Thesis

D.A.SpeeMSc Thesis

December 2008

Preparations for Making Back ContactedHeterojunction Solar cells

Supervisors:

Prof. Dr. Ruud E.I. Schropp

Dr. Yaohua Mai

Drs. Jan Willem Schuttauf

Dr. Hongbo Li

Nanophotonics 2009.01

ABSTRACT

Silicon heterojunction solar cells with a intrinsic thin layer (HIT cells) are com-bining the high efficiencies of crystalline silicon solar cells with the easy process-ing of amorphous silicon. The processes that are necessary to manufacture HITcells make this technology very suitable for producing interdigitated back con-tacted (IBC) cells. The cells discussed here were all made by plasma enhancedCVD, a deposition technique that enables easy deposition through a mask. Theaim of the research, which is partly described in this thesis, is optimizing theHIT structure and creating IBC cells using the HIT technology.

In this thesis several experiments, considering various aspects of (IBC) HITcells are discussed:

First of all we are looking for a better reproducibility and to extend thisresearch to other deposition systems. So far, we only made entire solar cells ina deposition system which is called the PILOT. This is a two chamber mediumsized (30 x 40 cm2) reactor, in which both PE CVD and HW CVD can beused. We want to extend this research to another deposition system called thePASTA. This is a multichamber deposition system with smaller (10 x 10cm2)chambers in which also both PE CVD and HW CVD can be used. Our workin the PASTA was started by investigating the deposition of p-layers. A p-layersuitable for the use in HIT cells was developed.

Furthermore three experiments on passivation of the p-n junction by anintrinsic a-Si:H layer were done. First the thickness of this layer was varied incells on textured wafers. An optimum i-layer thickness of 10 nm was found,which resulted in a cell with an efficiency of 16.4 %.

Investigating the optimum i-layer thickness on polished wafers, it was foundthat passivation properties of the i-layer were, in this case, very different. Be-cause the i-layer was suspected to grow epitaxially, a third experiment, varyingthe RF-power of the plasma, was done. Although an optimum in layer qualityand cell efficiency was found at an RF-power of 30W, none of the layers was ofvery good quality. Various possible reasons will be addressed in this thesis.

Because eventually the fabrication of IBC cells is desired, some trial depo-sitions were done and masks were designed. In addition to this some literaturestudies on IBC cells are included.

CONTENTS

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.1 a-Si/c-Si heterojunction solar cells . . . . . . . . . . . . . . . . . 51.2 Back contacted solar cells . . . . . . . . . . . . . . . . . . . . . . 61.3 This thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2. Heterojunction solar cells, theory and principles . . . . . . . . . . . . . 82.1 Semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2 Doping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.3 p-n junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.4 solar cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.4.1 practical solar cells . . . . . . . . . . . . . . . . . . . . . . 132.4.2 efficiency losses . . . . . . . . . . . . . . . . . . . . . . . . 14

2.5 heterojunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.6 back contacting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3. Production of layers and cells . . . . . . . . . . . . . . . . . . . . . . . 203.1 Plasma Enhanced Chemical Vapor Deposition . . . . . . . . . . . 203.2 PECVD systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.2.1 PILOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2.2 PASTA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.3 Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4. The characterization of a solar cell . . . . . . . . . . . . . . . . . . . . 234.1 RT mini . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234.2 Activation energy . . . . . . . . . . . . . . . . . . . . . . . . . . . 234.3 Solar simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234.4 Spectral response . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

5. Doped layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265.1 doped layers in the PILOT . . . . . . . . . . . . . . . . . . . . . 265.2 p-layers in the PASTA . . . . . . . . . . . . . . . . . . . . . . . . 27

6. Interface passivation by an intrinsic layers . . . . . . . . . . . . . . . . 286.1 passivation of the amorphous-crystalline interface . . . . . . . . . 286.2 i-layers on textured wafers . . . . . . . . . . . . . . . . . . . . . . 296.3 i-layer on polished wafers . . . . . . . . . . . . . . . . . . . . . . 31

6.3.1 Optimization of the i-layer thickness . . . . . . . . . . . . 326.3.2 Optimization of the rf-power . . . . . . . . . . . . . . . . 34

6.4 Conclusions and Recommendations . . . . . . . . . . . . . . . . . 37

Contents 4

7. Reproducibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397.1 Cleaning procedure: HF-dip . . . . . . . . . . . . . . . . . . . . . 39

7.1.1 Ultra pure water . . . . . . . . . . . . . . . . . . . . . . . 397.2 Reproducibility in the PILOT . . . . . . . . . . . . . . . . . . . . 40

7.2.1 Position in the plasma chamber . . . . . . . . . . . . . . . 407.2.2 Stability of the plasma . . . . . . . . . . . . . . . . . . . . 417.2.3 Opening of the plasma chamber . . . . . . . . . . . . . . . 41

7.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

8. Back contacted cells; deposition through a mask . . . . . . . . . . . . 428.1 IBC cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

8.1.1 Dimensions of the fingers . . . . . . . . . . . . . . . . . . 428.1.2 Front passivation and anti-reflection . . . . . . . . . . . . 43

8.2 Deposition through a mask . . . . . . . . . . . . . . . . . . . . . 438.2.1 Creating a practically feasible mask . . . . . . . . . . . . 448.2.2 Different cell configurations . . . . . . . . . . . . . . . . . 45

8.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

9. Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . 47

1. INTRODUCTION

If governments around the world stick with current policies, the world energydemand will be more than 50% higher in 2030 than today[1]. This would bedue to an ever increasing world population and the rapid economic growth incountries like India and China (together accounting for 45% of the increase indemand). In the most probable scenario fossil fuels will continue to dominatethe energy market. These trends will lead to two things: first the energy-relatedCO2 emission will grow by an estimated 57% between 2005 and 2030. RisingCO2 and other greenhouse-gas concentrations in the atmosphere are contribut-ing to higher global temperatures and changes in climate. Second there will bean increased reliance of oil and gas consuming countries on import, mostly fromthe Middle East and Russia. Both developments would be concerning [1].

Besides reducing the energy demand in all kinds of ways, a shift from fossilfuel to other forms of energy will be very important in avoiding future problems.These include nuclear energy as well as renewable energy sources. Globally,solar energy is probably the most promising form of renewable energy. Solarenergy has an enormous potential; the energy of the sunlight incident on earthexceeds the world energy consumption by more than 10.000 times [2, 3]. Thisenergy can be converted into useful forms in several ways. It can for instancebe used for heat production by solar collectors. However, the most elegant, andnowadays most useful, way of converting it is the photovoltaic energy conversion(PV), using the photoelectric effect, in which sunlight is directly converted intoelectricity. This technique of energy production has a very low environmentalimpact and brings no risks with it, in contrast to, for instance, nuclear energy.The efficiency of converting sunlight into energy is for this process typically 10-15%, which is more than 100 times higher than for the photosynthesis processin trees [3]!

The costs of energy produced by photovoltaic conversion are at this momentstill considerably higher than those for other forms of energy. This is the mainreason that it makes up only a tiny bit of the energy market today, despite allthe other advantages. In the future this will probably change and somewherebetween 2020 and 2030 the price of solar energy will be compatible with thatof other forms of energy, on the one hand because of the ever increasing fossilfuel prices and on the other hand because of the improvements in solar celltechnology [4].

1.1 a-Si/c-Si heterojunction solar cells

At present most solar cells are based on wafer technology, in which solar cellsare made of thin slices (0.2-0.5 mm) of crystalline silicon. In 2000 crystallinesilicon solar cells made up 86% of the PV market [5] and in 2006 even 93.5%

1. Introduction 6

[6]. Crystalline silicon solar cells can have very high conversion efficiencies butare very expensive for two reasons: first, silicon, although abundantly present,becomes very expensive when it has to be transformed into a crystalline formwith a very high purity. The costs become even higher because a lot of silicon islost when a single crystal is cut into thin wafers. Secondly, junctions are formedin a wafer by diffusion or ion-implantation, processes that need high tempera-tures (>900C) and consume a lot of energy [3]. Because of the high costs ofthese wafer based cells the so-called thin film technology was developed. Thinfilm solar cells can be made by relatively simple processes at low temperaturesand use up to a factor 1000 less silicon, with typical thicknesses between 200and 500 nm. Thin film silicon is deposited in an amorphous or microcrystallineform with the use of silane gas (SiH4). Although efficiencies of thin-film solarcells are ever increasing, they are still considerably lower than those of waferbased cells.

In the 1980’s the idea was launched to combine the high efficiency of crys-talline silicon with easy processing and low costs of amorphous silicon in asolar cell[3, 7]. The fact that the p-n junction is formed in a low tempera-ture process (mostly by PECVD) not only minimizes the thermal budget, butalso prevents the degradation of material quality: wafer quality can decreasesdrastically in high temperature steps of the conventional process[8]. This be-comes even more important nowadays, because industry tends towards the useof cheaper substrate material (multicrystalline silicon) in which thermal minor-ity lifetime degradation is even more critical[9]. Ever since the heterojunctionstructure was introduced in 1983[10], the efficiency steadily increased. On ofthe biggest challenges in increasing efficiency is to avoid recombination at thea-Si:H/c-Si interfaces[11]. In the early 1990’s heterojunction solarcells with athin intrinsic layer (HIT) were introduced. This was followed by various devel-opments such as the introduction of a back surface field (BSF), a highly dopedlayer at the back of the cell, which reduces recombination by repelling minoritycarriers from the back surface. Another development was the use of texturingto increase light scattering, which increases the path length and thus the ab-sorption probability of photons through the cell[3]. As a result of all of this, in1994, a heterojunction solar cell was made with an efficiency of 21%[12].

1.2 Back contacted solar cells

Although high efficiencies were achieved with HIT cells, the amorphous layers atthe front absorb short wavelength light and the best passivation is achieved withthicker layers, resulting in significant absorption losses in front emitter HIT cells.The realization of the full potential of heterojunction devices can be offered by astructure with both emitter and base contact at the back of the cell[13]: A way toincrease PV conversion efficiencies is represented by rear-junction, interdigitatedback contacted (IBC) solar cells, collecting photogenerated carriers entirely fromthe back of the cell. The advantage of this structure is that there is no contactgrid shading on the front which improves the short circuit current and theappearance of the solar cell[14]. In this way crystalline silicon solar cells withefficiencies of nearly 23% are already produced. However, the price of these isonly affordable to just a few niche applications[15].

Combining the IBC structure with the HIT technology has several advan-

1. Introduction 7

tages. First, thin wafers are attractive, since rear junction cells require a largediffusion length over device thickness ratio. Using low temperature depositioninstead of high temperature diffusions reduce thermal stress and bowing in thinwafers. Second, rear junction devices need low front surface recombination ve-locities, which can be achieved by deposited passivation layers. Finally, pattern-ing the rear, a central challenge in this structures, is easier in heterojunctions,since depositions are much easier masked than diffusions[16].

1.3 This thesis

In this thesis experiments on various aspects of heterojunction solar cells arediscussed, although the major part of it addresses experiments reducing recom-bination at the amorphous silicon/crystalline silicon interface by passivationwith a intrinsic amorphous silicon layer. After in chapters 2 to 4 theoreticalaspects and general remarks on the production and characterization of solarcells and silicon layers are discussed, chapters 5 to 7 are of a more specific andexperimental nature.

In chapter 5 the doped layers are discussed. With an eye on improving thestandard heterojunction solar cell fabrication process, a start was made withthe deposition of doped a-Si:H layers in a deposition system called the PASTA.In chapter 6 the a-Si/c-Si interface passivation by intrinsic a-Si:H layers is dis-cussed: Although good results were found on textured wafers with a diffusedBSF, it turned out to be difficult to achieve good passivation on polished wafers.Chapter 7 treats various reproducibility problems that we encountered duringthis research.

All the experiments discussed in chapters 5 to 7 concern conventional HIT-cells. However, they were performed with a view to develop a method to produceHIT IBC solar cells. In the last chapter a brief literature study on back contactedsolar cells, mask design as well as some trial depositions through a mask arediscussed.

2. HETEROJUNCTION SOLAR CELLS, THEORY ANDPRINCIPLES

In this chapter the theoretical aspects of heterojunction solar cells will be dis-cussed. Starting with the general properties of semiconductors, it will, via dop-ing, the p-n junction and heterojunctions, arrive at the theory of back contactedheterojunction solar cells. This makes it easier to see all features of the solarcells discussed in the last four chapters in the right context.

2.1 Semiconductors

From quantum mechanics we know, that, when the Coulomb potential of anisolated atomic nucleus acts on an electron, it can only have certain discreteenergy levels. These energy levels (relative to a reference energy E = 0) aregiven by:

En =−Z2m0q

4

8ε20h2n2

n ∈ N\0, (2.1)

where Z is the number of protons in the nucleus, m0 the free electron mass, q theelectron charge, ε0 the permittivity of free space and h Planck’s constant [17].When atoms are brought close together, to form a crystal, the Pauli exclusionprinciple ensures that all allowed electron energy levels have a slightly differentenergy. This results in the splitting of the original quantized energy levels of theisolated atom into many distinct, closely spaced energy levels: when the systemconsists of N atoms, the original energy level En splits into N allowed levels,which may, due to spin degeneracy, contain at most 2N electrons. Because thetotal extent of the levels originating from one distinct level is generally in theorder a few electron volts and the number of atoms in a crystal is very large,the separation between these levels is much smaller than the thermal energy ofan electron at room temperature. Thus the electrons will be able to move easilybetween these levels and we can speak of a continuous band of energies[18].These bands, bounded by minimum and maximum energies, may be separatedfrom other bands by forbidden-energy gaps, or overlap other bands. At T = 0K,thus when the electrons have no thermal energy, the lowest possible energy levelswill all be filled, the highest level that is filled in this situation is called the Fermilevel. Whether there are gaps and what is their size, in combination with theposition of the Fermi level, determines the electronic behavior of the material:In conductors, either the Fermi level lays within an energy band, or two bandsoverlap, ensuring that, at all temperatures, electrons can move through thesolid freely. In semiconductors and insulators, the Fermi level is located in themiddle of the band gap, which means that the lower one of the bands (thevalence band) is completely filled and the upper one (the conduction band) iscompletely empty at T = 0K. The valence band being completely full means

2. Heterojunction solar cells, theory and principles 9

Fig. 2.1: (a) Band representation of a semiconductor. (b) Energy density of allowedstates for electrons. (c) Probability of occupation of these states, accordingto the Fermi-Dirac distribution function. (d) Resulting energy distributionof electrons and holes. [20]

that electrons can not go to another state: they are bound to a specific nucleusand can not move through the solid. As soon as an electron is excited to theconduction band (a hole moves to the valence band at the same time) it canmove freely and thus conduct (hence the name conduction band). While theband gap size of insulators is generally large (typically 3-10 eV), the band gap ofsemiconductors is only 1-2 eV). The probability of the occupation of an allowedelectron energy level is given by the Fermi-Dirac distribution function

f(E) =1

1 + exp[(E − EF )/kT ], (2.2)

where T is the absolute temperature and k the Boltzmann constant. This meansthat at room temperature always some of the electrons are in the conductionband [19]. This is illustrated in Fig. 2.1 b and c. In an undoped (intrinsic) semi-conductor the amount of free electrons, which is determined by the temperature,will always equal the number of holes.

2.2 Doping

To increase the number of free electrons or holes, a semiconductor can be doped.For a material consisting of atoms with four valence electrons (group IV mate-rial), which silicon is, this can be done by adding atoms with either five or threevalence electrons to the material. These atoms act as substitutional impurities,meaning that they substitute for an atom of the host crystal, while the regularatomic arrangement of the material is maintained. These situations are shownin Fig. 2.2. When this is done with a group V atom, four of its valence electronswill form covalent bonds with the neighboring silicon atoms, but the fifth is ina different situation: it is not in the valence band, but because it is tied to anatom it is not free to move and not in the conduction band either. However, theenergy needed to ionize the atom and thus bring an electron into the conductionband, is, typically, much smaller than the band gap of the semiconductor[20].This results in an allowed energy level in the band gap close to the conduction


band. Atoms that, in this way, can easily donate an electron to the conduc-tion band are called donors, while semiconductors doped with donors are calledn-type semiconductors, because the conduction will be dominated by electrons(negative charge). In an analogous way a group III atom does not have enoughvalence electrons to form four covalent bonds and a hole is tied to the atom. Inthis case the energy needed to bring the hole into the valence band is relativelysmall compared to the band gap, which results in an allowed energy level closeto the valence band. These atoms are called acceptors (because of their accep-tance of an electron) and materials doped with them p-type semiconductors.Since the energy required to release a charge carrier from these dopant atoms

Fig. 2.2: Example of p-type silicon, doped with boron, a group III atom (left) andn-type silicon, doped with phosphorus, a group V atom (right).

is typically small (in the order of the thermal energy kT at room temperature),under regular conditions most of the dopant atoms will be ionized. When oneassumes that, for a n-type material, all of the dopant atoms are ionized andthat all electrons in the conduction band are donated by donors (which is avery good approximation at room temperature) the density of these electronswill be given by:

n = ND = NC exp[(EF − EC)/kT ], (2.3)

which can be written as:

EF − EC = kT ln[ND

NC], (2.4)

where ND is the density of donors and NC is the effective density of states inthe conduction band, which is constant for a fixed temperature. Similarly, forthe density of holes in the valence band of a p-type material:

p = NA = NV exp[(EV − EF )/kT ], (2.5)

andEV − EF = kT ln[

NA

NV]. (2.6)

When a semiconductor gets more heavily doped the energy difference betweenthe Fermi level and the conduction or valence band (activation energy) becomessmaller. The importance of making this difference as small as possible will beshown in the next section and measuring it will be discussed in section 4.2.


Fig. 2.3: Fermi levels in p-type doped silicon (left) and n-type doped silicon (right).

2.3 p-n junction

The essential part of a solar cell is a diode structure, or a p-n junction. Such ajunction can be formed by bringing a n-type material into contact with a p-typematerial. Due to diffusion, electrons from the n-type material, which are in theconduction band and are free to move, will move into the p-type material, whilethe holes that were in the p-type material, in the valence band, will move into then-type material. In the region close to the junction, which is called the depletionzone, these electrons and holes will recombine. Because of fixed charges in thenuclei of the dopant molecules, the disappearance of holes at the p-type sidecauses a net negative charge there, while the disappearance of electrons at then-type side of the junction causes a net positive charge. These charges cause anelectric field in the opposite direction of the diffusion current. When the currentthat results from this electric field cancels the diffusion current the material isin equilibrium. This situation is shown in Fig.2.4. An energy band diagram forthis situation can be easily drawn by recognizing that in thermal equilibriumthe Fermi level at both sides of the junction must be the same. It becomesclear from this diagram that the lower the activation energy of the n- as well asthe p-material is, the larger the ”built in voltage” VBI of the junction will be(see Fig.2.4(d)). Thus, in equilibrium, the current towards the junction, calledrecombination current (electrons and holes causing this current recombine atthe junction) and the so called generation current are equal. This last currentis caused by electron-hole pairs, that are created at the junction by thermalor photoelectric excitation and separated by the electric field at the junction.When a p-n junction is illuminated a lot of electrons and holes will be created,due to photoelectric excitation. In this case an excess of electrons on the n-sideand an excess of holes on the p-side is created, which, when the two sides areshort circuited by a wire, causes a recombination current through this wire. Thecurrent that is generated in these conditions is called the short circuit currentISC [2].

2.4 solar cells

The current through a wire can be used for energy production. To calculate theoutput power of a solar cell and from that the conversion efficiency, one can takea look at the so called I-V characteristics of a solar cell. In the ideal case a solarcell is a big flat diode, which means that the ideal diode law is valid. This law


Fig. 2.4: A p-n junction in thermal equilibrium: (a) Fixed charge distribution. (b)Distribution of electric field. (c) Potential distribution. (d) Energy banddiagram.[21]


Fig. 2.5: I-V characteristics of a diode in the dark and when illuminated.[20].

gives the current through an unilluminated diode as function of an externallyapplied voltage:

I(V ) = I0(exp[qV

kT]− 1), (2.7)

in which I0, the saturation current is given by:

I0 = A(qDen

2i

LeNA+

qDhn2i

LhND). (2.8)

Here A is the cross-sectional area of the diode, De and Dh the diffusion coef-ficients of electrons and holes, ni the intrinsic charge density of the semicon-ductor, Le and Lh the diffusion lengths of electrons and holes and NA and ND

the dopant concentrations at both sides of the junction[21]. The characteristicfor an unilluminated cell is shown in Fig. 2.5 (dashed line). The characteristicsfor a cell under illumination can now be obtained by using the superpositionprinciple: the characteristics of the unilluminated cell are shifted downwardsover a distance equal to the light generated current IL. The curve that is shownin Fig. 2.4, is given by:

I(V ) = I0(exp[qV

kT]− 1)− IL. (2.9)

In Fig. 2.4 it can be seen that, under illumination, there is a region in the I-Vcharacteristic (the fourth quadrant) in which power can be extracted from thecell. The power output for any point in this region is equal to the product of thecurrent and the voltage at that particular point. There is one point (Vmp,Imp),indicated in Fig. 2.4, that maximizes the power output and is, not surprisingly,called the maximum power point[20]. The conversion efficiency of the cell is nowsimply given by the output power divided by the power of the light incident onthe cell. The I-V characteristic of a solar cell will be discussed in more detail insection 4.3.

2.4.1 practical solar cells

The most simple form of a silicon solar cell consists of a planar diode with frontand back contacts and an anti reflection coating (see fig. 2.5). The base of the


Fig. 2.6: Schematic picture of a practical solar cell. For clarity the cell thickness isexaggerated.

solar cell is a lightly doped p- or n-type silicon wafer, which is already dopedduring the production. The more heavily doped (opposite to the wafer) thintop layer (emitter) is formed afterwards. The front contact usually consistsof a pattern of small lines (fingers) and one or more larger conductors, whichcover the front of the cell for 5-15%. The back contact can cover the wholesurface, because no light falls in here. The contacts are mostly made of silveror aluminium. The anti reflection coating is made of a transparent conductiveoxide (TCO). It has a thickness that is such that, by means of destructiveinterference, there is a minimum in reflection in the wavelength area in which thesolar spectrum has the highest intensity[21]. In this research indium tin oxide(ITO) with a thickness of 80 nm is used. Because the TCO is conductive, thelateral conduction in the top of the cell increases, which enhances the collectionof charge carriers at the front surface. In practice, in most solar cells a backsurface field (BSF) is included. This is a layer at the back of the cell, that ishighly doped with the same type of doping as the base of the cell. In this waya low/high (n − n+ or p − p+) junction is created, at which, just as at the p-njunction an electric field is present. This results in the repulsion of minoritycarriers from the crystal surface, decreasing the surface recombination at theback of the cell and thus improving the cell efficiency.

2.4.2 efficiency losses

The performance of a solar cell is limited by a number of factors that can bedivided in two categories: fundamental and technologically determined factors.The first fundamental limiting factor is spectral mismatch. This results fromthe fact that one semiconductor, with a fixed band gap, is used to convert lightwith a broad spectrum (see Fig.2.7). Photons with an energy lower than Eg

will not be absorbed and are lost. Of photons with an energy higher than Eg

the energy is only partially used because every photon creates just one electron-hole pair that represents a fixed energy of 1.1 eV. In this way 50% of the energyof the sunlight is lost. The second fundamental loss is unavoidable (Auger)


Fig. 2.7: Solar spectrum, the wavelength of 1100 nm, corresponding to the band gapenergy of crystalline silicon, is indicated.

recombination in the material. In this process an electron and hole recombine,transferring the excess energy to a third charge carrier. This process gives riseto an efficiency loss of 27%: the highest possible Voc is estimated around 800mV instead of 1100 mV corresponding to the electron-hole pair energy of 1.1 eV.Finally the performance is limited by the fact that the fill factor (FF), which isgiven by

FF =VocIsc

VmpImp, (2.10)

can at most be 86%. These three factors result in a theoretical maximum ef-ficiency of 50% x 73% x 86% = 31% for a crystalline silicon solar cell[21]. Animportant technologically determined loss factor is contact coverage: becausethe front contact collects the current its electric resistance needs to be as lowas possible, however in practice this means that the contact coverage increases,which causes a current loss. Where an optimum needs to be found in conven-tional solar cells, this problem can be avoided entirely by making back contactedcells, which will be discussed in section 2.6. A second factor is reflectance at thesurface of the cell. This can be decreased to a minimum by an anti reflectioncoating and texturing of the front surface, which both enhance the incoupling ofthe light. The third and most important factor is recombination, at the surfaceas well as in the bulk, which causes a loss in current as well as voltage. In practicerecombination in crystalline silicon (c-Si) is dominated by Shockley-Read-Hall(SRH) recombination, which is a two-step process that takes place via allowedenergy levels in the band gap (traps), caused by defects in the crystal. For thisprocess, energy states in the middle of the band gap (”deep” states) are mosteffective, since they have to be filled and emptied and both processes dependexponentially on the distance to the valence or conduction band[20, 21]. Allthis is valid for both recombination in the bulk and at the surface. A fourth lossfactor is the series resistance of a cell, which includes lateral resistance in theemitter, contact resistance (occuring during current transport from the emitterto the contact) and the resistance in the contacts itself. Finally shunting needs


to be mentioned as a loss factor: although an ideal diode virtually does notconduct at reverse bias and only starts to conduct above voltages of 0.3-0.4V,in practice shunting can occur through so-called Ohmic paths through the p-njunction, causing the diode to conduct at voltages at which it normally does not.Knowing all this, an equivalent circuit of a practical solar cell can be drawn (seeFig. 2.8).

Fig. 2.8: Equivalent circuit of a solar cell, showing the light induced current, darkcurrent (diode), series resistance and shunt resistance[21].

2.5 heterojunctions

In a conventional c-Si solar cells the p-n junction is created by thermal diffusionof doping atoms into the crystal. This process is expensive and consumes ahuge amount of energy. An alternative for a diffused p-n junction is an amor-phous silicon/crystalline silicon (a-Si/c-Si) p-n junction: an a-Si emitter can bedeposited using simple technology at low temperatures[22]. In this way high ef-ficiencies of c-Si solar cells can be combined with easy processing of amorphoussilicon. In a-Si the silicon atoms do not form an ordered matrix as in a crys-

Fig. 2.9: 2-dimensional representation of crystalline (left)and amorphous (right) sili-con, where the open circles represent hydrogen atoms [21].

tal, bond angle, length and strength are slightly varying (see Fig.2.9). Becausesome valence electrons in a-Si are unbound (dangling bonds), creating energystates in the band gap, hydrogen is incorporated during deposition, which willpassivate most of the dangling bonds. In this way hydrogenated amorphous sil-icon (a-Si:H)is made. Due to the presence of dangling bonds and hydrogen the


electronic properties of a-Si:H are not nearly as good as those of c-Si, resultingin much higher recombination rates[2].

Another difference between c-Si and a-Si is the band gap: c-Si has an indirectband gap (the highest energy state of the valence band and the lowest energystate of the conduction band have a different momentum), which means thatnot only a photon is needed to excite a electron into the conduction band,but also a phonon for conservation of momentum. This lowers the transitionprobability for this process drastically. Because a-Si has a direct band gap, theabsorption probability for photons is much higher than in c-Si[2]. Due to thehigh absorption coefficient a-Si layers can, or in the case of heterojunction solarcells must, be much thinner than c-Si layers (one wants as many photons aspossible to be absorbed in the c-Si because of the higher electronic quality).

Besides being direct instead of indirect, the band gap of a-Si is much largerthan that of c-Si. The exact size depends on several parameters, but it can be aslarge as 2.0eV (1.1eV in c-Si). This makes the band diagram of a heterojunctiondifferent from the band diagram shown in Fig.2.4, which belongs to a conven-tional homojunction cell. A band diagram for the structure discussed in thisthesis (p-a-Si:H/n-c-Si junction) is shown in Fig.2.10. The different band gaps

Fig. 2.10: Energy band alignment of two semiconductors with different band gaps,before (a) and after (b) junction formation.[3]


result in discontinuities in the bands at the amorphous/crystalline interface. Inthis figure ∆EV and ∆EC are the valence and conduction band discontinuities.The built-in voltage VD is distributed over both materials: VD = VD1+VD2. Al-though the band discontinuities can create barriers that hinder the transport ofcharge carriers through the interface it was found by van Cleef that this shouldnot be a problem as long as the doping in the amorphous layer and the banddiscontinuity in the valence band are high enough: the spike in the valence bandbecomes narrow and holes can easily tunnel through[3].

2.6 back contacting

While in a regular solar cell a trade off has to be made between lowering theseries resistance by increasing the contact coverage at the front of the cell, andkeeping shadowing losses as low as possible, this is avoided in so called backcontacted cells. This is a general definition for solar cells that have the boththe positive and the negative external contacts at the rear surface of the cell,which can be realized in varies ways[23]. One of these is the interdigitated backcontacted (IBC) cell, proposed by Lammert and Schwartz, in which both theemitter and the BSF including are at the back of the wafer[24]. This structure,which is shown in Fig. 2.11, provides an independent control of optimum opticalperformance at the illuminated front side and optimum electrical performancewith low series resistance on the back side and can be easily combined withheterojunction technology[25]. The performance of a IBC solar cell is strongly

Fig. 2.11: Schematic picture of a heterojunction IBC solar cell. For clarity the cellthickness is exaggerated.

dependent on the ratio of minority carrier diffusion length (material quality)over wafer thickness and front surface recombination velocity. The first needsto be high and the second low: most photons are absorbed in the front partof the wafer, while the emitter is at the back. On the other hand, decreasingthe wafer thickness will result in losses in the large wavelength area of the solarspectrum (low absorption coefficient)[23]. A second factor that is importantis the (relative) size of the n- and p-fingers. When the (relative) width of thep-type fingers, the emitter in this research, is increased, the holes, which are the


minority charge carriers, have to travel less far. However, the recombinationrate is higher at the p-type collection junction than at the n-type[16]. Thus atrade off in relative sizes has to be made.

A bonus of the contactless front surface of IBC cells is that this makes iteasier to cover the cell by anti reflection coatings or other extra layers, whichcan for instance be used for quantum cutting. Furthermore, the visual appear-ance of solar cells changes into a uniform black plane, something that is muchappreciated by architects and designers[23].

IBC heterojunction solar cells will be discussed in more detail in chapter 8.

3. PRODUCTION OF LAYERS AND CELLS

The method that was used for deposition of the doped and undoped layers ofamorphous silicon in this research, plasma enhanced chemical vapor deposition,will be briefly discussed in this chapter. This was done using two setups, thePILOT and the PASTA. The ITO layers were deposited using radio frequencymagnetron sputtering in a setup called the SALSA.

3.1 Plasma Enhanced Chemical Vapor Deposition

Chemical vapor depostion (CVD) is a method that can be used to depositlayers of semiconductor material from the gas phase. Gases are dissociated intochemically active species that contribute to the growth process. Whereas inregular CVD gases are thermally dissociated at a heated substrate, in plasmaenhanced CVD (PECVD) these active species are created by energetic electronsthat are present in a plasma (a conductive gas in which a part of the molecules isionized)[26]. The gas and substrate may remain at a relatively low temperature(below 250C). This makes it possible to deposit a-Si:H on a great variety ofsubstrates.

Plasmas can be generated by different power sources. In this research a radiofrequency (RF) power source with a frequency of 13.56MHz was used. The RF-power is coupled to a discharge between two electrodes. Silane is introducedbetween the electrodes and the substrate is placed on one of them. A schematicpicture of the system can be seen in Fig.3.1(a). Free electrons are needed toignite the plasma. Being accelerated in the applied electric field they gainenough energy to ionize gas molecules:

SiH4 + e− → SiH+3 + H + 2e−. (3.1)

Extra electrons (and ions) are created in the process: after the plasma is ignitedthe density of charged particles rapidly increases.

Due to their smaller mass, electrons move faster than ions and therefore willdisperse faster from a neutral region between the electrodes[26]. The result is acharge separation and thus an electric field. This effect is known as ambipolardiffusion: the potential of the central region is higher than that of the electrodes(although the plasma as a whole is electrically neutral). The potential profile,averaged over one RF period (this is the potential ions experience since theyare not able to follow the rapid oscillations of the applied electric field) is drawnin Fig 3.1(b). In the plasma itself the potential is constant (Vp). The regionsbetween the plasma and the electrodes are the sheaths: here the positive ionsare accelerated towards the electrodes. The electrons are repelled by the electricfield in the sheaths: all the dissociation and ionization processes take place in theplasma. Although the reaction given in equation 3.1 is essential for sustaining

3. Production of layers and cells 21

Fig. 3.1: Schematic lay out of a plasma (a) and the potential diagrams between equaland unequal sized electrodes (b), in the latter case a negative (the groundedelectrode is usually bigger) DC self bias Vdc is present on the powered elec-trode.

the plasma, the main growth precursors in the deposition process are radicals:an electron needs a minimum amount of energy of about 12 eV to ionize silane,however to dissociate silane into neutral radicals a kinetic energy of 10 eV issufficient. This means that the production rate of radicals is much higher thanthat of ions. Radicals are formed by the following reactions[2]:

SiH4 + e− → SiH3 + H + e− (3.2)

SiH4 + e− → SiH2 + 2H + e−. (3.3)

A growth radical which reaches the surface of the substrate attaches to one ofthe dangling bonds. The radicals can diffuse over the surface in order to find anenergetically favorable position. Eventually cross linking with the neighboringsilicon atoms, under the release of atomic hydrogen, results in film growth. Alast important process is the formation of negative ions:

SiH4 + e− → SiH−3 + H (3.4)

The negative ions formed this way are confined in the plasma bulk, where theywill react with positive ions. In this way (large silicon-hydrogen clusters can beformed) they are believed to play an important role in dust formation[26].

3.2 PECVD systems

In a PECVD setup an RF power generator delivers the power to generate theplasma. Such generators usually have an output impedance of 50Ω. This outputimpedance and the impedance of the rest of the system (plasma chamber withplasma) should match. When this is not the case , some of the input power is”reflected”. Transmitted and reflected power are measured with an in-line powermeter. To prevent mismatch a matching box is included in the network, whichadjusts the impedance of the network by adjusting two variable capacitors.The impedances being not well matched (reflected power is non zero) meansthat not as much power is coupled into the plasma as one would like (deposition

3. Production of layers and cells 22

conditions are not as one would like). Besides that, the system can get damaged,because in practice this means that standing waves occur in the network, causinglarge heat production at particular points[27, 28]. In this research two differentsystems were used.

3.2.1 PILOT

The PILOT is a two chamber PECVD deposition system. One of the chambersis for the deposition of intrinsic Si:H layers and the other one is for the depositionof doped Si:H layers. The i-layer chamber can also be used for hot wire CVD(HW CVD) It is designed for large area (30 cm x 40 cm) substrates. Centrallylocated between the two deposition chambers is the load lock and transportchamber. The function of the load lock is to prevent the entrance of air in thechambers when a substrate is loaded. The deposition chambers are at ultra highvacuum (UHV) and have a back ground pressure of 10−7mbar. The transportof the substrate from one chamber to another is done manually.

3.2.2 PASTA

The second setup we used is the PASTA (Process equipment for AmorphousSemiconductor Thin film Applications). It consists of five process chambers (ofwhich 3 are used for PECVD and 2 for HWCVD), a central chamber and a loadlock. The samples are transported from one chamber to another via the centralchamber. This is done by means of a robot arm. The PASTA is also a UHVsystem and has a back ground pressure of 10−8mbar[2].

3.3 Sputtering

The ITO layers used in this research were deposited by RF magnetron sputter-ing. In a sputtering process a release of atoms from a surface is induced by highenergetic ion bombardment: these ions penetrate into the target material whilethey lose their energy via multiple collisions. In this process surface atoms fromthe target material are ejected[29]. The ions are created in a plasma which isgenerated in Ar-gas by a RF ac voltage between two electrodes (see previoussection). The Ar+ ions that are produced are accelerated towards the targetby the electric field in front of the cathode. A plasma can already be sustainedat a pressure below 10−5mbar: in this way the sputtered atoms can reach thesubstrate, which is placed at the anode, without further collisions[2].

The system used is called the SALSA (Sputtering Apparatus for Light Scat-tering Applications). It consists of a load lock and one process chamber in whichfour target spots are available. The background pressure is below 10−7mbar.To produce the ITO in this research In2O3 : Sn2O3(10%) was used as targetmaterial.

4. THE CHARACTERIZATION OF A SOLAR CELL

The main techniques that were used in this research to optically and electricallycharacterize layers (reflection/transmission- , activation energy- and light/darkconductivity measurements) and entire solar cells (I-V and spectral responsemeasurements) are addressed. The measurements were performed using fourdifferent setups.

4.1 RT mini

The setup used for reflection/transmission (RT) measurements is called RT mini.Reflection and transmission spectra of a layer deposited on a flat transparentsubstrate (Corning glass) are simultaneously measured. As light source, a whitehalogen lamp is used. The reflectance and transmission are measured by twophotodiodes for different wavelengths between 380 and 1050nm. From the RTdata the thickness of the layer, real and imaginary parts of the refractive indexof the layer and the absorption coefficient can be calculated[2]. The absorptioncoefficient for different wavelengths can be used to calculate the (tauc) bandgap of the material.

4.2 Activation energy

To electrically characterize deposited layers, two silver contacts are evaporatedon top of the layer parallel to each other. When a voltage V is applied betweenthe contacts the resulting current I can be measured from which the dark currentσd can be determined by:

σd =Iw

V ld, (4.1)

in which d is the layer thickness, w the distance between the contacts and l thelength of the contacts. By measuring the temperature dependence of the darkconductivity at room temperature, the activation energy of the material can bedetermined following:

σd(T ) = σ0 exp−Ea

kT, (4.2)

in which σ0 is the conductivity prefactor and k is the Boltzmann constant[30].

4.3 Solar simulator

The electrical output of a solar cell can be characterized by analyzing the currentdensity versus voltage (J-V) measurement. These measurements are performedunder illumination of a Wacom dual beam solar simulator, which is calibratedto the AM1.5 spectrum, as well as in the dark. A xenon lamp provides the light

4. The characterization of a solar cell 24

in the UV and visible part of the spectrum and a halogen lamp the light in theinfrared part[2].

Ideal solar cells show the behavior of an equivalent circuit shown in Fig.2.8,consisting of a photocurrent source Jph, a diode, a series resistance and a parallelresistance, which is given by:

J(V ) = −Jph + J0(exp(e(V − JRs)

nkT)− 1) +

V − JRs

Rp, (4.3)

where J0 is the dark saturation current, n the diode quality factor (a numberbetween 1 and 2), k the Boltzmann constant and T the temperature[30]. Fromthe dark JV measurement (Jph = 0), J0 and n can be derived. The short circuitcurrent density (Jsc), the open circuit voltage (Voc), the fill factor (FF) and theconversion efficiency can all be determined form the light characteristics. Thecomputer program that is used for the analysis takes the inverse slope at J=0(V = Voc) as the series resistance:

Rs = (dJ

dV|J=0)−1. (4.4)

For the parallel resistance the inverse slope at V=0 (J = Jsc) is taken[2]:

Rp = (dJ

dV|V =0)−1. (4.5)

An example of an I-V characteristic is shown in Fig.2.5.Besides for the I-V characteristic of solar cells, the solar simulator can be

used to measure the light conductivity of a single layer: the same measurementas described in section 4.2 is performed, only now under illumination of anAM1.5 spectrum.

4.4 Spectral response

Additional information about solar cells can be obtained from spectral responsemeasurements: the response to light of different wavelengths is measured. In thespectral response setup a 1000 W high pressure Xenon lamp is used as a lightsource. A computer controlled monochromator creates a monochromatic lightbeam with a wavelength varying from 350 to 1000nm. Before every measurementa reference measurement is done to determine the photon flux (φph(λ)) for everywavelength, by means of a calibrated reference cell. The external collectionefficiency (ECE) is defined as the number of electrons per incident photon thatcontributes to the current density in the device. For a certain wavelength it isgiven by:

ECE = (Jph(λ)eφph(λ)

. (4.6)

This should in theory always result in a value between 0 and 1. From the ECEthe Jsc under AM1.5 conditions can be calculated as follows:

JSRsc = e

∫ECE(λ)φAM1.5(λ)d(λ). (4.7)

The calculated value correlates with the value for Jsc obtained from an AM1.5 I-V measurement, but, depending on, amongst others, the integration boundaries

4. The characterization of a solar cell 25

and the numerical integration procedure chosen, may deviate from this up to10%[31].

Spectral response measurements are particularly useful to obtain informationabout different parts of a cell: a low response at small wavelengths (UV andblue) indicates absorption losses in the front of the cell (ITO, amorphous p-or i-layer in this research), whereas a low response in the longer wavelengthsindicates losses in the back of the cell, which makes it a good way of investigatingthe effect of a BSF and the light trapping properties.

5. DOPED LAYERS

In this chapter doped layers, made in the PILOT, that were already optimizedand used in cells for this research are described. One experiment was done: asa start of developing a standard heterojunction recipe in the PASTA, a p-layerseries was deposited with varying TMB-flow.

5.1 doped layers in the PILOT

The built in voltage of a p-n (or n-n+) junction is dependent on the position ofthe fermi levels of the materials on both sides of the junction; the larger the dif-ference, the larger the built in voltage. Because of this it is desirable for dopedlayers to have an activation energy Ea as low as possible: to be as heavily dopedas possible. In the first place the Voc of a solar cell will be as high as possible,secondly a strong band bending prevents barriers, as a result of band offsets, tobe present at the junction (this argument is only valid for heterojunctions)[3].In practise a-Si:H is doped by introducing a dopant gas into the plasma duringPECVD. For doping with Boron (p-type) TMB (trimethylborate B(CH3)3 isused and for doping with phosporus (n-type) Phosphane (PH3). The dopinglevel in the deposited a-Si:H layer increases (Ea decreases) with increasing gasphase doping concentration. However at a certain level saturation occurs: thedoping in the deposited layer will remain constant[32]. On the other hand defectdensity in the a-Si:H increases with increasing gas phase doping concentration.Something which must be avoided because recombination in the layer will in-crease with it[33]. The TMB or PH3 flow thus has to be optimized (of coursethere are various other parameters like deposition temperature and pressure thatare important for layer quality, however discussing them is beyond the scope ofthis thesis).

The doped layers deposited in the PILOT were already optimized: the stan-dard p-layer that was used in this research is deposited with a plasma powerof 60W (45mW/cm2) a pressure of 1.8mbar and at a substrate temperature of240C. Gas flows used are 9.9, 100 and 13.98sccm for respectively SiH4, H2 andTMB. The best layer produced in this way has an Ea of 0.355eV[34]. Appliedin a solar cell made on a textured wafer the optimal thickness of the p-layer wasfound to be 20nm. Although, for the purpose of depositing an a-Si:H BSF, astandard n-layer was developed, this was, until now, never used in a cell. Forthis layer a PH3 flow of 5.2 is used, while all other deposition conditions are thesame as for the p-layer (except for the fact that of course no TMB is used). Thebest Ea we have measured for such a layer is 0.195eV.

5. Doped layers 27

Fig. 5.1: P-layer properties with varying TMB-flow during deposition.

5.2 p-layers in the PASTA

Because it is believed that the reproducibility in the PASTA is higher than inthe PILOT (the latter will be addressed in the next two chapters) and a goodstandard procedure is desired, it was decided to develop a standard process forthe production of heterojunction solar cells in the PASTA. As a start of this, ap-layer series with varying TMB-flows was measured. The layers were depositedat a pressure of 2.57 mbar at a substrate temperature of 180C with a plasmapower of 5 W (35mW/cm2).The TMB-flow was varied between 10 and 22 sccm,while the flows of SiH4 and H2 were kept constant at respectively 10 and 100sccm. The results of the series are shown in Fig.5.1. We see indeed that the Ea

decreases with gas phase doping concentration for low values, while there is anoptimum when saturation of the a-Si:H with Boron is reached. The optimum inlayer quality clearly lays at a TMB-flow of 18 sccm. At this value an Ea of 0.36eV is obtained, which is approximately the same as the standard p-layer in thePILOT. We see that the band gap, despite increasing doping concentrations,remains almost constant at around 1.94 eV. The light and dark conductivityfor this layer are also satisfying with values of 8.6 · 10−5 S/cm and 5.4 · 10−5

S/cm (in general a a conductivity larger than 10−5S/cm is regarded sufficientfor device quality material[35]).

6. INTERFACE PASSIVATION BY AN INTRINSIC LAYERS

Three experiments on the amorphous-crystalline interface passivation by an in-trinsic layer were performed. In the first experiment, the passivation effect ofi-layers with varying thickness was investigated on textured wafers with a dif-fused BSF. The optimum thickness was found to be 10 nm, which resulted in acell with an efficiency of 16.4%. In the second experiment the thickness of i-layers deposited under the same conditions as in the first experiment was variedon polished wafers without BSF. It was found that passivation properties of thesame i-layer were much worse on a polished surface. In the third series i-layersdeposited with varying rf-power densities were investigated.

6.1 passivation of the amorphous-crystalline interface

The surface of a crystalline semiconductor can be regarded as an abrupt dis-continuity of the crystal lattice, at which the density of defect states is muchhigher than in the bulk, due to impurities and unsaturated silicon bonds (dan-gling bonds). Causing deep states, close to mid gap inside the c-Si band gap,these states are very effective recombination centers. In c-Si based cells, car-rier recombination at the crystal surface is one of the major loss factors anda-Si/c-Si heterojunctions are very sensitive to the number of defects at theamorphous/crystalline interface[3]. The surface recombination can be reducedeither by reducing the number of surface states or by repelling one type of car-rier from the defective surface, since both electrons and holes are required forrecombination. Reducing the surface states is called surface passivation. One ofthe methods to do this is by deposition of an a-Si:H layer. To obtain a low defectdensity at the amorphous-crystalline interface it is important that during thedeposition process the c-Si surface states are saturated by silicon or hydrogenbonds. Although with doped a-Si:H layers minority carriers are shielded fromthe highly defective interface by an internal electric field, solar cells that havea thin undoped a-Si:H layer at the interface generally perform better. Such astructure was first reported on by Sawada et.al.[12]. It was thought that therelatively low Voc of heterojunction solar cells compared to that of conventionalhomojunctions was due to the high density of interface states, originating fromplasma damage, on the c-Si wafer and a high defect state density in the p-typea-Si:H layer. Based on these considerations they inserted a very thin intrinsicbuffer layer with a low defect density between the c-Si wafer and the a-Si:Hp-layer, which clearly improved the Voc as well as the FF[12]. This structurewas called HIT (heterojunction with thin intrinsic layer) and the technologywas even industrially developed [36]. Although the incorporation of the a-Si:Hi-layer at the hetero-interface has been confirmed to improve the solar cell per-formance, the role of the i-layer still remains ambiguous[37, 38]. Studies on the

6. Interface passivation by an intrinsic layers 29

passivating properties of the i-layer have not led to a complete understanding ofthe physical properties determining the passivation. Also the role of hydrogenis still unclear, although it is known to be very important for the saturation ofdangling bonds[3]. Most studies were done on optimizing the i-layer thickness,which has been shown to be one of the crucial parameters of a a-Si:H c-Si solar-cell. Fujiwara et al. showed that at the interface the SiH2 content in the a-Si:Hlayer in their deposition process is very high, which indicates a poor networkformation there. In the first few nanometers of the i-layer, this content dropsand a steady state is reached. They suggested that the i-layer must be at leastthick enough to reach this ”steady state region”[37]. In general it can be saidthat the passivation properties of the i-layer increase with thickness in the firstfew nanometers. However, by further increasing the thickness of the i-layer,the cell performance can decrease for a couple of reasons. First of all, whenthe layer becomes thicker the bulk recombination in the a-Si:H becomes moreimportant (the minority carrier diffusion length in a-Si:H can be a 100 to 1000times shorter than in c-Si[21]). Second, the light absorption in the a-Si:H layers(i-layer as well as the p-layer) increases rapidly with thickness due to the directbandgap, something one wants to avoid because of the high recombination ratecompared to the c-Si. This leads to a loss in the short circuit current. Third,and most important, when the i-layer becomes too thick it creates a barrier,because of its very low conductivity: the series resistance of the cell becomesreally high and an S-shape appears in the I-V curve. In most studies the op-timum i-layer thickness is found to be 3-5nm [37, 12, 39]. The quality of thepassivation at the a-c interface can be measured indirectly by measuring the Voc

of a cell: this is dependent on the total recombination in the cell, if all othercell properties are kept the same it is possible to compare the passivation withdifferent i-layers. Of all cells discussed in this chapter the a-Si:H layers weredeposited using standard 13.56 MHz RF PECVD in the PILOT (see chapter 3).

6.2 i-layers on textured wafers

The first series we measured was an i-layer thickness series. The cells in thisseries were made on n-type 2-5 Ωcm c-Si FZ-wafers with a diffused BSF andtexturing on the front. A standard a-Si:H p-layer with a thickness of 20nm wasused. Before depositing the i-layer the wafers were cleaned in a 1% HF in H2Osolution for two minutes to remove the native oxide. After finishing the front sideof the cell (thin intrinsic layer, p-layer, 80 nm ITO and silver contacts) the backside was once again cleaned for one minute in a 1% HF in H2O solution beforeevaporating the silver back contact. The i-layer was deposited at a substratetemperature of approximately 120C, an rf power density of 37mW/cm2 (50W)and a total pressure of 0.5 mBar. Pure silane was used with a flow rate of63.5 SCCM[34]. The results of this first series are shown in Fig 6.1 and 6.2.The optimum in i-layer thickness is found to be 10 nm, which is rather highin comparison with other groups. This could be explained by the fact thatit is difficult to deposit a uniform a-Si:H layer on a textured wafer comparedto a polished one[12]. To reach a homogeneous phase, a thicker i-layer mightbe necessary. On the other hand, the thickness of our i-layers is measured”timewise”: the thickness of i-layers with different (long) deposition times wasmeasured on glass. In this way a deposition rate was determined, which we


Fig. 6.1: IV-curves of the i-layer thickness series on textured wafers.

Fig. 6.2: Cell properties with different i-layer thicknesses for the first series.

extrapolated to short deposition times. This could be another important reasonfor the discrepancy between our results and that of other groups.

We see that with an i-layer of 10 nm the Voc of the cell increases from0.604mV (no i-layer) to 0.623mV and the FF from 0.703 to 0.777, resulting ina very good efficiency of 16.4%. Where Voc and FF steadily increase from noi-layer to an i-layer thickness of 10nm, with an i-layer of 15nm they decrease


again. With an i-layer of 20nm the FF totally collapses due to a high seriesresistance (the inverse slope of the I-V curve at V=Voc) and we see an S-shapeappear in the IV curve, which also means that the high Voc of 645 mV, foundwith an i-layer of 20nm, is not a useful value. This can be explained by the factthat a too thick i-layer creates a barrier, because intrinsic silicon is not veryconductive. In the Isc the trend of a decrease with increasing i-layer thickness

Fig. 6.3: Spectral response results for cells of the first series, for clarity only cells withi-layer thicknesses of 5, 10 and 15 nm are shown.

can be observed. To investigate where the loss in Isc, with increasing i-layerthickness, comes from, a spectral response measurement was done of which theresult can be seen in Fig. 6.3. It is clear that the loss in Isc can be observedin the short wavelength area. This is exactly what was expected, because theseare the wavelengths that are mainly absorbed in the front of the cell (they havea high absorption coefficient) and thus more and more in the highly defectivea-Si:H layer with increasing thickness of this layer.

6.3 i-layer on polished wafers

In the second and third experiment a series of i-layers on wafers without BSFwas made: after all we eventually wanted to deposit our own BSF and createa back contacted cell. Unfortunately the wafers without diffused BSF wereonly available without texturing (polished). This meant two parameters thatare crucial for the recombination in the cell were changed at the same time(on a polished wafer the passivating properties of one and the same i-layer areexpected to be different than on a textured wafer). This made it very difficultto compare the results to those of the first experiment


6.3.1 Optimization of the i-layer thickness

The second series was again a thickness series and the i-layers were depositedin exactly the same way as in the first experiment, only this time on polishedwafers without a BSF. The results are shown in Fig.6.4 and 6.5. It has tobe mentioned that the cells in these series were fabricated using suboptimalconditions, mainly in the cleaning procedure, to watch the trends. With thesame parameters as the cell in these series with an i-layer of 5 nm, a cell wasmade with an efficiency of 11.6%, a Voc of 581 mV, an Isc of 28.6 mA and a FFof 0.710.

In this series we did not find a ”regular” optimum. Looking at Fig.6.5 onewould expect an optimum around a thickness of 5 nm. At a thickness of 5nm clearly something was different compared to the other cells of the series:although the Voc is not lower than at 2.5 or 7.5 nm the FF and the Isc are. Thiscan not be due to the increasing thickness because at 7.5 nm these parametersagain increase. The optimum we have is at 2.5nm with a Voc of 484 mV and aFF of 0.692 resulting in an efficiency of 9.64%. At first it was suggested that

Fig. 6.4: Cell properties with different i-layer thicknesses.

the recombination at the back side of the cell, where no passivation layer waspresent at all, was underestimated (in the first series there was no passivation atthe back either, but then there was a BSF to prevent recombination). Althoughnow the difficulty of two changing parameters came into play and the effecton the Voc of the i-layer and the BSF were impossible to distinguish, we triedto get an impression of the influence of the BSF by comparing the spectralresponse of the best cell of this series with that of the first (see Figure 6.6).Although the spectral response cannot directly be related to the Voc of a cell,it is of course influenced by recombination in different parts of the cell. Itcould be seen that the BSF had indeed a big influence on recombination: in thelong wavelength area the Isc is much higher with the BSF indicating a strongly



Fig. 6.6: A comparison of ECE of the best cell of the first series and the best cell ofthe second series. Although the ECE of the first is higher at all wavelengths,as a result of the texturing, the largest difference can obviously be seen inthe long wavelengths. This shows the importance of the BSF to preventrecombination at the back of the cell.

reduced recombination rate at the back of the cell. However, this could notexplain that the gain in efficiency by introducing an i-layer in this experimentwas only 0.55% (9.09% without an i-layer and 9.64% with a 2.5nm i-layer),where it was 2% in the first experiment. From these results is was concluded


that the passivation properties of an i-layer deposited under the same conditionswere far less good on a polished wafer than on a textured wafer. One possibleexplanation for the bad passivation on polished wafers could be epitaxial growthof the i-layer (the texture preventing this in the first series).

6.3.2 Optimization of the rf-power

In the third experiment cells were made with i-layers deposited with differentrf-powers. This was again done on polished wafers without a BSF. The reasonfor this series was the fact that Fujiwara et al. found that epitaxial growthof the i-layer can influence the passivation of the heterojunction very badly[40].This epitaxial growth can occur when the rf-power becomes too low, while on theother hand passivation becomes better when the epi-Si/a-Si:H phase boundary isapproached (see Fig.6.7). Where at high rf-powers epitaxial growth is preventedby a more heavy ion bombardment of the wafer, displacing atoms in the crystalsites, this is not the case when the rf-power becomes too low. To be sure that

Fig. 6.7: Performance of the a-Si:H/c-Si solar cell, plotted as a function of the rf powerdensity, found by Fujiwara et al.. The dotted lines in the figure show the epi-Si/a-Si:H phase boundaries[40].

we could come close to this phase boundary, the substrate temperature wasraised, as our plasma was known to be stable only with a rf-power higher thanapproximately 16mW/cm2 (20W). Besides the fact that different rf-powers wereused and the substrate temperature was increased to 140C, all other conditionswere kept the same as in the first and second series. First the different i-layerswere deposited on glass with a longer deposition time of 15 min., resulting inlayers with at least a thickness of 90nm. This was mainly done to determine thedeposition rates for different rf-powers, which made it possible to keep the i-layerthickness in the cells more or less constant for different rf-powers. However, inthis way it was also possible to determine some properties, as light and darkconductivity, which are difficult to determine accurately for layers that are only


a few nanometers thick due to the low conductivity of i-layers. It has to besaid that at thicknesses between 90 nm and 500 nm measurements can stillbe unreliable due to surface effects. The results are shown Fig.6.8. With an

Fig. 6.8: Properties of 200 nm thick i-layers deposited with different rf-powers, thephotoresponse σlight/σdark is shown on a logaritmic scale.

decreasing plasma power we see that the quality of the intrinsic layer improves,in the sense that the activation energy approaches half the value of the band gapwhich means that there is a good doping efficiency and very little unintentionaldoping (by for instance oxygen) in the layer. The ratio of light conductivity anddark conductivity has an optimum at a rf-power of 30 W, indicating the lowestdefect density and least mid-gap energy states here, which should improve thepassivation properties of the i-layer. This optimum is also in good agreementwith findings of Maruyama et al., who find at a deposition temperature of150C the lowest defect density at deposition rates around 0.1 nm/sec.[41]. Onthe other hand, even the best layer, deposited at 30W rf-power only had aphotoresponse of 104, where device quality amorphous silicon should have aphotoresponse of at least 105[35]. The low photoresponse values are mainlycaused by the low light conductivity values, 7.55 · 10−7S/cm for the best layer(30W rf-power), where as the values for the dark conductivity were, exceptfor the layer deposited at 25W (1.7 · 10−9S/cm), reasonably good with valuebetween 10−11S/cm and 10−10S/cm.

It has to be mentioned that the layers were not annealed after deposition(cells in which the layers are applied are annealed for one hour at 160C afterfinishing). It was shown by Srinivasan et al. that annealing for 3-4 hours at150C could improve the light conductivity of a-Si:H layers deposited at lowtemperatures with a factor up to 103[42]. The result of the incorporation of thelayers in cells (all i-layers had a thickness of approximately 5nm) are shown inFig.6.9 and 6.10. The optimum lays at a rf-power of 30W, with an efficiency of9.67%. At first sight the efficiencies in Fig. 6.10 are in agreement both with the



Fig. 6.10: Cell properties with i-layer deposited with different rf-powers.

results of Fujiwara (see Fig.6.7) with a phase boundary between 20 W and 30 Wand with those of Maruyama et al. with the best quality layer resulting in thebest cell performance. However this is mainly due to the results for the FF andnot to those for the Voc. This means that the passivation did not improve witha better i-layer as one would expect. Actually it is even less good: the Voc ofthe best cell (0.446mV) is lower than that of one without any i-layer (0.457mV,


made in the second series): there is no observable passivation effect with anyoff the i-layers, which is most certainly due to a not high enough i-layer quality.

6.4 Conclusions and Recommendations

In the first experiment the performances of HIT-cells with different i-layer thick-nesses on textured wafers with a diffused BSF were investigated. The thicknessof the i-layer, which was deposited at a rf-power of 37mV/cm2, was found tohave an optimum around 10 nm. The passivation of this i-layer resulted in aincrease of Voc from 604 mV to 623 mV and an increase in FF from 0.703 to0.777 obtaining a very good conversion efficiency of 16.4%. The fact that theoptimum found was relatively high compared to the optimum found by othergroups (5 nm) can possibly be explained by the fact that it is difficult to growa uniform a-Si:H layer on as textured wafer and thus a greater thickness is re-quired to reach a uniform amorphous phase. Though another reason can bethat a discrepancy rises from the fact that the thickness of our layers could onlybe measured ”timewise”.

Because we eventually want to deposit our own BSF, a second experimentwas done to investigate the passivation properties of the i-layer with differentthicknesses on wafers without a diffused BSF. Unfortunately such wafers wereonly available without texturing which introduced the problem of changing twoparameters determining the total recombination in the cell: besides the differ-ence in BSF the i-layer is expected to passivate differently on a polished waferthan on a textured one. Because of the low increase in Voc and efficiency withthe introduction of any i-layer in the second experiment, it was indeed a conclu-sion that the passivation properties of one and the same i-layer are very differenton a textured wafer than on a polished one. A possible explanation of the badpassivation in this second series was that the layer grew epitaxially, somethingthat was found by Fujiwara et al. to be of very bad influence on the passivationproperties[40].

To investigate this, a third experiment was performed in which i-layers weredeposited using different rf-powers, while the possible epitaxial growth wasfound to occur only at low rf-powers. When the different i-layers were de-posited on glass, it was found that with decreasing rf-power the layer qualityimproved: less unintentional doping and less mid gap defect states (higher pho-toresponse value), although the latter results could be unreliable due to thesmall layer thicknesses. When incorporating the i-layers into cells it was foundthat, although the efficiency had an optimum in agreement with the data of Fu-jiwara et al., the Voc did not increase as much as one would expect. A possibleexplanation for this was a too low i-layer quality.

From these results it could be concluded, on the one hand, that it wouldbe wise to investigate the use of lower deposition rates (lower rf-powers), whichwas shown by Maruyama et al. to result in better layer quality and in betterpassivation and thus cell performance[41]. This should be tried at a lower tem-perature, hereby reducing the chance on growing an epitaxial layer drastically.On the other hand, all of the i-layers were found to be not of the quality thatis needed to passivate as well as we would like. This indicates that the i-layerquality should be checked more frequently than was done before, something thatcan only be done accurately by depositing it on glass with a sufficiently high


thickness.Remains the fact that, although the quality of the i-layers was found to be

low in the third experiment, the good results of the first experiment indicatethat the quality of the i-layer was good at that time. This means that thereproducibility is not as good as one would like. The reproducibility problemis in detail discussed in the next chapter and possible explanations are giventhere.

7. REPRODUCIBILITY

There were found to be reproducibility issues during a large part of the research.The cleaning procedure, which will be discussed first, is most probable to be thecause of this. However, in the experiment on i-layers deposited with differentplasma powers, the quality of the i-layers deposited in the PILOT was found tobe not as high as expected. Possible explanations will be mentioned in the secondpart. The main goal of this chapter is to investigate all steps in the process thatneed to be focussed on in the future for increasing the reproducibility: it is mainlyabout recommendations.

7.1 Cleaning procedure: HF-dip

Although it is very important that the layers and interfaces of a cell are notcontaminated in all stages of the production process, a particular crucial step isthe etching of a silicon wafer by hydro fluoric acid (HF): before the a-Si:H layers(and the back contact at the back of the cell) are deposited, the oxide has to beremoved from the wafer. This is done by etching the wafer in a HF solution (inultra pure water). Besides on the HF-concentration, the etch rate is dependenton the orientation and surface structure of the wafer: each type of wafer has itsown optimal recipe for etching[43]. The etching has a strong influence on thesurface structure: a too long etching time can make an atomic flat wafer rough:(111) facets appear. For a (100) wafer this means, that the number of danglingbonds increase. More places where unwanted atoms, as O, can connect to thewafer (and form efficient recombination centers) are created. For the texturedwafers in the first experiment of the preceding chapter a 2 minute etch in a 1%HF solution was found to be optimal. This recipe was used for the polishedwafers also, however, it was found by van Cleef that for standard (100) flatwafers the best recipe was a 1 minute 0.5% HF-dip[3]. It would be wise toinvestigate the etching conditions once again for the new polished wafers.

7.1.1 Ultra pure water

It was often suggested that a reproducibility problem could result from impu-rities in the ultra clean water. Not only is this water used for the HF-solutionsand to rinse the beakers used for etching, also for transport of the etched waferfrom the chemical room to the PILOT. This is to prevent the etched wafer fromquickly oxidizing again or getting contaminated on the way to the PILOT. Thequality of the ultra pure water is thus indeed an important factor for the perfor-mance of the produced cell. The quality is measured by means of the resistivityof the water (this should be 18.2MΩ ·cm), however in this way only the ion con-tent is measured and not the contamination with organic materials. Althougha UV light is included in the setup to dissociate organic material into charged

7. Reproducibility 40

components (eventually CO2), this might not be enough: it only works at lowconcentrations. For more certainty about the water quality it is recommendedto check it with an other method. The contamination could, for instance, bemeasured by high performance liquid chromatography (HPLC), which is a quickand simple method[44].

7.2 Reproducibility in the PILOT

Although the main suspect for bad reproducibility is the cleaning procedure,i-layers were found to be not as good as expected in the last experiment of thepreceding chapter. The reason for this must be looked for in the depositionconditions of the PILOT.

7.2.1 Position in the plasma chamber

The PILOT is designed for depositions on substrates which measure 30x40 cm2,but in this research only substrates of 10x10 cm2 were used. They were heldin a substrate holder with six slots. In general a substrate is placed in one ofthe slots in the middle (see Fig.7.1). An experiment was done to measure thehomogeneity of the deposition rate over these two slots. The layer was depositedwith a rf-power of 50 W and a high deposition time of more than 15 minutes wasused to deposit a thick enough layer. Afterwards the thickness was measuredwith the X-Y table setup, using a photodiode measuring the transmission oflight with a wavelength of 520 nm. The result is shown in Fig.7.1. The layer inthe slot closest to the point where the gas flows in has a homogeneous thickness(around 640nm). The layer in the second slot, on the other hand, is muchthinner and not homogeneous (550-490 nm). Given the facts that layer thickness

Fig. 7.1: Layer thickness over the two slots in the middle of the substrate holder ofthe PILOT. Deposition time was around 15 min. at a rf-power of 50W.

is an important parameter in HIT cell performance and that most thicknessesof layers included in cells are measured timewise (when deposited on a wafer itis impossible to measure the thickness afterwards), this should always be keptin mind.

7. Reproducibility 41

7.2.2 Stability of the plasma

Plasma stability can be a crucial factor for producing high quality layers: sub-strates can be damaged by exposure to non-uniform plasma conditions, such aselectric filed gradients [45]. Although it always takes some time for the plasmato stabilize after being ignited, the time that this takes can be greatly increased(by tens of seconds) by the fact that the matchbox is adjusting the impedanceof the system to that of rf-generator (see chapter 2). When depositing layersof only a few nanometers thick, deposition times are only a few seconds, whichmeans that the quality of entire layers can be heavily influenced by fluctuationin the plasma. To minimize this effect, the impedance of the matchbox should,before the start of the deposition, be set at a value that is close to the valuethat is needed in case of a stable plasma: it will probably take less time for theplasma (the matchbox) to reach stable conditions. Another possibility is to usefixed values that are set manually for the match box. In that case they shouldbe checked frequently.

7.2.3 Opening of the plasma chamber

A big advantage of the PILOT setup is that it is suitable for both hot-wireCVD (HWCVD) and PECVD. However, for switching from one to another,the vacuum chamber in which the passivating i-layers are deposited has to beopened. At this moment it is unknown (exactly) after how many depositionscontaminations, by for instance oxygen, have completely disappeared from theplasma chamber. It would be a good idea to, next time when the plasmachamber has been opened, immediately start doing depositions and measurethe quality of the layers. In this way it will be clear what deposition times areneeded to clean the chamber well enough to produce good quality layers again:at the moment this is only intuitively estimated.

7.3 Conclusions

Although at first site all steps in the process look optimized and standard (mostactually are), there are some parts that should be checked or optimized onceagain. It was found that an optimal HF-dip recipe is very crucial for solarcell performance; now that another type of wafer is used the recipe should beoptimized again. Second the quality of the ultra pure water that is used couldbe measured more thoroughly (at the moment organic material is expected notto be present, however, this can not be measured).

In the second part the reproducibility in the PILOT was considered. Itwas found that the deposition rates are very different in different slots of thesubstrate holder: something that should be paid attention to. Furthermore theeffect of plasma instability was discussed and possible methods to limit themto a minimum: the match box setting should be set as close as possible to thedesired values in advance. Finally the effect of opening the plasma chambershould be measured: opening is unavoidable, however, it would be useful toknow exactly how many depositions it takes to remove all the contaminationfrom the plasma chamber.

8. BACK CONTACTED CELLS; DEPOSITION THROUGH AMASK

The most important features of IBC cells that have to be taken into accountdesigning them, such as the dimensions of the fingers and the front passivation,are briefly addressed. To get an insight in the accuracy that can be achieveddepositing through a mask, before the masks were designed, some trial deposi-tions through existing masks were done. Finally, different possibilities for theconfiguration of the layers/fingers, having different advantages, are discussed.

8.1 IBC cells

Although HIT cells have demonstrated high Voc due to low surface recombina-tion and low emitter saturation currents, due to absorption losses in the a-Si:Hlayers, shading loss by metal grids and trade off between antireflection coatingsand series resistance by front TCO, a loss of Isc and efficiency is eminent[25].The IBC structure proposed by Lammert and Schwartz [24], where the emitterand contacts are on the back of the wafer, provides an independent control ofthe optical performance of the illuminating front side and the electrical perfor-mance (low series resistance) of the back side. Although IBC structures providethe possibility for high efficiencies, their implementation is difficult because ofseveral design constraints. First the ratio of the minority carrier diffusion lengthover the device thickness needs to be high, because most electron hole pairs arecreated in the front of the cell, while carriers are collected at the back. This canbe achieved by using very thin, high quality wafers[23]. For the same reason thefront surface recombination velocity has to be very low: the passivation needsto be very good. At third challenge is to find the optimal dimensions of theinterdigitated p and n strips. These last two issues will be addressed here.

8.1.1 Dimensions of the fingers

Considering the design of the p and n fingers, two things are important. Firstthe ratio of the p/n finger width. This aspect can be investigated with 2Dmodels: when the emitter (p finger in this research) width increases (relatively),Isc increases since the recombination of minority charge carriers decreases dueto a reduction in average travel length through the bulk. On the other handby increasing the emitter width the average travel length for majority carriersresults in a higher series resistance, which results in a decreasing FF[46]. Inthis case it is also important what the surface recombination velocities at theinterfaces of the BSF and emitter are. In our case they will probably be nearlyequal because a thin i-layer is included (see Fig.2.11). In practice most groupsuse a p/n finger width ratio between 2:1 and 10:3[15, 47, 46].

8. Back contacted cells; deposition through a mask 43

Fig. 8.1: Microscopic picture (of a detail) of a deposition through a mask. The inter-ference colors at the edges of the pattern indicate a change of thickness overa small distance.

The second issue is the width of the two fingers together (pitch). As longas this is much larger than the thickness of the used wafer (probably 300µm),it will be desirable to keep this width as small as possible, keeping the averagelateral carrier travel length through the wafer as short as possible. On the otherhand, with decreasing dimensions, in practice problems in the alignment willoccur which can easily result in an overlap of opposite fingers (and contacts)and in shunting.

8.1.2 Front passivation and anti-reflection

As most carriers are generated near the front surface of the cell and the collectingp-n junction is at the rear surface, they have to traverse the width of the cell tobe collected. If the carriers have a low average diffusion length most minoritycarriers can not be collected[15]. The average diffusion length of a cell is stronglydetermined (besides by the bulk lifetime of the wafer) by the front surfacepassivation. In models the efficiency of IBC cells even turns out to dependdirectly on the front surface recombination velocity[16].

One of the most effective ways of surface passivation is by a silicon nitride(SiNx) layer. When a deposited SiNx layer is shortly annealed, atomic hydrogenis released from it, which partly diffuses to the surface of the wafer, where itpassivates interface defects[48]. Apart from its good passivating properties, SiNx

layers can act as excellent ARCs, because of their tunable refractive index andlow extinction coefficient. Cells with a SiNx ARC (deposited by hot wire CVD)are shown to have a high blue response[48]. ”In house” device quality SiNx

layers are already available.

8.2 Deposition through a mask

To investigate deposition through a mask in practice before designing the actualmasks, some trial deposition through an existing mask were done. For thisexperiment a 0.2 mm thick mask was used with slits varying in width between


Fig. 8.2: Thickness of a layer deposited through a 0.5 mm wide slit (left) and resultsfor depositions through slits with various widths (right), widths of the stemand of the stem plus corona are indicated.

0.15 and 0.5 mm. The results are shown in Fig.8.2. There are two effects thatcan be noticed. First the deposition of a ”corona” at the masked area besides thearea where deposition was intended can be observed. The width of the coronaseems to be independent of the width of the slit (0.3mm). This effect can beexplained by the fact that, if there is a small space between the mask and thesubstrate, radicals diffuse underneath the mask (contrary to ions that move tothe substrate in a straight line due to the electric field in the sheath). Anotherpossibility is that particles are reflected multiple times between the mask andthe substrate[26]. Second there is a ”shadowing” effect; the layer close to theedges of the mask is thinner than in the middle of the slit, because particlesthat move in a more or less lateral direction are shielded from this area. Thecombination of these effects results in a deposition formed in the way depictedin Fig.8.2, instead of a sharp rectangular shape. Although patterns depositedin this way are believed to be accurate enough to manufacture an IBC cell, it isimportant that the used mask is as thin as possible (to keep the shadowing effectat a minimum) and rigid enough to enable a tight attachment to the substrate(reducing the corona size to a minimum).

8.2.1 Creating a practically feasible mask

When the interdigitated p and n strips are connected, two ”m-shaped” patternsare created(see Fig.8.3). It would be the most logical to deposit them in onestep. However, in a mask needed for this, the bars separating the fingers areattached only at one point to the rest of the mask. This does not allow for tightattachment to the substrate, especially when a thin, easily bendable, metal isused. Besides this it is known that after several depositions thin masks canbecome permanently bent due to thermal stress. To avoid this problem, it wasdecided to deposit the separate fingers without an interconnection, resulting ina much more rigid mask. They can be connected afterwards, by evaporatingtwo silver bars laterally over the ends of the fingers. A minor drawback in this isthat a extra evaporation is necessary, creating a extra opportunity for mistakes.It was decided to include three different pitch sizes in one mask to enable agood comparison (the deposition conditions will be the same for all three sizes).The sizes were chosen to be 1.25, 1.67 and 2.5mm. In this way always one cm2

could be filled. The ratio of p and n finger width is about 2:1. A example of


Fig. 8.3: When strips are connected at one end a m-shape is created.

Fig. 8.4: Example of a mask design with only fingers.

the designed masks is shown in Fig.8.4.The last thing that needs consideration is if separate masks for depositions

and silver evaporation are needed: is this necessary to prevent silver contami-nation in the p-n junction?.

8.2.2 Different cell configurations

The most conventional way of designing IBC cell is shown in Fig. 2.11. Althoughit is assumed that isolation between the interdigitated strips is necessary, it isnot confirmed that this is always the case[16]. Isolation can be achieved byvery accurate alignment or by additional process steps such as, for instance,photolithography or laser ablation. Besides that, good passivation of the areabetween the strips is difficult. It would be a good idea first to try manufacturingcells without any extra step for better isolation. It was also proposed to inves-


Fig. 8.5: Tunnel junction proposal. For clarity, the cell thickness is greatly exagger-ated.

tigate a tunnel junction structure. This structure is shown in Fig.8.4. After then+-layer is deposited the p-layer can be deposited over the entire cell, withoutany mask. In this way a difficult alignment procedure can be avoided. The car-riers collected at the n-n+ junction can tunnel through the p-layer (if it is thinenough) in the same way as is usually done in tunnel junctions in tandem andtriple cells[2]. In this design, the most crucial thing would be possible shuntingbetween opposite contacts directly through the p-layer.

8.3 Conclusions

Besides the used wafer, the two most important features of IBC cells are goodfront passivation and the design of the interdigitated p- and n-strips at the rearend. SiNx layers, that can be used for excellent front passivation are alreadyavailable. Considering the rear design, the ratio of p- and n-strips is importantand in most other groups a ratio between 1:2 and 3:10 is used. The pitch widthshould be as small as possible within the limits of practical alignment.

In trial depositions it was found that a corona is deposited underneath themask and that there is a shadow effect at the masks edges. To minimize theseeffects the masks need to be both thin and rigid. To combine these two features,masks were designed to deposit fingers and the connecting bars separately.

Because it is still uncertain how crucial the isolation between fingers is, fora start, cells should be made involving as few extra processing steps as possible.Another possibility is the use of a tunnel junction, thereby eliminating a difficultprocess step.

9. CONCLUSIONS AND RECOMMENDATIONS

• A p-layer recipe in the PASTA was developed, resulting in a layer withan activation energy of 0.36 eV, which is good enough to use it in cellproduction.

• On textured wafers an optimal i-layer thickness of 10 nm was found, re-sulting in a cell efficiency of 16.4%.

• On polished wafers, passivation was very different.

• The optimal i-layer thickness on polished wafers was 2.5 nm. Including itin a cell resulted in a efficiency increase from 9.09% to 9.64%

• Investigating the effect of varying rf-powers for the deposition of the i-layer, the best result was found at an rf-power of 22.3 mW/cm.

• In trial depositions it was observed that a corona is deposited underneaththe mask and that there is a shadow effect at the masks edges. to minimizethese effects the masks need to be thin and rigid.

• Mask where designed to deposit fingers and the connecting bars separately.

• The HF-dip recipe needs further investigation: each type of wafer has itsown optimal recipe.

• Matchbox settings are very crucial: because of the small layer thicknesses,deposition is finished before stable plasma conditions are reached.

BIBLIOGRAPHY

[1] World energy outlook. International Energy Agency, 2007.

[2] J. Schuttauf. Masterthesis, 2007.

[3] M. van Cleef. Amorphous crystalline silicon heterostructures & solarcells.PhD thesis, 1998.

[4] W. Hoffmann. Pv solar electricity industry: market growth and perspective.Solar Energy Materials & Solar Cells, 90, 2006.

[5] A. Goetzberger and C. Hebling. Photovoltaic materials, past, present,future. Solar Energy Materials & Solar Cells, 62, 2000.

[6] W. van Sark, G. Bransden, M. Fleuster, and M. Hekkert. Analysis of thesilicon market: Will thin films profit? Energy Policy, 35, 2007.

[7] D. Borchert, A. Gronbach, M. Rinio, and E. Zippel. Process steps for theproduction of large area (n)a-si:h/(p)c-si heterojunction solar cells. Pre-sented at the 20th European Photovoltaic Solar Energy Conference andExhibition, 6.-10. June 2005, Barcelona.

[8] T. Warabisako, T. Uematsu, S. Muramatsu, K. Tsutsui, H. Ohtsuka, Y. Na-gata, and M. Sakamoto. Optimization of thermal processing and devicedesign for high-efficiency c-Si solar cells. Solar Energy Materials and SolarCells, 48(1-4):137–143, 1997.

[9] D. Munoz, C. Voz, M. Fonrodona, M. Garin, A. Orpella, M. Vetter, J. Puig-dollers, R. Alcubilla, F. Villar, J. Bertomeu, et al. Characterization of bi-facial heterojunction silicon solar cells obtained by hot-wire CVD. Journalof Non-Crystalline Solids, 352(9-20F):1953–1957, 2006.

[10] K. OKUDA, H. OKAMOTO, and Y. HAMAKAWA. AmorphousSi/Polycrystalline Si Stacked Solar Cell Having More Than 12% ConversionEfficiency. Japanese Journal of Applied Physics, 22(9):L605–L607, 1983.

[11] M. Schmidt, L. Korte, A. Laades, R. Stangl, C. Schubert, H. Angermann,E. Conrad, and K. Maydell. Physical aspects of a-Si: H/c-Si hetero-junctionsolar cells. Thin Solid Films, 515(19):7475–7480, 2007.

[12] T. Sawada, N. Terada, S. Tsuge, T. Baba, T. Takahama, K. Wakisaka,S. Tsuda, and S. Nakano. High-efficiency a-Si/c-Si heterojunction so-lar cell. In Photovoltaic Energy Conversion, 1994., Conference Record ofthe Twenty Fourth; IEEE Photovoltaic Specialists Conference-1994, 1994IEEE First World Conference on, volume 2, 1994.

Bibliography 49

[13] UK Das, S. Bowden, M. Burrows, M. Lu, and RW Birkmire. Effect ofSurface Passivation on Si Heterojunction and Interdigitated Back ContactSolar Cells.

[14] M. Tucci, L. Serenelli, E. Salza, S. De Iuliis, LJ Geerligs, G. de Ce-sare, D. Caputo, and M. Ceccarelli. NOVEL SCHEME OF AMOR-PHOUS/CRYSTALLINE SILICON HETEROJUNCTION SOLAR CELL.gas, 260(250):750.

[15] KR McIntosh, MJ Cudzinovic, DD Smith, WP Mulligan, and RM Swanson.The choice of silicon wafer for the production of low-cost rear-contact solarcells. In Photovoltaic Energy Conversion, 2003. Proceedings of 3rd WorldConference on, volume 1, 2003.

[16] M. Lu, S. Bowden, U. Das, and R. Birkmire. A-SI/C-SI HETEROJUNC-TION FOR INTERDIGITATED BACK CONTACT SOLAR CELL.

[17] D.J. Griffith. Introduction to Quantum Mechanics 2nd edition. Reed Col-lege, 2005.

[18] R.S. Muller and T.I. Kamins. Device Electronics for Integrated Circuits 3rdedition. John Wiley & Sons, 2003.

[19] C. Kittel and H. Kroemer. Thermal Physics 2nd edition. W.H. Freemanand Company, 2003.

[20] M.A. Green. Solar Cells, Operating Principles, Technology and SystemApplications. University of New South Wales, 1992.

[21] R.E.I. Schropp, W.C. Sinke, and W.F. van der Weg. Zonnecelfysica. Uni-versiteit Utrecht, 1998.

[22] M. Farrokh Baroughi, R. Jeyakumar, Y. Vygranenko, F. Khalvati,and S. Sivoththaman. Fabrication and characterization of amorphoussi/crystalline si heterojunction devices for photovoltaic applications. J.Vac. Sci. Technol. A., 22(3), 2004.

[23] E. Van Kerschaver and G. Beaucarne. Back-contact Solar Cells: A Review.Progress in Photovoltaics: Research and Applications, 14(2):107–123, 2005.

[24] MD Lammert and RJ Schwartz. The interdigitated back contact solar cell:A silicon solar cell for use in concentrated sunlight. Electron Devices, IEEETransactions on, 24(4):337–342, 1977.

[25] M. Lu, S. Bowden, U. Das, and R. Birkmire. Interdigitated back contactsilicon heterojunction solar cell and the effect of front surface passivation.Applied Physics Letters, 91:063507, 2007.

[26] E.A.G. Hamers. Plasma deposition of hydrogenated amorphous silicon,plasma processes, ion fluxes and material properties. PhD thesis, 1998.

[27] H.C. Ritchey. Tuner topics, 1995. http://www.advanced-energy.com/upload/File/White Papers/ENG-WHITE16-270-02.pdf.

Bibliography 50

[28] R.A. Scholl. Forward and reflected powers. what do they mean?,1998. http://www.advanced-energy.com/upload/File/White Papers/SL-WHITE7-270-01.pdf.

[29] R.H.J. Franken. Transparent conducting oxide contacts and textured metalback reflectors for thin film silicon solar cells. PhD thesis, 2006.

[30] A. Gordijn. Microcrystalline silicon for thin-film solar cells. PhD thesis,2005.

[31] H. Li. Single and multijunction silicon based thin film solar cells on flexiblesubstrates with absorber layers made by hot-wire CVD. PhD thesis, 2007.

[32] A. Lloret, ZY Wu, ML Theye, I. Zawawi, JM Siefert, and B. Equer.Hydrogenated amorphous silicon p-doping with diborane, trimethylboronand trimethylgallium. Applied Physics A: Materials Science & Processing,55(6):573–581, 1992.

[33] K. Maydell, E. Conrad, and M. Schmidt. Efficient silicon heterojunctionsolar cells based on p-and n-type substrates processed at temperatures¡ 220C. Prog. Photovolt. Res. Appl, 14:289, 2006.

[34] J.W.A. Schuttauf, Y. Komatsu, L.J. Geerligs, Y. Mai, A. Bink, D.A.Spee,and R.E.I. Schropp. Emitter Optimization on a-Si:H/c-Si heterojunctionsolar cells for Isotextured Wafers . Presented at the 23rd European Photo-voltaic Solar Energy Conference, 1.-5. September 2008, Valencia.

[35] R.E.I. Schropp and M. Zeman. Amorphous and Microcrystalline SiliconSolar Cells: Modeling, Materials, and Device Technology. Kluwer AcademicPublishers, 1998.

[36] M. Tanaka, S. Okamoto, S. Tsuge, and S. Kiyama. Development of hitsolar cells with more than 21% conversion efficiency and commercializationof highest performance hit modules. In Photovoltaic Energy Conversion,2003. Proceedings of 3rd World Conference on, volume 1, 2003.

[37] H. Fujiwara and M. Kondo. Effects of a-si:h layer thickness on the perfor-mance of a-si:h/c-si heterojunction solar cells. Journal of Applied Physics,101, 2007.

[38] J. Damon-Lacoste, P. Roca i Cabarrocas, P. Chatterjee, Y. Veschetti,AS Gudovskikh, JP Kleider, and PJ Ribeyron. About the efficiency limitsof heterojunction solar cells. Journal of Non-Crystalline Solids, 352(9-20F):1928–1932, 2006.

[39] MWM van Cleef, JK Rath, FA Rubinelli, CHM van der Werf, REI Schropp,and WF van der Weg. Performance of heterojunction p microcrystallinesilicon n crystalline silicon solar cells. Journal of Applied Physics, 82:6089,1997.

[40] H. Fujiwara and M. Kondo. Impact of epitaxial growth at the heterointer-face of a-si/c-si solar cells. Applied Physics Letters, 90, 2007.

Bibliography 51

[41] E. Maruyama, A. Terakawa, M. Taguchi, Y. Yoshimine, D. Ide, T. Baba,M. Shima, H. Sakata, and M. Tanaka. Sanyo’s Challenges to the Develop-ment of High-efficiency HIT Solar Cells and the Expansion of HIT Business.In Photovoltaic Energy Conversion, Conference Record of the 2006 IEEE4th World Conference on, volume 2, 2006.

[42] E. Srinivasan, D.A. Lloyd, and G.N. Parsons. Dominant monohydridebonding in hydrogenated amorphous silicon thin films formed by plasmaenhanced chemical vapor deposition at room temperature. Journal of Vac-uum Science & Technology A: Vacuum, Surfaces, and Films, 15:77, 1997.

[43] H. Goldbach. Cleaning wafers(c-Si(FZ,CZ),mc-Si). 9-2-2004.

[44] C.G. Horvath, BA Preiss, and S.R. Lipsky. Fast liquid chromatography.Investigation of operating parameters and the separation of nucleotides onpellicular ion exchangers. Analytical Chemistry, 39(12):1422–1428, 1967.

[45] J.H. Soo, M. Spuller, M.S. Cox, M.J. Seamons, A. Al-Bayati, B.H. Kim,and H. M’Saad. Method for plasma processing, July 7 2006. US PatentApp. 11/483,951.

[46] DS Kim, V. Meemongkolkiat, A. Ebong, B. Rounsaville, V. Upadhyaya,A. Das, and A. Rohatgi. 2D-MODELING AND DEVELOPMENT OFINTERDIGITATED BACK CONTACT SOLAR CELLS ON LOW-COSTSUBSTRATES. In Photovoltaic Energy Conversion, Conference Record ofthe 2006 IEEE 4th World Conference on, volume 2, 2006.

[47] D. Diouf, JP Kleider, T. Desrues, and P.J. Ribeyron. Study of interdig-itated back contact silicon heterojunctions solar cells by two-dimensionalnumerical simulations. Materials Science & Engineering B, 2008.

[48] V. Verlaan, CH van der Werf, ZS Houweling, IG Romijn, AW Weeber,HF Dekkers, HD Goldbach, and RE Schropp. Multi-crystalline Si solarcells with very fast deposited (180 nm/min) passivating hot-wire CVD sili-con nitride as antireflection coating. PROGRESS IN PHOTOVOLTAICS,15(7):563, 2007.

LIST OF FIGURES

2.1 (a) Band representation of a semiconductor. (b) Energy densityof allowed states for electrons. (c) Probability of occupation ofthese states, according to the Fermi-Dirac distribution function.(d) Resulting energy distribution of electrons and holes. [20] . . . 9

2.2 Example of p-type silicon, doped with boron, a group III atom(left) and n-type silicon, doped with phosphorus, a group V atom(right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.3 Fermi levels in p-type doped silicon (left) and n-type doped silicon(right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.4 A p-n junction in thermal equilibrium: (a) Fixed charge distribu-tion. (b) Distribution of electric field. (c) Potential distribution.(d) Energy band diagram.[21] . . . . . . . . . . . . . . . . . . . . 12

2.5 I-V characteristics of a diode in the dark and when illuminated.[20]. 132.6 Schematic picture of a practical solar cell. For clarity the cell

thickness is exaggerated. . . . . . . . . . . . . . . . . . . . . . . . 142.7 Solar spectrum, the wavelength of 1100 nm, corresponding to the

band gap energy of crystalline silicon, is indicated. . . . . . . . . 152.8 Equivalent circuit of a solar cell, showing the light induced cur-

rent, dark current (diode), series resistance and shunt resistance[21]. 162.9 2-dimensional representation of crystalline (left)and amorphous

(right) silicon, where the open circles represent hydrogen atoms[21]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.10 Energy band alignment of two semiconductors with different bandgaps, before (a) and after (b) junction formation.[3] . . . . . . . 17

2.11 Schematic picture of a heterojunction IBC solar cell. For claritythe cell thickness is exaggerated. . . . . . . . . . . . . . . . . . . 18

3.1 Schematic lay out of a plasma (a) and the potential diagramsbetween equal and unequal sized electrodes (b), in the latter casea negative (the grounded electrode is usually bigger) DC self biasVdc is present on the powered electrode. . . . . . . . . . . . . . . 21

5.1 P-layer properties with varying TMB-flow during deposition. . . 27

6.1 IV-curves of the i-layer thickness series on textured wafers. . . . 306.2 Cell properties with different i-layer thicknesses for the first series. 306.3 Spectral response results for cells of the first series, for clarity

only cells with i-layer thicknesses of 5, 10 and 15 nm are shown. . 316.4 Cell properties with different i-layer thicknesses. . . . . . . . . . 326.5 Cell properties with different i-layer thicknesses. . . . . . . . . . 33

List of Figures 53

6.6 A comparison of ECE of the best cell of the first series and thebest cell of the second series. Although the ECE of the first ishigher at all wavelengths, as a result of the texturing, the largestdifference can obviously be seen in the long wavelengths. Thisshows the importance of the BSF to prevent recombination atthe back of the cell. . . . . . . . . . . . . . . . . . . . . . . . . . . 33

6.7 Performance of the a-Si:H/c-Si solar cell, plotted as a function ofthe rf power density, found by Fujiwara et al.. The dotted linesin the figure show the epi-Si/a-Si:H phase boundaries[40]. . . . . 34

6.8 Properties of 200 nm thick i-layers deposited with different rf-powers, the photoresponse σlight/σdark is shown on a logaritmicscale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

6.9 Cell properties with different i-layer thicknesses. . . . . . . . . . 366.10 Cell properties with i-layer deposited with different rf-powers. . . 36

7.1 Layer thickness over the two slots in the middle of the substrateholder of the PILOT. Deposition time was around 15 min. at arf-power of 50W. . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

8.1 Microscopic picture (of a detail) of a deposition through a mask.The interference colors at the edges of the pattern indicate achange of thickness over a small distance. . . . . . . . . . . . . . 43

8.2 Thickness of a layer deposited through a 0.5 mm wide slit (left)and results for depositions through slits with various widths (right),widths of the stem and of the stem plus corona are indicated. . . 44

8.3 When strips are connected at one end a m-shape is created. . . . 458.4 Example of a mask design with only fingers. . . . . . . . . . . . . 458.5 Tunnel junction proposal. For clarity, the cell thickness is greatly

exaggerated. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

Thesis

Documents

layer hit cells

contacted cells deposition

practical solar cells

entire solar cells

contacted ibc cells

si heterojunction solar

fabrication of ibc cells

deposition of p