Semiconductors and semimetals, volume 90

SERIES EDITORS
EICKE R. WEBER
Director
Fraunhofer-Institut
f€
ur Solare Energiesysteme ISE
Vorsitzender, Fraunhofer-Allianz Energie
Heidenhofstr. 2, 79110
Freiburg, Germany

CHENNUPATI JAGADISH
Australian Laureate Fellow
and Distinguished Professor
Department of Electronic
Materials Engineering
Research School of Physics
and Engineering
Australian National University
Canberra, ACT 0200
Australia


Academic Press is an imprint of Elsevier
32 Jamestown Road, London NW1 7BY, UK
525 B Street, Suite 1800, San Diego, CA 92101-4495, USA
225 Wyman Street, Waltham, MA 02451, USA
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK
First edition 2014
Copyright © 2014 Elsevier Inc. All rights reserved
No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and

retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publisher’s permissions policies and our
arrangements with organizations such as the Copyright Clearance Center and the Copyright
Licensing Agency, can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by
the Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and
experience broaden our understanding, changes in research methods, professional practices,
or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluating and using any information, methods, compounds, or experiments described
herein. In using such information or methods they should be mindful of their own safety and
the safety of others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of
products liability, negligence or otherwise, or from any use or operation of any methods,
products, instructions, or ideas contained in the material herein.
ISBN: 978-0-12-388417-6
ISSN: 0080-8784
For information on all Academic Press publications
visit our website at store.elsevier.com


CONTRIBUTORS
Christophe Ballif
Photovoltaics and Thin-Film Electronics Laboratory, Institute of Microengineering (IMT),
Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Neuch^atel, Switzerland. (ch2)
Stefaan De Wolf
Photovoltaics and Thin-Film Electronics Laboratory, Institute of Microengineering (IMT),

Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Neuch^atel, Switzerland. (ch2)
Antoine Descoeudres
Photovoltaics and Thin-Film Electronics Laboratory, Institute of Microengineering (IMT),
Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Neuch^atel, Switzerland. (ch2)
Bernhard Dimmler
Manz AG, Reutlingen, Germany. (ch3)
Giso Hahn
Department of Physics, University of Konstanz, Konstanz, Germany. (ch1)
Zachary C. Holman
School of Electrical, Computer, and Energy Engineering, Arizona State University, Tempe,
Arizona, USA. (ch2)
Sebastian Joos
Department of Physics, University of Konstanz, Konstanz, Germany. (ch1)

vii


PREFACE
The rapid transformation of our energy supply system to the efficient use of
renewable energies remains to be one of the biggest challenges of mankind
that increasingly offers exciting business opportunities as well. This truly
global-scale project is well on its way. Harvesting solar energy by photovoltaics (PV) is considered to be a cornerstone technology for this transformation process.
This book presents the third volume in the series “Advances in
Photovoltaics” in Semiconductors and Semimetals. This series has been
designed to provide a thorough overview of the underlying physics, the
important materials aspects, the prevailing and future solar cell design issues,
production technologies, as well as energy system integration and characterization issues. In this volume, three distinctly different solar cell technologies
are covered in detail, ranging from state-of-the-art crystalline silicon technology, the workhorse of the booming PV market, to one of the most
advanced technologies, silicon heterojunction cells, and to an overview of
thin film solar cell technologies. Therefore, this volume represents a cornerstone of “Advances in Photovoltaics,” as the first and the third chapter

together cover more than 98% of the current PV world market volume.
The second chapter provides a glimpse into the future of highly efficient
crystalline Si PV technologies that will allow further decrease in the cost
of PV-generated electricity available from premium modules with top performance produced at prices that will become competitive with present-day
low-cost PV modules. Following the tradition of this series, all chapters are
written by world-leading experts in their respective field.
In the past 2 years, since the introduction to the first volume of this series
has been written, the world PV market has undergone a decisive transformation. Huge production overcapacity, established especially in Asia,
resulted in rapidly declining prices, often to values beyond the production
costs, when fire sales of module supplies were the only way to generate desperately needed cash for financially stressed companies. Subsequently, many
companies went into insolvency, followed by either restructuring under
new ownership, often from abroad, or a complete shutdown of the production lines. The PV equipment manufacturers were especially hard hit, as they
had to survive several years practically without any new orders.

ix


x

Preface

Today we experience a new development: decreasing global production
capacity begins to meet further increasing PV market size, the growth of
which is fueled worldwide by the low cost of solar electricity. The consequence of this process will be the further decentralization of electricity supply, as PV systems increasingly allow owners of homes and industry to
produce electricity on their own roofs and free areas, to the benefit of energy
independence and the world climate, that desperately needs rapid further
market penetration of renewables to decrease the emission of climate gases.
GERHARD P. WILLEKE AND EICKE R. WEBER
Fraunhofer ISE, Freiburg, Germany

CHAPTER ONE

State-of-the-Art Industrial
Crystalline Silicon Solar Cells
Giso Hahn1, Sebastian Joos
Department of Physics, University of Konstanz, Konstanz, Germany
1
Corresponding author: e-mail address:

Contents
1. Introduction
1.1 History
1.2 General routes for cost reduction
1.3 PV market today
1.4 Basic structure of an industrial c-Si solar cell
2. Operation Principle of a c-Si Solar Cell
2.1 Band diagram
2.2 Solar cell parameters
2.3 Fundamental efficiency limit of an ideal c-Si solar cell
2.4 Two-diode model
2.5 Radiative recombination
2.6 Auger recombination
2.7 SRH recombination
2.8 Surface recombination
2.9 Recombination and saturation current density
2.10 Optical losses
3. The Basic Firing Through SiNx:H Process
3.1 Wafer washing, texturization, and cleaning
3.2 Phosphorus diffusion

3.3 Edge isolation
3.4 SiNx:H deposition
3.5 Metallization via screen-printing
3.6 Solar cell characterization
4. Recent Developments on Solar Cell Front Side
4.1 Wafer sawing
4.2 Alkaline wafer texturing
4.3 Front contact metallization
5. Advanced Emitter Formation
5.1 Improvement of homogeneous emitters
5.2 Selective emitters
6. Industrial PERC-Type Solar Cells
6.1 Dielectric rear side passivation

Semiconductors and Semimetals, Volume 90
ISSN 0080-8784
/>
#

4
4
5
7
9
10
10
12
13
14
14

15
16
17
18
18
19
20
22
25
25
27
33
34
34
35
35
40
41
42
51
52

2014 Elsevier Inc.
All rights reserved.

1


2

Giso Hahn and Sebastian Joos

6.2 Formation of local rear contacts
6.3 Boron–oxygen related degradation
6.4 State-of-the-art industrial PERC solar cells
7. Summary and Outlook
Acknowledgments
References

ABBREVIATIONS
A area
ALD atomic layer deposition
APCVD atmospheric pressure chemical vapor deposition
ARC antireflective coating
a-Si amorphous silicon
BSF back surface field
Bs substitutional boron concentration
cA,n (cA,p) Auger recombination coefficient for electrons (holes)
crad radiative recombination coefficient
c-Si crystalline silicon
Cz Czochralski
d layer/wafer thickness dBSF
D+ diffusion constant in the BSF
DI deionized
Dn (Dp) diffusion constant of electrons (holes)
E energy
ECV electrochemical capacitance voltage
EF (EFi) (intrinsic) Fermi energy level
EFG edge-defined film-fed growth
EFn (EFp) quasi-Fermi energy level of electrons (holes)

Eg band gap energy
Ephot photon energy
EQE external quantum efficiency
Et energetic position of the trap level
EVA ethylene vinyl acetate
FCA free carrier absorption
FF fill factor
FZ float zone
h Planck’s constant
HIT heterojunction with intrinsic thin-layer
I current
IBC interdigitated back contact
IPA isopropyl alcohol
IQE internal quantum efficiency
j current density
j0 saturation current density
j01 ( j02) saturation current density of the first (second) diode
j0e saturation current density of the emitter

54
57
59
60
62
62


State-of-the-Art Industrial Crystalline Silicon Solar Cells

jl light-generated current density

jsc short circuit current density
k Boltzmann’s constant
L+ diffusion length in the BSF
LFC laser fired contacts
Ln (Lp) diffusion length of electrons (holes)
LPCVD low pressure chemical vapor deposition
mono-Si monocrystalline Si
mpp maximum power point
mc-Si multicrystalline Si
n electron concentration
n+ (n++) (very) highly n-doped
n0 electron concentration in the dark
NA (ND) acceptor (donor) concentration
NA+ acceptor concentration in the BSF
nair (nSi, nSiN) refractive index of air (c-Si, SiN)
ni intrinsic carrier concentration
Nt trap density
Nts areal trap density at the surface
Oi interstitial oxygen
p hole concentration
p+ highly p-doped
p0 hole concentration in the dark
PECVD plasma-enhanced chemical vapor deposition
PERC passivated emitter and rear cell
PERL passivated emitter and rear locally diffused
PERT passivated emitter and rear totally diffused
pphot photon power density
PSG phosphor silicate glass
Psurf phosphorous surface concentration
Ptot total power loss

PV photovoltaic
q elementary charge
R recombination rate
RA Auger recombination rate
Rrad radiative recombination rate
Rs series resistance
Rs,tot total series resistance
RSRH Shockley-Read-Hall recombination rate
Rsh shunt resistance
Rsheet sheet resistance of the emitter
s (sn) (sp) surface recombination velocity (of electrons or holes)
sb surface recombination velocity at the backside
SCR space charge region
seff effective surface recombination
SIMS secondary ion mass spectrometry
SRH Shockley-Read-Hall
STC standard test conditions (1000 W/m2, AM1.5g spectrum, 25  C)
UMG upgraded metallurgical grade

3


4

Giso Hahn and Sebastian Joos

V voltage
vn (vp) thermal velocity of electrons (holes)
Voc open circuit voltage
Wp Watt peak (power of 1 W under STC)

α absorption coefficient
ΔEF splitting of quasi-Fermi levels
Δn excess charge carrier density
η conversion efficiency
Φ photon flux
λ wavelength
ρSi density of Si
ρ resistivity
σ n (σ p) capture cross section for electrons (holes)
τ A Auger lifetime
τ b bulk lifetime
τ eff effective lifetime
τ rad radiative lifetime
τ SRH Shockley, Read, Hall lifetime
τ minority charge carrier lifetime

1. INTRODUCTION
Solar cells fabricated based on crystalline Si (c-Si) generate electricity
from sunlight by absorbing photons and generating electron–hole pairs,
which are separated by a pn-junction. The pn-junction creates an electric
field in the semiconductor and the separated charge carriers have to leave
the solar cell via electrical contacts to perform work in an external circuit.
A solar cell in operation is therefore essentially an illuminated large area
diode, where emitter and base regions are contacted by metals to extract
the carriers.

1.1. History
The first c-Si solar cell operating using the principle described above was
reported in 1953 (Chapin et al., 1954), although research toward this
achievement dates back to the 1940s (e.g., Ohl, 1941; Shockley, 1950).

In the decades to follow, research was first directed toward application of
the photovoltaic (PV) effect in space (powering satellites) or for terrestrial
stand-alone systems. As for those applications the total cost of power generation was not the main issue, research was mainly driven by improving
the conversion efficiency η, which is the ratio between output power from
the PV device (generated from the solar cell or complete solar module) and


State-of-the-Art Industrial Crystalline Silicon Solar Cells

5

input power (impinging photon flux). The oil crisis in 1973 led to considerations to use PV also for terrestrial applications in larger scale as an alternative to fossil fuels. Since then a lot of R&D activities was focused on
reducing the cost of PV electricity generation to make it attractive for market penetration.
In research, a lot of progress was made in improving efficiency by developing new cell designs and applying novel processing steps, leading to efficiencies as high as 25% using standard test conditions (STC: 1000 W/m2
illumination, AM1.5g spectrum, 25  C) in 1999 (Zhao et al., 1999), indicating the efficiency potential of c-Si. This efficiency was reached on
extremely pure float zone (FZ) silicon and on small scale (4 cm2) without
the main part of the front side metallization grid being taken into account
for the efficiency measurement (so-called designated area measurement)
and using a very complex processing scheme. For most industrial applications, a full area measurement and cost-effective c-Si materials are of higher
interest. In addition, the number and complexity of processing steps needed
for cell fabrication has to be low, to allow a cost-efficient production. Here,
the main challenge for industrial c-Si solar cells becomes visible: there is a
trade-off between more complex processing on higher quality material allowing higher efficiencies, and less complex processing, e.g., in combination
with a lower c-Si material quality.

1.2. General routes for cost reduction
The lower efficiency for lower cost materials and less complex processing
might be advantageous cost-wise at cell level, but as there are also area
related cost factors at module and system level (e.g., costs for module glass
and installation), the question which route is more promising is not easy to

answer. Therefore, a lot of different technologies have been developed over
the past decades. This includes c-Si materials as well as solar cell fabrication
processes.
The Si feedstock of highest quality stems from the so-called Siemens
route using rods for Si production from the gas phase, which still accounts
for the majority of produced Si wafers for industrial solar cells, with fluidized
bed reactors as an alternative (Fabry and Hesse, 2012). So-called upgraded
metallurgical grade (UMG) Si can be produced with significantly less energy
needed per kg of fabricated Si, but a higher impurity concentration is the
consequence, with relatively high amounts of, amongst others, B and
P still present acting as doping elements in Si. This might cause problems
as after crystallization the material will be partly compensated, and due to


6

Giso Hahn and Sebastian Joos

different segregation coefficients of B and P their concentrations and therefore resistivity, influenced by the net doping, changes with ingot height
(Ceccaroli and Pizzini, 2012; Heuer, 2013).
For c-Si materials, three different material classes have been important for
PV in the past, as they have already been in industrial production in significant quantities. Monocrystalline Si (mono-Si) pulled using the Czochralski
(Cz) method shows the lowest amount of extended crystal defects (like, e.g.,
grain boundaries, dislocations, precipitates), but normally contains a high
amount of O, mainly in interstitial form (Oi) (Zulehner, 1983). Cast multicrystalline Si (mc-Si) can be produced in a more cost-effective way, but
contains due to the crystallization method used a higher amount of extended
crystal defects and impurities in interstitial or precipitated form, originating
mainly from the crucible wall and the crucible coating (Buonassisi et al.,
2006; Schubert et al., 2013). See Coletti et al. (2012) for an overview on
the role of impurities in c-Si for solar cells. For both methods, the crystallized

ingot has to be sliced in wafers for subsequent solar cell processing. To avoid
kerf and other Si material losses that easily amount to >50%, ribbon-Si techniques have been developed, crystallizing the Si wafer directly from the Si
melt (Hahn and Sch€
onecker, 2004). Of the three technology groups, ribbon
Si is the most cost-effective technique to produce wafers, but these wafers
normally show the highest defect densities, reducing the electronic quality of
the as-grown wafer.
Apart from Si wafer quality, solar cell process complexity is the other
main parameter determining the efficiency and cost structure of the solar
cell. In this contribution, focus is laid on industrial solar cell production,
but for a more complete picture also PV module and system aspects should
be considered. The heart of a solar module and every PV system is the solar
cell. The cells are stringed in series so that the same amount of current flows
through all cells in a string and the voltages of the cells add up. This makes
proper sorting of cells a necessity to ensure that cells of similar performance
end up in a string, as the cell with the lowest current at operation conditions
determines the current flowing through the string. Therefore, for all cells
not only the peak efficiency, but also a tight distribution of cell parameters
is important to facilitate sorting and matching of the cells. This means that in
industrial fabrication homogeneous Si wafer quality and stable processes
with large process windows are desired to minimize the spread of quality
in c-Si solar cell production.
In this chapter, an overview on industrial state-of-the-art c-Si solar cells
is given. As there is not only one industrial solar cell process, but a variety of
different processes applied for different cell designs, we will restrict the


7

State-of-the-Art Industrial Crystalline Silicon Solar Cells

overview on the most common cell architectures. Other cell designs already
used in industrial scale such as the interdigitated back contact (IBC), commercialized by company SunPower Corp. (Cousins et al., 2010), or the heterojunction with intrinsic thin-layer (HIT) concept pioneered by Sanyo
(now Panasonic) (Ballif et al., 2014) allow for the highest efficiencies in commercial c-Si solar cells on large area cells with lab cell record efficiencies up
to 25% on large area cells (Smith et al., 2014; Taguchi et al., 2013) and even
25.6% with a combined IBC-HIT approach (Panasonic, 2014), but the processes differ significantly from mainstream technology. Therefore, these
designs of very highly efficient c-Si solar cells will be treated in other chapters (e.g., Ballif et al., 2014).

1.3. PV market today
Figure 1.1 demonstrates the very dynamic growth of commercial PV over
the past decades, spanning more than four decades from around 1 MWp1 in
the early 1970s to >30 GWp in 2011. Annual growth rates over the past
10 years have been in the order of 50%, mainly driven by market stimulation
programs like, e.g., the renewable energy law with a guaranteed feed-in tariff in Germany. As the German feed-in tariffs have been adjusted recently
and the German PV market was the strongest worldwide, the growth slowed
down in 2012 and 2013. Strong growth in recent years allowed for a tremendous reduction in production cost due to scaling effects in mass production

PV-module power (MWp)

10,000

1000

100

10

1
1975

1980

1985

1990

1995

2000

2005

2010

Figure 1.1 Yearly production/shipment of solar modules. Data from PV News, Photon,
and Mehta (2014).
1

Watt peak (Wp) refers to the power generated under STC.


8

Giso Hahn and Sebastian Joos

as well as new and optimized processing technologies. This so-called
learning curve effect of PV resulted in an average module price reduction
of around 20% for every doubling of cumulated PV production (Nemet
and Husmann, 2012). The continuing reduction in processing costs
results in costs of a kWh generated by PV being now in the range of electricity generated from fossil fuels (depending on the installation site) (Kost

et al., 2013).
The market share of different PV technologies shown in Fig. 1.2 reveals
that c-Si still shows by far the highest market penetration, with thin film
technologies like amorphous Si (a-Si), CdTe and CuInxGa(1Àx)Se2
(CIGS) not really gaining market share above a 10–15% level. In contrast,
latest figures indicate an even further increasing market share for c-Si of
90% in 2013, with roughly 67% based on mc-Si and 23% on mono-Si
(Mehta, 2014). It is interesting to note that mono-Si lost market share to
mc-Si in the past decade. This can be explained by the huge production
expansion programs happening at most PV manufacturers in the past, as
mc-Si technology seems to be easier to ramp up and was the more costeffective way of production in the past. Whether this will hold true in
the future, with new cell designs allowing for higher efficiency approaching
the market, remains to be seen. The market share of ribbon-Si dropped to
almost zero as the two main technologies edge-defined film-fed growth
(EFG) and string ribbon are no longer on the market, due to the disappearing
of their production companies Schott Solar and Evergreen Solar as well as
EverQ, respectively.
100
90
Others
CIGS
CdTe
a-Si
Ribbon-Si
Multi-Si
Mono-Si

80
Technology(%)

70
60
50
40
30
20
10

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

0

Figure 1.2 Market share of different PV technologies. Data from PV News and Photon.


9

State-of-the-Art Industrial Crystalline Silicon Solar Cells

1.4. Basic structure of an industrial c-Si solar cell
A schematic of the basic structure for a typical state-of-the-art industrial c-Si
solar cell is shown in Fig. 1.3. The base is p-type material, moderately
B doped to a resistivity of around 1 Ω cm (NA ¼ 1.5Á1016 per cm3). The
emitter is n++-doped2 using P with high surface concentration ND > 1020
per cm3, and the front surface is textured to allow a better incoupling of
impinging photons (lower reflectivity). The emitter is covered by a thin
dielectric layer of H-rich silicon nitride (SiNx:H), acting as antireflective
coating (ARC), surface passivation layer, and reservoir of H. On the front,
the metallization finger grid is realized by Ag paste, fired through the SiNx:H
layer at high temperature. On the rear, a full area contact is realized by Al
paste, which forms an alloy with Si during the firing step, resulting in an
Al doped p+-region (around 1019 per cm3) at the rear after cool down to

room temperature (back surface field, BSF). To allow interconnection of
the individual cells for module integration using soldering, stripes or pads
of Ag/Al paste are used at the rear side, as Al is not solderable. The complete
cell thickness is around 180 μm (note that features shown in Fig. 1.3 are not
to scale). The formation of the respective regions of the cell will be dealt
with in more detail in the following sections.
The use of H-rich SiNx:H layers for PV (Morita et al., 1982) in the
so-called “firing through SiNx:H process” has been pioneered by Kyocera
(Kimura, 1984; Takayama et al., 1990) and Mobile Solar for their EFG
ribbon-Si material (Cube and Hanoka, 2005). In the 1990s, other companies
and research institutes like, e.g., IMEC (Szlufcik et al., 1994) and others developed the process further. The breakdown of costs for c-Si module production
in Fig. 1.4 reveals that wafer and module costs are the dominating factors.

hν
Ag

SiNx:H

n+
p-Si

Electron

p+

Hole

Al

Figure 1.3 Schematic basic structure of an industrial c-Si solar cell in cross section (not

to scale).
2

The superscripts + and ++ indicate a high and a very high doping concentration, respectively.


10

Giso Hahn and Sebastian Joos

Wafer
Cell production
Module
26%

36%

38%

Figure 1.4 Breakdown of c-Si PV module manufacturing costs. Data from Goodrich et al.
(2013).

Excellent early (e.g., Szlufcik et al., 1997) and more recent (e.g., Gabor,
2012; Neuhaus and Mu¨nzer, 2007) review papers on low-cost industrial c-Si
solar cell fabrication exist, forming the base of this chapter. Since then new
technologies have emerged, allowing for a reduction of costs as well as efficiency losses and therefore an increase of efficiency in mass production. To
tackle these losses, the next section will describe the physics involved in the
operation principle of a solar cell.

2. OPERATION PRINCIPLE OF A c-SI SOLAR CELL

2.1. Band diagram
The fundamental operation principle of a c-Si solar cell is visualized in the
band diagram shown in Fig. 1.5. The doping gradient due to the abrupt
change in doping concentration at the pn-junction results in electrons (free
majority carriers in the n-region) diffusing from the n-region into the
p-region and holes (free majority carriers in the p-region) diffusing into
the n-region. The remaining ionized doping atoms at lattice sites (positively
charged in the n-region, negatively charged in the p-region) form the space
charge region (SCR) extending into both sides of the pn-junction. The
electric field hinders the free carriers to completely diffuse into the regions
of opposite doping, when equilibrium between diffusion and drift current of
free carriers is reached. The built-up electric field causes bending of the
energy bands, with the Fermi energy EF as defined by the Fermi–Dirac function at a constant level (a horizontal line) in both regions.
Upon illumination, absorbed photons excite electrons from the valence
band to the conduction band via the internal photoelectric effect.


11

State-of-the-Art Industrial Crystalline Silicon Solar Cells

Energy

E
Conduction band

Electron

EF

hν

EF
Valence band
Metal

p-type Si

Hole
SCR

n-type Si

Metal

Figure 1.5 Schematic band diagram of a c-Si solar cell with pn-junction, space charge
region (SCR), photon absorption, charge carrier generation, and separation. Quasi-Fermi
levels and EF in the metal contacts are indicated as well.

Absorption of one photon therefore generates an electron–hole pair, as the
missing electron in the valence band is referred to as a hole. Free electrons
and holes can diffuse until they recombine or reach the SCR. Here, charge
carriers of different types are separated, electrons are accelerated into the
n-region, holes into the p-region. In case of illumination, the semiconductor
is not in thermal equilibrium anymore, and the relation for electron and hole
concentrations n0 and p0, respectively, as defined for thermal equilibrium
(without illumination or applied voltage)
n0 p0 ¼ n2i ,

(1.1)

(with intrinsic carrier concentration ni) is not valid anymore and becomes

np ¼ n2i exp


EFn À EFp
> n2i ,
kT

(1.2)

with n and p being electron and hole concentrations, respectively. As both
electron and hole concentrations are increased when the semiconductor is
illuminated, two separate Fermi–Dirac functions for each carrier type have
to be defined, with two resulting Fermi levels EFn and EFp referred to as
quasi-Fermi levels of electrons and holes.
Metal contacts with EF at roughly the same energetic position as for the
majority carriers in the contacted Si region can extract carriers from both
regions. The contact for the p-type region as depicted in Fig. 1.5 is ohmic,
whereas the n-type contact is of Schottky-type (energy barrier for electrons).


12

Giso Hahn and Sebastian Joos

The barrier can be overcome via tunneling, provided it is thin enough and
not too high.

2.2. Solar cell parameters
An ideal solar cell can be described by a 1-diode model and the j–V characteristic of an illuminated diode
 
!
qV
j ¼ j0 exp
(1.3)
À 1 À jl ,
kT
with current density j, saturation current density j0, elementary charge q,
Boltzmann’s constant k, and light-generated current density jl. j0 is defined as
j0 ¼

qDn n2i qDp n2i
+
,
Ln NA Lp ND

(1.4)

Current density/power density

wer

t po

pu
Out

Vmpp

Dark cu
r ve
Illuminate
d curve

with Dn (Dp) the diffusion constant of electrons (holes), NA (ND) the doping
density of acceptors (donors) and Ln (Lp) the minority charge carrier diffusion length of electrons (holes).
The resulting j–V curve is shown in Fig. 1.6. The maximum current
density at V ¼ 0 is the short circuit current density jjscj ¼ jl. The point of
maximum power density (mpp) is also indicated, with the fill factor FF
defined as

Voc
Voltage

Open-circuit voltage

Maximum power
point (MPP)
jmpp
jsc

Short circuit current density

Figure 1.6 Dark and illuminated j–V curve of a solar cell as well as output power in
dependence of voltage.


13

State-of-the-Art Industrial Crystalline Silicon Solar Cells

FF ¼

jmpp Vmpp
,
jsc Voc

(1.5)

resulting with the impinging photon power density pphot of photons with
energy Ephot in the efficiency
η¼

jsc Voc FF
:
pphot

(1.6)

2.3. Fundamental efficiency limit of an ideal c-Si solar cell

Energy

In a semiconductor with band gap Eg (1.12 eV at 25  C for c-Si), photons
with energy E > Eg can be absorbed, creating electron–hole pairs, while
photons with E < Eg cannot be absorbed and are transmitted, see Fig. 1.7.
Generation of electron–hole pairs by illumination is a non-equilibrium process with some of the carriers occupying states high in the conduction band
(electrons) and deep in the valence band (holes) directly after generation

depending on the photon energy. The generated electrons and holes reach
thermal equilibrium via collisions with other charge carriers or phonons
within the femtosecond (fs) range (thermalization). Afterward, they occupy
states close to the band edges according to Fermi–Dirac statistics. The maximum voltage reachable (open circuit voltage Voc of the cell) is limited by
splitting of the quasi-Fermi levels for electrons and holes ΔEF, with

hν

hν

1.

2.

ΔEF

qVmpp

3.

4.

Figure 1.7 Fundamental loss mechanisms for an ideal pn-junction based solar cell. 1.
Transmission Ephot < Eg, 2. Thermalization Ephot > Eg, 3. Quasi-Fermi level splitting
ΔEF < Eg, and 4. Voltage at mpp Vmpp < Voc.


14

Giso Hahn and Sebastian Joos

ΔEF < Eg. As the maximum power point (mpp) of the illuminated j–V curve
(Fig. 1.6) is between V ¼ 0 (maximum j ¼ jsc) and j ¼ 0 (maximum V ¼ Voc),
Vmpp is always < Voc. These four fundamental loss mechanisms limit the
maximum efficiency of an ideal c-Si solar cell under STC to 29.4%
(Richter et al., 2013).

2.4. Two-diode model
A real solar cell can be described by an equivalent circuit containing two
diodes, with the addition of series resistance Rs, shunt resistance Rsh and a
second diode accounting for recombination in the SCR with an ideality factor generally assumed to be 2 (Fig. 1.8).

 !

 !
qðV À jRs Þ
qðV À jRs Þ
ðV À jRs Þ
j ¼ j01 exp
À jl :
À1 + j02 exp
À1 +
kT
2kT
Rsh
(1.7)
Contributions to Rs are ohmic resistive losses in emitter, base, and metallization as well as the contact resistance between semiconductor and metal.
Finite Rsh values are caused by alternative current paths short circuiting the
diode (e.g., around the cell’s edge, by a damaged emitter or current paths
through the SCR).

Apart from ohmic losses, recombination of generated charge carriers can
occur, limiting performance of the solar cell.

2.5. Radiative recombination
Radiative recombination refers to direct band-to-band transitions of an electron from the conduction band to the valence band while emitting a photon.
It is the inverse process of photon absorption. The generated excess charge
carrier density Δn with
n ¼ n0 + Δn and p ¼ p0 + Δn

j01

(1.8)

RS

j02

jI
RSh

Figure 1.8 Equivalent circuit of a real pn-junction solar cell.


15

State-of-the-Art Industrial Crystalline Silicon Solar Cells

can be reduced due to recombination of charge carriers with a recombination rate R defining the lifetime τ of excess charge carriers
τ¼

Δn
:
R

(1.9)

c-Si is an indirect band gap semiconductor. In addition to an electron (in the
conduction band) and a hole (in the valence band), a phonon is necessary for
the band-to-band transition to occur due to conservation of momentum.
Therefore, this mechanism is not probable and can normally be neglected
in c-Si. With the radiative recombination coefficient crad, the net rate Rrad
for this type of recombination becomes3
À
Á
(1.10)
Rrad ¼ crad np À n2i ,
resulting for low injection (Δn much lower than doping concentration4) in
the radiative lifetime
τrad¼

1
crad p0

(1.11)

for p-doped material.

2.6. Auger recombination
Instead of creating a photon, the energy of the recombination process can be
used to excite another existing free charge carrier (an electron in the conduction band or a hole in the valence band). This charge carrier thermalizes

after excitation toward the band edge, converting the recombination energy
into phonons. With the Auger recombination coefficients cA,n and cA,p for
electrons and holes, respectively, the Auger recombination rate reads
À
Á
À
Á
RA ¼ cA, n n np À n2i + cA, p p np À n2i :
(1.12)
As above, for low injection we obtain the Auger lifetime for p-doped
material
τA ¼
3

4

1
cA, p p20

:

(1.13)

Note that we are only interested in the recombination rate of the excess charge carriers (therefore
np À n2i , subtracting recombination occurring also in thermal equilibrium).
At room temperature, all dopants are assumed to be ionized (NA ¼ p0 in p-type material), and therefore
Δn ( p0 for low injection.


16

Giso Hahn and Sebastian Joos

Auger recombination as a three-particle process is only relevant for high
doping concentrations >1017 per cm3 in standard industrial solar cells.

2.7. SRH recombination
Energy levels in the band gap can trap free charge carriers and cause a very
effective recombination mechanism, especially when their energetic position is close to mid-gap. This type of recombination was formulated by
Shockley, Read, and Hall (Hall, 1952; Shockley and Read, 1952), using statistics of capture and emission of free carriers and is therefore referred to as
SRH recombination. Its recombination rate
À
Á
np À n2i
RSRH ¼
(1.14)
τp ðn0 + n1 + ΔnÞ + τn ðp0 + p1 + ΔnÞ
with




1
1
Et À EFi
EFi À Et
τp ¼
, τn ¼
, n1 ¼ ni exp
, p1 ¼ ni exp

,
kT
kT
Nt vp σp
Nt vn σn
(1.15)
includes the trap density Nt of the energy levels in the band gap, the thermal
velocity of electrons and holes (vn, vp) and the capture cross sections of the
trap for electrons and holes (σ n, σ p). Et is the energetic position of the trap
level and EFi the position of the Fermi level in intrinsic c-Si. The SRH
lifetime
τSRH ¼

τp ðn0 + n1 + ΔnÞ + τn ðp0 + p1 + ΔnÞ
p0 + n0 + Δn

(1.16)

for p-type material (p0 ) n0), low injection (p0 ) Δn), and trap energy level
at mid-gap (Et ¼ EFi) reads
τSRH ¼ τn ¼

1
Nt vn σn

(1.17)

and is inversely proportional to the trap density as well as the thermal velocity and capture cross section of the minority carriers (electrons in p-type
material).
All recombination channels are acting in parallel, and the resulting bulk

lifetime τb is given by


State-of-the-Art Industrial Crystalline Silicon Solar Cells

1
1
1
1
¼
+
+
:
τb τrad τA τSRH

17

(1.18)

2.8. Surface recombination
At the crystal surface, dangling bonds5 are responsible for a multitude of
defect levels distributed throughout the band gap. In analogy to the SRH
recombination formalism in the bulk of the crystal, a lifetime of the charge
carriers at the physical surface can be derived using areal instead of volume
densities of charge carriers and traps. For p-type material in low injection,
this results in
sn ¼ Nts vn σn ,

(1.19)

with the areal density of traps at the surface Nts, and sn being referred to as the
surface recombination velocity s of electrons (minority carriers in p-type
material) in units of cm/s.
The influence of surface recombination on the observable effective lifetime can be expressed by a surface lifetime τs (Aberle, 1999)
1
1
1 1
¼ + ¼ + α 2 Dn ,
τeff τb τs τb

(1.20)

with α a solution of the transcendental equation (wafer thickness d)
tan

αd
s
,
¼
2 αDn

(1.21)

which can be approximated with (Sinton and Cuevas, 1996)
τs %

d
d2
:
+

2s Dn π2

(1.22)

For reasonably good surface passivation with s <1000 cm/s, the second
term can be neglected and
1
1 2s
¼ + :
τeff τb d
5

(1.23)

Dangling bonds are generally reconstructed bonds where the lengths and angles differ from their standard values in the c-Si bulk.


18

Giso Hahn and Sebastian Joos

2.9. Recombination and saturation current density
Recombination reduces the maximum current density jsc of the solar cell, as
only minority charge carriers generated within roughly one diffusion length
on either side of the pn-junction reach the junction and are injected into the
region on the opposite side of the junction. But from Eq. (1.3) also strong
influence of j0 on Voc can be seen, as for j ¼ 0
 



kT
jl
kT
jl
Voc ¼
:
+1 %
ln
ln
j0
j0
q
q

(1.24)

As the diffusion lengths of both types of carriers in Eq. (1.4) are linked to
recombination via the lifetime of the minority charge carriers
Ln, p ¼

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Dn, p τeff ,

(1.25)

maximizing the effective lifetimes in emitter and base is crucial for improving solar cell performance. Effective lifetime is affected by bulk lifetime and
surface recombination velocity (Eq. 1.23), therefore good solar cells should
combine a high τb (low recombination in bulk and emitter) and good surface
passivation on emitter and base to reduce s.

2.10. Optical losses
If all impinging photons with Ephot > Eg were absorbed in the solar cell, with
all of these photons contributing to the extracted current density, the maximum jsc would be around 44 mA/cm2 under STC. Apart from recombination losses described above, another fraction is lost due to optical losses.
These losses include reflection at the front side (metal grid and ARC),
absorption in the metal and ARC, absorption via free carrier absorption
(FCA)6 and photons not being absorbed in c-Si (mostly long wavelengths
photons7) leaving the cell.
The different loss mechanisms are visualized in Fig. 1.9, where they are
separated into optical and electrical losses.
6

7

Free carrier absorption is the absorption of a photon by an electron in the conduction band or a hole in
the valence band without generation of additional free carriers. It is important in highly doped areas
(emitter and BSF).
The absorption coefficient in c-Si with indirect bandgap leads to an absorption coefficient strongly
varying with wavelength, leading for photons with wavelengths >1000 nm to absorptions lengths
>200 μm.


19

State-of-the-Art Industrial Crystalline Silicon Solar Cells

Shadowing loss
(total reflection on metal)

Incident
photon flux F

solar spectrum
AM1.5g

ARC reflection loss
(mainly short wavelengths)
Back reflection
(mainly long wavelengths)

Ag
SiNx:H

n+

Carrier loss emitter & SCR

ARC absorption loss
(mainly short wavelengths)

p-Si

Free carrier
absorption

Free carrier
absorption

Carrier loss bulk

Final carrier
flow jsc/q

Carrier loss BSF

BSF

Al

Rear absorption loss

Figure 1.9 Visualization of the conversion of photon flux into carrier flow in a standard
industrial p-type Si solar cell with the optical and electrical losses as indicated.

3. THE BASIC FIRING THROUGH SiNx:H PROCESS
As already mentioned in the introduction, most industrial solar cells
today are fabricated based on a so-called “firing through SiNx:H” process
(Fig. 1.3). Therefore, in this section we will describe this process in its basic
form as it was developed in more detail (compare with, e.g., Neuhaus and
Mu¨nzer, 2007; Szlufcik et al., 1997), before alternatives and improvements
will be dealt with in the next sections.
Generally, for every process step there are two options, inline or batch
processing. Inline processing offers the possibility to fabricate solar cells with
a minimum of handling steps and a smaller footprint due to the lack of storage room necessary for partially processed cells. On the other hand, not all
processing steps can easily be performed inline and batch processing allows
for more freedom in optimization. The first example of a complete true
inline processing fabrication of solar cells was RWE Schott Solar’s
SmartSolarFab in 2002. Nowadays, cell processing is normally done by a
mixture of inline and batch processing equipment, as the throughput of
machines used for the different steps is not the same. In addition, if single
machines are not operational or have to be maintained, not the complete
production is halted, but other parts within cell fabrication can continue

to produce. Therefore, often several machines of the same type work in parallel to increase throughput and minimize the risk of bottlenecks.


20

Giso Hahn and Sebastian Joos

3.1. Wafer washing, texturization, and cleaning
After crystallization, mono-Si and mc-Si wafers are sliced out of the Si ingot
using wire saws, containing slurry with abrasives for cutting into the Si
(Dold, 2014). This leaves, apart from contaminants, saw damage on both
sides of the Si wafer with a depth in the range of up to 10 μm (depending
on sawing conditions). After wafer washing, this saw damage has to be
removed, as the disturbed region of the crystal (cracks, dislocations) is of
poor electronic quality.
For mono-Si, this is done in an alkaline wet chemical solution of KOH
and isopropyl alcohol (IPA) at temperatures of around 80  C. The KOH
solution etches the Si while the alcohol masks the surface randomly. Etching
is anisotropic, with the result that the most densely packed crystal planes in
c-Si have the slowest etch rate (the (111)-planes). If the wafer is (100)oriented, the four (111) orientations in the diamond lattice of c-Si will randomly form square-based upright pyramids (Fig. 1.10). These pyramids very
effectively reduce the reflectivity of the surface and therefore increase the
incoupling of photons into c-Si. The etching reaction can be summarized as
Si + 2H2 O + HOÀ ! HSiO3À + 2H2

(1.26)

and consists of oxidation of Si, formation of a solvable salt and dissolving the
salt in water (Neuhaus and Mu¨nzer, 2007).
The surface is increased after random pyramid texturing by a factor of
$1.7, which has consequences for surface passivation and saturation current

densities of the emitter and the SCR.
mc-Si does not offer a well-defined grain orientation at the wafer surface,
as the grains are randomly distributed. Therefore, other texturing solutions
had to be developed. Standard is an acidic solution based on HF and HNO3
without further additives (Einhaus et al., 1997; Hauser et al., 2003). The texture attacks the Si surface first at areas where not all Si bonds are perfectly
saturated. Therefore, the saw damage is needed for a non-uniform attack
of the surface. Existing surface defects like cracks are widened and a
“worm-like” structure is formed (Fig. 1.10). Once the saw damage is etched
away, the textured surface starts to flatten again for prolonged processing
times, as sharp edges are rounded. Four to five micrometer removal of Si
per side is normally enough to remove the saw damage and obtain a low
reflectivity.8 The etching reaction takes place in two steps, an oxidation
8

Note that the maximum depth of saw damage can be up to around 10 μm, but as predominantly the
damaged areas are attacked, less overall removal of Si is needed.