Semiconductors and semimetals, volume 90

6.2 Formation of local rear contacts 54ABBREVIATIONS A area ALD atomic layer deposition APCVD atmospheric pressure chemical vapor deposition ARC antireflective coating a-Si amorphous sil

SERIES EDITORS

EICKE R WEBER

Director

Fraunhofer-Institut

f€ur Solare Energiesysteme ISE

Vorsitzender, Fraunhofer-Allianz Energie Heidenhofstr 2, 79110

Freiburg, Germany

CHENNUPATI JAGADISHAustralian Laureate Fellow

and Distinguished Professor

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

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)

The rapid transformation of our energy supply system to the efficient use ofrenewable energies remains to be one of the biggest challenges of mankindthat increasingly offers exciting business opportunities as well This trulyglobal-scale project is well on its way Harvesting solar energy by photovol-taics (PV) is considered to be a cornerstone technology for this transforma-tion process.

This book presents the third volume in the series “Advances inPhotovoltaics” in Semiconductors and Semimetals This series has beendesigned to provide a thorough overview of the underlying physics, theimportant materials aspects, the prevailing and future solar cell design issues,production technologies, as well as energy system integration and character-ization issues In this volume, three distinctly different solar cell technologiesare covered in detail, ranging from state-of-the-art crystalline silicon tech-nology, the workhorse of the booming PV market, to one of the mostadvanced technologies, silicon heterojunction cells, and to an overview ofthin film solar cell technologies Therefore, this volume represents a corner-stone of “Advances in Photovoltaics,” as the first and the third chaptertogether cover more than 98% of the current PV world market volume.The second chapter provides a glimpse into the future of highly efficientcrystalline Si PV technologies that will allow further decrease in the cost

of PV-generated electricity available from premium modules with top formance produced at prices that will become competitive with present-daylow-cost PV modules Following the tradition of this series, all chapters arewritten by world-leading experts in their respective field

per-In the past 2 years, since the introduction to the first volume of this serieshas been written, the world PV market has undergone a decisive transfor-mation Huge production overcapacity, established especially in Asia,resulted in rapidly declining prices, often to values beyond the productioncosts, when fire sales of module supplies were the only way to generate des-perately needed cash for financially stressed companies Subsequently, manycompanies went into insolvency, followed by either restructuring undernew ownership, often from abroad, or a complete shutdown of the produc-tion lines The PV equipment manufacturers were especially hard hit, as theyhad to survive several years practically without any new orders

ix

Today we experience a new development: decreasing global productioncapacity begins to meet further increasing PV market size, the growth ofwhich is fueled worldwide by the low cost of solar electricity The conse-quence of this process will be the further decentralization of electricity sup-ply, as PV systems increasingly allow owners of homes and industry toproduce electricity on their own roofs and free areas, to the benefit of energyindependence and the world climate, that desperately needs rapid furthermarket penetration of renewables to decrease the emission of climate gases.

GERHARDP WILLEKE ANDEICKER WEBER

Fraunhofer ISE, Freiburg, Germany

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: giso.hahn@uni-konstanz.de

2.3 Fundamental efficiency limit of an ideal c-Si solar cell 13

Semiconductors and Semimetals, Volume 90 # 2014 Elsevier Inc.

6.2 Formation of local rear contacts 54

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

c A,n ( c A,p ) Auger recombination coefficient for electrons (holes)

c rad radiative recombination coefficient

c-Si crystalline silicon

ECV electrochemical capacitance voltage

E F ( E Fi ) (intrinsic) Fermi energy level

EFG edge-defined film-fed growth

E Fn (E Fp ) quasi-Fermi energy level of electrons (holes)

E g band gap energy

E phot photon energy

EQE external quantum efficiency

E t energetic position of the trap level

EVA ethylene vinyl acetate

FCA free carrier absorption

IBC interdigitated back contact

IPA isopropyl alcohol

IQE internal quantum efficiency

j current density

j 0 saturation current density

j 01 ( j 02 ) saturation current density of the first (second) diode

j saturation current density of the emitter

j l light-generated current density

j sc short circuit current density

k Boltzmann’s constant

L+ diffusion length in the BSF

LFC laser fired contacts

L n ( L p ) 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

n 0 electron concentration in the dark

N A ( N D ) acceptor (donor) concentration

N A acceptor concentration in the BSF

n air ( n Si , n SiN ) refractive index of air (c-Si, SiN)

n i intrinsic carrier concentration

p 0 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

p phot photon power density

PSG phosphor silicate glass

P surf phosphorous surface concentration

P tot total power loss

PV photovoltaic

q elementary charge

R recombination rate

R A Auger recombination rate

R rad radiative recombination rate

R s series resistance

R s,tot total series resistance

R SRH Shockley-Read-Hall recombination rate

R sh shunt resistance

R sheet sheet resistance of the emitter

s (s n ) ( s p ) surface recombination velocity (of electrons or holes)

s b surface recombination velocity at the backside

SCR space charge region

s eff 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

V voltage

v n ( v p ) thermal velocity of electrons (holes)

V oc open circuit voltage

Wp Watt peak (power of 1 W under STC)

α absorption coefficient

ΔE F splitting of quasi-Fermi levels

Δn excess charge carrier density

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 electricityfrom sunlight by absorbing photons and generating electron–hole pairs,which are separated by a pn-junction The pn-junction creates an electricfield in the semiconductor and the separated charge carriers have to leavethe solar cell via electrical contacts to perform work in an external circuit

A solar cell in operation is therefore essentially an illuminated large areadiode, where emitter and base regions are contacted by metals to extractthe carriers

1.1 History

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

In the decades to follow, research was first directed toward application ofthe photovoltaic (PV) effect in space (powering satellites) or for terrestrialstand-alone systems As for those applications the total cost of power gen-eration was not the main issue, research was mainly driven by improvingthe conversion efficiencyη, which is the ratio between output power fromthe PV device (generated from the solar cell or complete solar module) and

input power (impinging photon flux) The oil crisis in 1973 led to erations to use PV also for terrestrial applications in larger scale as an alter-native to fossil fuels Since then a lot of R&D activities was focused onreducing the cost of PV electricity generation to make it attractive for mar-ket penetration.

consid-In research, a lot of progress was made in improving efficiency by oping new cell designs and applying novel processing steps, leading to effi-ciencies as high as 25% using standard test conditions (STC: 1000 W/m2illumination, AM1.5g spectrum, 25
C) in 1999 (Zhao et al., 1999), indi-cating the efficiency potential of c-Si This efficiency was reached onextremely pure float zone (FZ) silicon and on small scale (4 cm2) withoutthe main part of the front side metallization grid being taken into accountfor the efficiency measurement (so-called designated area measurement)and using a very complex processing scheme For most industrial applica-tions, a full area measurement and cost-effective c-Si materials are of higherinterest In addition, the number and complexity of processing steps neededfor 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 atrade-off between more complex processing on higher quality material all-owing higher efficiencies, and less complex processing, e.g., in combinationwith a lower c-Si material quality

devel-1.2 General routes for cost reduction

The lower efficiency for lower cost materials and less complex processingmight be advantageous cost-wise at cell level, but as there are also arearelated cost factors at module and system level (e.g., costs for module glassand installation), the question which route is more promising is not easy toanswer Therefore, a lot of different technologies have been developed overthe past decades This includes c-Si materials as well as solar cell fabricationprocesses

The Si feedstock of highest quality stems from the so-called Siemensroute using rods for Si production from the gas phase, which still accountsfor the majority of produced Si wafers for industrial solar cells, with fluidizedbed reactors as an alternative (Fabry and Hesse, 2012) So-called upgradedmetallurgical grade (UMG) Si can be produced with significantly less energyneeded per kg of fabricated Si, but a higher impurity concentration is theconsequence, 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

different segregation coefficients of B and P their concentrations and fore resistivity, influenced by the net doping, changes with ingot height(Ceccaroli and Pizzini, 2012; Heuer, 2013).

there-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 icant 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 highamount of O, mainly in interstitial form (Oi) (Zulehner, 1983) Cast mul-ticrystalline Si (mc-Si) can be produced in a more cost-effective way, butcontains due to the crystallization method used a higher amount of extendedcrystal defects and impurities in interstitial or precipitated form, originatingmainly from the crucible wall and the crucible coating (Buonassisi et al.,2006; Schubert et al., 2013) See Coletti et al (2012)for an overview onthe role of impurities in c-Si for solar cells For both methods, the crystallizedingot has to be sliced in wafers for subsequent solar cell processing To avoidkerf and other Si material losses that easily amount to>50%, ribbon-Si tech-niques have been developed, crystallizing the Si wafer directly from the Simelt (Hahn and Sch€onecker, 2004) Of the three technology groups, ribbon

signif-Si is the most cost-effective technique to produce wafers, but these wafersnormally show the highest defect densities, reducing the electronic quality ofthe as-grown wafer

Apart from Si wafer quality, solar cell process complexity is the othermain parameter determining the efficiency and cost structure of the solarcell 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 solarcell The cells are stringed in series so that the same amount of current flowsthrough all cells in a string and the voltages of the cells add up This makesproper sorting of cells a necessity to ensure that cells of similar performanceend up in a string, as the cell with the lowest current at operation conditionsdetermines the current flowing through the string Therefore, for all cellsnot 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 inindustrial fabrication homogeneous Si wafer quality and stable processeswith 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 ofdifferent processes applied for different cell designs, we will restrict the

overview on the most common cell architectures Other cell designs alreadyused in industrial scale such as the interdigitated back contact (IBC), com-mercialized by company SunPower Corp (Cousins et al., 2010), or the het-erojunction with intrinsic thin-layer (HIT) concept pioneered by Sanyo(now Panasonic) (Ballif et al., 2014) allow for the highest efficiencies in com-mercial 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 even25.6% with a combined IBC-HIT approach (Panasonic, 2014), but the pro-cesses differ significantly from mainstream technology Therefore, thesedesigns of very highly efficient c-Si solar cells will be treated in other chap-ters (e.g.,Ballif et al., 2014)

1.3 PV market today

Figure 1.1demonstrates the very dynamic growth of commercial PV overthe past decades, spanning more than four decades from around 1 MWp1inthe 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 stimulationprograms like, e.g., the renewable energy law with a guaranteed feed-in tar-iff in Germany As the German feed-in tariffs have been adjusted recentlyand the German PV market was the strongest worldwide, the growth sloweddown in 2012 and 2013 Strong growth in recent years allowed for a tremen-dous reduction in production cost due to scaling effects in mass production

as well as new and optimized processing technologies This so-calledlearning curve effect of PV resulted in an average module price reduction

of around 20% for every doubling of cumulated PV production (Nemetand Husmann, 2012) The continuing reduction in processing costsresults in costs of a kWh generated by PV being now in the range of elec-tricity generated from fossil fuels (depending on the installation site) (Kost

et al., 2013)

The market share of different PV technologies shown inFig 1.2revealsthat c-Si still shows by far the highest market penetration, with thin filmtechnologies like amorphous Si (a-Si), CdTe and CuInxGa(1x)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 of90% 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 tomc-Si in the past decade This can be explained by the huge productionexpansion programs happening at most PV manufacturers in the past, asmc-Si technology seems to be easier to ramp up and was the more cost-effective way of production in the past Whether this will hold true inthe future, with new cell designs allowing for higher efficiency approachingthe market, remains to be seen The market share of ribbon-Si dropped toalmost 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 asEverQ, respectively

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

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-Sisolar cell is shown in Fig 1.3 The base is p-type material, moderately

B doped to a resistivity of around 1Ω cm (NA¼1.51016 per cm3) Theemitter is n++-doped2using P with high surface concentration ND>1020per cm3, and the front surface is textured to allow a better incoupling ofimpinging photons (lower reflectivity) The emitter is covered by a thindielectric layer of H-rich silicon nitride (SiNx:H), acting as antireflectivecoating (ARC), surface passivation layer, and reservoir of H On the front,the metallization finger grid is realized by Ag paste, fired through the SiNx:Hlayer at high temperature On the rear, a full area contact is realized by Alpaste, 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 toroom temperature (back surface field, BSF) To allow interconnection ofthe 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 completecell thickness is around 180μm (note that features shown inFig 1.3are not

to scale) The formation of the respective regions of the cell will be dealtwith in more detail in the following sections

The use of H-rich SiNx:H layers for PV (Morita et al., 1982) in theso-called “firing through SiNx:H process” has been pioneered by Kyocera(Kimura, 1984; Takayama et al., 1990) and Mobile Solar for their EFGribbon-Si material (Cube and Hanoka, 2005) In the 1990s, other companiesand research institutes like, e.g., IMEC (Szlufcik et al., 1994) and others devel-oped the process further The breakdown of costs for c-Si module production

inFig 1.4reveals that wafer and module costs are the dominating factors

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.

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-Sisolar cell fabrication exist, forming the base of this chapter Since then newtechnologies have emerged, allowing for a reduction of costs as well as effi-ciency losses and therefore an increase of efficiency in mass production Totackle these losses, the next section will describe the physics involved in theoperation 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 theband diagram shown in Fig 1.5 The doping gradient due to the abruptchange in doping concentration at the pn-junction results in electrons (freemajority carriers in the n-region) diffusing from the n-region into thep-region and holes (free majority carriers in the p-region) diffusing intothe n-region The remaining ionized doping atoms at lattice sites (positivelycharged in the n-region, negatively charged in the p-region) form the spacecharge region (SCR) extending into both sides of the pn-junction Theelectric field hinders the free carriers to completely diffuse into the regions

of opposite doping, when equilibrium between diffusion and drift current offree carriers is reached The built-up electric field causes bending of theenergy bands, with the Fermi energy EFas defined by the Fermi–Dirac func-tion at a constant level (a horizontal line) in both regions

Upon illumination, absorbed photons excite electrons from the valenceband to the conduction band via the internal photoelectric effect

36%

26%

38%

Wafer Cell production Module

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

Absorption of one photon therefore generates an electron–hole pair, as themissing electron in the valence band is referred to as a hole Free electronsand holes can diffuse until they recombine or reach the SCR Here, chargecarriers of different types are separated, electrons are accelerated into then-region, holes into the p-region In case of illumination, the semiconductor

is not in thermal equilibrium anymore, and the relation for electron and holeconcentrations n0 and p0, respectively, as defined for thermal equilibrium(without illumination or applied voltage)

to be defined, with two resulting Fermi levels EFn and EFp referred to asquasi-Fermi levels of electrons and holes

Metal contacts with EFat roughly the same energetic position as for themajority carriers in the contacted Si region can extract carriers from bothregions The contact for the p-type region as depicted inFig 1.5is ohmic,whereas the n-type contact is of Schottky-type (energy barrier for electrons)

Hole

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 E F in the metal contacts are indicated as well.

The barrier can be overcome via tunneling, provided it is thin enough andnot too high.

2.2 Solar cell parameters

An ideal solar cell can be described by a 1-diode model and the j–V acteristic of an illuminated diode

2 i

LpND

with Dn(Dp) the diffusion constant of electrons (holes), NA(ND) the dopingdensity of acceptors (donors) and Ln(Lp) the minority charge carrier diffu-sion length of electrons (holes)

The resulting j–V curve is shown in Fig 1.6 The maximum currentdensity at V¼0 is the short circuit current density jjscj¼jl The point ofmaximum power density (mpp) is also indicated, with the fill factor FFdefined as

Current density/power density

Illumina ted cur v

Short circuit current density

Open-circuit voltage

Maximum power point (MPP)

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

In a semiconductor with band gap Eg(1.12 eV at 25
C for c-Si), photonswith energy E>Eg can be absorbed, creating electron–hole pairs, whilephotons with E<Egcannot be absorbed and are transmitted, seeFig 1.7.Generation of electron–hole pairs by illumination is a non-equilibrium pro-cess with some of the carriers occupying states high in the conduction band(electrons) and deep in the valence band (holes) directly after generationdepending on the photon energy The generated electrons and holes reachthermal equilibrium via collisions with other charge carriers or phononswithin the femtosecond (fs) range (thermalization) Afterward, they occupystates close to the band edges according to Fermi–Dirac statistics The max-imum voltage reachable (open circuit voltage Vocof the cell) is limited bysplitting of the quasi-Fermi levels for electrons and holes ΔEF, with

ΔEF<Eg As the maximum power point (mpp) of the illuminated j–V curve(Fig 1.6) is between V¼0 (maximum j¼jsc) and j¼0 (maximum V¼Voc),

Vmpp is always<Voc These four fundamental loss mechanisms limit themaximum 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 twodiodes, with the addition of series resistance Rs, shunt resistance Rshand asecond diode accounting for recombination in the SCR with an ideality fac-tor generally assumed to be 2 (Fig 1.8)

Apart from ohmic losses, recombination of generated charge carriers canoccur, limiting performance of the solar cell

can be reduced due to recombination of charge carriers with a tion rate R defining the lifetimeτ of excess charge carriers

recombina-τ ¼Δn

c-Si is an indirect band gap semiconductor In addition to an electron (in theconduction band) and a hole (in the valence band), a phonon is necessary forthe 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 Rradfor this type of recombination becomes3

τrad ¼ 1

cradp0

(1.11)for p-doped material

2.6 Auger recombination

Instead of creating a photon, the energy of the recombination process can beused to excite another existing free charge carrier (an electron in the con-duction band or a hole in the valence band) This charge carrier thermalizesafter excitation toward the band edge, converting the recombination energyinto phonons With the Auger recombination coefficients cA,nand cA,pforelectrons and holes, respectively, the Auger recombination rate reads

Auger recombination as a three-particle process is only relevant for highdoping concentrations >1017

per cm3in standard industrial solar cells

2.7 SRH recombination

Energy levels in the band gap can trap free charge carriers and cause a veryeffective recombination mechanism, especially when their energetic posi-tion is close to mid-gap This type of recombination was formulated byShockley, Read, and Hall (Hall, 1952; Shockley and Read, 1952), using sta-tistics of capture and emission of free carriers and is therefore referred to asSRH recombination Its recombination rate

τpðn0+ n1+ΔnÞ + τnðp0+ p1+ΔnÞ (1.14)with

τSRH¼τpðn0+ n1+ΔnÞ + τnðp0+ p1+ΔnÞ

for p-type material (p0n0), low injection (p0Δn), and trap energy level

at mid-gap (Et¼EFi) reads

veloc-All recombination channels are acting in parallel, and the resulting bulklifetimeτ is given by

τb

τrad+ 1

with the areal density of traps at the surface Nts, and snbeing referred to as thesurface recombination velocity s of electrons (minority carriers in p-typematerial) in units of cm/s

The influence of surface recombination on the observable effective time can be expressed by a surface lifetimeτs(Aberle, 1999)

life-1

τeff

¼τ1b+ 1

τs

¼τ1b+α2

stan-2.9 Recombination and saturation current density

Recombination reduces the maximum current density jscof the solar cell, asonly minority charge carriers generated within roughly one diffusion length

on either side of the pn-junction reach the junction and are injected into theregion on the opposite side of the junction But from Eq.(1.3)also stronginfluence of j0on Voccan be seen, as for j¼0

Voc¼kT

q ln

jl

j0+ 1

improv-2.10 Optical losses

If all impinging photons with Ephot>Egwere absorbed in the solar cell, withall of these photons contributing to the extracted current density, the max-imum jscwould be around 44 mA/cm2under STC Apart from recombina-tion 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)6and photons not being absorbed in c-Si (mostly long wavelengthsphotons7) leaving the cell

The different loss mechanisms are visualized inFig 1.9, where they areseparated into optical and electrical losses

6 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).

7

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.

3 THE BASIC FIRING THROUGH SiNx:H PROCESS

As already mentioned in the introduction, most industrial solar cellstoday 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 basicform as it was developed in more detail (compare with, e.g.,Neuhaus andMu¨nzer, 2007; Szlufcik et al., 1997), before alternatives and improvementswill be dealt with in the next sections

Generally, for every process step there are two options, inline or batchprocessing Inline processing offers the possibility to fabricate solar cells with

a minimum of handling steps and a smaller footprint due to the lack of age room necessary for partially processed cells On the other hand, not allprocessing steps can easily be performed inline and batch processing allowsfor more freedom in optimization The first example of a complete trueinline processing fabrication of solar cells was RWE Schott Solar’sSmartSolarFab in 2002 Nowadays, cell processing is normally done by amixture of inline and batch processing equipment, as the throughput ofmachines used for the different steps is not the same In addition, if singlemachines are not operational or have to be maintained, not the completeproduction is halted, but other parts within cell fabrication can continue

stor-to produce Therefore, often several machines of the same type work in allel to increase throughput and minimize the risk of bottlenecks

par-Al BSF

SiNx:H

p-Si

Ag

n+

ARC absorption loss

(mainly short wavelengths)

ARC reflection loss (mainly short wavelengths)

Shadowing loss (total reflection on metal)

Free carrier absorption

Rear absorption loss

Back reflection (mainly long wavelengths)

Carrier loss BSF Carrier loss bulk

Free carrier absorption

Final carrier flow jsc /q

Carrier loss emitter & SCR

3.1 Wafer washing, texturization, and cleaning

After crystallization, mono-Si and mc-Si wafers are sliced out of the Si ingotusing wire saws, containing slurry with abrasives for cutting into the Si(Dold, 2014) This leaves, apart from contaminants, saw damage on bothsides of the Si wafer with a depth in the range of up to 10μm (depending

on sawing conditions) After wafer washing, this saw damage has to beremoved, as the disturbed region of the crystal (cracks, dislocations) is ofpoor electronic quality

For mono-Si, this is done in an alkaline wet chemical solution of KOHand isopropyl alcohol (IPA) at temperatures of around 80
C The KOHsolution etches the Si while the alcohol masks the surface randomly Etching

is anisotropic, with the result that the most densely packed crystal planes inc-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 ran-domly form square-based upright pyramids (Fig 1.10) These pyramids veryeffectively reduce the reflectivity of the surface and therefore increase theincoupling of photons into c-Si The etching reaction can be summarized as

Si + 2H2O + HO! HSiO3 + 2H

and consists of oxidation of Si, formation of a solvable salt and dissolving thesalt 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 currentdensities 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 solutionshad to be developed Standard is an acidic solution based on HF and HNO3without further additives (Einhaus et al., 1997; Hauser et al., 2003) The tex-ture attacks the Si surface first at areas where not all Si bonds are perfectlysaturated 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 etchedaway, the textured surface starts to flatten again for prolonged processingtimes, as sharp edges are rounded Four to five micrometer removal of Siper side is normally enough to remove the saw damage and obtain a lowreflectivity.8The etching reaction takes place in two steps, an oxidation8

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.

3Si + 4HNO3! 3SiO2+ 3H2O + 4NO (1.27)followed by etching of the SiO2

3SiO2+ 18HF! 3H2SiF6+ 6H2O: (1.28)Afterward, the thin porous Si layer at the surface is etched off in (cold)KOH The remaining reflectivity is significantly higher than for randompyramids, therefore it is not used for mono-Si (Fig 1.11) Acidic texturingcan be done elegantly inline, as texturing time is in the range of only around

2 min (depending on temperature) (Hauser et al., 2004; Neuhaus andMu¨nzer, 2007)

Figure 1.10 SEM images of textured c-Si surfaces for mono-Si using KOH/IPA (left) and mc-Si using an acidic texture solution (right).

400 600 800 1000 1200 0.0

0.1 0.2 0.3 0.4 0.5 0.6 0.7

Wavelength (nm)

Bare Si

Si textured Bare Si AR coated

Si textured and AR coated

Figure 1.11 Reflectivity of bare Si and alkaline-textured mono-Si with and without SiN :H ARC.

After texturing, the wafers are thoroughly cleaned as the next step is thediffusion taking place at high temperatures Impurities present on the wafersurface could diffuse in to the wafer causing recombination and thereforeloweringτb Cleaning normally consists of rinsing in deionized (DI) water,cleaning in HCl, DI water rinsing, etching in HF to form a hydrophobicsurface, followed by a short dip in DI water and drying.

3.2 Phosphorus diffusion

In this step, the heart of the solar cell, the pn-junction, is formed The twomost common ways to form the P-doped emitter will be described in thefollowing

In most cases, the in-diffusion of P into c-Si takes place in a quartz tubefurnace N2is directed into a bottle (bubbler) containing POCl3, which isliquid at room temperature POCl3molecules are transported with the N2flow into the quartz tube, where the wafers are located in quartz containers(boats) with spacing between the wafers at temperatures around 800–850
C

O2is added and on the wafer surface P2O5is formed according to

4POCl3+ 3O2! 2P2O5+ 6Cl2, (1.29)where the formed Cl2 provides an additional cleaning effect on the wafersurface The O2flow also oxidizes the Si surface, and the resulting SiO2layertogether with the P2O5 forms the so-called phosphor silicate glass (PSG)layer acting as the diffusion source P diffuses into c-Si and the diffusioncoefficient depends strongly on doping concentration, as the level of EFdetermines the amount of vacancies present in the material For

P concentrations well above 1019per cm3at diffusion temperature, a ent diffusivity is observed due to the existence of double negatively chargedvacancies in large amounts, forming a mobile quasi-particle with ionized P+.For P concentrations below 1019per cm3, single negatively charged vacan-cies dominate the diffusion mechanism This results in the characteristickink-and-tail shaped profile of P diffusion in c-Si whereby the tail is formeddue to the “normal” diffusion mechanism involving vacancies (Fair and Tsai,

differ-1977) During diffusion, time, temperature, and gas flows have an influence

on the diffusion profile formed To increase (double) the throughput, wafersare often loaded in the quartz boats back-to-back

As an alternative to quartz tube POCl3diffusion, a liquid P-containinglayer can be deposited on the wafer surface (mainly diluted H3PO4), e.g., byspraying Wafers then move horizontally through a conveyor belt firing

furnace As for this inline technique, the time allowed for diffusion is limiteddue to throughput and length of the furnace possible, diffusion temperaturesare normally higher than for POCl3diffusion, resulting in steeper P dopingprofiles Higher doping concentrations normally result in lower emitterquality and more Auger recombination, increasing the saturation currentdensity contribution of the emitter j0e In addition, surface passivation isinfluenced by doping concentration with better passivation quality possiblefor lowly doped surfaces (Cuevas et al., 1996).

Only P atoms on Si lattice sites are electrically active dopants The surfaceconcentration of P in c-Si for an unlimited source is given by the solubilitylimit in the range of 3–6 1020

per cm3 between 800 and 900
C withhigher values for higher temperatures (Trumbore, 1960) Apart from theelectrically active P atoms, interstitial P or P-containing clusters can form,increasing the amount of P present in Si especially close to the surface(Fig 1.12) (Bentzen et al., 2006a) The almost flat shape of the P dopingprofile with electrically active P concentration above 1020 per cm3 is alsoreferred to as “dead layer,” as this layer is highly recombination active.Although the high surface concentration of P close to the surface is limitingthe electronic quality, it seems to be needed for the formation of a good lowresistivity contact with the front metal Ag paste during the firing step

An important parameter of the emitter is its conductivity, as charge riers have to flow laterally toward the collecting finger grid As the emitter is

car-a very thin lcar-ayer (usucar-ally well below 1μm thick), a sheet resistivity is definedfor a uniformly doped layer as

Figure 1.12 P profiles of identical P diffusions in c-Si measured by ECV (electrically active concentration) and SIMS (total concentration) The solubility limit at diffusion temperature according to Bentzen et al (2006b) and Solmi et al (1996) is also indicated.

whereρ is the resistivity of the layer with thickness d

For non-uniformly doped layers as is the case for a diffusion profile, sheetresistivity calculates according to

Rsheet¼Ð 1

d 0

3.2.1 Phosphorus diffusion gettering of impurities

During P diffusion, SiO2is formed at the wafer surface For the formation ofSiO2Si, atoms have to leave their lattice sites, and a flux of Si interstitials Siiisgenerated In addition, diffusing P atoms change position with a Si atom onlattice site (Sis), further increasing the concentration of Sii These Si inter-stitials themselves can change position with an impurity atom (M) located atlattice sites via a “kick-out reaction”

In interstitial form, the impurity atom is mobile and can diffuse throughthe crystal toward a location with higher solubility Such a location of highsolubility can, e.g., be the highly doped region containing precipitated

P close to the surface or the PSG (Bentzen et al., 2006a) Such regionsare called getter sinks Due to the presence of regions with high solubilityfor impurity atoms, a concentration gradient is formed toward the sink,resulting in more and more impurity atoms moving toward the sink, leaving

a cleaner region behind This self-cleaning process is called gettering ofimpurities Depending on the location of the getter sink we distinguishbetween external gettering (e.g., at the crystal surface, where the impuritiescan be removed) or internal gettering at extended defects in the Si bulk (e.g.,

at grain boundaries, dislocations, or precipitates) The gettering process ingeneral can be divided into three phases: (1) freeing the impurity from its(bonded) position by supplying an activation energy, (2) diffusion of theimpurity in the wafer, (3) capture of the impurity at the gettering location.Depending on the specific mechanism present, it is distinguished between

relaxation induced, segregation induced, and injection induced gettering,where also combinations of these different mechanisms are possible (Kangand Schr€oter, 1989; Seibt and Kveder, 2012).

It could be shown that for back-to-back diffusion the positive getteringeffect is less pronounced, as only one effective external getter sink is available(Schneider et al., 2005) The same can be assumed for inline diffusion, as thediffusion source is applied only on one side.9

An elegant way to remove the rear side emitter is by inline etching of therear side with the wafer floating on the etch solution containing H2O,

HF, HNO3, and H2SO4 As the PSG etch can be performed inline as well,both steps can be combined in a single inline wet bench (Delahaye et al.,2004; Melnyk et al., 2005) Thereby care has to be taken that the emitter

on the front side is not attacked by the etch solution or the atmosphere taining reactive species

con-3.4 SiNx:H deposition

To further minimize reflection losses at the front side, an ARC based onSiNx:H is deposited on the front.10 For normal incidence of photons,destructive interference is reached if the thickness d of the ARC is

with nairand nSirefractive indices of air and c-Si being the materials aboveand below the ARC, respectively With nSi(600 nm)¼4, on cell level theoptimum refractive index would be nSiN¼2 As the cells are later encapsu-lated into modules using ethylene vinyl acetate (EVA) with refractive indexaround 1.5, a slightly higher nSiNwould be optimum for solar cells in mod-ule application With Eqs (1.33) and (1.34) typical ARC thicknesses are75–80 nm for nSiN¼2, if reflectivity of photons with λ¼600–650 nm(the maximum in photon flux of the AM1.5g spectrum) should beminimized.

Apart from the ARC effect, SiNx:H is also suited for surface passivation

of the n++P-doped emitter Defects located close to the c-Si/SiNx:H face in the SiNx:H layer provide fixed positive charges in the order of 1012–

inter-1013per cm2(Aberle, 1999; Lamers et al., 2012) The minority charge riers (holes in n-type material) are repelled from the surface due to Coulombrepulsion while majority carriers are attracted, and therefore recombination

car-is lowered.11This surface passivation mechanism is referred to as field effectpassivation in contrast to chemical passivation, where the reconstruction ofchemical bonds lowers the density of energy levels in the bandgap Chemicalpassivation is also present for SiNx:H layers, but remaining defect densitiesare usually higher than for SiO2layers, which in turn have a lower density offixed charges

If the SiNx:H layer contains significant amounts of H, this H can bereleased during the firing step and diffuse into the c-Si bulk (Hahn et al.,2004; Jiang et al., 2003) Here, H can passivate bulk defects and drasticallyimprove material quality (Duerinckx and Szlufcik, 2002; Hahn et al., 2010).This is a crucial step especially for mc-Si material with high defect densities.The most common technology to deposit SiNx:H layers is plasma-enhanced chemical vapor deposition (PECVD) There are direct and remoteplasma techniques available A direct plasma system usually operates at lowfrequency of 40 kHz The wafer forming one electrode is in contact with theplasma, and accelerated ions can bombard the wafer surface leading to a cer-tain surface damage.12 These systems usually operate in batch mode Inremote plasma systems, operating usually at high frequencies around13.56 MHz, the plasma is spatially separated from the wafer, and a linearplasma source is used with microwaves supplying the excitation Remoteplasma systems are usually operating in inline geometries with the wafers

lying on trays being transported through the reactor underneath the linearplasma source For both techniques, pressure is around 0.1–1 mbar anddeposition temperature is between 300 and 450
C, depending on technol-ogy Precursor gases used are SiH4and NH3, and the SiH4/NH3ratio deter-mines the stoichiometry and therefore refractive index and absorptioncoefficient of the resulting SiNx:H Higher SiH4/NH3ratios (Si-rich layers)lead to higher refractive indices and higher absorption (Nagel et al., 1998).

As absorption in the ARC is unwanted, usually a compromise between mum nSiN (2.3–2.4 for module application) and low absorption is found,with nSiN(600 nm)¼2.0–2.1

opti-Before PECVD SiNx:H with high throughput became available for PV

at the end of the 1990s, TiO2was often used as ARC for industrial c-Si cells.Compared to SiNx:H, a higher refractive index without significant absorp-tion was possible, but TiO2layers showed poor surface passivation qualitiesand contained no H needed for bulk passivation The method used for depo-sition was atmospheric pressure chemical vapor deposition (APCVD).13

3.5 Metallization via screen-printing

Screen-printing of metal pastes for PV application is a very robust methodalready introduced in 1975 (Ralph, 1975) and can be used as an inline pro-cess.14 A conveyor belt transports the wafer onto the printing chuck

A screen consisting of a mesh of wires partly covered with an emulsion isthe mask for the metallization process Metal paste is printed through theopenings in the emulsion through the mesh of wires onto the wafer lyingunder the screen The screen is positioned on top of the wafer with awell-defined distance between screen and wafer (the snap-off distance).The paste is placed on top of the screen and a squeegee moving horizontallywithout pressure on the screen fills the openings of the mesh uniformly withpaste In the next horizontal movement of the squeegee over the screen, it ispressed onto the screen with a defined pressure, pressing the screen locallyagainst the wafer surface and pushing the paste from the filled areas onto thewafer surface The screen snaps off from the wafer after the passing of thesqueegee because of the screen tension (Fig 1.13) After printing, the wafer

is transported into a drying furnace for evaporation of the volatile ingredients

of the paste at temperatures of around 150–200
C to avoid smearing of thepaste when it is flipped over to metallize the other surface Due to the13

See Richards (2004) for a review on TiO 2 and other dielectrics for use of ARC.

14

See also Holmes and Loasby (1976) and Neuhaus and Mu¨nzer (2007) for more details.

tension put on the screen during every printing step, the lifetime ofthe screens is limited to several thousand printing steps as they wear outwith time.

3.5.1 Front side metallization

For front side metallization, the following criteria have to be met by thepaste used: (1) low contact resistance to c-Si, (2) low specific resistance inthe printed structure, (3) no junction shunting, (4) good aspect ratio (height

to width ratio) of the fingers, (5) good adhesion to c-Si, (6) opening ofSiNx:H layer provided, and (7) solderability for cell interconnection inthe module Ag containing paste (70–80%weight) is used, as Ag is highly con-ductive and therefore allows for good conductivity in the printed metal fin-gers Additional components are glass frits containing PbO, B2O3, and SiO2(1–10% ), which are responsible for locally dissolving the SiN :H layer

Squeegee

Frame

Solar Cell

Closed mesh Open mesh

Metal paste

Vacuum chuck

Figure 1.13 Screen-printing of metal paste.

as well as for a good adhesion Also present are organic binders(15–30%weight), influencing the rheology of the paste which has to below enough to ensure that a continuous finger is formed and high enough

to keep a high aspect ratio (Neuhaus and Mu¨nzer, 2007)

During the firing step at temperatures around 800
C, the electrical tact between Ag and c-Si is established Early detailed studies for Ag frontcontact formation by Ballif et al (2002, 2003) and Schubert et al.(Schubert, 2006; Schubert et al., 2002, 2004) led to the following picture(Fig 1.14) Below 600
C organic components burn out (A) At higher tem-perature, the contact is formed as first the PbO melts, wets, and etches theSiNx:H layer (B) The Ag particles with sizes of severalμm sinter togetherand form a conductive film Then a redox reaction between PbO and Siforms Pb (C) The liquid Pb starts to melt Ag (D), and the Ag/Pb melt reactswith Si, etching inverted pyramids locally into the c-Si surface (E) Oncooling down Ag recrystallizes on (111)-Si planes, forming isolated contactpoints to the emitter (F) The recrystallized Ag points at the c-Si surface caneither be in direct contact with the sintered Ag layer, or the glass layer iso-lates them from each other If there is no direct contact established, contactresistance depends strongly on the thickness of the glass layer formed inbetween Thin layers can be tunneled through, with small (nanoscale) metalprecipitates of Ag and/or Pb/Bi providing additional hopping sites for elec-trons The thickness of the isolating glass layer is a very crucial parameter forachieving low contact resistance in case of no direct connection between Agcrystallites and sintered Ag layer Therefore, firing parameters are veryimportant, with too high peak temperature resulting in thicker glass layersand too low peak temperature resulting in not completely opened SiNx:Hlayers (Schubert, 2006)

con-As the contact is not formed everywhere underneath the Ag tion printed onto the wafer, the contact resistance is significantly higher thanfor contacts established, e.g., via evaporation of Ag directly on c-Si n+emit-ters Typical values for contact resistance of screen-printed Ag on c-Si are1–10 mΩ cm2

metalliza-(Schubert, 2006), while values for evaporated contacts are

in the range of 100–200μΩ cm2

(Fischer, 1994)

3.5.2 Rear side metallization

For rear side metallization, Al containing pastes containing Al powder, glassfrit, organic binders, and solvents are used The lower conductivity of Alcompared to, e.g., expensive Ag does not play a role as long as the contact

is formed on the full area rear side Another very important advantage of Al is

the fact that Al is an acceptor in c-Si and can form a good ohmic rear contact

in combination with a highly p+-doped layer, the BSF (Mandelkorn andLamneck, 1972) The BSF is formed by alloying during the firing step afterdrying of the paste (L€olgen, 1995)

A description of the formation process is given in Huster (2005) TheAl/Si phase diagram depicted in Fig 1.15 shows the composition of theAl/Si melt dependent on melt temperature Upon heating up during the fir-ing step, Al starts to melt at 660
C The Al2O3shells around the Al spheresstay in shape, but liquid Al can penetrate through the oxide shell locally andgets in contact with the c-Si surface and other Al particles Si is dissolved intothe Al melt at that temperature until according to the phase diagram the meltcontains around 17% Si As the volume in the stable oxide shells stays con-stant, the amount of Al leaving the shell covering the c-Si surface corre-sponds to the volume fraction of Si entering the shells to form the correctconcentration according to the phase diagram During further increase oftemperature, more and more Si is dissolved in the melt Assuming a peakfiring temperature of 800
C, the melt contains around 27% of Si Duringcooling down from peak temperature, a part of the Si has to leave the meltaccording to the phase diagram This Si recrystallizes at the c-Si/melt inter-face During recrystallization, a small amount of Al is incorporated into therecrystallized Si lattice according to the solid solubility of Al in Si at that

Silver

Figure 1.14 Schematic contact formation for Ag screen-printing on n++emitters After Schubert (2006)

temperature This accounts for the p+doping of the recrystallized Si layer inthe range of 1018–1019per cm3, the BSF The recrystallization stops whenthe temperature of the melt reaches the eutectic point at 577
C The solid-ified Al/Si melt then has eutectic composition with 12% Si This holds truefor both the solidified melt in the oxide shells as well as for the film directly

on top of the recrystallized c-Si (BSF) As solubility of Al in Si decreases frompeak temperature toward 577
C, the BSF contains a doping gradient withhigher Al concentration at the p/p+interface

The thickness of the resulting BSF in dependence of the amount of Aldeposited can be calculated according to

0 20 40 99.98 400

600 800 1000

%at silicon

Solubility of Al in Si Liquidus curve L (T)

T5 = 700 ⬚C T6 <577 ⬚C

Figure 1.15 Phase diagram of Al/Si (top) with characteristic stages during rear side tact firing (bottom) Al/Si data from Krause et al (2011) and solubility data from Murray and McAlister (1984) and Yoshikawa and Morita (2003) After Huster (2005)

con-The p/p+ interface forms a high/low junction, and similar to thepn-junction the doping gradient is the source for an electric field causingslight band bending close to the rear metal contact (seeFig 1.5) This results

in repelling of electrons (minority carriers) and therefore less recombination

at the semiconductor/metal interface in analogy to surface passivation via thefield effect But as the BSF region is highly doped, its electronic quality(minority carrier lifetime and diffusion length) is quite low The effect ofBSF parameters on surface passivation can be calculated via (Godlewski

et al., 1973)

seff¼NA

N+ A

NA+(high concentration gradient) are beneficial Therefore, B can be mixed

to the Al paste to increase p+doping of the BSF, as B has a higher solubility

in Si compared to Al (Rauer et al., 2013)

As cells have to be interconnected for module integration and the Al rearside is not solderable, a small fraction of the c-Si rear side is metallized with

Ag or Al/Ag paste pads or busbars These regions have to be kept as small aspossible to avoid unnecessary recombination, as in these regions no BSF isformed and seffis significantly higher.15On the other hand they have to belarge enough to assure reliable soldering

3.5.3 Co-firing step

Contact formation is realized in a conveyor belt furnace with opticalheating The simultaneous contacting of emitter and base in the so-calledco-firing step is a very critical part of the process, as it has to be optimizedfor several purposes During this step, BSF and rear contact are formed,while in addition on the front side the Ag metallization has to be firedthrough the SiNx:H layer, and H has to be released from the SiNx:H layerinto the c-Si bulk to passivate crystal defects As optimum parameters for15

Surface recombination velocity of a pure c-Si/metal contact is in the range of 106cm/s.

some of these steps are going in different direction of the parameter space, acompromise has to be found Ag and Al paste as well as emitter profile andSiNx:H layer therefore have to be tuned to match their optimum firingparameters as good as possible and allow for a good overall end result.

InFig 1.16, a typical firing profile is shown (1) the first plateau visiblerefers to the temperature of 660
C, where Ag starts to melt (latent heat).The peak temperature of around 800
C (2) is kept for only a few seconds.During cooling down, another plateau at the eutectic point of 577
C can beseen (heat of crystallization) (3)

If edge isolation was not already performed earlier in the process (by gle side etching or other methods like, e.g., plasma edge isolation), laser edgeisolation close to the cell’s edge at the front side is another option Here, theemitter is locally removed by laser heating, and a groove is formed (Emanuel

sin-et al., 2001; Schneiderl€ochner et al., 2003) Hereby active cell area is lost,slightly compromising the current and therefore efficiency (Hauser

et al., 2001)

3.6 Solar cell characterization

After solar cell processing, the j–V characteristics of every cell are measured,with Voc, jscand FF determining the conversion efficiency (Eq.1.6) Fromthese, the current density and voltage at mpp are determined and solar cellsare classified into bins according to current density classes under mpp con-ditions to avoid current mismatch in the string of the module when cells areFigure 1.16 Typical firing profile Data from Huster (2005)

connected in series In addition, all cells are inspected visually to assure thatonly solar cells with the same color end up in a module to achieve homo-geneous optical appearance Also the reverse break through characteristic ofthe diode is checked by applying reverse bias of 12–15 V to avoid the for-mation of hot spots, which might cause dangerous local heating (hot spots)

of the cell This could happen when, e.g., one cell or part of it is shaded inthe module and current is flowing in reverse direction through the string,leading to destruction of the cell

j–V characteristics are normally measured using a halogen flash lamp thatcan provide a constant power of 1000 W/m2for around 50 ms, with thelight intensity checked by a monitor cell The actual temperature of thesolar cell is measured and the voltage is corrected to the temperature ofSTC, 25
C As for STC, the spectrum should be AM1.5g, which is neverachieved exactly by the flasher system, certified reference solar cells have to

be used for calibration to minimize the effect of spectral mismatch The ibration cell should exhibit an external quantum efficiency (EQE) very sim-ilar to the cell to be measured to avoid errors, especially when introducingnew cell designs (Herguth et al., 2011) Contacting and measurement arerealized using the four point probe setup, with several probes for each bus-bar simulating the situation after tabbing of the cells for interconnection inthe module

cal-Solar cell parameters are normally determined under STC conditions,but solar modules under realistic operation conditions in the field mightoperate at significantly different temperatures Depending on the location

of operation, solar cell temperatures well above 25
C are reached, reducingthe efficiency mainly according to the temperature behavior of the voltage.This should be kept in mind when optimizing not only the solar cell, but thecomplete PV system for an optimized energy output

4 RECENT DEVELOPMENTS ON SOLAR CELL FRONT SIDE4.1 Wafer sawing

The standard technology for sawing of wafers is still slurry-based cutting ofwafers using a steel wire as described in the previous section As the wire has

a diameter of around 120μm, resulting in kerf loss of around 140 μm, andwafer thickness is typically 180μm, around 40–50% of Si is lost during thisstep The slurry used for cutting and cooling normally contains SiC particlesfor cutting and other additives contaminating the Si kerf Therefore, thecontaminated Si kerf cannot be easily recycled and is lost

A lot of effort went into the development of alternative sawing processes,with diamond wire sawing as the most promising alternative Here, the wire

is coated with small particles (diamond-plated wire), and the wire itself is thecutting source No additional slurry is needed, and water is used as a coolingagent In this way, Si kerf can in principle be separated and recycled.After diamond wire cutting, the surface damage of the wafer is differentcompared to slurry-based cutting (Buchwald et al., 2013) The damage isgenerally not as deep as for slurry-based cutting, but is less homogeneous

In addition, the surface exhibits ripples, which might have an influence

on mechanical stability and processing steps following later during solar cellprocessing

4.2 Alkaline wafer texturing

Wafer texturing based on alkaline solutions with IPA as an additive showslimitations in throughput, cost, and reliability As IPA has a boiling point of

82
C, texture bath temperature is limited to around 80
C This causesthroughput issues, as the underlying chemical texture reaction increases withtemperature In addition, IPA constantly evaporates during texturing.Therefore, it has to be replenished to assure the correct composition ofthe texture bath

A lot of research was carried out to find alternatives to KOH/IPA texturesolutions with industrial relevance Some of them try to minimize the con-sumption of IPA (seeBasu et al., 2013for an overview), others try to avoid itcompletely IPA free texturing recipes/processes (see, e.g.,Moynihan et al.,2010; Ximello et al., 2009) using additives/alcohols with higher molecularweight have the advantage that higher texture bath temperatures can beused, speeding up the texturing process and leading to more homogeneoustexturing results This is of high interest especially for wafers sawn with thenew diamond wire technique or treated with different washing/cleaningsolutions prior to texturing, as standard KOH/IPA texture solutions oftenfail to texture these wafers reliably An alternative to avoiding IPAcompletely might be the application of a closed chamber texture bath, withthe texturing process being sped up by application of vacuum pulses andautomatic recycling of evaporated IPA (Ximello et al., 2011)

4.3 Front contact metallization

Although screen-printing is still the most reliable and cost-effective way offorming contacts in solar cell mass production, the standard Ag screen-printing technology has several drawbacks and challenges