards 22%,
[34] Anon www.pvtech.org/news/q_cells_sets_two_new_world_records_for_ focused on technology-related constraints, Energy Procedia 55 (2014) 17–21.
multi_crystalline_and_quasi_mono_sol. [54] G. Fischer, K. Strauch, T. Weber, M. Mü ller, F. Wolny, R. Schiepe, A. Fü lle,
[35] B. Tjahjono, M. J.Yang, V. Wu, J. Ting, J. Shen, O. Tan, T. Sziptalak, B. Liu, H.-P. F. Lottspeich, S. Steckemetz, E. Schneiderloechner, K.-H. Stegemann,
Sperlich, T. Hengst, B. Beilby, K.-C. Hsu, Optimizing Celco cell Technology in H. Neuhaus, Simulation based development of industrial PERC cell production
one year of ass production, in: Proceedings of the 28th European Photovoltaic beyond 20.5% efficiency, Energy Procedia 55 (2014) 425–430.
Solar Energy Conference and Exhibition, 2013. [55] B. Min, H. Wagner, M. Mü ller, D.-H. Neuhausand P.P. Altermatt, Incremental
[36] Jade Jones, PERC ,solar cell technology gaining ground in 2014, GTM Research, efficiency improvements of mass-produced PERC cells up to 24%, predicted
August 14, 2014. solely with continuous development of existing technologies and wafer
[37] Finlay Colville, PERC capacity hits 2.5 GW and offers new technology buy cycle materials, EUPVSEC, in: Proceedings of the 30th European Photovoltaic Solar
option, August 14, 2014. Energy Conference and Exhibition, September 2015.
[38] Anon., Efficiency enhancements to define solar PV technology roadmap for the [56] A. Fell, K.R. McIntosh, P.P. Altermatt, G.J.M. Janssen, R. Stangl, A. Ho-Baillie,
next five years, Solarbuzz, October 14, 2014. 〈http://www.solarbuzz.com/ H. Steinkemper, J. Greulich, M. Mü ller, B. Min, K.C. Fong, M. Hermle, I.
news/recent-findings/efficiency-enhancements-define-solar-pv-technology- G. Romijn, M.D. Abbott, Input parameters for the simulation of silicon solar
roadmap-next-five-years-acco〉. cells in 2014, IEEE J. Photovolt. 5 (2015) 1250–1263.
[39] Anon. International Technology Roadmap for Photovoltaic (ITRPV), Sixth [57] S. Wilking, C. Beck, S. Ebert, A. Herguth, G. Hahn, Influence of bound hydrogen
Edition, April 2015 (www.itrpv.net). states on BO-regeneration kinetics and consequences for high-speed regen-
[40] N. Huang and A. Hwang, First-Tier China solar cell makers expected to have eration processes, Sol. Energy Mater. Sol. Cells 131 (2014) 2–8.
1.0-1.2GWp PERC capacity, DIGITIMES, Feb. 2015. (www.digitimes.com/pda/ [58] D.C. Walter, B. Lim, K. Bothe, R. Falster, V.V. Voronkov, J. Schmidt, Lifetimes
a20150226PD205.html). exceeding 1 ms in 1-Ω cm boron-doped Cz-silicon, Sol. Energy Mater. Sol.
[41] C. Kranz, S. Wyczanowski, U. Baumann, K. Weise, C. Klein, F. Delahaye, Cells, 131, (2014) 51–57.
T. Dullweber, R. Brendel, Wet chemical polishing for industrial type PERC solar [59] N. Nampalli, B. Hallam, C. Chan, M.D. Abbott, S.R. Wenham, Evidence for the
cells, Energy Procedia 38 (2013) 243–249. role of hydrogen in the stabilization of minority carrier lifetime in boron-
[42] E. Cornagliotti, AUruena, J. Horzel, J. John, L. Tous, D. Hendrickx, V. Prajapati, S. doped Czochralski silicon, Appl. Phys. Lett. 106 (2015) 173501.
Singh, R. Hoyer, F. Delahaye, K. Weise, D. Queisser, H. Nussbaumer and J. [60] P. Mints, Priced to sell, PV Magazine, Vol. 6, 2015, 108-109.
Poortmans, How much rear side polishing is required? A study on the impact [61] J. Libal and R. Kopecek, N-type silicon solar cell technology: ready for take off?
of rear side polishing in PERC solar cells, in: Proceedings of the 27th European 17 March 2015 (and comments) . 〈http://www.pvtech.org/guest_blog/n_type_
Photovoltaic Solar Energy Conference and Exhibition, 2012. silicon_solar_cell_technology_ready_for_take_off〉 (accessed 25.06.15).
[43] C-W. Chang, P-T. Hsieh, T. Fang, J-L. Lue and C-H. Wu, The effects of rear-side [62] Z. Xiong, Z. Zhang, H. Ye, S. Fu, P. P. Altermatt, Z. Feng, P.J. Verlinden, High
morphology on PERC cells by alkaline etching, 28th European Photovoltaic Performance Multicrystalline Wafers with Lifetime of 400 μs at Industrial
Solar Energy Conference and Exhibition, 2013. Scale, in: Proceedings of the 42nd IEEE Photovoltaic Specialists Conference,
[44] M. Dhamrin, S. Suzuki, M. Matsubara, Y. Nisho and H. Tada, Ultrasonic map- New Orleans, June 14–19, 2015.
ping of voids generated during the LBSF formation of PERC solar cells, 29th
European Photovoltaic Solar Energy Conference and Exhibition, 2014.
[45] L-C. Cheng, M-C. Kao, H-H. Huang, P-S. Huang and L-W. Cheng, 21% p-type
industrial PERC cells with homogeneous emitter profile and thermally grown