Advanced Amorphous Silicon

5Advanced Amorphous Silicon Solar Cell Technologies

Miro Zeman

Delft University of Technology

5.1INTRODUCTION

The?rst amorphous silicon layers were reported in1965as?lms of‘silicon from silane’deposited in a radio frequency glow discharge[1].Ten years later,Walter Spear and Pe-ter LeComber from Dundee University reported that amorphous silicon had semiconducting properties.They demonstrated that the conductivity of amorphous silicon can be manipulated by several orders of magnitude by adding some phosphine or diborane gas to the glow discharge gas mixture[2].This was a far reaching discovery since until that time,it had generally been thought that amorphous silicon could not be made n type or p type by substitutional doping. It was not recognized immediately that hydrogen plays an important role in the newly made amorphous silicon doped?lms.In fact,amorphous silicon suitable for electronic applications, requiring doping,is an alloy of silicon and hydrogen.Electronic grade amorphous silicon is therefore called hydrogenated amorphous silicon(a-Si:H).

The successful doping of amorphous silicon created tremendous interest in this material for two reasons.First,the material had several interesting properties that opened up many opportunities for semiconductor device applications.For example,due to the high absorption coef?cient of a-Si:H in the visible range of the solar spectrum,a1micrometer(μm)thick a-Si:H layer is suf?cient to absorb90%of the usable solar energy.Second,the glow discharge depo-sition technique,also referred to as plasma enhanced chemical vapour deposition(PECVD), enabled the production of a-Si:H?lms over a large area(larger than1m2)and at a low tem-perature(100to400?C).The low processing temperature allows the use of a wide range of low cost substrates such as glass sheet,metal or polymer foil.The a-Si:H is simply doped and alloyed by adding the appropriate gases to a source gas,usually silane.These features have made a-Si:H a promising candidate for low cost thin?lm solar cells.At present,besides solar cells,this material is used for thin?lm transistors in?at panel displays and photoconductive layers in electrophotography.

Since the?rst a-Si:H solar cell Carlson and Wronski made in1976,which had an energy conversion ef?ciency of2.4%[3],a-Si:H solar cell technology has improved considerably, and today,it is capable of producing solar cells with initial ef?ciencies exceeding15%[4]. Today,amorphous silicon solar cell technology is a mature thin?lm solar cell technology that in2003already delivered modules with a total output power of25.8MW p[5].

Thin Film Solar Cells: Fabrication, Characterization and Applications Edited by J. Poortmans and V. Arkhipov C 2006 John Wiley & Sons, Ltd. ISBN: 0-470-09126-6

174THIN FILM SOLAR CELLS

5.2OVERVIEW OF AMORPHOUS SILICON SOLAR CELL

TECHNOLOGY DEVELOPMENT AND CURRENT ISSUES

5.2.11970s

Carlson and Wronski announced that they had made the?rst experimental a-Si:H solar cell at the RCA Laboratory in1976[3].This single junction p-i-n a-Si:H solar cell deposited on a glass substrate coated with transparent conductive oxide(TCO)and aluminium back contact exhibited a2.4%conversion ef?ciency.In order to increase the output voltage of a-Si:H solar cells,the concept of a stacked(also called multi-junction)solar cell structure was introduced [6].A key step to industrial production was the development of a monolithically integrated type of a-Si:H solar cell[7].Using the monolithic series integration of a-Si:H solar subcells,a desired output voltage from a single substrate can easily be achieved.In1980,the integrated type a-Si:H solar cells were commercialized by Sanyo and Fuji Electric and applied in consumer electronics such as calculators and watches.

5.2.21980s

Much research in the?eld of a-Si:H solar cells was devoted to developing and optimising a-Si:H based alloys in the1980s.A p type hydrogenated amorphous silicon carbide(a-SiC:H) was incorporated in solar cells as a low absorbing layer,usually denoted as a window layer[8]. Hydrogenated amorphous silicon germanium(a-SiGe:H)became an attractive low bandgap material for stacked solar cells[9].Surface textured substrates were introduced to enhance optical absorption[10].The laboratory cells reached an initial ef?ciency in the range of11to 12%.The next generation of a-Si:H modules came on the market in the second half of the 1980s and was aimed at off grid power generation.These modules were single junction p-i-n a-Si:H solar cells,produced mainly in a single chamber batch process.The typical area of the modules ranged from0.1to0.3m2and they were aimed to deliver a power of around14W (stabilized ef?ciency up to5%).However,this promising technology suffered from some setbacks that gave it a bad reputation:pronounced initial degradation due to illumination, insuf?cient protection and framing of these modules against moisture,which resulted in the corrosion of contacting electrodes.

5.2.31990s

In the1990s,the main research and manufacturing efforts were directed towards achieving 10%stabilized module ef?ciency and a high throughput process.Several companies optimized and implemented an a-SiGe:H alloy in tandem(BP Solar[11],Sanyo[12],Fuji Electric[13]) and triple junction(United Solar[14])solar cell structures.The main characteristics of a-Si:H modules developed in the1990s were a multijunction solar cell structure,improved encapsu-lation and framing.Lightweight frames from organic materials that provided better protection against corrosion substituted the aluminium frames.The module area reached1m2and the total area stabilized module ef?ciency was increased to6–7%.The improved environmental protection of the modules enabled the producers to guarantee more than20years of power

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES175 output.At the end of the20th century,the annual total production capacity for amorphous sil-icon single and multijunction modules reached around30MW p.The focus on the application of the modules shifted from off grid to building integrated applications.

Hydrogenated microcrystalline silicon deposited by the low temperature PECVD technique emerged in this period as a new candidate for the low bandgap material in multijunction a-Si:H based solar cells.The University of Neuch?a tel introduced a micromorph tandem solar cell in 1994,which comprized an a-Si:H top cell and aμc-Si:H bottom cell[15].The promising poten-tial of the micromorph cell concept was soon demonstrated by the fabrication of micromorph tandem and triple solar cells with stabilized ef?ciencies in the range of11to12%[16,17],and Kaneka Corporation started the development of micromorph module production technology [17].The introduction and implementation ofμc-Si:H in thin?lm silicon solar cells shifted attention to increasing the deposition rate.Several new deposition techniques[18]started to be investigated and developed for fabricating absorber layers at high deposition rates(10to 20?A/s),such as very high frequency and microwave PECVD,hot wire CVD,and expanding thermal plasma CVD.

5.2.4After2000

Research has concentrated on understanding and improving light trapping techniques,where surface textures as well as new TCO materials play a crucial role.This activity has resulted in the commercialization of novel deposition techniques for ZnO as an alternative TCO material for SnO2[19,20].Several deposition machine manufacturers have started developing commercial production machines for the fabrication of thin?lm silicon solar cells[21,22].Today the most advanced a-Si:H production lines are characterized by fully automated facilities and large area deposition over more than1m2,with an annual production capacity in the range of 10MW p to30MW p(Mitsubishi Heavy Ind.10MW p,Kaneka Corporation20MW p,United Solar30MW p).

5.2.5Current technology issues

In order to increase the competitiveness of a-Si:H modules on the market,several cost-to-performance aspects of the a-Si:H solar cell technology are of importance,which can be pided into the following performance and production related issues:

1.Increase of the conversion ef?ciency of a-Si:H solar cells.Based on fundamental considera-

tions,major performance improvement is expected in the near future from an increase in the current of thin?lm silicon solar cells[23].This increase has to result from improved light management schemes such as light trapping and reduction of light absorption losses.For solar cells deposited on glass plates,also called superstrate type cells,the development of a TCO front electrode material with an optimal surface morphology that results in improved light scattering properties is essential.Essential for solar cells deposited on(?exible)opaque carriers,often denoted as substrate type solar cells,is improvement of the texturing and re?ectivity of the back contact.Ongoing attention has to be paid to further improvement of the optoelectronic quality of a-Si:H and a-SiGe:H absorbers,the doped layers and the interfaces between the doped layers and intrinsic absorbers.

176THIN FILM SOLAR CELLS

2.Elimination of the light induced degradation known as the Staebler–Wronski effect[24].

This effect is responsible for a decrease in the initial performance of an a-Si:H solar module of typically15–30%.Full understanding of the Staebler–Wronski effect in a-Si:H based materials is necessary for fabricating a-Si:H with improved stability against light exposure.

The light induced degradation of the modules can be suppressed by using thin absorber layers.However,the use of the thin absorbers strongly depends on the implementation of ef?cient light trapping techniques in the solar cells,which have to provide for suf?cient absorption in these layers.

3.The deposition rate of absorber layers is required to be10to20?A/s in order to limit

the investment in the a-Si:H deposition machine,which is strongly re?ected in the cost of the modules.The central question regarding the deposition rate is how to avoid the increased light induced degradation of?lms deposited at elevated deposition rates[25].

In addition to the radio frquency(rf)PECVD technique,several deposition techniques are being intensively investigated capable of fabricating absorber layers with suf?cient quality at high deposition rates,such as very high frequency PECVD,hot wire CVD,and expanding thermal plasma CVD.

4.The choice of mass production technology.Although the deposition of a-Si:H based layers

is the most important part of solar cell fabrication,complete production includes several fabrication steps which substantially contribute to the total cost of the solar module.These include the deposition of the TCO front electrode,the deposition of the multilayer back electrode,laser scribing for the subcell series connection,and encapsulation and framing.

The solar cell structure and module design determine the choice of sequence of fabrication steps and the deposition and processing techniques to be applied.At present,there are three major approaches to depositing a-Si:H based layers:the one chamber batch process,the multichamber process,and the roll-to-roll process.The advantages and disadvantages of particular a-Si:H production systems are discussed in reference[18].The general trend is to increase the substrate size,which reduces the cost per unit area,by lowering the relative contribution of the edges.The experience gained in the display industry regarding the deposition of a-Si:H on large area substrates is being transferred to solar cell production.

The general requirements for the production machines include suf?cient reliability of the deposition process,high production uptime,high yield and the right choice of procedure for cleaning the process chambers.

5.Lowering of material costs.The material costs contribute considerably to the overall cost

of a-Si:H modules.A substantial part is formed by the cost of the substrate carrier,the glass plate or high temperature resistant polymer foil.Therefore,cheaper thin metal foils that also allow the use of the continuous roll-to-roll technology are a preferable choice.In case of?exible modules,usually a relatively thick?uoropolymer type encapsulant is applied in order to guarantee a module lifetime of20years.The encapsulant dominates the cost of the module and the development of a cheap encapsulant is therefore one of the most important cost issues.The choice of substrate carrier determines the acceptable process temperatures and the sequence of processing steps.The choice of gasses for the deposition of a-Si:H based layers,their purity and the gas utilization also have?nancial consequences.For example, the use of germane for multijunction solar cells with a-SiGe:H absorbers can substantially increase the material costs.

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES177 5.3HYDROGENATED AMORPHOUS SILICON

In order to understand the design and operation of a-Si:H based solar cells,which are different from those for crystalline silicon solar cells,we will summarize the structural and material properties of a-Si:H and compare them to those of single crystal silicon.Some widely used techniques to characterize a-Si:H are also described in this section.

5.3.1Atomic structure

Figure5.1illustrates the difference in atomic structure between single crystal silicon and a-Si:H.Figure5.1a shows the structure of single crystal silicon schematically.Each silicon atom is covalently bonded to four neighboring atoms.All bonds have the same length,and the angles between the bonds are equal.The number of bonds of an atom with its immediate neighbors in the atomic structure is called the coordination number or coordination.Thus,in single crystal silicon,the coordination number for all silicon atoms is four;we can also say that silicon atoms are fourfold coordinated.A unit cell can be de?ned,from which the crystal lattice can be reproduced by duplicating the unit cell and stacking the duplicates next to each other.Such a regular atomic arrangement is described as a structure with a long range order.

Figure1b illustrates that a-Si:H does not exhibit a structural order over a long range. Nevertheless,there is a similarity in atomic con?guration on a local atomic scale,where most silicon atoms have covalent bonds with four neighbors.Though a-Si:H lacks the long range order,it has the same short range order as single crystal silicon.This conclusion about the bonding structure in a-Si:H has been obtained from X-ray diffraction measurements[26].The small deviations in bonding angles and bonding lengths between the neighboring atoms in a-Si:H lead to a complete loss of the locally ordered structure on a scale exceeding a few atomic distances.The resulting atomic structure of a-Si:H is called a continuous random network.

(a)(b)

unpassivated

dangling bond

Figure5.1Schematic representation of the atomic structure of(a)single crystal silicon,(b)hydro-genated amorphous silicon.

178THIN FILM SOLAR CELLS

Due to the short range order,the common semiconductor concept of the energy state bands, represented by the conduction and valence bands,can still be used in a-Si:H.

The larger deviations in bonding angles and bonding lengths between the neighboring atoms in a-Si:H result in so-called weak or strained bonds.When enough energy is available, for example in the form of heat,weak bonds can easily be broken.This process leads to the formation of defects in the atomic network.We note that in the continuous random network, the de?nition of a defect is modi?ed with respect to the crystal structure.In a crystal,any atom that is out of place in a lattice forms a defect.In the continuous random network,an atom cannot be out of place.Because the only speci?c structural feature of an atom in the continuous random network is the coordination to its neighbors,a defect in a-Si:H is a coordination defect [26].This happens when an atom has too many bonds or too few.In a-Si:H,defects are mainly silicon atoms that are covalently bonded to only three silicon atoms(threefold coordinated)and have one unpaired electron,a so-called dangling bond.Since this con?guration is the dominant defect in a-Si:H,the defects in a-Si:H are often related to dangling bonds.Some dangling bond defects are depicted in Figure5.1b.Another defect con?guration is a silicon atom bonded to ?ve silicon atoms(?vefold coordinated).This con?guration is referred to as a?oating bond.

A silicon atom representing this?oating bond defect is indicated in Figure5.1b by a dotted circle.

In pure a-Si(amorphous silicon that contains no other atoms than silicon),there is a large concentration of about1021defects per cm3in the amorphous atomic structure.Material with such a large defect density cannot be used for a functional device.When amorphous silicon is deposited in such a way that hydrogen can be incorporated in the atomic network (as by glow discharge deposition from silane),then hydrogen atoms bond with most of the silicon dangling bonds.Strong silicon–hydrogen bonds are formed,which are illustrated in Figure5.1b.Hydrogen passivation of dangling bond defects reduces the defect density from about1021cm?3in pure a-Si to1015–1016cm?3in a-Si:H,i.e.less than one dangling bond per one million silicon atoms.It is this material,an alloy of silicon and hydrogen,in which substitutional doping has been demonstrated and which is suitable for electronic applications.

5.3.1.1Electron spin resonance

An experimental technique that can provide information on the microscopic structure of defects in semiconductors,including the amorphous ones,is electron spin resonance(ESR)[27]. Electron spin resonance measurements on a-Si:H have identi?ed a single type of defect that is associated with a neutral dangling bond[26,28].The ESR is regarded as an experimental standard for determining defects in a-Si:H,and the ESR results are considered unambiguous. However,the sensitivity of this method is limited for thin?lms with a low spin density and the method only gives information about paramagnetic defects,i.e.defects with an unpaired electron.For this reason,ESR can underestimate the defect density,because the charged dangling bonds do not possess an unpaired spin signal.Therefore,the results from ESR strongly depend on the Fermi level position,which determines the electron occupation of defects. 5.3.1.2Hydrogen characterization in hydrogenated amorphous silicon

Since hydrogen plays an important role in defect passivation,the incorporation and stability of hydrogen in a-Si:H has been the topic of intensive research.Infrared absorption spectroscopy

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES179 is widely used to provide information about Si–H x bonding con?gurations in a-Si:H[26]. Three characteristic infrared absorption bands are observed in a-Si:H:a peak at640cm?1,a doublet at840–890cm?1,and absorption peaks in the range of2000–2200cm?1.The peak at640cm?1re?ects the rocking mode of hydrogen covalently bonded in all possible bonding con?gurations,i.e.silicon mono(x=1),di(x=2),and trihydride(x=3)and polymeric (Si–H2)n bonding con?gurations,and therefore this peak is used to determine the hydrogen content in a-Si:H[18].The doublet at840–890cm?1is assigned to the dihydride wagging mode.A peak around2000cm?1is assigned to the stretching mode of the isolated Si–H bonds(also referred to as the low stretching mode(LSM)in the literature)and a peak in the range of2060–2160cm?1includes contributions from the stretching mode(referred to as the high stretching mode(HSM))of Si–H bonds at internal surfaces,e.g.voids,dihydride and trihydride bonds.A‘microstructure parameter,’denoted as R?,is determined from the LSM and HSM absorption peaks.The R?is widely used to characterize the microstructure in the a-Si:H network as it roughly indicates two different‘phases’,namely a dense network and a fraction of the network containing voids.The microstructure parameter is de?ned as:

R?=

I HSM

I LSM+I HSM

(5.1)

where I HSM and I LSM are the integrated absorption strength of the LSM and HSM,respec-tively.In general,device quality a-Si:H contains less than10atomic%of hydrogen and is characterized by R?<0.1.

Hydrogen diffusion and evolution measurements help to characterize hydrogen motion, trapping and evolution in a-Si:H[29].Nuclear magnetic resonance(NMR)gives information about the local atomic environment in which the hydrogen atoms reside[30].Recently,it has been reported,based on NMR experiments,that molecular hydrogen forms up to40%of the total hydrogen content in a-Si:H[31].

5.3.2Density of states

An essential component for determining the distributions and concentrations of charge carriers in a semiconductor material is information about the energy distribution of states,often called the density of states.For an ideal intrinsic silicon crystal,the valence band and the conduction band are separated by a well de?ned bandgap,E g,and there are no allowed energy states in the bandgap.Due to the long range disorder in the atomic structure of a-Si:H,the energy states of the valence band and the conduction band spread into the bandgap and form regions of states that are called band tails.In addition,the defects introduce allowed energy states that are located in the central region between the valence band and conduction band states.This means that there is a continuous distribution of density of states in a-Si:H and that there is no well de?ned bandgap between the valence band and the conduction band.

The energy states in which the charge carriers can be considered as free carriers are described by wave functions that extend over the whole atomic structure.These states are nonlocalized and are called extended states.The disorder in a-Si:H causes the wave functions of the tail and defect states to become localized within the atomic network.These states are called localized states.Consequently,mobility that characterizes transport of carriers through the localized states is strongly reduced.This feature of a sharp drop in the mobility of carriers

180THIN FILM SOLAR CELLS

in the localized states in comparison to the extended states is used to de?ne the bandgap in a-Si:H.This bandgap is denoted by the term mobility gap,E mob,because the presence of a considerable density of states in the mobility gap is in con?ict with the classical concept of a bandgap without any allowed energy states.The energy levels that separate the extended states from the localized states in a-Si:H are called the valence band,E V,and the conduction band, E C,mobility edges.The mobility gap of a-Si:H is larger than the bandgap of single crystal silicon and has a typical value between1.7eV and1.8eV.

5.3.3Models for the density of states and recombination–generation

statistics

In general,the energy distribution of states in a-Si:H is characterized by three different regions: (i)extended states above the mobility edge of the conduction band,(ii)extended states below the mobility edge of the valence bandand(iii)localized states between the mobility edges.The continuous distribution of the localized states is a superposition of the conduction and valence band tail states and the defect states.

In Figure5.2,we present a standard model of the density of states distribution.In this model,the valence and conduction band states are represented by a parabolic dependence on energy that merges with exponentially decaying valence and conduction band tail states. The defect states are represented by two equal Gaussian distributions,which are shifted from each other by the correlation energy,U.The correlation energy is assumed to be constant and positive.As mentioned earlier,dangling bonds are considered the dominant defect in a-Si:H.A dangling bond can be in three charge states:positive(D+),neutral(D0)and negative(D?).An imperfection with three possible charges,such as a dangling bond,is represented in the band diagram by two energy levels E+/0and E0/?,which,depending on the position of the Fermi level,characterize the charge occupation of the imperfection.These two energy levels are

Figure5.2The standard model for density of states in a-Si:H.

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES181 called the transition energy levels.The two Gaussian distributions,D+/0and D0/?,represent the energy distributions of states corresponding to+/0and0/?charge transitions of dangling bonds,respectively.Since the dangling bonds are represented by both the donor like(+/0) and acceptor like(0/?)states,dangling bonds are called amphoteric defects.

The Gaussian distribution that is used to describe the defect states in a-Si:H re?ects the concept that the structural disorder results in a distribution of states rather than in states lo-cated at a particular energy level.However,the Gaussian representation of defect states does not contain any information about the origin of the defect states.The defect pool theory based on the weak bond–dangling bond conversion model[32]has attracted a lot of atten-tion because it can successfully describe the defect state distribution in a-Si:H[33–35].The theory assumes that the energy of the defect state can take a range of values due to the in-herent disorder of the amorphous network and that the defects can be formed in different charge states.The resulting total defect density of states is the sum of three energy distri-butions,D h,D z,and D e,which correspond to positive,neutral and negative defects.The defect pool model predicts that the total number of dangling bonds in the intrinsic a-Si:H in-creases when the Fermi level shifts from the midgap towards the mobility edges of the bands. The position of the Fermi level also determines the energy dependence of the defect state distribution.

The energy states in the bandgap act as trapping and recombination centers and therefore strongly affect many electronic properties of a-Si:H and the performance of a-Si:H devices.In contrast to crystalline semiconductors,in which the recombination process is typically domi-nated by a single energy level in the bandgap,in a-Si:H,contributions from all bandgap states to the recombination–generation(R–G)rate are included.In order to model the recombination process through the single level states,such as localized tail states,Shockley—Read–Hall R–G statistics[36]is used and Sah and Shockley multilevel R–G statistics[37]are applied for the amphoteric defect states.A detailed analysis and comparison of the modeling approaches of the R–G rate in a-Si:H is given in reference[18].

A large number of experimental techniques have been applied to obtain information about the density of states in a-Si:H[26].In particular the distribution of localized tail and defect states is of interest.There is no direct method to obtain the energy distribution of states in a-Si:H. The energy distribution of states is determined indirectly from measurements of optical and electrical properties of a-Si:H?lms or from properties of a space–charge region that is formed at a-Si:H interfaces.In order to provide an understanding of procedures to determine the density of states,we?rst discuss the optical and electrical properties of a-Si:H.

5.3.4Optical properties

The optical properties of a-Si:H are usually characterized by the absorption coef?cient,the refractive index,and the value of the optical bandgap.

Figure5.3a shows the typical absorption coef?cient of a-Si:H as a function of photon energy. In Figure5.3b,the absorption coef?cient of a-Si:H is plotted with the absorption coef?cient of a-SiGe:H,p type a-SiC:H,and crystalline silicon.In the visible part of the solar spectrum, a-Si:H absorbs almost100times more than crystalline silicon.This means that a1μm thick a-Si:H layer is suf?cient to absorb90%of the usable solar energy.In practice,the thickness of a-Si:H solar cells is around than0.3μm,that is about1000times thinner than a typical single crystal silicon cell.

182THIN FILM SOLAR CELLS

(a)

(b)

0.5

1.0

1.5

2.0 2.5

3.0

Photon energy [eV]

A b s o r p t i o n c o e f f i c i e n t [1/c m ]

2.48 1.240.830.620.50

0.41

Wavelength [micrometers]10510410210110-110610310

10-2

A b s o r p t i o n c o e f f i c i e n t [1/c m ]

2.48 1.240.830.620.50

0.41

Wavelength [micrometers]10510410210110-110

6

10310

10

-2

Figure 5.3(a)Absorption coef?cient of a-Si:H as a function of photon energy,(b)absorption coef?cient as function of photon energy for a-Si:H,p type a-SiC:H and a-SiGe:H fabricated at Delft University of Technology.The absorption coef?cient of c-Si is shown for reference.

Due to the lack of translational symmetry of a unit cell in the structural network of a-Si:H,the law of crystal momentum conservation is relaxed in a-Si:H,and it behaves like a direct semiconductor.The optical absorption coef?cient,α(E),is therefore determined by optical transitions involving all pairs of occupied and unoccupied electronic states that are separated by the same photon energy E .For this reason,optical absorption measurements are widely used to determine the density of states distribution in a-Si:H.The absorption coef?cient curve is of fundamental importance for evaluating the quality of both amorphous and microcrystalline silicon ?lms.

As indicated in Figure 5.3a,the absorption spectrum of a-Si:H exhibits three regions.In region A,absorption occurs predominantly by transitions between the extended states of the valence and conduction bands.The absorption coef?cient in this region is higher than 103–104cm ?1and is commonly determined by re?ection-transmission spectroscopy mea-surement.Region B extends from α～1to 103cm ?1and is characterized by an exponential dependence of the absorption coef?cient on the photon energy.This region is called the Urbach edge.When assuming that the optical matrix element is energy independent in this region,as experimentally observed [38],the Urbach edge re?ects the transitions between the valence and conduction band tail states.The absorption coef?cient in region B can be ?tted to

α=α0exp (E /E 0),

(5.2)

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES183 whereα0is a constant and E0is the Urbach energy that characterizes the exponential slope of the energy dependence.Because the conduction band tail state distribution is narrower than the valence band tail state distribution,the Urbach energy re?ects the slope of the exponential region of the valence band tail.A typical value for device quality a-Si:H?lm is E0≤50×

10?3eV.Region C,where the absorption coef?cient is less than1cm?1,is associated with the transitions involving the defect states.Regions B and C are denoted as subbandgap absorption since the absorption coef?cient re?ects transitions involving states within the bandgap.

From the absorption coef?cient of a-Si:H based materials,the so-called optical bandgap is determined.The optical bandgap is a useful material parameter for comparing the light absorption properties of a-Si:H based materials.Generally,the higher optical bandgap of a material,the less light it will absorb.The optical bandgap,E opt,is determined by extrapolating a linear part of the following function[α(E)×n(E)×E]1/(1+p+q)versus the photon energy E toα(E)=0,forα≥103cm?1:

(α(E)×n(E)×E)1/(1+p+q)=B(E?E opt)(5.3) whereα(E)is the absorption coef?cient,n(E)is the refractive index,p and q are constants that describe the shape of the density of extended states distribution for the conduction band and valence band,respectively,and B is a prefactor.When the density of state distribution near the band edges has a square root energy dependence(p=q=1/2),as is commonly the case in crystalline semiconductors,Equation(5.3)describes the so-called Tauc plot[26],and the corresponding optical bandgap is called the Tauc optical gap.When the distribution near the band edges is assumed to be linear(p=q=1),E opt is called the cubic optical gap.The Tauc gap of device quality intrinsic a-Si:H is in the range of1.70to1.80eV;usually0.1to 0.2eV smaller than this is the cubic gap of the same material.The optical bandgap increases with increasing hydrogen concentration in the?lm[39].

The refractive index of a-Si:H shows a maximum of almost5.0around425nm and then decreases as a function of wavelength in the region of interest(350nm to900nm).The refractive index slightly decreases above900nm.For device quality a-Si:H,the refractive index at900nm is above3.6.

5.3.5Electrical properties

The electrical properties of a-Si:H are usually characterized in terms of dark conductivity, photoconductivity and mobility lifetime product.Measuring these properties is a standard approach to obtain information about the quality of a-Si:H material for application in solar cells.

5.3.5.1Dark conductivity

The dark conductivityσd of device quality intrinsic a-Si:H is less than1×10?10 ?1cm?1. To determine it and its activation energy,a very low current is measured,in the order of picoamperes.Such measurements are usually carried out on samples with two1to2cm long coplanar metal electrodes deposited less than1mm apart from each other on a single a-Si:H layer on highly resistive glass,such as Corning1737.Care has to be taken that moisture or

184THIN FILM SOLAR CELLS

diffusing impurities do not affect the measurement of current.Therefore,the measurement is usually taken in vacuum or in an inert atmosphere and before the measurement,the sample is annealed at150?C for half an hour to evaporate all moisture present on the surface of the ?lm.A voltage of typically100V is applied to a sample with an a-Si:H layer of about1μm thick in order to obtain a current of tens of picoamperes through the sample that can be reliably measured.

The dark conductivity is determined as:

σd=I

U w

ld

(5.4)

where U is the applied voltage,I is the measured current,l is the length of the electrodes (～1to2cm),w is the distance between the electrodes(0.5to1mm),and d is the thickness of the?lm.

5.3.5.2Activation energy of the dark conductivity

The measurement of the temperature dependence of the dark conductivity is used to evaluate the activation energy of the dark conductivity,E A,which gives a good approximation of the position of the Fermi level in a-Si:H?lms.The temperature dependence of the dark conductivity σd(T)is described as

σd(T)=σ0exp(?E A/kT)(5.5) whereσ0is a conductivity prefactor,T the absolute temperature and k Boltzmann’s constant. From the slope of the Arrhenius plot,which in this case is the relationship between log(σd(T)) and1/T,the activation energy is determined.In combination with the optical bandgap,the activation energy is a good measure to evaluate the presence of impurities in the?lm.The impurities often act as dopants,and even a small concentration of impurities,1×1017cm?3 of O or N,causes a shift of the Fermi level.For undoped a-Si:H,the activation energy should be higher than0.80eV.

The low dark conductivity of undoped a-Si:H is a result of the low mobility of charge carriers and the high mobility gap of a-Si:H.This is also re?ected by the high activation energy of the dark conductivity.The mobility of the charge carriers in the extended states of a-Si:H is about two orders of magnitude lower than in single crystal silicon.Typically,intrinsic a-Si:H is characterized by electron mobility values of10to20cm2V?1s?1,and the hole mobility is 1to5cm2V?1s?1.

5.3.5.3Photoconductivity

The photoconductivity can be determined by illuminating the same samples as used for the dark conductivity measurement with appropriate light.Often the AM1.5light spectrum with an incident power of100mW cm?2is used.With these conditions,the photoconductivity of device quality undoped a-Si:H,calculated from the photocurrent similarly to Equation(5.4),should be higher than1×10?5 ?1cm?1.The ratio of the photoconductivity and dark conductivity is called the photoresponse.This parameter gives an indication of the suitability of a material

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES185 for use as a photoactive layer in a solar cell.A good photoresponse value for a-Si:H is higher than105.

Photoconduction is a complex process of generation,transport and recombination of excess photogenerated carriers.The optical generation rate of carriers G depends on the absorption coef?cient,α,and the quantum ef?ciency for carrier generationηg.The latter represents the number of electron–hole pairs generated by one absorbed photon.When we assume that the current in a-Si:H is dominated by electrons,transport and recombination are characterized by the extended state mobilityμof electrons and their lifetime,τ.Its photoconductivity can be written as:

σph=qμ n=qμτG(5.6) where q is the unit charge and n the concentration of photogenerated electrons.The average optical generation rate over the whole thickness,d,of the?lm is related to the absorbance,A, which can be calculated from the Lambert–Beer absorption formula:

A= 0(1?R)(1?exp(?αd)),(5.7)

where 0is the incident photon?ux density and R the re?ectance of the air–a-Si:H interface. Neglecting the spectral dependence of the re?ectance,the average generation rate can be approximated by:

G=ηg A

d

=ηg

0(1?R)(1?exp(?αd))

d

(5.8)

By combining Equations(5.6),(5.7)and(5.8),the photoconductivity can be expressed as:

σph=qμτηg 0(1?R)(1?exp(?αd))

d

(5.9)

The quantum ef?ciency mobility lifetime productηgμτis a useful and often used?gure of merit that combines the photoabsorption,transport and recombination in an a-Si:H?lm. This quantity is determined by measuring photocurrent when illuminating the sample with a monochromatic light of a relatively long wavelength.At such a wavelength,the absorption coef?cient is small,which results in almost uniform carrier generation in the a-Si:H?lm. Usually,600nm light is chosen as the probe c8ab0301a6c30c2259019ebcing Equation(5.9)combined with the geometry factors of the sample as in Equation(5.4),theηgμτproduct is obtained as:

(μτηg)600=

I ph w

qUl 0(1?R)(1?exp(?αd))

(5.10)

When we assume thatηg=1in amorphous silicon,it means that one absorbed photon generates one electron–hole pair and the mobility lifetime product at600nm for device quality a-Si:H isμτ≥1×10?7cm2V?1.

5.3.5.4Ambipolar diffusion length

Theμτproduct determined from the photoconductivity measurement characterizes the charge carriers that dominate the transport,i.e.the electrons in amorphous silicon.However,solar

186THIN FILM SOLAR CELLS

cell performance is often determined by the transport properties of minority carriers,which

are holes in a-Si:c8ab0301a6c30c2259019ebcmonly the steady state photocarrier grating(SSPG)technique is used

to determine the ambipolar diffusion length in amorphous silicon[40,41],from which the

mobility lifetime product of the holes is calculated.

When an excess of photogenerated carriers is not distributed uniformly throughout a semi-

conductor,diffusion of the excess carriers takes place.The excess carriers will diffuse in the

semiconductor until they recombine.In the absence of an electric?eld,the photogenerated

electrons and holes diffuse in the same direction.This process is called ambipolar transport.

The average distance that the excess carriers can diffuse due to ambipolar transport before

being annihilated is de?ned as the ambipolar diffusion length.The ambipolar diffusion length

is therefore a good?gure of merit for applying a semiconductor material as a photoactive layer

in a solar cell.In case of intrinsic amorphous silicon,ambipolar transport is determined by the

mobility of the less mobile charge carriers,holes.

The same samples as for the conductivity measurements can be used for the SSPG technique.

The principle of the SSPG technique is based on the creation of a steady state interference

pattern in the concentration of photogenerated carriers in the a-Si:H?lm.This concentration

pattern is made by illuminating the sample with two interfering beams generated from one

laser.The concentration fringes are parallel to the electrodes of the sample.The photocarriers

diffuse from highly populated regions to regions of lower concentration,which leads to a

reduction of the amplitude of the modulated carrier population.This reduction depends on the

period of the grating pattern and results in a change in the photoconductivity of the sample,

which is measured perpendicular to the grating fringes.

In the SSPG experiment,the laser beam is split into two beams,where the intensity of one

beam I1is larger than that of the second beam I2.First,the sample is illuminated with both

beams which interfere.The weak excitation beam is chopped and the resulting photocurrent is

measured using a lock-in ampli?er.The obtained signal is proportional toσg?σ(I1),whereσg represents the photoconductivity in the presence of the light interference pattern andσ(I1)is

the photoconductivity due to the illumination with only beam I1.Then the sample is illuminated

again,but this time the beam with the intensity I1is incoherent with the second beam.The

obtained signal is proportional toσg(I1+I2)?σ(I1).The ratio between these signals de?nes the parameterβ:

β=

σg?σ(I1)

σ(I1+I2)?σ(I1)

(5.11)

The parameterβdepends on the ambipolar diffusion length L amb and the grating period in the following way(as derived in reference[40]):

β=1?

2γγ20

[1+(2πL amb/ )2]2

(5.12)

In Equation(5.12),γis a power exponent in the relation of the photoconductivity with the generation rate(σph∝Gγ),andγ0is a factor that characterizes an interference contrast that can change due to,for example,partial coherence between the beams or light scattering (0<γ0≤1).The ambipolar diffusion length is a?tting parameter in the dependence ofβon the grating period and is obtained either from the slope of a plot ?2versus(1?β)?1//2, or from the intersection of the line with the ?2axis.The exponentγis determined from a

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES187 plot of the photoconductivity as a function of the light intensity,which is varied using neutral density?lters.

5.3.6Determination of density of states

In the following,we present widely used methods to evaluate the density of states in thin amorphous and microcrystalline silicon layers.

5.3.

6.1Optoelectrical methods

A large group of methods is based on correlating the optical absorption in a-Si:H?lms or devices with the density of states distribution.The subbandgap absorption especially is of major interest since it re?ects transitions involving the localized states within the bandgap. However,the subbandgap absorption is weak and therefore indirect methods,which are based on measurement of some secondary effect,are used to determine the absorption coef?cient. In photothermal de?ection spectroscopy[42],the de?ection of a probe laser beam re?ects a change in the refractive index of a medium which is in contact with the a-Si:H?lm.The change of the refractive index depends on the amount of heat generated by the absorption of monochromatic light in the a-Si:H?lm and dissipated from the?lm into the medium.Other techniques,such as constant photocurrent method[43],dual beam photoconductivity[44], and the recently introduced Fourier transform photocurrent spectroscopy[45]are based on measurement of the spectral dependence of the photoconductivity.

5.3.

6.2Photothermal de?ection spectroscopy

Photothermal de?ection spectroscopy(PDS)is based on the conversion of a fraction of the absorbed photon energy in a?lm into thermal energy(heat),which dissipates and causes a change in the index of refraction of a medium adjacent to the surface of the?lm.By probing the medium’s refractive index change with a laser beam,one can relate the probe beam de?ection to the optical absorption of the?lm.A sample with a-Si:H?lm is immersed in an optically transparent and thermally conductive liquid.A chopped,monochromatic beam,which is called a pump beam,illuminates the sample.The de?ection signal is measured by a position sensitive detector,which is connected to a lock-in ampli?er.

The attractive feature of this technique is its high sensitivity:absorbance values ofαd≈10?5can be measured on thin?lms.For a-Si:H?lms with a typical thickness of1μm,an absorption coef?cient as low as10?1cm?1can be determined.The PDS technique is sensitive to surface states,and for this reason,PDS results can overestimate the bulk density of states.

5.3.

6.3Constant photocurrent method

The constant photocurrent method(CPM)is based on measurement of the photoconductivity as described in the previous section.In CPM measurement,the steady state photocurrent is measured as a function of photon energy in the subbandgap region.In this region,the absorption

188THIN FILM SOLAR CELLS

is weak,which means thatαd 1.The exponential function can be expanded to a power series and exp(?αd)≈1?αc8ab0301a6c30c2259019ebcing this approximation in Equation(5.9),the photoconductivity can be expressed as:

σph=qμτηg 0(1?R)α(5.13) Equation(5.4)describes the relation between the photocurrent I ph and the photoconductivity σc8ab0301a6c30c2259019ebcbining Equations(5.4)and(5.13),the photocurrent can be expressed as:

I ph∝ 0αηgμτ(5.14)

The basic idea of CPM measurement is to ensure that the termηgμτis kept constant during measurement.This is achieved by keeping the photocurrent constant while changing the photon ?ux density.The constant photocurrent means that the positions of the quasi-Fermi levels for holes and electrons in the bandgap,which determine the number of recombination centres,do not change during the experiment.During the whole measurement,the carrier lifetime is,in this way,kept constant.When we assume that the mobility of the carriers and the generation quantum ef?ciency are not spectrally dependent,then the absorption coef?cient is dependent only on the incident photon?ux:

αCPM(E)≈C CPM

0(E)(5.15)

where C CPM is an energy independent constant and 0(E)the number of photons necessary to keep the photocurrent constant.The relative absorption coef?cientαCPM is calibrated to the ab-solute absorption coef?cient,which is determined from re?ection–transmission measurement. This original CPM method was later adapted in order to measure the optical(photocurrent) absorption spectrum directly in absolute units without additional calibration and undisturbed by interference fringes[46].

5.3.

6.4Dual beam photoconductivity

The basic idea of dual beam photoconductivity(DBP)measurement is that in addition to a chopped probe beam of monochromatic light,a second light source is used to illuminate the sample.This additional‘bias’light keeps the splitting of the quasi-Fermi levels constant during the measurement.The ac fraction of the photocurrent is measured using a lock-in ampli?er and depends on both the monochromatic photon?ux and the absorption coef?cient of the a-Si:H ?lm[44].The absorption coef?cient can be extracted from the measured photocurrent and photon?ux density by:

αDBP(E)≈C DBP I ph(E)

0(E)(5.16)

where C DBP is an energy independent constant and 0(E)the photon?ux density.As in the case of the CPM,the relative absorption coef?cientαDBP has to be normalized to the absolute absorption coef?cient.The advantage of the DBP measurement in comparison to the CPM is a shorter measurement time since it is not necessary to adjust the photon?ux density of monochromatic light in order to keep the photocurrent constant.

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES189 5.3.6.5Fourier transform photocurrent spectroscopy

Recently,the Fourier transform photocurrent spectroscopy method[45]has been introduced for determining the sub-bandgap absorption.A Fourier transform infrared(FTIR)spectrometer is used as the interferometer and an a-Si:H sample as the external detector.The spectrometer can be equipped with a globar and white light source in order to measure in0.4to1.8eV spectrum range.The sample,which can be a thin?lm deposited on a substrate or a complete solar cell, is connected to an electrical circuit with a low noise voltage source and a current preampli?er. The photocurrent of the sample is ampli?ed,and an A/D converter digitizes the output of the preampli?c8ab0301a6c30c2259019ebcing a computer,the signal is?nally translated from the time domain to the frequency domain by Fourier transformation.The FTIR signal from the sample is normalized to the FTIR signal from a spectrally independent detector.The advantage of this method is its sensitivity to low photon energies;the measurement of the subbandgap absorption is extended from～0.8eV,typical for the CPM and DBP measurement,to0.4eV.The measurement time is strongly reduced and is only a few minutes.

5.3.

6.6Determination of defect concentration from absorption coef?cient

Using the subbandgap region of the absorption coef?cient(see Figure5.3a),the defect con-centration in a-Si:H is determined using a simple procedure.First,the absorption due to the defect states is determined as:

αdef=α?α0exp(E/E0)(5.17) whereα0and E0are obtained from the?t to the exponential absorption in region B.Assuming that the optical matrix element is constant in this region,αdef is related to the defect density concentration,N d,by[42]:

N d=K d

αdef·d E(5.18)

where K d is a correlation factor that depends on the measurement method used to deter-

mine the absorption coef?cient in region B and C.For two widely used approaches,pho-

tothermal de?ection spectroscopy and the photoconductivity methods,the correlation factor

has been determined to be K d=7.9×1015cm?2eV?1[42]and K d=1.6×1016cm?2eV?1 [47],respectively.Another approach often encountered in the literature is to correlate the

value of the absorption coef?cient at1.2eV to the density of states in the following way

[48]:

N d=2.4?5.0×1016αCPM(1.2eV)or N d=1.2?2.5×1016αPDS(1.2eV)(5.19)

Information about the energy dependence of the density of states in the mobility gap can be extracted from the absorption coef?cient using the deconvolution approach[43].In this approach,the energy distribution of the density of states is extracted by matching the simulated absorption coef?cient to the experimental one.

190THIN FILM SOLAR CELLS

5.3.

6.7Space charge methods

Several methods use the properties of a space charge region that is formed at a-Si:H interfaces. The?eld effect technique[49],deep level transient spectroscopy(DLTS)[50]and space charge limited current[51,52]are the main representatives of these techniques.

5.3.

6.8Deep level transient spectroscopy

Deep level transient spectroscopy(DLTS)is an extremely sensitive technique and different from ESR measurements in that it also detects energy levels from nonparamagnetic defects.It is routinely used to determine energy levels and defect concentrations in semiconductors,and it is perhaps the most common technique for measuring deep levels in crystalline semiconductors.It has been demonstrated[50,53]that in a-Si:H,DLTS is also a valuable method for evaluating the electronic density of states.Originally,a capacitance version of DLTS was used to characterize lightly doped a-Si:H?lms.Undoped a-Si:H?lms can be characterized using a charge version of DLTS(Q-DLTS)[54].

The Q-DLTS sample is usually a metal/oxide/semiconductor(MOS)structure consisting of a1μm thick a-Si:H layer deposited on a highly doped n+type crystalline silicon substrate, which acts as a back contact.For successful Q-DLTS experiments on undoped a-Si:H,a very thin insulating layer has to be created in the surface region of a-Si:H.This insulating layer strongly reduces the leakage current of the sample,which is then negligible with respect to charge transients,and enables shifting of the Fermi level in the a-Si:H?lm with an applied bias voltage.An Al(semitransparent)layer usually forms the top electrode.By applying bias voltage pulses to the MOS sample,the Fermi level is shifted towards the conduction band mobility edge and the states in the gap of a-Si:H are?lled with charge carriers.After each ?lling pulse,the transient current in the external circuit is measured as a function of temperature. The charge emitted from the occupied trap states is determined by integrating the measured current.The charge released at a speci?c temperature is proportional to the concentration of states at a speci?c energy in the mobility gap of the a-Si:H material.This technique is suitable for investigating the evolution of the gap states distribution due to light or particle induced degradation,providing new insights in the complex behaviour of a-Si:H under light or particle exposure[55].

Table5.1summarizes the criteria for device quality intrinsic amorphous silicon.

5.3.7Metastability

Inherent to a-Si:H are changes in its electronic properties under light exposure.This is known today as the Staebler–Wronski effect[24].Since the observation of the Staebler–Wronski effect, a large effort has been put into obtaining an understanding of the processes that cause the light induced structural and optoelectronic changes in a-Si:H[56–59].An essential feature of the light induced effects on a-Si:H?lms and solar cells is that most of the effects are‘metastable,’which means they are reversible and can be removed by annealing at temperatures above 150?C.

Light soaking is thought to lead to the creation of additional dangling bond defects,which is regarded as the principal cause of the Staebler–Wronski effect.The increase of the density of

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES

191

Table 5.1Requirements for device quality a-Si:H and a-SiGe:H ?lms for solar cells Property

a-Si:H a-SiGe:H Dark conductivity [ ?1cm ?1]<5×10?10<5×10?9AM1.5conductivity [ ?1cm ?1]>1×10?5

>5×10?6

Urbach energy [meV]<47<55Activation energy [eV]≈0.8≈0.7Bandgap,Tauc [eV]<1.8～1.45Bandgap,cubic [eV]

<1.6 1.32Absorption coef?cient (600nm)[cm ?1]≥3.5×104≥1×105Absorption coef?cient (400nm)[cm ?1]≥5×105≥6×105Density of defect states

(CPM,DBP)methods [cm ?3]≤1×1016≤1×1017ESR method [cm ?3]

≤8×1015Mobility-lifetime product (600nm)[cm 2/V]≥1×10?7

H content [at.%]

9–1110–15Microstructure parameter <0.1

<0.2Ge content [at.%]

40

defects in a-Si:H due to light soaking is demonstrated in Figure 5.4,which shows the change in the absorption coef?cient determined by the DBP technique.The sample was illuminated using a He–Ne red laser (λ=633nm)with an intensity of ～40mW cm ?2.Shown by Figure 5.4,the subbandgap absorption increases with illumination time in the photon energy range between 0.8and 1.4eV ,which re?ects an increase in the defect density.An important aspect of light induced defect generation in a-Si:H is that it saturates.The saturation value of the defect density near room temperature has been found to be ～2×1017cm ?3and is almost independent of illumination intensity and sample temperature up to 70?C [60].

1x10

-2

1x10-1

1x10

1x101

1x102

A b s o r p t i o n c o e f f i c i e n t (c m -1)

Energy (eV)

Figure 5.4The change in subbandgap absorption coef?cient of a-Si:H due to light soaking.

192THIN FILM SOLAR CELLS

0.00.5

1.0

ΔQ (n C )Temperature (K)

Figure 5.5The Q-DLTS signal of a-Si:H after light soaking for different exposure times.

Figure 5.5presents the evolution of the Q-DLTS signal measured after light soaking,which gives information about the energy distribution of the defect states.Light soaking was effected with a He–Ne red laser (λ=633nm)with an intensity of ～250mW cm ?2.The time evolution of the Q-DLTS spectrum shows complex behaviour.At a low temperature the signal disappears,while at 370K a peak grows signi?cantly with increasing illumination time.The Q-DLTS response around 450K does not seem to be in?uenced by moderate light soaking.These three components in the Q-DLTS signal are related to the positively charged,D h ,neutral,D z ,and negatively charged,D e ,defect state distributions,respectively,as predicted by the defect pool model [34].The arrows in Figure 5.5indicate the peak positions of D h ,D z ,and D e gap state distributions in the Q-DLTS spectra.The Q-DLTS measurements indicate that a substantial amount of positively charged defects is removed,the concentration of negatively charged defects remains unchanged and additional neutral defects representing dangling bonds [61]are c8ab0301a6c30c2259019ebcing the Q-DLTS technique,new insights in the origin and behaviour of different types of defects in a-Si:H have been obtained recently [55].In this work,it is proposed that in addition to dangling bonds,other types of defects exist in a-Si:H that play an important role in the Staebler–Wronski effect.Positively charged states above midgap are related to a complex formed by a Si dangling bond and a hydrogen molecule.The origins of negatively charged states below midgap are attributed to the ?oating bonds.Nevertheless,the Staebler–Wronski effect remains a complex phenomenon in a-Si:H and many unresolved issues still remain.The exact role of hydrogen,weak Si–Si bonds and Si–H bonds and complexes in the creation of the metastable defects is still under c8ab0301a6c30c2259019ebcputer simulations of an a-Si:H network support these investigations [62,63].Still,there is no commonly accepted model for the metastable defect creation in a-Si:H that is able to explain all experimental observations [58].

5.3.8Hydrogenated amorphous silicon from hydrogen diluted silane

So far,the effort to fabricate more stable a-Si:H using the PECVD technique has demon-strated the bene?cial effect of hydrogen dilution of the source gas on the quality of a-Si:H.

ADV ANCED AMORPHOUS SILICON SOLAR CELL TECHNOLOGIES193

Figure5.6TEM photograph of1μm thick silicon layer grown in the protocrystalline regime. Solar cells with a-Si:H absorbers prepared with hydrogen dilution of silane showed a better performance after light exposure than their conventional undiluted counterparts[64,65].Hy-drogen diluted a-Si:H has received a lot of attention since then and owing to several speci?c properties,this material is now described in the literature as protocrystalline silicon[66].The hydrogen–silane dilution ratio(R=H2/SiH4)roughly de?nes the deposition conditions at which the so-called protocrystalline Si:H growth regime occurs.In the protocrystalline Si:H growth regime,unique evolutionary growth behaviour is observed,when the?lm evolves from an amorphous phase into?rst a mixed amorphous–microcrystalline and subsequently into a single microcrystalline phase.This behaviour is clearly demonstrated in Figure5.6, which shows a transmission electron microscopy photograph of a layer of about1μm thick deposited on a glass substrate at R=30.It is important to note that protocrystalline Si:H material is de?ned as being fully amorphous,and once the transition from the amorphous to the mixed phase occurs,the?lm is no longer considered to be protocrystalline.Based on in situ real time spectroscopic ellipsometry studies,amorphous-to-microcrystalline silicon phase diagrams have been determined,which predict the thickness at which the transition from amorphous to microcrystalline material for a given R occurs[67].For lower dilutions (R<10),?lms are invariably amorphous;this means that there is no evolutionary growth that is characteristic for the protocrystalline growth regime(R>10).For R<10,a critical thickness is observed beyond which the growing amorphous silicon surface becomes rough. The improved stability and quality of protocrystalline Si:H layers compared to conventional a-Si:H deposited from pure silane or using a low hydrogen dilution is attributed to an improved structural order.An enhanced medium range order is determined by measuring the full width at half the maximum( 2θ)of the?rst X-ray diffraction(XRD)peak for a-Si:H,which is centred at a scattering angle of2θ～28.5?.A typical value for a-Si:H is 2θ～6?,which decreases for protocrystalline Si:H to 2θ～5?[68].For the protocrystalline growth regime, an important practical feature is that the phase of the growing?lm strongly depends on the substrate[66].Under protocrystalline growth conditions,local epitaxial growth is favoured on crystalline silicon substrates.For this reason,it is dif?cult,in solar cells,to grow intrinsic protocrystalline Si:H layers directly on doped microcrystalline silicon layers without incorpo-rating a conventional a-Si:H interlayer.On amorphous silicon,the nucleation of crystallites is suppressed.The phase diagrams are important tools for optimization of amorphous and mi-crocrystalline silicon solar cells,as they help to keep the growth of the?lm within the desired phase.

本文来源：https://www.bwwdw.com/article/2uul.html