SolarCells_Varios

462
Silicon Solar Cells, Crystalline SANTO MARTINUZZI 1 ,ABDELILLAH SLAOUI 2 ,JEAN-PAUL KLEIDER 3 ,MUSTAPHA LEMITI 4 ,CHRISTIAN TRASSY 5 ,CLAUDE LEVY -CLEMENT 6 ,SE ´ BASTIEN DUBOIS 7 ,REMI MONNA 7 ,YVES VESCHETTI 7 ,ISABELLE PE ´ RICHAUD 8 ,NAM LE QUANG 9 ,JED KRAIEM 10 1 Formerly University of Marseilles, Marseilles, France 2 CNRS-INESS, Strasbourg, France 3 LGEP, Paris, France 4 INL-INSA, Lyon, France 5 CNRS, Grenoble, France 6 CNRS, Thiais, France 7 INES-CEA, Le Bourget du Lac, France 8 University of Marseilles, Marseilles, France 9 Photowatt International S.A, Bourgoin Jailleu, France 10 Apollon Solar, Lyon, France Article Outline Glossary Definition Introduction The Silicon Feedstock (Polysilicon) Silicon Single Crystals, Multicrystals, Ribbons, and Sheet Growth Techniques Future Directions Abbreviations Bibliography Glossary Antireflection coating layer (ARC) Layer deposited on the illuminated surface of a cell which reduces the reflection of the light. Dislocations Lack or additional line of atoms in the crystal. Gettering process Extraction of unwanted impurities and trapping in a region of a wafer which does not contribute to the photocurrent. Grain boundary (GB) Border zone of two adjacent crystalline grains. Homojunction (heterojunction) Part of a semicon- ductor diode which separates a p-type region from an n-type one made with the same material (in a heterojunction, the materials are different). Minority carrier lifetime t and diffusion length L Dur- ation and distance run by an electron in excess (generated by the sunlight) in p-type silicon or by a hole in excess in n-type silicon. Minority carrier diffusion length L Distance run by carriers in excess. Multicrystalline silicon Large grained polycrystalline material, a few mm to 1 cm in size. Passivation Mechanism which reduces the electrical activity of crystallographic defects, of surfaces, of interfaces, and of unwanted impurities. Precipitates Agglomeration of impurity atoms within the crystal. Solar cell conversion efficiency h Ratio of the electrical power output to the sunlight power input. Texturization Chemical or physical technique which increases the roughness of the surface in order to reduce the light reflection and the diffusion. Wafer Trench of silicon cut out from an ingot, 200 mm in thickness. Definition Solar cells are sources of electrical energy when they are illuminated by solar radiations. They deliver to a load a photocurrent and a photovoltage. First used for space applications, solar cells were progressively taken into account for terrestrial applications. The main problems to solve are, today, the photovoltaic energy cost which is too high and the conversion efficiency which is limited. Most of the cells are based on a p–n junction made with a p-type semiconductor and an n-type semiconductor. When both materials are the same, the cell is based on a homojunction. When the materials are different, the cell is based on a heterojunction. One silicon cell, 15.6 15.6 cm 2 , can deliver 7–9 A under 0.6 V only, and for this reason, the cells are connected in modules in order to provide a substantial electrical power (about 100–250 W). The semiconductor silicon is known as an extremely pure material; 9 N purity level is required for electronic components. However, for solar cells, a 6 N purity level could suffice, and the use of polycrystalline materials is acceptable. Wafers were initially cut out from monocrystalline and also of multicrystalline ingots grown from electronic-grade 226 Silicon Solar Cells, Crystalline C. Richter et al. (eds.), Solar Energy, DOI 10.1007/978-1-4614-5806-7, # Springer Science+Business Media New York 2013 Originally published in Robert A. Meyers (ed.) Encyclopedia of Sustainability Science and Technology, # 2012, DOI 10.1007/978-1-4419-0851-3

Transcript of SolarCells_Varios

  • Bibliography

    226 Silicon Solar Cells, CrystallineGlossary

    Antireflection coating layer (ARC) Layer deposited

    on the illuminated surface of a cell which reduces

    the reflection of the light.

    Dislocations Lack or additional line of atoms in the

    crystal.

    Gettering process Extraction of unwanted impurities

    and trapping in a region of a wafer which does not

    contribute to the photocurrent.

    Grain boundary (GB) Border zone of two adjacent

    crystalline grains.

    Homojunction (heterojunction) Part of a semicon-

    ductor diode which separates a p-type region

    from an n-type one made with the same material

    (in a heterojunction, the materials are different).

    C. Richter et al. (eds.), Solar Energy, DOI 10.1007/978-1-4614-5806-7,Silicon Solar Cells, Crystalline

    SANTO MARTINUZZI1, ABDELILLAH SLAOUI2, JEAN-PAUL

    KLEIDER3, MUSTAPHA LEMITI4, CHRISTIAN TRASSY5, CLAUDE

    LEVY-CLEMENT6, SEBASTIEN DUBOIS7, REMI MONNA7, YVES

    VESCHETTI7, ISABELLE PERICHAUD8, NAM LE QUANG9, JED

    KRAIEM10

    1Formerly University of Marseilles, Marseilles, France2CNRS-INESS, Strasbourg, France3LGEP, Paris, France4INL-INSA, Lyon, France5CNRS, Grenoble, France6CNRS, Thiais, France7INES-CEA, Le Bourget du Lac, France8University of Marseilles, Marseilles, France9Photowatt International S.A, Bourgoin Jailleu, France10Apollon Solar, Lyon, France

    Article Outline

    Glossary

    Definition

    Introduction

    The Silicon Feedstock (Polysilicon)

    Silicon Single Crystals, Multicrystals, Ribbons, and

    Sheet Growth Techniques

    Future Directions

    Abbreviations# Springer Science+Business Media New York 2013

    Originally published in

    Robert A. Meyers (ed.) Encyclopedia of Sustainability Science and Technolospace applications, solar cells were progressively taken

    into account for terrestrial applications. The main

    problems to solve are, today, the photovoltaic energy

    cost which is too high and the conversion efficiency

    which is limited. Most of the cells are based on a pn

    junction made with a p-type semiconductor and an

    n-type semiconductor. When both materials are the

    same, the cell is based on a homojunction. When

    the materials are different, the cell is based on

    a heterojunction. One silicon cell, 15.6 15.6 cm2,can deliver 79 A under 0.6 V only, and for thisreason, the cells are connected in modules in order to

    provide a substantial electrical power (about

    100250 W). The semiconductor silicon is known as

    an extremely pure material; 9 N purity level is required

    for electronic components. However, for solar cells,

    a 6 N purity level could suffice, and the use of

    polycrystalline materials is acceptable. Wafers were

    initially cut out from monocrystalline and also of

    multicrystalline ingots grown from electronic-gradeMinority carrier lifetime t and diffusion length L Dur-

    ation and distance run by an electron in excess

    (generated by the sunlight) in p-type silicon or by

    a hole in excess in n-type silicon.

    Minority carrier diffusion length L Distance run by

    carriers in excess.

    Multicrystalline silicon Large grained polycrystalline

    material, a few mm to 1 cm in size.

    Passivation Mechanism which reduces the electrical

    activity of crystallographic defects, of surfaces, of

    interfaces, and of unwanted impurities.

    Precipitates Agglomeration of impurity atoms within

    the crystal.

    Solar cell conversion efficiency h Ratio of the

    electrical power output to the sunlight power input.

    Texturization Chemical or physical technique which

    increases the roughness of the surface in order to

    reduce the light reflection and the diffusion.

    Wafer Trench of silicon cut out from an ingot,

    200 mm in thickness.

    Definition

    Solar cells are sources of electrical energy when they are

    illuminated by solar radiations. They deliver to a load

    a photocurrent and a photovoltage. First used forgy, # 2012, DOI 10.1007/978-1-4419-0851-3

  • 227Silicon Solar Cells, Crystallinethe lack of toxicity of this element, and, last but not the

    least, to its very high abundance in the Earths crust.

    Moreover, the huge knowledge accumulated in the

    electronic device research and industrial production is

    a source of continuous progress.

    However, the cost is always a problem, and a lot of

    efforts in the field of research and engineering

    are devoted to reduce the cost of the peak watt Wp,

    i.e., the cost of the electrical power produced when the

    cells are illuminated by the sun under optimal

    conditions. Such a cost depends on the cost of mate-

    rials, of the cell processing steps, and of the

    module fabrication, but it depends also markedlyfeedstock, but progressively, solar-grade feedstocks

    could be used. The physical, the chemical, and the

    electrical properties of such wafers must be investigated

    in order to be qualified for cell production. The struc-

    ture of the solar cells can be very basic, i.e., one pn

    junction with simple metallization and antireflection

    coating. However, in order to increase the conversion

    efficiency, which tend to reduce the cost of the

    produced electrical energy, advanced concepts must

    be used in industrial production like selective emitters,

    new metallization techniques, and new cell structures

    like interdigitated back contact and back junction

    cells and heterojunction cells.

    Introduction

    Silicon solar cells were used, for the first time, as

    electrical energy sources for satellites. The first one

    was Vanguard launched 52 years ago. The technical suc-

    cess of the photovoltaic panels was so clear that their use

    in terrestrial applications was proposed at the end of the

    1960s. Several solutions have been considered to decrease

    the cost of the photovoltaic systems and energy. Thin film

    cells, like CdSCu2S or CdTeCu2 Te structures, were

    investigated at the end of the 1960s, but they were not

    convincing. Silicon solar cells based on monocrystal-

    line, on multicrystalline wafers, and on hydrogenated

    amorphous silicon thin films appear, during the 1980s,

    as a possible solution. Indeed, although new thin films

    cells have been developed during the last 20 years, at the

    end of 2010, crystalline silicon solar cells share87% ofthe worldwide photovoltaic market. This success is due

    to the stability of the photovoltaic silicon structures, toon the cell conversion efficiency. Therefore, there

    was a competition between monocrystalline and

    multicrystalline solar cells during the last 20 years,

    and today, both type of cells share approximately the

    same part of the worldwide market. The conversion

    efficiency of commercial monocrystalline cells achieve

    1824%, while that of multicrystalline cells, whichincreases continuously, is by 23% absolute less higher.

    Although multicrystalline wafers contain large

    densities of imperfections like grain boundaries, dislo-

    cations, and impurities, they lead to acceptable solar

    cells because the processing steps needed to make the

    devices (phosphorus diffusion, aluminumsilicon

    alloying, deposition of a hydrogen-rich antireflection

    coating) improve the electrical properties of the wafers,

    enhancing the minority carrier lifetime and diffusion

    length, unless the cost of the cells be increased.

    An intense research activity was developed in the

    field of impurity gettering and passivation, which

    contributes to a clear understanding of the influence

    of metallic elements, distinguishing the role of fast

    and slow diffusers in silicon. It was clearly shown

    that the extended crystallographic defects, by

    themselves, are not very detrimental to the electrical

    properties of the wafers, and it is the interaction

    between impurities and crystallographic defects which

    impair the electrical properties.

    From a material point of view, R&D efforts must

    lead to a very abundant production of a low-cost

    solar-grade silicon feedstock which can be used in the

    monocrystalline ingots pullers, as well as in the casting

    of multicrystalline ingots. The crystalline and the

    electrical quality of the ingots, as well as that of

    the wafers, must be improved and controlled. In order

    to save a pure and crystallized material, the thickness of

    the wafersmust be reduced to less than 120 mm.To avoidthe sawing of ingots, maybe ribbons or sheets could

    replace the wafers cut out from a crystalline ingot.

    Solar cell structures are still improving at an

    industrial production level because well-investigated

    concepts are included in the processing steps. Such con-

    cepts like the selective emitter, the back contact, and

    back junction cells, the use of heterojunctions in place

    of homojunctions, present definite advantages which

    contribute to reach conversion efficiencies higher than

    20% at an industrial level and 24% at a laboratory scale.

  • The Purification of MG-Si

    reaction with HCl at 575 K, as indicated by the overall

    reaction:

    Si s 3HCl g ! SiHCl3 g H2 g 2Obviously, other chlorides are also produced, like

    SiCl4.

    Trichlorosilane is generally transported to

    polysilicon production plants where it can be purified

    to a very high level by fractional distillation because

    its boiling point is higher than those of other

    chlorides [1]. Boron, phosphorus, and metallic atom

    concentrations are reduced to less than 1 ppba (9 N

    level).

    (Notice that a large par of SiCl4 can be converted

    into SiHCl3 by simple processes.)

    A lot of extraction techniques have been proposed

    in the past [2]; however, today, the Siemens process is

    dominant [3]. It is based on the bell-jar reactor system

    and on elaborated techniques to recover the chemical

    by-products from the reactors. Inside the reactor

    228 Silicon Solar Cells, CrystallineThe required purification results from two different

    ways: the chemical or liquid routes and the direct

    metallurgical routes.

    The Chemical Route The requirements of the semi-

    conductor industry appear extremely hard to be

    satisfied but, fortunately, chlorine when purchased as

    HCl and MG-Si are inexpensive, and that open the

    way for the use of trichlorosilane SiHCl3 (TCS).

    TCS is a very pure precursor for the deposition of

    EG polysilicon by the chemical vapor deposition tech-

    nique (CVD). Indeed, MG-Si is transformed mainly

    (90%) into trichlorosilane SiHCl3 (TCS) by theSuch results are close to the theoretical limits of

    silicon solar cells based on single junction, and the way

    is open for new cell structures, like tandem cells and new

    silicon-based materials (nanomaterials, intermediate

    band materials, nanowires, quantum dots).

    Crystalline silicon solar cells are made with wafers

    that are cut out frommonocrystalline ormulticrystalline

    ingots after some processing steps. Ingot growth requires

    very pure silicon feedstock, although the purity level is

    lower than that needed for electronic devices. The fol-

    lowing chapters describe successively the feedstock pro-

    duction, the ingot growth and properties, the

    conventional and advanced cell structures.

    The Silicon Feedstock (Polysilicon)

    The Metallurgical Silicon

    Metallurgical-grade silicon (MG-Si) is produced by the

    carbothermic reduction of pure quartz and quartzite

    rocks (silica), as indicated by the overall reaction:

    SiO2 s 2C s ! Si l 2CO g 1The reaction occurs in an electric arc furnace and

    requires a huge amount of electrical energy.

    MG-Si is99% pure (2N), andmajor impurities areAl, Fe, and C. MG-Si is mainly used in the aluminium

    and iron industries (60%) and for the production ofsilicones (25%). Only a few part (15%) is highlypurified in order to reach the electronic-grade (EG-Si),

    at least 9 N pure, required by the semiconductor indus-

    try, or 67 N needed for solar cells.chamber, gaseous trichlorosilane is dissociated at

    575C in elemental silicon thanks to resistance-heatedfilaments. Several hundreds of reactors are located in

    one silicon plant in order to produce several thousands

    of metric tons a year. A schema of the Siemens-type

    bell-jar reactor is given by Fig. 1. TCS is carried by

    preheater

    Quartzbell jar

    Siliconcore

    polysilicon

    TCS + H2exhaust Electric energy

    Silicon Solar Cells, Crystalline. Figure 1

    The Siemens bell-jar reactor

  • instance, Wacker Chemie is contributing to polysilicon

    supply by expanding its annual capacity to more than

    25,000 metric tons [6]. Probably, the worldwide

    amount will be larger than 150,000 metric tons in 2010.

    TheMetallurgical Route For solar cells, a solar-grade

    silicon (SoG-Si), less pure than EG-Si, could be accept-

    able at a 67 N level. Downgrading the Siemens process

    to make it less energy consuming was proposed, but

    such a technique does not enable to reach the capacity

    production for GWp installation of solar cells.

    Another approach is to upgrade MG-Si more

    cheaply with metallurgical methods, e.g., slagging,

    leaching, or impurity segregation combined with

    directional solidification, as proposed by Crystal

    Systems Inc. [7], leading to the so-called upgraded

    metallurgical silicon (UMG-Si). Investment cost and

    energy consumption are drastically reduced; the mate-

    rial could be very abundant, but the degree of impurity

    is higher, and refining processes are needed, especially

    to reduce the concentrations of phosphorous and

    Silicon Solar Cells, Crystalline. Figure 2

    Schema of the fluidized bed reactor

    229Silicon Solar Cells, Crystallinehydrogen, comes in contact with filaments heated at

    1,100C, and then, silicon is deposited in form of rods.The sizes of a 40-metric tons/year capacity reactor are

    1.2 m in diameter and2.5 m in height. The reactorscan dissipate thousands of kilowatt of heat, accept and

    exhaust thousands of cubic meters of hydrogen per

    hour, use corrosive HCl, and may produce 1,000 kgof silicon in a single run. The electrical energy

    consumption is 120160 kWh/kg.

    The Siemens process is very efficient; however, the

    plants are very expensive and need huge investment as

    a 1,000 t/y production plant requires 100 M.Moreover, the chemical products are hazardous

    (SiHCl3; SiH4) or explosive (SiH2Cl2), and a lot of

    cautions must be taken. These plants are in fact more

    suited to the silicon production for the electronic

    device industry than for a mass production devoted

    to solar cells.

    Other techniques to produce polysilicon have been

    investigated like the fluidized bed reactor (FBR), the

    vapor to liquid deposition process, or the free-space

    reactor. The more efficient is FBR, investigated by REC

    and Wacker Chemie [4]. In this technique, silicon

    fluoride instead of MG-Si is converted to monosilane

    SiH4, and then, a silicon seed is dropped into the

    reactor in a flow of monosilane and hydrogen gases,

    as shown by Fig. 2. At the heated zones of the

    reactor, the gaseous flow dissociates to solid silicon in

    form of pellets. The theoretical advantages of this

    process are the possible continuous operation as well

    as lower capital and electricity costs. The produced

    polysilicon is less pure than that given by the Siemens

    process and could be a good candidate for photovoltaic

    devices.

    Polysilicon could also be obtained through decom-

    position of silane (SiH4), as indicated by the overall

    reaction:

    SiH4 g ) Si l 2H2 g 3The advantages are that the temperature needed for

    decomposition is lower than for TCS, the conversion

    efficiency is higher, and corrosive compounds are not

    formed [5].

    It is difficult to give the amount of polysilicon

    produced in 2010 because all the production plants

    have increased their capacity in order to satisfy the

    requirements of the photovoltaic industry. ForSilicon seedintroduction

    heater

    Effluent gas

    hydrogen

    SiCreactorwall

    Reactor gaz inlet +Granules removal

  • Promising results have been obtained with the Elkem

    Solar Silicon (ESS) [18, 19] as well as with the Sintef

    Solsilc silicon [8] when the material is blended with EG

    polysilicon (ESS results from the reaction between

    highly pure raw materials like carbon black and high-

    purity quartz).

    Several other techniques of refining MG-Si have

    been proposed, such as slagging or segregation steps.

    However, the dopant concentrations, especially

    boron, are generally too high (above 5 1017 cm3).The mixing of EG and UMG feedstock was used

    successfully and more recently by IBM [20]. Details of

    the processes are generally missing, and information

    may only come from the laboratory scale about plasma

    torch process [16] used by FerroPem. In such

    a process, the feedstock is introduced and melted in a

    crucible, and an argon plasma gas, containing other

    230 Silicon Solar Cells, Crystallineboron (n- and p-type dopants) which are difficult to

    eliminate, in order to avoid a compensation of silicon

    and a reduction of minority carrier lifetime.

    The metallurgical refining techniques include

    raw material selection, segregation, leaching, and slag

    treatment. However, the most critical point remains the

    dopant concentration.

    High-purity precursors have been used in the frame

    of the Sintef Solsilc project [8]. Ultrapure silica is

    obtained by selection of high-purity quartz or

    by oxidation of gaseous silicon chloride. This silica is

    reduced by black carbon obtained by gaseous route in

    an arc furnace. However, the high-purity silicon

    obtained is polluted by SiC, which has to be removed

    by several solidification steps.

    The choice of pure quartz and pure reducer,

    generally charcoal, is the first step to obtain MG-Si

    with a low content of boron and phosphorus. At this

    stage, metallic pollutants are not the most critical

    because they can be reduced afterward by segregation

    or acid leaching, what is not the case of boron and,

    to a lower extent, of phosphorus. Slag treatment can be

    used to reduce boron [9]; it is the route developed by

    Elkem.

    Another way to remove boron or phosphorus is to

    use plasma treatment or electron beam evaporation

    under low pressure. The first studies on plasma

    treatment have been carried out at laboratory scale by

    a French research group [10]. But it is the Kawasaki

    research group, in the frame of a NEDO project, which

    has developed a four-stage process. This process used

    electron beam for phosphorus removal, arc plasma

    with steam added argon for boron removal, and two

    segregation steps [1113]. The boron and phosphorus

    concentrations were reduced to 0.2 ppma.

    Alemany et al. have developed, at laboratory scale,

    a process involving an inductive plasma torch with the

    electromagnetic stirring of the molten silicon [14, 15].

    The reactive gas was hydrogen and oxygen.

    This process has been implemented by FerroPem in

    the frame of the Photosil project [16].

    Several companies and academic laboratories

    [6, 17] are currently involved with the development of

    solar-grade silicon by different approaches. For

    instances, Elkem Solar uses in house production of

    MG-Si followed by three sequential purification

    steps to reduce impurity level of critical elements.reactive species, is created by induction. The plasma is

    blown onto the surface of liquid silicon, which is

    continuously renewed by electromagnetic stirring.

    Impurities like B, C, Al, and Ca are partly removed

    and volatilized. It is less easy to remove phosphorus

    atoms, even by evaporation, and the best solution is to

    lower the phosphorus content at the beginning of the

    entire process. After the plasma treatment, the

    concentration of boron is reduced by a factor 23.

    Following such treatment, the boron concentration

    can be adjusted to less than 2 ppma. Figure 3 shows

    a picture of the plasma torch in operation.

    Silicon Solar Cells, Crystalline. Figure 3

    Purification of melted UMG-Si by the plasma torch

  • the top of the rod, and the rod is rotated in order to

    reduce the inhomogeneities. Figure 4 shows a diagram

    of the pulling process.

    During the growth process, two basic effects are

    operating, crystallization and purification. Indeed,

    impurities, with higher solubility in the liquid

    compared to that in the solid phase, are transported

    with the molten zone and can be accumulated at the

    extremities of the grown crystal, as a consequence of

    the impurity segregation phenomena, resulting of their

    distribution coefficients between the solid and the

    liquid silicon phases, given in Table 1.

    The repeatability of the process, the lack of crucible,

    and the use of a neutral atmosphere lead to a quasi-

    perfect single crystal (practically without dislocations)

    of extreme high purity, provided the silicon rod is itself

    of high purity. These basic steps have not changed since

    they were first established in the 1950s [2224], and

    if FZ silicon crystal sizes have considerably increased,

    such crystals are always the best from a crystallographic

    as well as from a purity level point of view. Indeed,

    today, commercial FZ silicon ingots achieve15 cm in

    231Silicon Solar Cells, Crystallinedescribed in Ref. [1]), while the properties of

    multicrystalline silicon will be detailed in section

    Physical, Chemical and Electrical Properties of mc-Si

    Wafers.

    Float-Zone-Grown Monocrystals (FZ)

    The FZ pulling method, described in details by Dietze

    et al. [21], has the great advantage to be crucible free.

    A rod of polysilicon, containing a monocrystalline seed

    at its bottom, experiences a local fusion by means of an

    RF coil. The molten zone is moved from the bottom toNotice that a marked advantage for refined UMG-Si

    is that the processes for producing solar cells are similar

    to those used for current silicon material but with

    different parameters to be applied, as will be shown later.

    Silicon Single Crystals, Multicrystals, Ribbons,

    and Sheet Growth Techniques

    Crystalline siliconwafers used for solar cells are cut from

    monocrystalline float zone (FZ), from Czochralski

    grown (Cz), from multicrystalline (mc-Si) ingots, or

    from ribbons. Such crystals are grown from a melt

    contained in various crucibles, except for FZ crystals.

    Cz silicon is mainly used for manufacturing highly

    integrated low-power devices, especially memories,

    whereas FZ silicon is mainly used for power devices

    and photo detectors. Both crystals can be used

    for high-efficiency solar cells, especially FZ silicon.

    In 2009, 37% of the photovoltaic market wasshared by single-crystalline cell modules, 45% bymulticrystalline cell modules, and 2% by ribbon-based cell modules.

    The crystalline ingots are cut into thin wafers

    (180 mm) by means of a diamond blade or a wiresaw, while from ribbons wafers, crystalline ingots are

    cut out by means of a laser.

    In the following paragraphs, a few basic properties

    of single crystals are given (they are well known andAdditional problems in the material come from

    light elements like C and O, as well as from slow

    diffusers like Ti or Al.

    Researches have been carried out recently in Europe

    and Japan to develop cleaning and crystallization pro-

    cesses for metallurgical silicon feedstock and wafers. Due

    to confidential restrictions, only few results are known.Polycrystallineingot

    Molten silicon

    RF coil

    Grown singlecrystalline material

    Single crystalline seed

    Silicon Solar Cells, Crystalline. Figure 4

    Diagram of the FZ growth process (Courtesy of UNSW)

  • ie

    ib

    04 Bismuth 7 104

    104 Oxygen 1.25

    02 Sulfur 105

    101 Manganese 10 5

    102 Iron 8 106

    04 Cobalt 8 106

    101 Nickel 3 105

    01 Tantalum 107

    102

    Argon

    Si melt

    SiO

    Single crystal

    Seed

    Quartz crucible

    Susceptor

    Bottle neck

    232 Silicon Solar Cells, Crystallinediameter, 60 cm in height, and the purity level ishigher than 10 N (less than 1 ppba, except for oxygen

    and carbon). In FZ single crystals, the minority carrier

    diffusion length and lifetime can reach a few mm and

    a few ms, respectively.

    Czochralski-Grown Monocrystals (Cz-Si)

    Invented in 1918, the crucible pulling of single crystals

    according to the Czochralski (Cz) technique [25] was

    applied first to germanium crystals and later to silicon

    [26]. As shown in Fig. 5, polysilicon is melted in a pure

    quartz crucible, and a seed crystal just touches the top

    Silicon Solar Cells, Crystalline. Table 1 Distribution coeffic

    silicon (From Ref. [1])

    Impurity Distribution coefficient Impurity Distr

    Lithium 102 Indium 4 1Copper 4 104 Thallium 1.7 Silver 106 Carbon 6 1Gold 2.5 105 Germanium 3.3 Zinc 105 Tin 1.6 Cadmium 106 Nitrogen 7 1Boron 8 103 Phosphorus 3.5 Aluminium 2 103 Arsenic 3 1Gallium 8 103 Antimony 2.3 of the liquid before it is then slowly retired in order that

    liquid silicon solidifies close to the seed. The slow

    vertical ascent of the solidified silicon gives rise to

    a single-crystalline ingot, which emerge from the

    melt. To obtain a dislocation-free material, a crystal

    neck must be grown which becomes dislocation-free

    after few cm because dislocations grow laterally out

    of the crystal. Then, the diameter is enlarged, and the

    crystal growth is finished without dislocations by

    reducing the diameter to zero at the bottom of the

    crystal.

    Like for FZ pulling, the Cz process is

    a crystallization and a purification step, and a large

    part of the impurities contained in the polysilicon

    and coming from the silica crucible remains in the

    melt, except oxygen which the distribution coefficient,

    given in Table 1, is higher than 1.nts k between liquid and solid phases of some impurities in

    ution coefficient Impurity Distribution coefficientFortunately, silicon monoxide evaporates easily

    from the melt. Nevertheless, oxygen concentration

    can achieve up to 20 ppma in the crystals, which, in

    Heater

    Argon + SiO +CO

    Silicon Solar Cells, Crystalline. Figure 5

    Diagram of a Cz single-crystal growth puller. Notice the

    counter rotation of the pulled single crystal and of the

    crucible

  • 233Silicon Solar Cells, Crystallinethe reader could consult the chapters 68 in Handbook

    of Semiconductor Silicon Technology [1].

    Albeit they are close to perfection, FZ and Cz

    crystals contain imperfections like swirl defects. Such

    defects are more concentrated in Cz crystals due to

    oxygen-related defects coming from an oxygen super

    saturation and a slightly higher metallic impurity level.

    Nevertheless, such crystals are used to make integrated

    circuits for electronic device thanks to internal

    gettering effects [27] and to the formation of

    a precipitate-denuded zone in the region where the

    components are realized. Large-area efficient solar

    cells can also be obtained, provided that the material

    does not experience high-temperature anneals

    (above 900C) which could lead to the formation ofoxygen-related precipitates and stacking faults.

    In Cz single crystals, minority carrier diffusion

    lengths and lifetime are higher than 400 and 200 ms,respectively.

    However, in p-type Cz wafers, the high concentra-

    tion of oxygen led to the formation of boronoxygen

    complexes under sunlight, and this will be a drawback

    for the high-efficiency cells as will be shown later.

    Topsil in Denmark, MEMC in USA and in Italy, as

    well as Siltronic in Germany and France are the more

    important producers.

    Multicrystalline Silicon (mc-Si) Ingot Growth

    For a massive production of large-size cells, mc-Si cast

    ingots are the material of choice. In 2007, more than 4

    GWp of mc-Si modules have been installed. In 2010,

    mc-Si modules shared more than 45% of the 24 GWp

    produced.

    If single-crystalline silicon wafers are a very mature

    product, which the properties are well controlled,

    mc-Si ingots suffer from a large variation of basicfact, are supersaturated at room temperature.

    To enhance the removal of SiO and to prevent

    a contamination by CO, the furnace is purged by a

    strong argon stream. Today very large commercial

    single-crystalline ingots are produced by the Cz

    techniques up to 30 cm in diameter and 1.52.5 m in

    height. Ingots, 45 cm in diameter, are announced. The

    properties of these single crystals have been widely

    investigated, and it is impossible to give an exhaustive

    bibliography in this chapter. For detailed information,properties like crystalline defect and impurity concen-

    trations. Each ingot may have various properties,

    depending on the growth techniques, and in a given

    ingot, the basic properties of the wafers are depending

    on their position in the ingot. This is why the crystal-

    line, the chemical, and the electrical properties of mc-Si

    wafers will be detailed in section Physical, Chemical

    and Electrical Properties of mc-Si Wafers.

    In order to decrease the cost of the material and to

    increase the size of solar cells, a lot of researches were

    devoted to the production of large grain size (few mm

    to few cm) crystalline silicon in which the concentra-

    tion of impurities could be below the 0.01 ppma.

    Square cells, up to 21 21 cm2, could be realized.Such a material was frequently labeled solar-grade

    silicon because its utilization is restricted to the

    production of solar cells. In addition to metallic impu-

    rities of carbon and, at a lesser extent, of oxygen and

    nitrogen, mc-Si wafers contain extended crystallo-

    graphic defects like grain boundaries (GBs), disloca-

    tions, and twin boundaries. Moreover, metallic

    precipitates are frequently formed at extended defects

    and within the grains. When the carbon and oxygen

    concentrations are both higher than 10 ppma, a severe

    crystallographic degradation occurs with the formation

    of grit structures. In these grit structures, it is found

    with high concentrations of C, O, and SiC, and the

    grain size becomes very small. Such regions cannot

    contribute to the photocurrent of solar cells, but

    today, such bad regions are practically eliminated by

    a better control of the growth conditions.

    The first material produced at an industrial level

    was the so-called Silso, developed by Wacker/

    Heliotronics [28] in Germany at the end of the 1970s.

    It was followed by other companies like Photowatt in

    France with the Polix in USA and in Japan. More

    than 220 bibliographic references of papers and patents

    can be found in Ref. [30], for a period going from 1970

    to 1980 only. Today, mc-Si is a quasi-standard product

    produced worldwide, especially in China and Taiwan,

    which the growth is based on the directional solidifica-

    tion, and most of the research and development

    activities are concentrated in Germany, Japan, and

    more recently in China.

    The basic principle is the following. In a quartz

    crucible, heated by graphite heaters and coated with

    Si3N4, silicon (EG-Si or silicon waste coming from the

  • is the quasi-monocrystal or the monocast silicon. This

    point will be developed at the end of the chapter in

    Future Directions.

    Wafer Sawing Wafers are commonly cut out of the

    ingots or of the bricks bymeans of wire saws using a SiC

    loose abrasive, which enables the production of

    200 mm thick slices. This technique possesses a highthroughput, but up to 50% of the crystallized and

    purified silicon is lost during process. In order to

    reduce this waste, 100 mm diameter wires can be usedwith fine SiC particles in the slurry [35]. The sawing

    can also be improved, employing diamond grains fixed

    onto the surface of the wires by means of a resin [36],

    because the sawing speed is increased by 2.5 while

    the thickness variation is decreased by 3. Moreover,

    after sawing, the slurry could be reused because it

    does not contain SiC particles like in the conventional

    process. Obviously, kerf-loss-free techniques are also of

    a great interest, as that proposed recently by IMEC

    group [37].

    Directional solidification of multicrystalline silicon ingot.

    A temperature gradient results from heat extraction

    through the crucible bottom

    234 Silicon Solar Cells, Crystallinemicroelectronic industry) is melted. Then, a vertical

    temperature gradient between the bottom and the top

    of the melt is established in order that the bottom

    solidifies first and the top solidifies last. Crystallization

    begins at the contact of the crucible bottom when its

    temperature decreases below the fusion point. After

    a few cm from the bottom of the crucible, quasi-

    columnar grains grow more or less vertically. Larger

    blocks can be obtained up to 650 kg in weight and up to

    35 cm in height [29, 30].

    The temperature gradient results from a mobile

    heating RF core or, more simply, by the so-called heat

    exchange method [29], in which the bottom of the

    crucible is cooled by a gas flow. Other methods of

    cooling use a natural cooling down due to the

    removing of the thermal insulation of the crucible

    bottom or, more originally, the action of an infrared

    radiation transparent bottom [31]. The process time

    for one cycle, from loading the crucible to unloading,

    is usually around 48 h. Figure 6 summarizes the

    directional solidification process.

    The dominant (in concentration) impurities in

    mc-Si are carbon, oxygen, and nitrogen due to the

    contact with the crucible followed, at a lesser extent,

    by metallic atoms.

    Notice that larger ingots can be produced by the

    electromagnetic continuous casting into a cooled

    crucible. This technique was first developed by Sumco

    in Japan, formerly Sumitomo-Sitix [32], and then in

    France by EMIX [33, 34]. The principle of the electro-

    magnetic continuous casting is to feed continuously

    a copper cold crucible in which alternative

    electric current is flowing. This current creates an

    induced current in the silicon, and the Joule effect

    heats the material up to the melting point. Simulta-

    neously, the induced magnetic field and current create

    a Lorentz force close to the surface of the melt that

    prevents contact with the cold crucible wall (a gap of

    some hundreds of mm exists). The process begins byheating a graphite piece which is pulled downwhen the

    Si melt is large enough. The pulling rate is typically

    50 mm s1.The blocks are cut into columnar bricks, which are

    wire-sawed into thin wafers (200 mm), then cleanedand texturized.

    Notice that a new product will be proposed which

    combines the advantage of mono- and multicrystals: itheat

    crucible

    Solid silicon

    Liquid silicon

    Silicon Solar Cells, Crystalline. Figure 6

  • 235Silicon Solar Cells, CrystallineSi melt

    Ribbon

    Die

    Silicon Solar Cells, Crystalline. Figure 7

    Schema of the growth of an edge-defined film-fed (EFG)Ribbons The sawing of the ingots wastes at least 40%

    of an expensive, pure, and crystalline material

    (the subproduct powder is too much contaminated to

    be reused). To avoid this wasting, the growth of shaped

    crystalline ribbons from silicon melt, which has been

    already investigated in the 1970s, became of interest

    during the last 20 years [3840].

    Ribbons have been prepared mainly by various

    techniques such as edge-defined film-fed growth

    (EFG), edge-supported pulling called also String Rib-

    bon (SRT), ribbon on graphite substrate (RGS), and

    ribbon on sacrificial carbon template (RST). EFG and

    SRT techniques have, today, reached the production

    level. Such techniques are recognized to enable the

    direct production of flat wafers after laser cutting.

    The EFG ribbons, which the growth technique is

    described in Fig. 7, can be pulled in a multiple ribbon

    growth furnace, for instance, in shape of nonagon tubes

    with a side width of 156 mm or in form of 12-face

    tubes with a side width of 125 mm. The length of the

    tube can be as long as 6.5 m [41, 42]. The EFG ribbon is

    developed by RWE Schott Solar.

    ribbonIn the STR ribbon [4346], two high-temperature

    Ribbon

    Silicon melt

    Seed

    Filament orString

    Silicon Solar Cells, Crystalline. Figure 8

    Schema of the growth of a string (STR) ribbonresistant wires (strings) are pulled vertically through

    a shallow siliconmelt, and the molten silicon spans and

    freezes between the strings, as shown by Fig. 8.

    The process is continuous: long strings are used, the

    melt is replenished, and the silicon ribbon is cut to

    length for further processing, without interrupting

    growth. The STR ribbon was developed by Evergreen

    Solar.

    RGS ribbons that consist mainly in melting of

    silicon powders possess the advantage of a higher

    throughput production, but it remains to a pilot line

    production due to contamination problems [47].

    These horizontal growth processes, described by Fig. 9

    which was developed at ECN, the Netherlands, have the

    highest throughput rate. However, these processes

    appear to be bound to the production of thick wafers,

    typically over 400 mm, with a poor crystalline texture,and, in turn, lower cell conversion efficiencies.

    Moreover, extended chemical treatments are necessary

    to eliminate surface corrugations and backside

    contamination of the sheets.

    The RST ribbon growth process [48] is based upon

    the crystallization of two opposite silicon films drawn

  • of carbon and also of metallic impurities coming from

    the dies. There are also large densities of intragrain

    defects like dislocations and twins. Dislocations are

    detrimental, but the minority carrier lifetime is higher

    in twin lamellae lying parallel to the growth direction.

    The main impurity is carbon, with a substitutional

    concentration of 1018 cm3, exceeding the solubility

    Throughput (cm2/min) >300 165 55 (4 ribbons)

    Thickness (mm) 80 300 280

    236 Silicon Solar Cells, Crystallinealong by a carbon ribbon substrate, which is pulled

    upward through a siliconmelt (see Fig. 10). The carbon

    ribbon is passed vertically through a slot at the bottom

    of a crucible (silica or carbon), which contains the melt.

    As it emerges from the melt upper surface, this ribbon

    shapes the freestanding silicon meniscus and is coated

    on both sides with silicon layers.

    After growth, the carbon ribbon is burnt in an

    oxygen-containing gas at high temperature, upon

    which the self-supported silicon sheets are released

    and ready for the fabrication of the solar cells. The

    substrate participates in the elimination of the latent

    heat of crystallization, which allows relatively high pull

    rates. It shapes the freestanding freezing meniscus,

    which ultimately yields essentially flat silicon sheets

    free of grooves. However, thermoelastic stresses, gener-

    ated by the difference in thermal expansion coefficients

    between the silicon films and the carbon substrate, set

    a limit to the minimal thickness of the silicon films at

    Si melt

    Ribbon

    Substrate

    Silicon Solar Cells, Crystalline. Figure 9

    Schema of the growth of a ribbon on graphite substrate

    (RGS)around 4050 mm.Table 2 gives the throughput rate (cm2/min) and

    compares the actual thickness for vertical ribbon

    growth. Although not in production yet, the RST pro-

    cess, developed by Solar Force in France, is the only

    ribbon process which can simultaneously achieve thin

    silicon films with relatively flat and smooth surfaces

    and a high throughput rate. Typically, at a pull rate of

    10 cm/min, the thickness of the silicon films is 80 mm(or below).

    In all ribbon technologies, there is a very high-

    temperature gradient at the solidliquid interface

    related to the high growth speed. Wafers cut out of

    ribbons are multicrystalline and contain large amountsPull direction

    High pull

    Extendedgrowth front

    Silicon Solar Cells, Crystalline. Figure 10

    Schema of the RST ribbon growth process

    Silicon Solar Cells, Crystalline. Table 2 Throughput rate

    and actual thickness for vertical ribbon growth processes

    Process RST EFG STRlimit at the melting temperature.

    In addition to the material saving (there is no

    kerf-loss), ribbons posses another advantage: the

    energy pay back of the cells is lower than that of other

    crystalline cells because the pulling processes are less

    energy consuming.

    Silicon Sheets Another technique to produce directly

    large-size flat wafers proposed by the Sharp group in

    1997 is the crystallization on dipped substrate (CDS).

    The basic principle of CDS technology consists of to

    dip a cold refractory substrate into molten silicon. The

    silicon crystallizes uniformly; then, the substrate is

    pulled out of the molten silicon; a multicrystalline

  • crystals, multicrystalline silicon wafers contain

    grain boundaries (GBs) and intragrain defects like

    subgrain boundaries, dislocations, and twin bound-

    aries, as shown by Fig. 12. Due to a higher impurity

    concentration, precipitates could be formed within the

    grains or at extended defects.

    Notice that the grain growth is columnar, and this is

    a great advantage because in the wafers cut perpendic-

    ularly to the ingot growth direction, there is only one

    High porosity Substrate

    Silicon Solar Cells, Crystalline. Figure 11

    Microphotography of an epitaxial layer before detachment

    Silicon Solar Cells, Crystalline. Figure 12

    Photography of a polix mc-Si wafer after alkaline etching

    237Silicon Solar Cells, Crystallinesilicon sheet is detached from the substrate and cut by

    a laser to 15.6 15.6 cm2 size. The main advantage ofthe CDS is a high production throughput which

    achieves 1,825 cm2/min [49].

    The wafer molding has also been studied by some

    laboratories, and the molded wafers have been realized

    on graphite mold coated with SiC or Si3N4.

    The research efforts are focused today on the

    improvement of molded wafers photovoltaic quality

    and on the reduction of their thickness. Moreover,

    directional crystallization is obtained by means of

    a reusable oriented seed.

    Transfer Layer To produce thin substrates of mono-

    crystalline silicon, less than 50 mm, a new technologicalroute has been considered: the transfer of thin film

    [5053]. It is based on the use of a sacrificial layer of

    porous silicon on which grows by epitaxy, the active

    layer of silicon, as described by Fig. 11. Then, the thin

    film is transferred on a low-cost substrate and depends

    on the architecture of the cell considered. The starting

    silicon substrate can then be reused after cleaning its

    surface. This avoids the sawing step.

    Porous silicon (PS) is formed using a specific

    anodization cell. Indeed, in order to simplify the

    further transfer step, porous silicon is formed on

    the whole surface. Electrochemical anodization of

    (100) silicon wafers in HF solution is realized on the

    entire surface. A double-phase etching process generates

    two superposed PS layers: a top layer with low porosity

    (2023%) allowing high-quality crystal growth and

    a buried layer with high porosity (6570%) for the

    further detachment of the epilayer [54, 55].

    Physical, Chemical, and Electrical Properties of

    mc-Si Wafers

    The huge number of publications and conferences

    dedicated to the properties of mc-Si wafers is so high

    that it is impossible to respect an exhaustive list in the

    references. Moreover, progresses are so fast that only

    some historical and recent results will be given (233

    references can be found in Ref. [30] covering the results

    up to 1980).

    Extended Crystallographic Defects and Impurities

    In addition to the whole defects found in singleEpitaxial layer = 40 m

    Low porosity restructured

  • a polix ingot [69]. As shown by Fig. 13, there is an

    increase of t at the vicinity of GBs which could beexplained by the trapping of impurities (metal; oxygen)

    by these defects.

    The electrical activity of extended crystallo-

    graphic defects is related to the presence of shallow

    or deep energy levels in the silicon band gap.

    A simple distinction could be made by cooling the

    wafer and investigating their electrical activity by elec-

    tron or light beam induced current (LBIC) contrast. If

    the contrast increases when T decreases, deep energy

    levels have to be considered, generally related to

    metallic impurities. If the contrast is irrespective of

    the variations of T, shallow levels are dominant

    [7072]. Figure 14a, b gives an example of such vari-

    ation. Low-temperature scan maps reveal the pres-

    ence of extended defects in raw samples which are

    poorly recombining in the initial material. Such

    defects are sleeping defects which are activated

    238 Silicon Solar Cells, Crystallinegrain between the front and the back surface.

    Consequently, the photogenerated carriers do not

    have to cross grain boundaries before to be collected

    by a junction.

    The large grains have various orientations; the

    dominant one is (111). GB types vary from one grain

    to another, and one can distinguish small-angle GBs

    with a misorientation level from 0 to 10 fromlarge-angle GBs with larger misorientation angles

    [56, 57]. In these large grained materials, dislocations

    are the more harmful intragrain defects. Dislocations

    aremore or less homogeneously distributed or agglom-

    erated in clusters, which frequently induce metallic

    atom precipitation [58].

    Crystallographic defects are harmful to the electrical

    properties of the material, mainly in terms of minority

    carrier lifetime t or diffusion length L. However, bythemselves (without any decoration by impurities),

    their influence is weak, and it is the interaction between

    the crystallographic defects and segregated or precipi-

    tated impurities (light elements; metallic atoms), which

    generates the more harmful recombination centers.

    The effect of twin boundaries has also been

    extensively studied. Exact twinning (the S3 twin) hasno electrical effect because all bonds are saturated and

    there is no distortion of the bond angles across the twin

    boundary. More deviated twins show distorted

    reconstructed bonds and a dislocation structure,

    and they have an electrical effect, as was shown for S9twins. Subgrain boundaries formed essentially by

    dislocations [59] are very detrimental in terms of

    minority carrier lifetime reduction.

    From an electrical point of view, rain boundaries

    (GBs) are characterized both by an electrical potential

    barrier which reduces the mobility of the majority

    carriers and energy levels in the band gap which

    reduce the lifetime t (or diffusion length L) of theminority carriers. However, electron beam induced

    current (EBIC), as well as light beam induced current

    (LBIC), scan maps indicate that this electrical activity

    is inhomogeneous from 1 GB to another one and also

    along a given GB [6062]. In fact, this activity

    depends strongly on the segregation of metallic

    atoms or of light elements. In large grained wafers,

    the electrical influence of GBs is limited, while that

    of subgrain boundaries and of dislocations is dom-

    inant, especially when they are decorated byimpurities. This was very well demonstrated first

    by Sopori [63] by means of diffusion length measure-

    ments as a function of dislocation density determined

    by etch pit counting, later confirmed by El Ghitani

    [64, 65], and more recently by Warta [66, 67]. Never-

    theless, the role of transition metallic impurities is

    dominant [68]. An illustration of the impurity defect

    interaction at GBs is given by the lifetime scan map

    around GBs in a raw wafer cut out from the top of

    tau (s)11

    9.7

    8.4

    7.1

    5.8

    4.5

    3.2

    2

    1000 m

    Silicon Solar Cells, Crystalline. Figure 13

    Minority carrier lifetime scan map around a GB in a wafer

    cut out from the top of a large polix ingot: segregation of

    impurities improves the grains near the grain boundary

  • 239Silicon Solar Cells, Crystalline.70

    .52

    .35after an annealing at T > 500C and becomerecombining after processing steps.

    Like in the FZ and Cz growth processes, liquid and

    solid phases are present during the solidification of the

    multicrystalline ingots, and in addition to the crystal-

    lization, a marked purification by impurity segregation

    occurs. Impurities which are contained in the silicon

    melt or which have been introduced by contact with the

    .17

    1 mm

    .70

    .52

    .35

    .17

    1 mm

    a

    b

    Silicon Solar Cells, Crystalline. Figure 14

    (a) Light beam induced current contrast scan map at 300 K

    of a typical raw polix sample. Few intragrain defects appear

    in the picture because the contrast is weak. (b) Same

    picture at 100 K of the sample scanned in a. Due to the

    presence of deep levels in the gap, probably associated to

    metallic impurities, both grains in the right of the picture

    appear in white. Due to the presence of shallow levels,

    a marked contrast appears at the extended defects in the

    left part of the picturecrucible walls are progressively accumulated at the very

    top of the future ingot. The segregation efficiency is

    directly related to the distribution coefficient k0 given

    in Table 1. Metallic impurities which the k0 value is

    below 103 tend to remain in the liquid phase, whileoxygen accumulates preferentially in the solid phase

    which solidifies first. Nevertheless, metallic impurities

    and light elements are present in the crystallized ingot

    in which they can be dissolved and precipitated in the

    homogeneous regions of the grains like at extended

    defects. As a consequence of all these imperfections,

    mc-Si ingots and wafers are characterized by

    a pronounced in homogeneity, and along a given

    ingot, the macroscopic and microscopic electrical

    properties of the wafers are not the same. It was

    found that the very bottom and the very top of the

    ingots must be discarded; the mean values of lifetime

    t and diffusion length L increase from the bottom tothe middle of the ingot and then decrease to the top

    [73]. Figure 15 shows the variation of L as a function of

    the height in a typical brick cut out of a large polix

    ingot. In virgin p-type wafers, the minority carrier

    properties vary very widely, diffusion lengths are

    found between 50 and 200 mm, and the lifetimes arein the range 320 ms. All materials need improvementsbefore to be used to make efficient solar cells;

    fortunately, improvements come from the solar cell

    processing steps.

    Material Improvement Techniques The electrical

    quality of the wafers, evaluated by the measurement of

    lifetime t or diffusion length L, could be improved by twobasic treatments: hydrogenation and impurity gettering.

    Hydrogen is able to in-diffuse rapidly as an atomic

    species H+, at T 450C, and can interact with dan-gling bonds and impurities [7476]. It was verified that

    deep-level transient spectroscopy (DLTS) signal of

    some impurities disappears after a hydrogen

    in-diffusion and reappears after a short anneal at

    T 500C. However, H+ ions tend to recombine toform hydrogen molecules which diffuse slowly, and the

    passivation depth is limited to a few tens of micrometer

    below the surface.

    Hydrogenation can be carried out by different

    techniques such as the immersion in direct or remote

    plasma, or by low-energy ion implantation, using

    a Kaufmann source. All these techniques have been

  • lon

    eig

    tte

    240 Silicon Solar Cells, Crystallinedescribed by M. Stavola [76]. However, the best

    0 5 10

    30405060708090

    100110120130140150160170

    Diff

    usio

    n Le

    ngth

    (m

    )

    Bottom 1022 cm3), and orthorhombicprecipitates of SiP are formed near the surface.Dislocations are formed at the interface between the

    15 20 25g the ingot=> Top

    RawP diff. 850C 30 mnP diff. 900C 2h

    ht (in cm) in a polix ingot for raw and phosphorus diffused

    ring treatment)precipitates and the cubic silicon matrix, leading to

    a relaxation gettering in the n+ layer. Moreover, due

    to the molar volume expansion developed by these

    precipitates, self-interstitials are injected in the bulk

    in which they interact with dissolved and precipitated

    impurities. Substitutional metallic impurities are trans-

    ferred in interstitials sites, oxygen-related precipitates

    are shrunken, and, provided they are fast diffusers,

    metallic atoms diffuse through the wafer and can be

    trapped in the n+ layer.

    On the other hand, a thermal treatment at

    T 850C for at least 30 min of a mc-Si wafer coveredwith a thick Al layer (1 mm thick) can lead to anefficient gettering as well. Indeed, at T > 570Ca liquid eutectic alloy is formed and vacancies are

    injected in the bulk. In the liquid alloy, the solubility

    of most metallic impurities is ten to thousand times

    higher than in the solid silicon [81], and fast metallic

    diffusers can be trapped easily. This is a pure segrega-

    tion gettering with an efficiency which depends on the

    diffusion coefficient of impurities at the annealing

    temperature. The best results are obtained for T in the

    range 700900C.

  • dislocation cores, at dislocation clusters, and at GBs.

    These precipitates can, in turn, trap metallic impurities

    and become recombining. This is probably why

    the wafer properties are degraded after anneals at tem-

    peratures higher than 900C.The improvement of the wafers is enhanced when

    phosphorus gettering, aluminium gettering, and

    hydrogenation are applied successively or combined,

    provided that phosphorus diffusion is applied first

    [8288]. Certainly, they induce complementary

    effects on the different defects and impurities

    contained in mc-Si wafers. For instance, the diffusion

    length scan map of a multicrystalline sample which

    experienced long phosphorus diffusion and then long

    aluminium gettering is given by Fig. 17. Very high

    values are found, indicating that the material has been

    strongly improved due to the extraction of impurities.

    241Silicon Solar Cells, CrystallineAfter each preceding treatments, t and L substan-tially increase, and such improvements are observed,

    irrespectively of the position of the wafers along the

    ingots, as shown by Fig. 15.

    If the gettering treatment can extract fast diffusers,

    it is practically ineffective toward slow diffusers which

    remain in the wafers and limit the increase of L. From

    the values of t0 (initial bulk lifetime) and of tdif(after phosphorus diffusion at 850C for 30 min), it ispossible to deduce a lifetime tMi related to the intersti-tial metallic atoms (Mi) which are extracted by

    gettering during the phosphorus diffusion, according

    to the expression:

    1

    tMi 1

    t0 1

    td i f5

    Assuming that the captured cross-section of the

    recombination centers related to these interstitial

    metal atoms is 1015 cm2 and that SRH statisticscould be used, it is possible to evaluate approximately

    their concentration [Mi]. The variation of [Mi] along

    the height of the ingot is given by Fig. 16. In wafers cut

    out of large (310 kg) polix ingot [73], it is found, after

    phosphorus diffusion, that [Mi] 1013 cm3 in thebottom, 1012 cm3 in the central part of ingots,and 5 1012 cm3 in the top. Half of these impurityamounts are due to iron, which the identification and

    the evaluation of the concentration as interstitial atoms

    [Fei] in p-type mc-Si are possible by means of the

    dissociation of the FeB complexes.

    It is important to evaluate also the concentration

    variations of recombination centers due to impurities

    precipitated or segregated by extended defects and

    also to slow diffusers. This variation could be given

    by that of the reverse of tget (after phosphorus diffu-sion at 900C for 2 h) because it could be reasonablyassumed that most of fast diffusers are removed what-

    ever is their position in the wafer (interstitial atoms,

    easily dissolved precipitates in grains and at GBs). It is

    observed that the concentration of such recombina-

    tion centers is more marked in the bottom and in the

    top than in the central part of the ingot. This is

    a consequence of the precipitation and segregation at

    extended defects of impurities coming from the cru-

    cible floor and of the accumulation of impurities in

    the top, linked to the duration of the solidification

    process.Figure 15 illustrates also the increases of the

    diffusion length after a long phosphorus diffusion,

    which is a gettering step, and shows that the gettering

    is very efficient in the lower middle of the brick because

    this part is protected from the back diffusion of impu-

    rities coming from the top and sufficiently far from the

    crucible floor.

    Due to the growth conditions of the mc-Si ingots,

    oxygen atoms are also present at concentration levels

    larger than those deduced from infrared spectroscopy.

    Indeed, with such a technique, interstitial oxygen

    atoms are detected while precipitated or aggregated

    ones are not. Oxygen precipitation occurs certainly at

    00.20.40.60.8

    11.21.41.61.8

    5 10 15 20 25h (cm)

    (Mi)

    (1013

    cm

    3 )

    Silicon Solar Cells, Crystalline. Figure 16

    Computed variation of the concentration of interstitial

    metallic atoms [Mi] along a polix ingot

  • hic

    0

    242 Silicon Solar Cells, CrystallineFortunately, the preceding treatments are

    processing steps which are included anyway in most

    of the preparation techniques leading to industrial

    mc-Si solar cells. For instance, for cells made with

    wafers of Fig. 12, diffusion length Ln values reached

    Silicon Solar Cells, Crystalline. Figure 17

    Minority carrier diffusion length Ln scan map of a sample w

    aluminium gettering; both treatments were carried out at 90

    at the grain and twin boundariesup to 300 mm [73]. These strong improvements explainwhile conversion efficiencies higher than 16% can be

    achieved, despite that the initial material properties

    were very poor compared to those of single crystals.

    Probably, it seems today that the major part of the

    improvement is due, after the phosphorus diffusion,

    to the aluminium gettering developed during the rapid

    thermal anneal at700Cwhich gives rise to the ohmicback contact [89] and to the back surface field effect, as

    will be shown in section Single Junction p-type Silicon

    Based Solar Cells (n+-p-p+ Structure).

    n-Typemc-Si Although the first solar cells have been

    made with n-type single-crystalline silicon, most of

    the industrial production is based on p-type material,

    essentially because the processing steps are simple and

    relatively not expensive. However, n-type material

    possesses some remarkable advantages. First, there

    are little or no boronoxygen complexes and no

    light-induced degradation. Secondly, the minority

    carrier capture cross-sections of metallic impurities,frequently found in processed silicon, are markedly

    smaller for holes than for electrons. This property of

    n-type siliconwas clearly explained [90] on the basis of

    the formation of donor trap levels in the gap which

    are positively charged or neutral and can be strongly

    400m

    302m

    205m

    107m

    100 m

    h experienced first a phosphorus gettering and then an

    C for 2 h. Notice the high values of Ln in the grains and alsoattractive for electrons or poorly attractive for holes.

    Experimental results have reported exceptional

    values of lifetime in raw n-type wafers cut from

    ingots made by Italsolar and then by Deutsche

    Solar or from ribbons. Indeed, Cuevas et al. [91]

    and Libal et al. [92] have found lifetime values of

    several hundreds of ms in raw materials and values inthe range of ms in wafers gettered by weak phosphorus

    diffusion. A lot of papers have investigated the role of

    impurities and have confirmed the interest of n-type

    material [9395]. For example, interstitial ironwhich is

    one of the more harmful contaminant in mc-Si wafers

    is a good example because the capture cross-sections

    for electrons and for holes, reported by Istratov [96],

    are sn = 5 1014 cm2 and sp = 7 1017 cm2,respectively. Assuming that Schockley-Read-Hall

    (SRH) recombinations occur and that interstitial

    iron is the dominant impurity, high lifetime values

    are expected for holes in n-type iron-contaminated

    silicon. The same tendency is observed for several

    other metallic impurities except for chromium for

  • on where UMG feedstock was obtained and on its

    impurity concentrations as well as on the nature of

    these impurities. This is why we give more details on

    the results obtained for ingots made with plasma torch

    refined UMG-Si [103, 104].

    The wafers cut out from the bottom and the

    middle of the ingots are p-type and contain high con-

    centrations of Al, Cu, and Zn (other impurities are

    below the detection limit of the high resolution glow

    discharge mass spectroscopy). Interstitial oxygen and

    substitutional carbon atom concentrations detected by

    Fourier transformed infrared spectroscopy (FTIR) are

    about 5 1017 cm3 and 7 1017 cm3, respectively.In these wafers, electron diffusion length Ln values are

    in the range 3060 mm because, probably, B, P, and alsoAl concentrations [105, 106] are too high. Obviously,

    other impurities, like Ti, are present.

    In the top of the ingot, wafers are n-type. The

    advantage of n-type silicon is again very clear: the

    243Silicon Solar Cells, Crystallinewhich the capture cross-sections are closed in both n-

    and p-type silicon, but this impurity cannot be present

    in sufficiently large concentrations in the wafers after

    the end of the crystallization and after a gettering treat-

    ment [97]. Although the diffusion coefficient of elec-

    trons is three times higher than that of holes, it is

    expected that diffusion lengths of minority carriers, in

    such a simple configuration, are neatly higher in n-type

    than in p-type silicon (provided the nature and the

    concentration of impurities is the same in both types

    of wafers). Oxygen precipitates are also less harmful in

    n-type silicon because their surfaces are expected to

    have a hole repelling positive charge [98]. Obviously,

    there are other sources of recombination centers, like

    precipitates, and the reality is more complex.

    Another remarkable advantage of n-type silicon

    results from the lifetime of minority carrier variation

    with the injection level: it does not decrease too much

    with this level, suggesting that n-type base cells will be

    more efficient at low illumination levels than p-type

    base cells.

    As an example of the typical high electrical quality

    of the n-type wafers, Fig. 17 shows a light beam induced

    current scan map (LBIC) at l = 960 nm, with localvalues of the hole diffusion length Lp, of a region which

    contains GBs and dislocations, especially in the center

    of the picture (Fig. 18).

    Such interesting properties can be useful for solar

    cells, as will be described later.

    Refined UMG-mc-Si To open the way to the huge

    production of silicon solar cells which could

    be predicted for 2013 [99], another approach is

    to cast refined upgraded metallurgical silicon

    feedstock (UMG-Si). After crystallization by direc-

    tional solidification, the wafers look like the conven-

    tional ones (large grains, same intragrain defects).

    Problems come from the high concentrations of dop-

    ant and of light elements, mainly carbon. The dopant

    concentrations are close to 2 ppma, and due to the

    distinct distribution coefficients of boron and phos-

    phorus atoms, the material is compensated, and in

    the top of the ingot, the conduction type changes

    from p- to n-type. Concentrations of carbon are close

    to 20 ppma in the top of the ingots, as a consequence of

    the carbothermic reduction and process contamina-

    tion. Precipitation of SiC filaments occurs frequentlyin the upper part of the silicon block [100] as shown by

    Fig. 19.

    These filaments can be detrimental because they

    cause damage during the sawing and can form ohmic

    shunts [101, 102]. Blends of electronic grade (EG) and

    of upgrade metallurgical (UMG) feedstock have given

    unexpected good results because L values are found

    higher than 100 mm in raw wafers [1821]. However,such results cannot be generalized; they are depending

    350

    275

    120

    1.0

    .75

    .50

    .25

    100 m

    Silicon Solar Cells, Crystalline. Figure 18

    Light beam induced current contrast scan map of a raw

    n-type silicon wafer. Some local values (in mm) of minority

    carrier diffusion length Lp are indicated. Notice the high

    values of Lp, even in the central zone where the density of

    extended defects are high

  • 244 Silicon Solar Cells, Crystallinelight beam induced current (LBIC) contrast is more

    marked than for p-type wafers, and the mean values of

    electron diffusion length Lp reach 100 mm in the rawwafers, as shown by the scan map of Fig. 20.

    When n-type wafers experimented phosphorus

    diffusion and aluminium alloying, the mean values

    of Lp achieve 180 mm. Light beam induced current(LBIC) scan maps of n-type wafers look like those of

    conventional p-type mc-Si wafers; the grain boundary

    0 100 m

    Silicon Solar Cells, Crystalline. Figure 19

    SiC filaments observed in UMG-Si wafers (Courtesy of Dr

    Moller-TU-Freiberg)contrast is marked while intragrain defects are

    weakly active.

    However, a difference with conventional mc-Si

    could be noted: there is a weak improvement of Lpafter short phosphorus diffusion. Probably, that is due

    to the presence of slow diffusers like Al and Zn and

    other ones which have not been detected by HR-GDMS

    but which are certainly contained in the material.

    Notice that the preceding results could be impaired

    when the concentrations of oxygen and carbon are higher

    than 8 1017 cm3, as frequently observed in the top ofthe ingots. As reported by Pizzini [107], a detrimental

    coprecipitation of these elements could occur.

    Single-Junction p-Type Silicon-Based Solar Cells

    (n+-p-p+ Structure)

    Solar cells made with crystalline silicon wafers have

    been investigated for a long time, and in 2010, theyshare at least 83% of the total photovoltaic market

    (45% for mc-Si cells), although the part of thin filmcells is increasing. This success is explained by

    the relative low cost and/or by the high conversion

    efficiency of these devices.

    The conversion efficiency of solar cells depends on

    the quality of silicon wafer, i.e., on the values of the

    minority carrier lifetime t and diffusion length L, andon the cell structure design. Particularly, the passiv-

    ation of the surface, the surface texturization, the

    antireflection coatings, as well as the metallization

    techniques are also of a paramount importance because

    they strongly influence the short circuit photocurrent

    density Jsc, the open circuit photovoltage Voc, and the

    350m

    262m

    175m

    87m

    1 mm

    Silicon Solar Cells, Crystalline. Figure 20

    Hole diffusion length scan map Lp in a raw n-type refined

    UMG-Si waferfill factor FF.

    Notice that the conversion efficiencyZ of a solar cellis given by:

    Jsc Voc FFP

    6

    where P is the incident sunlight power per cm2 (W/cm2).

    The Basic Structure Figure 21 shows the schema of

    a very basic solar cell structure made on p-type

    silicon. Most of such conventional cells used an n++-p

    or an n+-p junction which results most often from the

    in-diffusion of phosphorus at 850C from POCl3through the front surface. The depth of the n+ layer

    is typically 0.30.5 mm; its sheet resistance is inthe range 4090 O/square; phosphorus surface

  • 245Silicon Solar Cells, Crystallineconcentration can exceed 1021 cm3 in the case of ann++-p junction. A dead layer is thus formed at the

    emitter surface, which does not contribute to the

    photocurrent. Recombination in this heavily doped

    region causes a high dark current which impairs the

    photovoltage. This very high doping level makes easier

    the formation of a good ohmic contact with the metal-

    lic front grid used as top electrode; however, it is not

    easy to reduce the carrier recombination at such

    a surface although a silicon oxide or a silicon nitride

    layer is deposited or grown on it.

    The back surface is also a source of recombination.

    To reduce this recombination activity, the back surface

    doping level is enhanced (p+-type), totally or locally,

    by means of aluminiumsilicon alloying or boron

    diffusion in order to develop a repulsive back surface

    field (BSF) for electrons. The edges of the cell must be

    discarded in order to eliminate parasitic shunting

    paths, and a single or a double antireflection coating

    (ARC) layer is deposited on the front surface before or

    ARC (oxyde +SiN) n++ or n+

    p- type -Si BSF

    p+ zone

    Silicon Solar Cells, Crystalline. Figure 21

    Diagram of the very basic structure of a p-type silicon cell.

    Metallization are drawn in black (Figure is not drawn to

    scale)after the metallization step.

    Notice that the processing steps used to make the

    cells, e.g., phosphorus diffusion, aluminiumsilicon

    alloying, deposition of a hydrogen-rich silicon nitride

    layer always improve the performances of the cells,

    especially when the base is made with multicrystalline

    silicon.

    Antireflection Coating and Texturization The opti-

    cal properties of silicon play a major role in the design

    and operation of solar cells; the light absorption and

    the reflectance of the front surface must be taken

    into account. The reflectance is 30% for wavelengthsl < 1.1 mm and increases to 60% for l 0.4 mm[108]. Thus, to minimize the optical losses, deposition

    of an antireflective coating or/and texturization of

    the silicon surface are essential.

    Antireflection Coatings Reflectance is reduced by

    employing an antireflective coating (ARC) on the

    silicon front surface. Predominantly, single-layered

    ARCs are generally used industrially. The optical

    thickness of the ARC should be equal to a quarter of

    the wavelength lmin at which zero reflectanceoccurs (i.e., reflected waves must be out of phase

    by p) [109]:

    nARCdARC lmin4

    7

    The refractive index of the layer, nARC, has to be

    equal toncSi

    pfor the crystalline silicon (c-Si) cell in

    air. As nc-Si varies from 4.28 to 3.76 between 0.5

    and 0.7 mm, the refractive index of an optimizedARC should be, for instance, equal to 1.9 at 600 nm.

    For a glass-encapsulated cell, the refractive index of

    the glass being nglass = 1.5 instead of 1 for air; the

    optimal index value for the ARC will be 2.3. Such

    ARC reduces the sunlight lost by reflectance on

    polished silicon from 30% down to 10% and from

    20% to 5% in air and under a protecting glass,

    respectively.

    Antireflection coatings used in the industry are

    dielectric materials such as TiO2 (n = 2.3) [110] and

    SiNx:H (whose refractive index varies between 1.9 and

    2.3) [111]. For laboratory cells, two complementary

    dielectric layers are sometimes applied to broaden the

    spectral band reduction of the reflectance.

    TiO2 was first introduced in the early 1970s

    [112, 113] and presents numerous advantages such

    as high chemical stability, insulating properties when

    stoichiometric, good mechanical properties. TiO2thin films can be deposited at low temperatures

    (400C) by atmospheric pressure chemical vapordeposition (APCVD) or by spray pyrolysis [114].

    Nontoxic and noncorrosive liquid precursors, such as

    tetraisopropyl titanate (TPT), are used as a deposition

    source. At such temperatures, the metastable anatase

    crystalline phase is formed which exhibits an optimal

    refractive index for glass-encapsulated silicon solar

    cells, as well as excellent transmittance. A thin silicon

  • less than the cell thickness. Thirdly, long-wavelength

    photons which are reflected from the rear surface will

    encounter an angled silicon front surface, improving

    their chance of being internally reflected [119].

    This process is referred to as light trapping and gives

    an improved response to infrared light, especially for

    thin Si substrates.

    There are two types of texturization. When the

    etching rate varies with crystallographic direction,

    the texturing is anisotropic, while when the material

    etches in all directions at same rate, the texturing is

    isotropic.

    Anisotropic Alkaline Etching A geometrical texture

    (V-shaped grooves texture, random pyramidal struc-

    tures) at the front surface of silicon can be done

    mechanically, for instance using specially developed

    dicing saws or (electro)chemically, by etching the

    silicon in an alkaline or acidic solution [120].

    60

    246 Silicon Solar Cells, Crystallinedioxide (SiO2) layer could be grown before deposition

    of TiO2 in order to provide passivation of the front

    silicon surface.

    SiNx:H deposition by PECVD was invented in 1965

    [115]. The amorphous a-SiNx:H films are typically

    deposited using plasma-enhanced chemical vapor

    deposition (PECVD), at temperatures of less than

    450C with deposition rates as high as 1.7 nm/s.The properties of a-SiNx:H films are extremely

    dependent on the deposition system, process condi-

    tions, and gas composition. Thus, when using dilute

    silane and ammonia reactant gases, the best surface

    passivation results (see Surface Passivation. The Selec-

    tive Emitter Concept) from stoichiometric a-SiNx:H

    films that have a refractive index of 1.95, while the

    Si-rich films deposited from pure silane and ammonia

    possess a refractive index of 2.3 [114]. Effective

    reflectance of 11% is observed on a planar siliconsurface, as shown by Fig. 22.

    SiNx:H coating is presently widely used in the

    industry as the efficiency of the solar cells is strongly

    increased.

    Notice that porous silicon exhibiting a refractive

    index intermediate between those of silicon and air

    has been widely investigated as an AR coating in the

    1990s [116, 117].

    Silicon Surface Texturization Light incident normally

    to the silicon is transmitted to an amount of70% intosilicon while 30% is reflected away from theflat surface as shown by Fig. 23a. Texturing the front

    surface of a solar cell increases the short-circuit current

    due to three distinct mechanisms. All of which are

    related to the fact that the incident photons strike the

    cell surface at a specific angle [108, 118]. Firstly, some

    light rays will be reflected from one angled surface

    merely to strike another as shown by Fig. 23b,

    resulting in an increased probability of absorption,

    and therefore reduced reflection. Minimal facet angles

    of 30 and 54 are required for double- andtriple-bounce reflectance, respectively. For the double

    bounce case shown in Fig. 23b, the light reflected away

    from the silicon is lost. However, if the silicon is

    encapsulated under glass, it is possible to totally

    reflect the light arriving at the glassair interface,

    above the critical angle of 42, whereby it willbe redirected toward silicon yet again for possible

    re-incidence. Critical angles for total internal

    reflectance at siliconair interface are 1617.Secondly, photons refracted into the silicon will

    propagate at an angle causing them to be absorbed

    closer to the junction than would occur with a planar

    surface. This is especially relevant in material with

    minority carrier diffusion lengths comparable to or

    50

    40

    30

    20

    10

    0

    Ref

    lect

    ance

    , R (%

    )

    400 500 600 700Wavelength, (nm)

    800 900 1000 1100

    air/TiO2/planar Si (Rw =8.5%) air/glass/TiO2/planar Si (Rw =11.7%) air/glass/TiO2/textured Si (Rw =5.8%)

    Silicon Solar Cells, Crystalline. Figure 22

    Effective reflectance of planar and textured emitter

    surfaces with TiO2 single-layer antireflection coating (From

    [114], courtesy of Dr. BS Richards)

  • h

    ar

    247Silicon Solar Cells, CrystallineSi surface

    h

    (i)

    (t)

    (r)

    Si bulk

    a b

    Silicon Solar Cells, Crystalline. Figure 23

    The influence of surface texture on light reflection: (a) plan

    reflected lights, (t1) and (t2) absorbed lightsThe first solar cell using the alkaline etching process

    was reported in 1974 [121]. It is a well-established

    texturing method for single-crystal silicon solar cell.

    The wet etching process can be globally represented

    by the overall reaction:

    Si 2H2O 2OH ! Si OH 2O2 2 2H2 8

    Etching rates in NaOH or KOH depend on the

    crystallographic orientations; the (100) planes being

    etched more rapidly than (111) planes, which are thus

    revealed. (111)-oriented surfaces are flat etched, with

    some triangular features due to the presence of terraces,

    while (100)-oriented surfaces exhibit random square-

    shaped pyramids [122] made of (111) facets, as shown

    by Fig. 24.

    An anisotropic etched on (100) plane forms a facet

    angle of 54.7 on a (100)-oriented surface [123].

    a

    Silicon Solar Cells, Crystalline. Figure 24

    SEM images in plane view of NaOH-textured silicon: (a) (100)-(i)

    (r1)

    (r2)

    (t1)(t2)

    surface; (b) textured surface. (1) incident light, (r1) and (r2)This angle satisfies the requirements for both external

    reflection control and total internal reflections.

    The reflectivity decreases from 34% for a mirror-

    polished silicon surface to 11% for an alkaline-textured

    silicon wafer [126], as shown in Fig. 24.

    A promising alternative solution is quaternary

    ammonium hydroxide etchings, typically tetramethy-

    lammonium hydroxide (TMAH) [124, 125]. This solu-

    tion is clean (room compatible, nontoxic, and easy to

    handle). A low weighted reflectance of 2.7% can be

    obtained with a combination of TMAH pyramidal

    texture and SiNx:H AR coating, as shown in Figs. 25

    and 26 [126].

    Anisotropic alkaline texturing does not work very

    well on multicrystalline substrates due to the different

    crystallographic grain orientations. A number of

    different surface geometries will determine the

    b

    and (b) (111)-oriented crystals

  • interaction of incident light with the surface and will

    contribute to the overall reflectance of the

    multicrystalline wafer. The bare saw etched wafers

    have reflectance in air equivalent to polished silicon

    because the tilt angles of the dominant etch facets

    exposed are lower than the 30 required for double-bounce reflection. The (111)-oriented surfaces are flat

    etched, with triangular plateaus as predicted by Sopori

    [109], while (100)-oriented surfaces exhibit the upright

    random pyramids. By measuring the reflectivity of

    individual grains, it was found that (100) and

    (111) grains exhibit minimum and maximum effective

    reflectivity between 400 and 1,100 nm of 11.7%

    and 34.2%, respectively, while grains with other

    orientations exhibit intermediate values [127]. How-

    ever, the anisotropic texturization is still currently used

    in multicrystalline silicon industry because under

    encapsulation, both saw damage and textured etched

    multicrystalline wafers couple light more effectively

    into the silicon.

    Isotropic Texturization Numerous methods such as

    chemical, electrochemical, defect etching, or based on

    more sophisticated techniques (e.g., reactive ion etch-

    ing (RIE), lithography, laser scribing, and mechanical

    grooving) have been investigated.

    Silicon Solar Cells, Crystalline. Figure 25

    SEM picture of a typical texturized silicon surface using

    TMAH texturization

    50

    45

    40

    35

    30

    25

    20

    15

    10

    5

    0

    Polished +

    Polished (W

    TMAH Tex

    TMAH Tex

    vel

    Hem

    isph

    eric

    al re

    flect

    ance

    (%)

    ie

    248 Silicon Solar Cells, Crystalline350 450 550 650Wa

    Silicon Solar Cells, Crystalline. Figure 26

    Weighted reflectance of different silicon surface morpholog SiN ARC (WR 10,2%)R 41%)

    turing (WR 13%)turing + SiN ARC (WR 2,7%)

    ength (nm)750 850 950 1050

    s, with or without ARC and TMAH texturing

  • Plasma-enhanced dry chemical etching at atmo-

    spheric pressure is a promising alternative to isotropic

    wet etching, especially when combined with similar pro-

    cess technologies, to provide a continuous in-line

    processing steps. Plasma texturing is particularly appro-

    priate for advanced solar cell structures, for wafers pro-

    duced without surface damage such as Si ribbons and

    epitaxial layers on low-cost Si substrates. It allows

    decoupling the properties of the front and rear surfaces

    Silicon Solar Cells, Crystalline. Figure 27

    SEM plan view image of multicrystalline silicon surface

    after isotropic acid etching (From [130], Courtesy of Dr.

    R Einhaus)

    oxide

    p-type

    rear metal contact

    platedmetal

    n++n+

    p+

    Silicon Solar Cells, Crystalline. Figure 28

    Passivated emitter and buried front contact solar cell

    (Courtesy of UNSW)

    249Silicon Solar Cells, CrystallineAmong these methods, the acid texturization

    method has been the most developed and is currently

    used in industry. The RIE method is very attractive

    because it allows a better automatized solar cell

    processing, avoiding the use of wet techniques.

    In search of a simple method with high throughput,

    low-cost investment, and compatible with thin wafers,

    Sarti et al. proposed the wet acidic texturing that tends

    to etch isotropically the multicrystalline silicon wafers,

    resulting in features with rounded surfaces [128].

    The acidic water solution consists of a mixture of nitric

    acid, hydrofluoric acid (HFHNO3), water, and

    eventually some additives. The overall reaction is the

    following:

    3Si 4HNO3 18HF! 3H2SiF6 8H2O 4NO9

    The etching rate as well as the surface morphology

    depends strongly on the chemical composition of the

    texturization bath but not on the crystallographic

    orientation. Etch textures vary between rough, concave-

    shaped tub structures, and smooth surfaces. The rough

    texture is of interest for solar cell industry [129, 130].

    The texturing process starts on as-cut multi-Si

    wafers. The saw damages trigger the etching process.

    Thus, the removal of saw damage and the surface

    texturing are being done simultaneously in one short

    single chemical step. The resulting etch pits of 110 mmin diameter are uniformly distributed over the wafer

    surface irrespective of crystallographic orientation,

    giving a homogenous reflectance over the surface of

    the wafer and the absence of steps between grains, as

    shown in Fig. 27.

    The acidic isotexturing results in lower reflection than

    traditional anisotropic etching on multicrystalline mate-

    rial, as shown in Fig. 28, and better conversion efficiency

    [130132]. Short-circuit current density increases of

    up to 1.5 mA/cm2 have been measured.

    In laboratory, the masked acidic etching has been

    developed to produce hexagonally symmetric honey-

    comb surface texture. This texture reduces reflection

    loss as well as substantially increasing the cell effective

    optical thickness by causing light to be trapped within

    the cell by total internal reflection. Multicrystalline

    solar cells with record efficiency of 19.8% have been

    obtained [130, 131].

  • 250 Silicon Solar Cells, Crystallinewhich are both texturized with the wet etching.

    A nontextured rear surface appears to be needed to reach

    low effective recombination in advanced solar cells.

    First report on reactive ion etching (RIE) texturization

    of Si in c