Pavel_Babal_PhD_thesis.

Pavel_Babal_PhD_thesis.

P. Babál

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft; op gezag van de Rector Magnificus prof. ir. K. C. A. M. Luyben; voorzitter van het College voor Promoties in het openbaar te verdedigen op woensdag 29 oktober 2014 om 10:00 uur door

Pavel BABÁL

In

ž

inier, Slovensk

á

Technick

á

Univerzita geboren te Bratislava, Slowakije

Dit proefschrift is goedgekeurd door de promotor:

Prof. dr. M. Zeman

Copromotor:

Dr. ir. A. H. M. Smets

Samenstelling promotiecommissie:

Rector Magnificus,

Prof. dr. M. Zeman,

Dr. ir. A. H. M. Smets,

Prof. dr. J.A. La Poutre,

Prof. dr. L.D.A. Siebbeles,

Prof. dr. E. Vlieg,

Dr. W. Soppe,

Dr. A. Gordijn,

Prof. dr. E. Charbon,

voorzitter

TechnischeUniversiteit Delft, promotor

Technische Universiteit Delft, copromotor

Technische Universiteit Delft

Technische Universiteit Delft

Radboud University Nijmegen

Energie Centrum Nederland

Forschungszentrum J

ü

lich GmbH

Technische Universiteit Delft, reservelid

Copyright © 2014 P. Bab

á l

All rights reserved.

No part of this material may be reproduced, stored in a retrieval system, nor transmitted in any form or by any means without the prior written permission of the copyright owner.

ISBN 978-94-6203-682-6

Cover photo credits: Marek Bab

á l

Printed by: CPI - Koninklijke Wohrmann Print Service

To Katerina

1.1

1.2 THIN-FILM SOLAR CELLS AND THEIR ADVANTAGES

1.3 REFLECTING LAYERS IN THIN-FILM SILICON SOLAR CELLS

1.3.1

1.3.2

Desired Intermediate Reflector Properties & Simulations

Zinc oxide as a reflecting layer

1.3.3

1.3.4

Silicon oxide as a reflecting layer

Advanced concepts of reflecting layers

1.4 THE OBJECTIVE OF THIS THESIS

2 PROCESSING AND CHARACTERIZATION OF THIN-FILM SILICON

2.1

2.1.1

2.1.2

2.1.3

Plasma Enhanced Chemical Vapor Deposition (PECVD)

Magnetron Sputtering

Thermal and electron beam evaporation

2.1.4 Reactive Ion Etching (RIE)

2.2 SOLAR CELL CHARACTERIZATION

2.2.1

2.2.2

External Parameters

External Quantum Efficiency (EQE)

2.3 SINGLE LAYER CHARACTERIZATION

2.3.1 Raman spectroscopy

2.3.2 Fourier transform infrared spectroscopy

3 SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

INTRODUCTION

3.2 EXPERIMENTAL DETAILS

DEVELOPMENT

3.3.1

3.4 NANOSTRUCTURE ANALYSIS

3.4.1

3.4.2

3.4.3

3.4.4

Single layer properties

TEM analysis

Raman spectroscopy analysis

Fourier transform infrared spectroscopy analysis

X-ray photoelectron spectroscopy comparison

3.5 DISCUSSION AND CONCLUSION

VII

27

27

39

45

48

62

67

30

31

31

39

15

20

21

23

23

24

15

15

17

18

19

20

9

10

12

4

7

7

8

1

2

4 SOLAR CELL APPLICATION

4.1

RESULTS

4.2.1

4.2.2

N-doped nc-SiOx:H as a back reflector

P-doped nc-SiOx:H as a p-layer

4.3 CONCLUSION

5 INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

5.1.1

5.1.2

Choice of materials and ASA simulations

Bragg reflector on glass

5.1.3 Cell integration

5.2 CONTROL OF INTERFACE TEXTURING

5.2.1

5.2.2

Influence of front TCO texture

Etched zinc oxide as textured intermediate reflector

5.2.3 Mechanical polishing of the intermediate reflector

5.3 CONCLUSIONS

6.1 NANOSTRUCTURE

6.1.1 Crystalline phase

6.1.2 Amorphous phase

6.2 CELL INTEGRATION

6.3 INTERMEDIATE REFLECTOR CONCEPTS

6.3.1 Distributed Bragg Reflectors

6.3.2 Intermediate reflector texturing

6.4 RECOMMENDATIONS

BIBLIOGRAPHY

SUMMARY

SAMENVATTING

113

113

114

115

116

116

116

117

117

119

73

74

74

74

79

80

83

93

97

97

101

84

84

88

106

110

ACKNOWLEDGEMENTS

127

129

133

135

137

VIII

The enormous dependency on fossil fuels as a source of energy

[1] is the cause of many global problems. Their uneven distribution around the world has been a source of political and economic problems, limiting development and causing (armed) conflicts [2]. In addition, the abundance of fossil fuels is limited and even with new deposits being discovered they will eventually run out. Therefore, it is highly important to replace these sources of energy with other (practically) unlimited, widely available, inexpensive, and safe alternatives.

One of these alternatives is solar energy converted directly to electricity via photovoltaic (PV) cells (more commonly known as solar cells). An amount of 1.13 x 10

18

kWh of solar radiation reaches the earth’ surface every year, where 19.6 x 10

12

kWh per year is the global electric power consumption (in 2010) [3]. Therefore, even with increasing global electricity consumption, if only a fraction of this immense amount of solar energy could be converted, the electrical energy problems that humanity faces now and will face in the future could be solved in a sustainable way. world, have no access to grid connected electric energy [4],[5]. Battery systems charged by solar cells are the most promising and cost effective solution for providing off-grid electric energy. They can provide these

1. INTRODUCTION

Figure 1.1: Different technologies for the production of electricity according to their global installed capacity in 2011 [7].

remote locations with electricity at a reasonable price without having to wait for the construction of expensive grid infrastructure. This will provide the local populations new opportunities for development and education [6].

Solar cells are seen as one of the most promising alternative, renewable, and eco-friendly sources of energy and are by far the most newly installed type of electricity generation in recent years (Fig. 1.1).

Figure 1.1 shows that two times more new PV capacity was installed in 2011 than new wind or gas capacity. This high PV capacity of new installations was partially due to subsidized feed-in tariffs. However, prices of PV systems are dropping and coming near to or - in most locations – are equaling grid prices (grid parity) [8]. For PV technology to become successful on a large scale, the materials used in solar panels must satisfy strict requirements. These materials need to be abundant

(Fig. 1.2), cheap, and non-toxic. Their processing and recycling should be cheap as well.

1.1 Solar cells types

To date, there are various PV technologies on the market and many are being researched and developed. Each type of PV technology has its specific advantages and disadvantages, such as low material costs, low production costs and high solar to electrical power conversion efficiency.

To date, the most widespread PV technology is based on bulk waferbased crystalline silicon (c-Si) cells. They dominate the market with an

84% share in 2012 [7] and are predicted to dominate the market in the

2

1.1 SOLAR CELLS TYPES

Figure 1.2: The abundance of elements in the upper earth’s crust plotted against their atomic number [9].

decade to come (Fig. 1.3). Commercial crystalline silicon modules have efficiencies between 16-20%. Taking into account their production costs, the levelized cost of the generated energy (in this example in the United

States; non-dispatchable) is 144.3 $/MWh while conventional sources of electric energy such as coal or natural gas are substantially cheaper, with

100,1 and 67,1 $/MWh respectively [10]. In addition, the conversion efficiency of such laboratory scale c-Si single-junction solar cells is approaching its theoretical limit. Often the Shockley Queisser limit is used with a value of ~30%, however, it only takes into account radiative recombination which is not the dominant charge carrier recombination mechanism for a indirect bandgap material as silicon [11]. By taking into account Auger recombination as well, the theoretical limit drops to 29,4%

[12]. Current state of the art crystalline solar cells are close to this limit, with 25,6% and 25% for HIT and PERL cells respectively [13], [14].

Thin-film (TF) PV technology, compared to bulk silicon solar cells, use much less material with a thickness of on average just a few micrometers. The consumption of material in TF solar cells is therefore much lower than in 200 micrometer-thick c-Si wafers and TF cells in general use cheaper processing methods. However, TFs still face many challenges, where the most important one is raising the conversion efficiency. The module efficiency of TF technology on the current market is lower than c-Si modules, being between 10-13%. Yet still their annual production capacity is predicted to grow in the coming years (Fig. 1.3).

Some TF PV technologies produce more energy from each module per invested amount (described as the price per Watt peak), however, other costs need to be taken into account as well.

3

1. INTRODUCTION

Figure 1.3: Production capacity of c-Si and TF cells in the coming years [7].

Next to module costs, the costs of a PV system contain non-modular costs, often called the balance of system costs (BOS). These costs include, for example, the structural frame, inverter, installation, etc., and currently account for ~68% of the entire PV system cost [15]. The cost of a PV system per watt decreases with increasing installed capacity as shown in the learning curve in Figure 1.4. However, the non-modular costs decrease at a slower rate than the module costs per installed power.

From about 2 GW of installed production capacity the limiting factors for price reduction become the non-modular costs. Therefore, due to the lower efficiency of TF PV modules, they need more area to produce this energy compared with c-Si modules and that raises their non-modular costs. The fact that the costs of the whole PV system usually scale with the area of the system and not with the installed electrical power further underlines the necessity for higher conversion efficiencies of TF solar cells.

1.2 Thin-film solar cells and their advantages

There are a variety of TF PV technologies based on different semiconductors, including germanium, III-V semiconductors, cadmium telluride (CdTe), copper indium gallium selenide (CIGS) and organic materials. III-V semiconductor solar cells are expensive cells with high conversion efficiencies from 25-44% for lab-scale cells [17]. CdTe and

CIGS laboratory scale cells have conversion efficiencies between 15-

18%. CdTe and CIGS solar modules also have reasonably high efficiencies

(11-13%) but their cost heavily depends on the cost of the elements that

4

1.2 THIN-FILM SOLAR CELLS AND THEIR ADVANTAGES

Figure 1.4: The learning curve of PV modules and systems comparing their price per watt with cumulative installations [16].

constitute them [18], [19]. Especially Indium and Telluride are scarce for production up-scaling. Neither of these materials meets the desired prerequisites needed for successful large scale PV implementation, being cheap, abundant, and non-toxic materials. An additional shortcoming of

CdTe and CIGS is the significant loss in efficiency when laboratorysized cells are up-scaled to modules (~35%).

One way of increasing TF cell efficiency is by stacking cells on top of each other, making a multijunction cell (Fig. 1.5). In case of two stacked cells, they are called tandem cells (Fig. 1.5a, b). Tandem cells take advantage of better utilization of bandgap energy and spectrum using two different bandgap materials in the same solar cell. Gallium arsenide based multijunction devices are the best demonstrators of this concept, being the most efficient solar cells to date (efficiencies of over 40% under solar concentration and laboratory conditions [20]).

TF Silicon PV technology uses the concept of multijunctions as well. The most common combination of materials constituting a tandem cell are a-Si:H and nc-Si:H. It combines the advantages of the lowcost TF deposition technology with higher spectrum utilization than in single junction TF Si cells. Due to spectral mismatch between the solar spectrum and the absorber material bandgap, the bandgap of one PV cell is not enough to capture all incoming photons. Thermalization and nonabsorption constitute the greatest losses in solar cells. Tandem cells are an easy way to absorb a greater portion of the solar spectrum with the

5

1. INTRODUCTION a-Si:H p-i-n

a

a-Si:H / nc-Si:H p-i-n / p-i-n

b

nc-Si:H / nc-Si:H / a-Si:H n-i-p / n-i-p / n-i-p

c

a-Si:Ge:H / a-Si:Ge:H / a-Si:H n-i-p / n-i-p / n-i-p

d

nc-Si:H / a-Si:Ge:H

/ a-Si:H n-i-p / n-i-p / n-i-p

e

Figure 1.5: State-of-the-art laboratory scale multijunction TF silicon solar cells and their stable efficiencies.

added advantage of a high output voltage. Currently, laboratory scale single junction and a-Si:H/nc-Si:H tandem cells have a stable efficiency of 10,7% and 12,3% respectively (Fig. 1.5a, b). To-date the most state of the art TF silicon devices are triple junction cells achieving an initial efficiency of 16,3% and stable efficiency of 13,4 % at laboratory scale

(Fig. 1.5e, c) [21], [22].

From the TF materials, TF Si is the only material that meets all the prerequisites of a PV material. It is cheap, abundant and easily processable

[23]. However, the efficiency of TF Si modules is still low compared with other PV technologies, being around 10%. The usual thickness of TF Si double-junction solar cells studied in this thesis is around 3

μ m. This thickness is a compromise of deposition rate and as large as possible light absorbance of the layers. In addition, larger thicknesses reduce the charge carrier collection efficiency because of the relatively poor charge transport in these materials. Therefore, to enhance light absorption, light trapping methods are introduced into the cell. These methods include, for example, the application of anti-reflective coatings, textured interfaces, and back reflectors. Besides their lower costs, other advantages of TF

Si cells include the possibility to be flexible, lighter weight, and ease of integration into other products. A relatively small difference between laboratory-scale cell and module efficiency exists, showing the mature status of the technology.

6

1.3 REFLECTING LAYERS IN THIN-FILM SILICON SOLAR CELLS

Hydrogenated amorphous Si (a-Si:H) solar cells are drift based devices, in comparison to c-Si which are diffusion based devices. To successfully extract charge carriers from the cell, an intrinsic silicon layer has to be sandwiched between a doped p- and n-layer which create an electric field in the intrinsic layer. This intrinsic absorber layer is defect rich and its electric properties are, to a certain extent, metastable. The conversion efficiency of an a-Si:H solar cell decreases after exposure to light. This effect is called the Staebler-Wronski effect (SWE) and was first described in 1977 [24]. The efficiency reaches a steady state after about 1000 hours exposure to illumination. The typical relative degradation of the cell efficiency upon light soaking is between 10% to

25% for a-Si:H cells [25]. Nc-Si:H cells practically do not suffer from light induced degradation. As the generation of light induced metastable defects is a bulk process, the SWE is less profound for thinner layers.

1.3 Reflecting layers in thin-film silicon solar cells

Due to the usage of thinner and more stable TF a-Si:H absorber layers, it is necessary to include reflecting layers at the back of the junctions to enhance the optical path of light through the absorber layers. A method of light trapping in multijunction cells, other than using a back reflector, is incorporating between each junction an intermediate reflector (IR). The basic function of any IR in a multijunction cell is the back reflection of incident photons with energies higher than the bandgap of the absorber layer above it with the aim of increasing spectral utilization. The IR is an additional tool that helps in matching the currents of the individual junctions. Note that the tandem cell is a series-connected device in which

J

SC

is determined by the lowest current coming from a junction. To date, a-Si:H/nc-Si:H tandem cells have been studied for almost two decades.

Optimization of these cells brought intense research in the field of IRs.

During that time, many interesting IR materials, concepts, simulations, and results were achieved which will be shortly summarized.

1.3.1 Desired Intermediate Reflector Properties & Simulations

The development of device grade nc-Si:H [26], [27] led to a new type of TF Si double junction, the a-Si:H/nc-Si:H tandem. A logical next step was the integration of IRs into tandems [28]. A well-functioning IR needs to meet certain requirements. These requirements include a high reflectance of low-wavelength photons (<600 nm, in case of a-Si:H/ nc-Si:H tandems) and a high transparency for high-wavelength photons

(>600 nm). This can be achieved by properly tuning the thickness of low

7

1. INTRODUCTION refractive index film (for high index contrast with the silicon absorber layers) and high band gap materials. Optical analysis of IRs with a small refractive index (n < 2.0) revealed possibilities for significant improvements of the top cell short-circuit current density ( J

SC

) (>25%) or thickness reductions of the top cell absorber layer (>50%) [29]. The material should also have minimal parasitic absorption.

If the IR material is p- or n-doped it can serve as the p- or n-layer in the active part of the junction. Another important quality of IR materials is high transversal conductivity, as the tandem cell is a series connected 2-terminal device. As-low-as-possible activation energy of doped layers is desired to reduce device voltage loss as the IR forms the tunnel recombination junction (TRJ) in multijunction cells. The effects of IRs inside tandem cells have been extensively simulated, confirming the necessity of IRs to obtain tandem cells with high performance [30]

[31][32]. Simulations proved to help in the optimization process of IR parameters for cell application.

As mentioned earlier, the two junctions of a tandem cell absorb different wavelengths of light. This light needs to be scattered for optimum absorption. Optimal light scattering for the top- and bottom junction requires different textured surfaces for the top- and bottom junctions respectively. In case of subsequent nc-Si:H layer deposition, the nanocrystals in an IR can serve as a seed layer providing high quality nc-Si:H material growth.

Combining all these requirements, an IR can have a theoretical quadruple functionality, serving as the reflecting, doped, texturing, and seed layer. Only a limited number of materials exhibit most or all of these properties. In addition, the material of the IR should be cheap and abundant. In this section, the most commonly used materials will be discussed.

1.3.2 Zinc oxide as a reflecting layer

Aluminum-doped zinc oxide (ZnO:Al) and boron-doped zinc oxide (ZnO:B) are commonly used as a front TCO. These TCOs have several favorable properties: high electron mobility and therefore high conductivity, wide bandgap (3.3 eV), and low refractive index (n~2 at

600 nm) which results in high reflectance at an interface with silicon layers (n

Si

~4). ZnO:Al is commonly deposited by sputtering or a low pressure chemical vapor deposition (LPCVD). Disadvantages of ZnO for use as an IR include its limited doping ability. The material is also deposited by different methods than the other layers in the TF Si cell, complicating the flow chart of cell processing.

8

1.3 REFLECTING LAYERS IN THIN-FILM SILICON SOLAR CELLS

ZnO was the first material considered as an IR in an a-Si:H/nc-

Si:H tandem cell [28]. The thickness of the IR was between 30-80 nm, providing decent reflective properties. Since then, it has been a popular material for use as an IR [33]. At EPFL in Neuchatel, the research was broadened to applying a 1.5 μ m thick asymmetric IR made of ZnO:Al in a tandem cell [34]. The asymmetric IR (with a random textured surface) proved to be effective, enhancing the J

SC

of the top cell, without the need of increasing its absorber layer thickness. The scattering effects at the asymmetric IR/Si interface provide the optimum light in-coupling into the top a-Si:H solar cell thanks to the 300 nm lateral feature size. This allows a thinner a-Si:H junction to be deposited, which boosts the stabilized tandem cells efficiency, due to reduced light induced degradation of the thinner a-Si:H absorber layer. The initial efficiency of this laboratoryscale cell was reported at 11.2% and stable at 9.8% [34].

1.3.3 Silicon oxide as a reflecting layer

The first successful integration of silicon oxide (SiO

X

) into thin-film silicon solar cells was reported by the Fuji company [35]. Yamamoto et al. was the first to apply SiO

X

as an IR [36]. The beneficial properties of this material sparked great interest and led to significant research efforts in SiO

X

-based reflecting layers from many groups around the world [37]–[39]. In the past few years, silicon oxide became widely used in thin-film silicon PV and has been applied in the most state of the art solar cells with the best performance to date [21], [40], [41].

Silicon oxide has also been applied in hetero- junction solar cells based on monocrystalline Si wafers [42].

Similar as ZnO used as an IR, silicon oxide improves the top cell current by reflecting back light because of a smaller refractive index compared to silicon. Doped silicon oxide has greater optical transparency than the standard a-Si:H doped layers due to its higher band gap. It serves as a better TRJ in tandem cells because of a high transversal conductivity and lower activation energy compared with doped a-Si:H layers. In single junction cells it also improves electron collection [43].

Silicon oxide has several advantages over ZnO. It has a lower lateral conductivity, preventing potential activation of shunt paths between the two sub-cells when used as an IR. Silicon oxide can be deposited in the same chamber as the rest of the silicon-based layers of the cell and it can be doped to a desired level in-situ. It can serve as a seed layer with its tunable nanocrystalline structure, hence it has been termed as hydrogenated nanocrystalline silicon oxide (nc-SiO

X surface texturing possibilities of nc-SiO

X

:H are limited.

:H). However,

9

1. INTRODUCTION

The properties of nc-SiO

X

:H with respect to plasma enhanced chemical vapor deposition (PECVD) conditions have been extensively studied [43]–[45]. In general it can be concluded that:

1. Adding more CO

2

into the gas mix is accompanied by more oxygen incorporation leading to a lower refractive index and higher band gap while the conductivity drops.

2. The material can be doped by incorporating doping gases into the gas mix, for instance PH

3

or B

2

H

6

. N-doping in general is considered more efficient than p-doping since the presence of boron in the plasma is detrimental for silicon crystal formation.

3. A minimal pressure and H

2 phase in the material.

flow is needed to achieve a crystalline

Developing p-doped nc-SiO

X to n-doped nc-SiO

X

:H is a more complicated task compared

:H as the presence of boron in the plasma influences film growth, crystallinity, and properties more profoundly than phosphorus

(see section 3.3). After the successful application of nc-SiO

X

:H as an n-layer, it was used as a p-layer, as well showing reduced parasitic absorption in reference to a conventional p-doped silicon carbide layer

[46], [47]. The material showed improved refractive index matching between the TCO and absorber layer, thus improved antireflective properties [48], [49]. The integration of a nc-SiO

X been shown to increase the open-circuit voltage ( V

OC

:H p-layer has

) [50]. From the processing point of view, the application of p-doped nc-SiO

X

:H simplifies the p-layer in comparison to the use of silicon carbide; only one layer is necessary as there is no need for a buffer layer with a higher band gap that prevents electrons in the i-layer to diffuse into the p-layer [51].

1.3.4 Advanced concepts of reflecting layers

To improve the absorption enhancement in tandem solar cells even further, many advanced concepts are being explored. In case of single junction cells, the best back reflectance was achieved with a combination of nc-SiO

X

:H, ZnO:Al, and silver back reflector [52]. This configuration of two (or more) optically active materials stacked on top of each other can be considered as a photonic crystal or Bragg reflector. Photonic crystals in general are extremely appealing for incorporation into solar cells as reflective layers [53] and have been studied in the past [32],

[54]–[56]. Bragg reflectors are categorized as 1D photonic crystals.

High reflectance can be achieved with two materials with different

10

1.3 REFLECTING LAYERS IN THIN-FILM SILICON SOLAR CELLS

Figure 1.6: A 1D photonic crystal sandwiched between the top a-Si:H and bottom nc-Si:H cell of a tandem.

refractive indexes forming a stack (Fig. 1.6). The bigger the refractive index contrast, the greater is the reflection. The characteristic parameter for the Bragg-stack of a homogeneous film is the optical film thickness h opt

= n

1

· d

1

= n

2

· d

2

of both layers. By tuning the thickness of each layer and the number of periods that are repeated, reflectance up to

100% in a narrow wavelength range (also called the photonic band gap) can be achieved. To broaden this range, modulated photonic crystals can be applied [57]. A Bragg stack with a photonic band gap that spectrally matches the absorption edge of a-Si:H can enhance the overall absorption in the top cell up to a factor of 1.22 when compared to a solar cell without an IR [56]. Most of the previous work on Bragg stacks serving as IRs was simulation work. Therefore the practical application is of great interest and will be described in detail in section 5.1.

The two junctions constituting the a-Si:H/nc-Si:H tandem have different band gaps and therefore absorb light of different wavelengths.

To increase absorption, light is scattered at textured interfaces when entering the solar cell. The texture is adopted by all layers deposited on top of it. As this texture may induce scattering beneficial for the top cell absorption, it may seem flat for light reaching the bottom cell. Due to limited scattering in the bottom cell the J

SC

does not increase. Therefore, after the top cell deposition, it would be beneficial for the interface texture to change. Obermeyer et al. simulated a double structure diffractive interlayer for the p-i-n configuration [58]. They concluded that for good scattering in a 200 nm thick top cell the period size of surface features should be 300-400 nm and 1200-1600 nm for the bottom cell.

With these ideal feature periods, a current gain of up to 20% is predicted.

It is beneficial to increase the texture period size of the bottom cell not just for gaining more J

SC

, but as well for better nc-Si:H material quality resulting in higher V

OC and fill factor ( FF ). Defective regions are known to form above the sharp valleys of the small texture features [59], [60].

11

1. INTRODUCTION

Increasing the feature size or flattening the surface, as was shown by

Boccard [61] can prevent these defective regions from forming, thus improving V

OC

and FF.

In case of an n-i-p configuration, the native texture of ZnO was used to scatter light in the top cell, as shown by Soderstrom et al. [34]. At first, the deposition substrate had large features beneficial for bottom cell scattering. After the deposition of the bottom cell, when growing the

ZnO IR, these initial features were canceled out and taken over by the native texture of the ZnO from an LPCVD process. This native texture has roughly 300 nm periods, beneficial for top cell scattering. This work proved how beneficial it is to re-scatter light between the two junctions.

As this was done in an n-i-p configuration, in section 5.2 an asymmetric

IR will be shown in a p-i-n cell.

1.4 The objective of this thesis

Different materials can be used as reflective layers. In this thesis, the main focus is on doped nc-SiO

X

:H which has a combination of good optical properties (low refractive index, high bandgap) and electrical properties

(high conductivity, low activation energy). TF Si solar cells are mostly made by PECVD from silane and hydrogen gas. Nc-SiO

X

:H layers can be deposited in the same chambers as the rest of the solar cell, making it a versatile material to work with. Nc-SiO

X

:H layers can be characterized for various optical, electrical, and nanostructural properties. Details of the nc-SiO

X

:H layer and solar cell processing and measurements will be explained in Chapter 2.

It has been established that nc-SiO

X

:H works well as doped reflective layers in TF Si solar cells. However, deeper insight between the optimized optical and electrical properties of nc-SiO

X

:H and the nanostructure is unknown. In Chapter 3, the nanostructural nature of nc-SiO

X

:H is studied in detail to reveal the relation between the optical and electrical properties and its nanostructure. It is examined in detail whether both the amorphous and crystalline phases play an important role in the material.

The following questions about the amorphous phase will be answered: is the amorphous tissue homogeneous or heterogeneous? What is the composition and is the quality of the amorphous tissue related to the optimized optical and electrical properties? The following questions about the crystalline phase will be answered: what (minimum) fraction of crystalline material is optimal? What is the optimum size of the crystal grains? Finally, how universal are the processing and nanostructure in relation to optical and electrical properties? All the above mentioned topics are covered in Chapter 3.

12

1.4 THE OBJECTIVE OF THIS THESIS

P- and n-doped silicon oxide development and nanostructure study

Chapter 3

Single junction cell results with p- and n-doped silicon oxide

Intermediate reflector development and tandem incorportion

Chapter 4 Chapter 5

Figure 1.7: Schematic overview of topics covered in this thesis and their depicted positions in a tandem cell.

The developed nc-SiO

X

:H layers have device grade optical and electrical properties, but whether they can improve device efficiency in reference to solar cells without nc-SiO

X

:H will be answered in Chapter

4. The performance of single junction a-Si:H cells with the integrated device grade materials will be shown. N-doped nc-SiO

X

:H layers

(Fig. 1.5, yellow layer) are applied as the n-layer and back reflector with Ag. P-doped nc-SiO

X

:H (Fig. 1.5, red layer) are applied as the p-layer, serving as the window layer. The n-doped nc-SiO

X

:H improves the spectral response over the entire spectrum. What is causing this unexpected effect will be answered in Chapter 4.

The n- and p-doped nc-SiO

X

:H meet in the middle of a tandem cell.

They are primarily doped layers but by adjusting their thickness and refractive index they serve as a functional IR. Can higher tandem cell efficiency be achieved by increasing the number of IR layers, making

Bragg reflectors? In Chapter 5 this question will be answered as well as how can Bragg reflectors be efficiently designed and what rules apply to their design? How critical is the thickness and refractive index of each layer for the desired optical performance? Can simulation software design an effective Bragg reflector?

The second part of Chapter 5 will be about interface texturing which brings up the following questions: Which materials can easily be textured on a cell without damaging it? Is it possible to separate the textures between the two junctions of a tandem cell to achieve optimal Asahi-like

13

1. INTRODUCTION scattering in the top cell and optimal scattering and defect-free growth for the bottom cell? Can this be achieved with wet-etching steps? Can this be achieved with a combination of polishing/wet etching? All these issues will be covered in Chapter 5.

14

To test the newly-developed silicon oxide material and the IR concepts, complete thin-film silicon solar cells need to be deposited and characterized. This chapter will describe all the processing methods and equipment necessary to make the silicon oxide layers and TF Si solar cells as discussed in this thesis. The applied contact configuration and measurement of a solar cell will be described as well.

2.1 Solar cell processing

2.1.1 Plasma Enhanced Chemical Vapor Deposition (PECVD)

Thin-film layers can be grown by Chemical Vapor Deposition (CVD).

Precursor gasses react and decompose on a substrate (usually glass) surface producing the deposited layer. Normally, gas temperatures gasses. Using a plasma, the temperature to dissociate precursor gases can be much lower as the energetic free electrons and ions provide the necessary energy to break the chemical bonds.

To ignite a plasma a rapidly alternating electric field is applied between two electrodes. Atoms and/or molecules are ionized in the

2. PROCESSING AND CHARACTERIZATION OF THIN-FILM

SILICON LAYERS AND SOLAR CELLS plasma. Highly energetic electrons collide with gas molecules initiating the chemical reaction. This process is called Plasma Enhanced Chemical

Vapor Deposition (PECVD). When the plasma is generated between two electrodes with an alternating bias, PECVD is referred to as a capacitively coupled plasma. Ionization events produce new electrons and ions to maintain the plasma. For example, for a-Si:H layer growth, the electrons in the plasma have energies typically ranging between 0 and 30 eV. The power provided by the induced alternating electric field may not always be sufficient to free electrons to start up the plasma. Often a spark needs to be used to provide highly energetic electrons to ignite the process chain reaction.

The SiH

4

molecule can dissociate into Si x

H radicals and ions. The y dissociated ions and radicals can be deposited on the grounded substrate electrode. Figure 2.1 gives a scheme of a typical PECVD setup. Inside the chamber, the substrate is placed between two parallel capacitively coupled electrodes. The generator is set to radio frequency (RF: 13.56

MHz) for the growth of a-Si:H, silicon oxide and silicon carbide layers, while for the nc-Si:H (intrinsic) layers a very high frequency (VHF:

40.68 MHz) is used. A matching box with two capacitors is used to regulate the reflected power. The heater sets the substrate temperature by a pump system. The in-flow of precursor gasses are controlled by the gas system consisting of mass flow controllers. SiH

4

is the main gas used to deposit all layers in combination with either CO

B

2

H

6

2

, CH

4

, H

2

, PH

3

, and

. The main advantages of RF-PECVD compared to other deposition techniques are:

• Variety of available substrates: from glass to flexible foils

• Large deposition area possible

• Effective doping of layers

• Cost-effective on industrial scale

2.1.1.1 Amigo cluster tool

The samples for this thesis were prepared in a six-chamber cluster tool from Elettrorava called the AMIGO. Each chamber is designated for the deposition of different types of silicon alloys. There is one chamber for p-doped silicon alloys, one for n-doped silicon alloys, one for intrinsic a-Si:H, one for intrinsic nc-Si:H, one chamber for special silicon alloys

16

2.1 SOLAR CELL PROCESSING

Figure 2.1: Configuration of a PECVD setup and the components involved.

like silicon carbides and nitrides, and a sputtering chamber for ZnO:Al.

The deposition gases in each chamber are injected through a showerhead electrode. All of the chambers operate at RF biasing frequency except the chamber dedicated to intrinsic nc-Si:H which operates at VHF: 40,68

MHz. To preserve the necessary vacuum conditions, all six chambers are connected through a transport chamber. This transport chamber is equipped with a robot arm which takes samples from a load lock. The load lock chamber has a capacity to handle 5 substrates.

2.1.2 Magnetron Sputtering

Sputtering is a Physical Vapor Deposition (PVD) technique. A target is bombarded by ions which sputter species (atoms, molecules, nanoparticles) of material. These species deposit on a substrate. This technique is used to deposit aluminum doped zinc oxide (ZnO:Al) that serves either as the front Transparent Conductive Oxide (TCO) contact in a solar cell or as the IR. A RF generator applies an alternating electric field between the material target and substrate. As a result, a fraction of an inert gas, in this case argon, is ionized. These positive argon ions are accelerated towards the target and sputter species from the target.

These species are deposited onto the substrate. To avoid any drastic increase of temperature of the target due to the ion bombardment, the target is cooled with water.

17

2. PROCESSING AND CHARACTERIZATION OF THIN-FILM

SILICON LAYERS AND SOLAR CELLS

Figure 2.2: Schematic diagram of a sputtering process.

In order to increase the ion-current density, a magnetic field is applied perpendicular to the electric field and the electrons are confined near the surface of the electrode. The combination of sputtering and focusing of electrons using a magnetic field is thus referred to as magnetron sputtering. A scheme of the general sputtering technique is shown in

Figure 2.2. The sixth chamber of the AMIGO cluster tool is equipped with this deposition technique.

2.1.3 Thermal and electron beam evaporation

Thermal and electron beam evaporation are physical vapor deposition

(PVD) methods used to deposit thin layers of material, mainly metals used as BRs and contacts in both single layers and complete devices. In these techniques a (metal) target is heated up until it begins to evaporate.

Evaporated metal then reaches a cold substrate where it deposits. This process is performed under a vacuum environment to minimize the contamination with other substances. In this work, thin films of silver

(Ag) that act as back reflectors are deposited using this method. On top of the Ag, chromium (Cr) and aluminum (Al) are deposited. Ag is covered to prevent it from oxidizing in air. Cr is separating Ag and Al because in the annealing step these two metals can diffuse into each other.

Another approach is the thermal evaporation technique used to deposit

Ag. In this PVD approach, high electrical current is passed through a metallic boat that contains the metal, warming it up to its evaporation

18

2.1 SOLAR CELL PROCESSING

Figure 2.3: Schematic drawing of the thermal evaporation technique (a) and the electron beam evaporation technique (b).

temperature (Fig. 2.3a). The Cr and Al are evaporated using the electron beam evaporation technique which heats the evaporant located in a crucible by using a magnetically focused beam of high energy electrons

(Fig. 2.3b). The advantage of this method is that the temperature is not limited by the melting point of the filament, allowing the evaporation materials with high evaporation temperatures. The evaporants are present in a crucible which is cooled by water during evaporation. In this work all metal evaporation was done using the Provac PRO500S.

It is a single chamber high-vacuum system equipped with a 4-pocket

10 kV electron gun and a boat for thermal evaporation. Pumping cycles and deposition recipes are full programmable. Desired contact sizes are created by masking.

2.1.4 Reactive Ion Etching (RIE)

Reactive ion etching (RIE) is used to selectively remove thin-film layers. In this thesis it was used mainly for removing (etching) deposited nc-SiO

X

:H from around the metal contact to achieve a better defined size of the contact area and reduce additional current collection.

RIE is a technique very similar to PECVD, except gases with large molecular weights are used. These gases etch atoms away from the layer surface instead of depositing on it. These gases are ionized between two parallel plates by RF bias. Because of the greater impulse moment of the ions, they are not deposited, but sputter atoms from the thin film. The setup used is a single chamber Alcatel system with load lock. Etching gases CF

4

and CHF

3

in a He carrier gas were used at a pressure of 0,5 mbar. A plasma was ignited with these gases the same way as in PECVD

(Fig. 2.1).

19

2. PROCESSING AND CHARACTERIZATION OF THIN-FILM

SILICON LAYERS AND SOLAR CELLS

2.2 Solar cell characterization

Solar cells were processed by PECVD deposition of all silicon alloys on a TCO substrate, metal contact deposition via evaporation, and eventual

RIE for better contact area definition. The solar cells are then measured to determine their external parameters and spectral response.

2.2.1 External Parameters

The external parameters are a set of measured characteristics used to describe the electrical behaviour of a solar cell. The main external parameters are: The short circuit current density (

J

SC circuit voltage ( V

OC

), the open

), the fill factor ( FF ) and the efficiency (η). These m 2 of the AM 1.5 spectrum. In addition, the series resistance ( R

S parallel (shunt) resistance (

R

P

) and the

) can be determined using a J-V curve.

A typical J-V curve that relates these variables is shown in Figure

2.4. The intersections with the J=0 and V=0 axis of the curve are marked by the

(

V

V

OC

and J

SC

. There is a certain current ( J m

) and voltage

) where the output power is maximum. The shape of the J-V curve m greatly affects the resulting maximum power output. The FF is used to characterize the“rectangleness”of the J-V curve defined by the ratio between the maximum power and the product of J

SC is as follows:

and V

OC

. The equation

(2.1)

The efficiency of a solar cell can be defined in terms of the external parameters as:

(2.2) where P in

is the power density of the incident solar radiation.

All the external parameters are obtained using the PASAN Flash solar simulator, which provides the standard AM 1.5 solar spectrum and has an automatic probing stage. The J-V characteristics of solar cells exposed to light pulses of 4 ms are measured. Solar cells reported in this work were deposited on 2,5 x 5 cm substrates using two different contact configurations. The first configuration accommodates about fourteen

20

2.2 SOLAR CELL CHARACTERIZATION

Figure 2.4: J-V curve.

Figure 2.5: The components of an EQE setup.

individual cells of 4x4 mm. The second configuration has four cells of

1x1 cm. The measured FF and V

OC trustworthy. In contrast, the

J

SC

using the PASAN are considered

is not reliable due to the imprecise determination of the cell area and discrepancies of the light source in reference to the AM 1.5 spectrum. The J

SC

was taken from the EQE measurement as this technique is independent on the light source or contact area.

2.2.2 External Quantum Efficiency (EQE)

The spectral response or the external quantum efficiency (EQE) of a solar cell is defined as the percentage of photons incident on the at its terminals. In the PVMD group, the setup used is formed by a light source, a monochromator and a current amplifier (Fig. 2.5). The

21

2. PROCESSING AND CHARACTERIZATION OF THIN-FILM

SILICON LAYERS AND SOLAR CELLS

Figure 2.6: An example of a spectral response measurement of a tandem cell measured with the EQE setup.

solar cell is exposed to monochromatic light and the photo-generated current is measured. The current is integrated over the wavelength and multiplied by the AM 1.5 spectrum, giving the

J

SC

. The

J

SC

is measured under short-circuit conditions (no bias voltage). In addition, in some experiments, EQE was measured using a reverse bias of -1V to ensure that most of the light excited charge carries are collected. In this manner, comparing EQE at no bias and reverse bias gives an indication of the charge carrier recombination in a solar cell and shows its potential of optical performance.

When measuring the EQE of a tandem cell (Fig. 2.6), both series connected junctions need to be producing light excited electron-hole pairs in order to collect current. In the standard EQE configuration it is not possible to extract the spectral response of each junction. For this purpose, bias light is incident onto the cell. To measure the top cell, infrared bias light is used as it passes through the top cell and is solely absorbed in the bottom cell. To measure the bottom cell, blue light is used, which is solely absorbed in the top cell. This way the biased cell is conductive and the photocurrent of the non-biased cell can be measured.

Since the bias light intensity is constant, only the charge carriers in the non-biased junction, generated by absorption of light transmitted through the monochromator, are detected.

22

2.3 SINGLE LAYER CHARACTERIZATION

Figure 2.7: The Raman spectrum between 150-750 cm constitutes the I

521

/I

480

ratio as a measure of crystallinity.

-1

with the black and red Gaussians showing the amorphous and crystalline phonon modes respectively. The ratio of the area of these Gaussians

2.3 Single layer characterization

2.3.1 Raman spectroscopy

Raman spectroscopy is a fundamental method, by which the structure and composition of materials is investigated. The crystalline volume fraction of the nc-Si:H based thin-films can be determined by Raman spectroscopy. Upon interaction with valence electrons, most photons scatter elastically. This spectroscopy technique is based on the small fraction that scatters inelastically. The inelastically scattered photons have a frequency that is slightly shifted compared to the initial frequency before scattering. This shift in energy is called the Raman Effect (shift).

As probing light a monochromatic light source is used (usually a laser).

The recorded Raman shift spectrum reflects the phonon density of states. Within the detected spectral range the transverse-optic modes of crystalline and amorphous silicon are visible at 521 and 480 cm -1 respectively (Fig. 2.7).

The Raman spectra of the samples were measured with a Renishaw

InVia, grating 1800 lines/mm Raman microscope with a 180° back scattering geometry. It can measure Raman shifts from 250 to 3000 cm

-1

. The phonon modes for determining the crystallinity of silicon samples can be found in a scan between 150 to 750 cm

-1

(Fig. 2.7).

The microscope is equipped with a 25 mW Argon laser operating at 514

23

2. PROCESSING AND CHARACTERIZATION OF THIN-FILM

SILICON LAYERS AND SOLAR CELLS

Figure 2.8: Scheme of the Raman spectroscopy setup.

nm and a Helium-Neon laser with a wavelength of 633 nm operating at the same power and spot size (Fig. 2.8). The intensity of the lasers can be regulated from 0,001 to 100%, where 5% was most commonly used. This beam is focused on the sample via mirrors and lenses. The incident laser beam is scattered by the sample and is then redirected back into a spectrophotometer. The elastic scattered light is filtered out by a notch filter. The remaining inelastically scattered light is diffracted into a spectrum and collected by a spectrometer (Fig. 2.8). There it is converted into electric signal which is later processed by a computer to determine the Raman shift spectrum. The resulting spectrum is fitted with

Gaussians from which the crystalline content in nc-SiO

X via the I

521

/I

480

ratio.

:H is estimated

2.3.2 Fourier transform infrared spectroscopy

Fourier Transform Infrared Spectroscopy (FTIR) is the most widely used infrared spectroscopic method. Various oxygen-related bonding configurations in nc-SiO

X

:H can be distinguished with this method.

Vibration modes of non-symmetric bonding configurations are detected in the infra-red absorption spectra. The absorption intensity related to a certain mode is a measure for the number of associated bonds present in the material. The measurement method uses a Michelson interferometer and a beam splitter. The source beam is split into two beams by the beam splitter (Fig. 2.9). One split beam is reflected off a stationary mirror and the other from a moving mirror. The two beams interfere and pass through the sample into the detector. The absorption spectrum is obtained by Fourier-transforming the complex interference of these two superimposed beams. Multiple measurements are taken to reduce the

24

2.3 SINGLE LAYER CHARACTERIZATION

Figure 2.9: Configuration of a FTIR setup and the components involved.

noise in the measurement [62].

In this technique, light having a black body radiative spectrum from an incandescent lamp is directed onto a film deposited on a c-Si wafer

(prime wafer, 500 μ m). The c-Si wafer is transparent for wavenumber range of interest (500–2200 cm

-1

). Therefore, glass is not used as a substrate since it has a high absorption coefficient in this range. Part of the radiation is transmitted and some is absorbed by the film. The transmittance spectrum of the bare substrate and the substrate with a deposited film is measured. The resulting transmittance spectrum of only the film is obtained by dividing the spectrum of the substrate with film by the spectrum of the bare substrate. The transmittance is measured as a function of the position of the moving mirror (see Fig.

2.9) in the interferometer. This measured interferogram is converted into a spectrum as a function of frequency using a Fourier Transform.

The resulting spectrum shows the absorption peaks of the deposited film versus the wavenumber (Fig. 2.10).

The absorption peaks are linked to certain vibration modes (bending, wagging, and stretching). FTIR can provide information on the chemical bonding and indirectly provide information on the quality of a film and its components. It can help in developing and optimizing a material with certain desired properties [62]. An overview of the stretching modes

25

2. PROCESSING AND CHARACTERIZATION OF THIN-FILM

SILICON LAYERS AND SOLAR CELLS

Figure 2.10: Sample FTIR absorption spectrum.

Table 2.1: Assignments of stretching modes.

Wavenumber [cm

-1

]

1050

1106

1135

2100

2140

2180

2250

Assignment

(Si-O-Si)Si

X

stretching

(single oxygen)

Interstitial oxygen (O

2i

)

(Si-O-Si)O

X

stretching

(oxygen rich)

Si-H stretching on void surfaces

Si-H

X

stretching /

OSi-H stretching

O

2

Si-H stretching

O

3

Si-H stretching examined in this thesis is shown in Table 2.1.

The setup used to perform FTIR measurements is a Thermo Electron

Corp model Nicolet 5700. The infrared light source is an Ever-Glo lamp

(9600 – 20 cm

-1

). A KBr beam splitter (7400 – 350 cm

-1

) and DTGS detector (6400 – 200 cm

-1

) are used. The sample compartment contains two holders mounted on a shuttle which automatically switches between the reference wafer and the wafer with the deposited film during the measurement.

26

3.1 Introduction

In current state-of-the-art multijunction TF Si solar cells, silicon oxide layers form a vital part of the devices. These bi-functional layers are used as doped layers and (intermediate) reflector layers for light management [21], [22], [28], [41], [45], [52], [63]. The low mobility of holes in the intrinsic layer requires the holes to be collected at the front side of the cell where most carriers are generated due to light absorption. For that reason, the p-layer is at the front side of the cell.

To couple as much light as possible into the absorber layer, the p-layer needs to have minimal reflection and parasitic absorption. This can be achieved with silicon oxide which is a low refractive index (n) material.

Its refractive index has values between that of a-Si:H and the TCO to achieve a welcome refractive index grading (n a-Si:H

> n

SiOx

> n

TCO

). On the contrary, the reflection of the n-layer serving as the BR needs to be high. This can be as well achieved via a low n silicon oxide layer. Due to a larger refractive index mismatch between nc-SiO

X layers, nc-SiO

X

:H and the silicon

:H layers reflect light in the desired spectral range of

500-700 nm back into the top cell. Directing more light into the intrinsic a-Si:H absorber layer is an important optimization tool to achieve current

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION matching with the other cell(s) in the multijunction. In addition, it allows the i-layer to be kept reasonably thin in order to minimize the impact of light-induced degradation [24]. Apart from these strict optical criteria, the silicon oxide layers need to be sufficiently conductive to transport charges to the TCO and metal contacts.

Silicon oxide is a material which is cheap, abundant, and easy to process.

It helps to reduce parasitic absorption losses in doped layers because of its higher bandgap in reference to silicon layers. As mentioned in section

1.3.3, silicon oxide can be multifunctional as it serves as a reflective and doped layer. In addition the p-doped silicon oxide can serve as a seed layer for the growth of an intrinsic nanocrystalline absorber layer in a p-i-n configuration. To what extent is the presence of crystalline grains in silicon oxide necessary and what grain properties are required to achieve good optical and electrical properties? Similarly, what is the importance of the amorphous tissue and its role in the functioning of the material to have good optical and electrical properties?

As vitally important as nc-SiO

X

:H is, not much is known about its nanostructure. In this thesis, four possible cases of heterogeneous nanostructure are considered that could potentially represent the real nanostructure, as depicted in Figure 3.1. This heterogeneous material is currently interpreted to be a matrix with a crystalline phase made of crystalline silicon (c-Si) grains and an amorphous phase of a-Si:H and amorphous silicon oxide (a-SiO

X

:H) tissue (Fig. 3.1(d)). Can doped amorphous silicon oxide work just as well (Fig. 3.1(a))? Is the presence of crystalline Si grains (Fig. 3.1(b)) or amorphous silicon (Fig. 3.1(c)) necessary? There are reports of the nc-SiO

X

:H material possessing purely silicon filaments aligned in the layer growth direction being responsible for good transversal conductivity [47], [64]. These filaments are surrounded by isolating amorphous silicon oxide tissue having very low refractive index values. Films with such nanostructure show anisotropy in the conductive properties, where the lateral conductivity is low and the transversal conductivity high. When integrated as part of an IR in a tandem cell, the lower lateral conductivity has the potential to quench possible interconnections between shunts in the top and bottom cell. Therefore, the doped nc-SiO

X

:H has an advantage over zinc oxide

(ZnO) which facilitates this shunting of tandem cells due to its high lateral conductivity.

In this chapter, the dependence of conductivity on the nanostructure is thoroughly studied. Silicon filament-like structures have been observed in the p-type material studied here as well. Using TEM and

Raman spectroscopy, a small amount of the crystalline silicon phase was detected. The question arises about what is the role of the crystalline

28

3.1 INTRODUCTION

Figure 3.1: Planar depiction of four cases of silicon oxide nanostructure: (a): pure amorphous silicon oxide, (b) crystalline silicon directly embedded in amorphous silicon oxide, (c) regions of amorphous silicon embedded in amorphous silicon oxide, (d) crystalline silicon surrounded by amorphous silicon embedded in amorphous silicon oxide.

grains in the silicon filaments. Samples under varying conditions were deposited observing that not all nc-SiO

X

:H material has the same crystalline properties. High values of lateral conductivity were achieved in samples with varying crystallinity, showing isotropic behavior. In addition, the conductivity of nc-SiO

X

:H still differs between samples with similar crystallinity and crystalline grain size. Therefore, the (lateral) conductivity of nc-SiO

X

:H is of great interest as a parameter to study the nanostructure of nc-SiO

X

:H. The conductivity depends on the presence of the crystalline phase as well as the properties of a-Si:H tissue in the filaments and the properties of a-SiO

X

:H tissue.

This chapter presents a thorough study of the nanostructure of nc-

SiO

X

:H using different measurement methods. First, the relation between measured optical and electrical single layer parameters and the deposition parameters will be studied (section 3.3). To improve the insights into the nanostructure of the nc-SiO

X

:H, the properties of the crystalline silicon grains were studied in detail by TEM imaging (section 3.4.1) and Raman

Spectroscopy (section 3.4.2). The elemental composition and bonding configurations of the a-SiO

X

:H tissue and the quality of the a-Si:H tissue were studied using FTIR (section 3.4.3) and XPS (section 3.4.4) measurements. Layers deposited at various deposition parameters were integrated into solar cells. The performance of these cells is analyzed in the next chapter.

29

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

3.2 Experimental details

The nc-SiO

X

:H based films were deposited on Corning Eagle XG glass using the AMIGO RF-PECVD cluster tool (section 2.1.1). Samples were deposited in series in which the RF power, pressure, temperature, and gas flows of B

H

2

, and CO

2

2

H

6

(for p-doped material), PH

3

(for n-doped material)

were varied (Tables 3.1, 3.2). For each series, only one parameter is varied while the others are kept constant at values used to process the device grade material. The deposition time was kept constant. Consequently the film thickness varies from sample to sample but is mostly around 250 and 120 nm for n- and p-doped nc-SiO

X

:H, respectively. The best solar cell performance was obtained with device grade nc-SiO

X

:H materials processed under the conditions listed in electrode gap of 14 mm, and bias frequency set at 13.56 MHz as well. For electrical characterization purposes aluminum contacts were evaporated

(section 2.1.3) onto the films. Next, they were annealed for 30 minutes interface.

Table 3.1: Variation of deposition parameters of p-doped nc-SiO

X deposition conditions are shown in red.

:H. The device grade material

Parameter

Power

Pressure

Temperature

CO

2

:SiH

4

SiH

4

:B

2

H

6

H

2

:SiH

4

Variation & best material

0.08 –

0.29

- 0.5 W/cm 2

1.4 – 2.2 - 3 mbar

120 –

180

- 240°

1.17 –

1.75

- 3.5

250 –

400

– 700

0 -

213

– 250

Table 3.2: Variation of deposition parameters of n-doped nc-SiO

X deposition conditions are shown in red.

:H. The device grade material

Parameter

Power

Pressure

CO

2

:SiH

4

SiH

4

:PH

3

H

2

:SiH

4

Variation & best material

0.06 –

0.07

- 0.14 W/cm 2

0.5 –

1.25

- 2.75 mbar

1 –

1.6

- 2.2

25 –

42

- 125

0 –

100

- 150

30

3.3 MATERIAL DEVELOPMENT

The single nc-SiO

X

:H layers were characterized by reflectiontransmission (RT; page 34), conductivity measurements (σ; page 35) in a lateral contact configuration, activation energy ( E a

; page 35), Raman spectroscopy (section 2.3.1), and X-ray photoelectron spectroscopy

(XPS; page 60). Simultaneously with the films on glass, films were deposited on c-Si wafers for FTIR measurements (section 2.3.2). Both planar view and cross sectional images of the device grade materials were made using TEM imaging (page 43).

3.3 Material development

3.3.1 Single layer properties

First, the observed dependence of the optical and electrical properties of nc-SiO

X

:H on the processing parameters is shortly summarized. The requirements for an ideal reflector layer based on nc-SiO

X values of activation energy ( E a

) and refractive index ( high value of band gap and transverse conductivity (σ t

:H are low n) combined with a

). Note again, that the measured conductivity is in the lateral direction (page 35). In Figure

3.2 and 3.3, these measured parameters are plotted against the main deposition parameters of n-doped and p-doped nc-SiO

X

:H, respectively.

For each deposition an optimum level of conductivity was found. The activation energy showed lowest values for conditions with lower power, pressure, and PH lowest E a

E a

3

:SiH

4

ratio settings. For the CO

2

:SiH

4 important in tandem cells when nc-SiO

X

ratio series the

was found between ratios of 1-1.6. In single junction cells, the

is not a critical parameter affecting cell performance, but

E a

becomes

:H is applied as an IR to form tunnel recombination junction for minimum loss in V

OC

.

The optical parameters as bandgap and refractive index are developing as expected with each deposition parameter except the PH

3

:SiH

4

ratio.

In that case it has to be noted that the explored parameter range was small (Fig. 3.2). Table 3.3 shows a summary of the dependence of the material properties of nc-SiO

X

:H as a function of the chosen deposition parameters. Figures 3.2 and 3.3 show that the optical and electrical parameters desirable for an IR are in competition. The device grade material was therefore found as a compromise between the optical and electrical requirements (Table 3.2). This tradeoff between optical and electrical properties was also observed by others [42], [44], [64], [66].

31

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.2: The measured optical and electrical properties vs. the deposition parameters of n-doped nc-SiO

X

:H.

32

3.3 MATERIAL DEVELOPMENT

Figure 3.3: The measured optical and electrical properties vs. the deposition parameters of p-doped nc-SiO

X

:H.

33

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Reflection-Transmission Spectroscopy (RT)

Reflection-transmission spectroscopy (RT) is a technique used to obtain the reflectance, transmittance, and absorption of thin films as a function of wavelength. From the measured R(λ) and T(λ), the refractive index ( n), and extinction coefficient ( k) can be obtained. In addition, these values are fitted to a predefined optical model of a glass substrate with a thin a-Si:H film on top to determine the film thickness and Tauc band gap ( E g

) of a-Si:H.

The spectral measurement ranges from 1.17 – 2.95 eV (375

– 1060 nm). The measurement technique consists of the following steps: A 50W halogen lamp shines light on a film deposited on Corning

7059 glass. The reflected and transmitted spectra are detected with separate silicon photodiodes. Absorption is determined form the relationship A(λ) = 1–T(λ)–R(λ). Finally, Scout software via a predefined model performs the calculations of the thickness and optical constants of the film on glass by fitting the measured spectra [91].

Table 3.3: Summary of the dependence of trends in material properties of n-doped nc-SiO

↑=increasing, ↓=decreasing, s=strong, w=weak.

X

:H;

Parameter

Power ↑

Pressure ↑

CO

2

flow ↑

PH

3

flow ↑

σ l

↓s none

↓s

n

600

↑s

↓w none

E a

↑ none none

E g

↓w

↑ none

Dep. Rate

↓w

↑ none

Table 3.4: Summary of the dependence of trends in material properties of p-doped nc-SiO

X

↑=increasing, ↓=decreasing, s=strong, w=weak.

:H;

Parameter

Power ↑

Pressure ↑

CO

2

flow ↑

B

2

H

6

flow↑

SiH

4

flow ↑

H

2

flow ↑

Temperature ↑

σ l

↓s

↓s none

↓w

n

600

↓w

↓s

↑s none

E a

↑s

↑s

↑w

↑s

E g

none

↑w

↓s

Dep. Rate

↑s

↑s

↑s

↓s

↓s

34

3.3 MATERIAL DEVELOPMENT

Dark Lateral Conductivity and Activation Energy

Conductivity (σ) is the ability of a material to pass electric charge. The activation energy ( E a

) is a measure of the energy difference between the Fermi level and the conduction band edge for electron transport (valence band for hole transport) [23] (Fig

3.a).

Figure 3.a: A band diagram depicting the p-i-n device with the doped layers. For intrinsic layers the

E

E a

and band gap shown in the a

is half the value of the band gap.

In calculating E a

, the sample has coplanar aluminum contacts with well defined dimensions and gap evaporated on its surface.

Electrical conductivity is obtained by:

(3.1) where R is the resistance of the layer, d the distance between the electrodes,

L the length of the electrodes and t the film thickness (Fig. a voltage (1–100 V, depending on whether the sample is doped or intrinsic) is applied onto the electrodes for each measurement.

The experimental setup measures the variation of the electrical conductivity of the material at different set temperatures. First the

35

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Table 3.b: A sample with parallel contacts for conductivity measurements. The dimensions for calculating conductivity are indicated.

taking a current measurement at each step. The activation energy is then obtained by fitting the current-temperature relationship to:

(3.2) where σ is the measured electrical conductivity, σ

O k b equation can be rearranged to extract the E a

:

is a constant,

is the Boltzmann constant and

T is the absolute temperature. The

(3.3)

A linear relation between the temperature and the natural logarithm of the electrical conductivity, where

E a

is the slope, is evident.

This device grade n-doped nc-SiO

X

:H material performs the best in a solar cell when applied as an IR or in combination with silver as a back reflector (BR). It is worth to note that n for this material (measured at

600 nm) is among the higher values obtained ( n=2.6-3). This comes from the trade-off between optical and electrical properties. The welcome drop in E a

and rise in conductivity is accompanied by a fall in the bandgap (Fig. 3.4). The same trends were noticed for the p-doped nc-SiO

X

:H as well (Fig. 3.5). The conductivity of the n-doped material

36

3.3 MATERIAL DEVELOPMENT

(

Figure 3.4: Measured material properties of n-doped silicon oxide: decreasing of activation energy

E a

) and bandgap ( E g

) and increasing of conductivity with increasing refractive index ( lower refractive index (higher bandgap) is accompanied with lower conductivity. n ). The welcome

37

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

(

Figure 3.5: Measured material properties of p-doped silicon oxide: decreasing of activation energy

E a

) and bandgap ( E g

) and increasing of lateral conductivity with increasing refractive index ( nm; same trends as in case of the n-doped silicon oxide.

n ) at 600

38

3.4 NANOSTRUCTURE ANALYSIS is almost two orders of magnitude lower than that of the p-doped. The difference in thickness between the materials could be an effect, as the n-doped material was on average 100 nm thicker. The conductivity has device grade values over a larger range of values of refractive index for the p-doped nc-SiO

X

:H compared to the n-doped nc-SiO

X

:H. This suggests that reasonable conductivity can be kept with varying optical properties over a large range. The search for device grade p-doped nc-

SiO

X

:H resulted in optimized material with a refractive index of around

2.7. However it is believed that device grade p-doped nc-SiO even lower values of refractive index can be developed.

X

:H with

The relation between bandgap and the refractive index has similar trends for both materials. The E a

however is generally higher for the p-doped nc-SiO

X

:H. To achieve low

E a

in p-doped material was more complicated and challenging in reference to the n-doped material. This processing challenge has its origin in the competition between the boron content and crystallinity. The more boron present in the material, the harder it is to achieve the desired minimum crystallinity to ensure good conductivity. This was not observed for the n-doped nc-SiO

X the variation of PH

3

:H as

flow does not substantially affect the crystallinity.

Further optimization must be devoted to achieve a lower E a index with good conductivity.

and refractive

3.4 Nanostructure analysis

3.4.1 TEM analysis

The most direct way to identify the heterogeneous nanostructure of the nc-SiO

X

:H films is using TEM imaging. May it be reminded that all

TEM images shown in this thesis are of device grade p- and n-doped nc-SiO

X

:H material.

First, we discuss the TEM analysis of n-doped nc-SiO

X images of n-doped nc-SiO

X

:H. TEM

:H (Fig. 3.6) reveal the presence of crystalline silicon grains of approximately 5 nm in size in the matrix. These grains with random crystal orientations are agglomerated more densely close to the substrate than close to the surface. Since the material has a similar structure in the lateral as well as in the transversal direction it can be presumed that the transversal conductivity is similar to the measured lateral conductivity (Fig. 3.6). These observations imply that the crystalline grains cannot solely be responsible for the charge transport in this n-type nc-SiO

X

:H. Therefore the quality of the a-SiO

X

:H tissue is of great importance in determining the conductivity properties and will be discussed in section 3.4.3.

39

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

(a) (b)

Figure 3.6: TEM images of the best performing n-doped nc-SiO silicon surrounded by amorphous SiO

X is a planar view.

X

:H showing ~5 nm grains of crystalline

:H. Image (a) is a cross-section of a nc-SiO

X

:H layer; image (b)

(b)

Figure 3.7: a) Bright field TEM image of cross-section of device grade p-doped nc-SiO

X

:H (above). The dark filaments are from crystalline material. The inlet shows the SAD pattern. The first inner 5 rings are Si (111), Si (022), Si (113), Si (004), and Si (133). (b) A high resolution TEM image on the filaments

(indicated by red contours) showed that each filament consists of many small nanocrystals which are randomly oriented.

As for the p-doped nc-SiO

X

:H, silicon appears to be ordered in filaments, which is in agreement with earlier mentioned observations in other studies [47],[64]. Figure 3.7(a) shows a bright field (BF) TEM image of the cross-section of the device grade material. The dark contrasts in the film are silicon crystals which are arranged in filament-

40

3.4 NANOSTRUCTURE ANALYSIS

(a)

(b)

Figure 3.8: (a) EF TEM image of the cross-section of the device grade p-doped nc-SiO

X

:H showing silicon filaments in a tree-like structure (bright contrast) and silicon oxide tissues (dark contrast). (b)

EF TEM image in high magnification, Silicon crystals with random orientations are circled in red.

like structures growing perpendicular to the substrate. Figure 3.7(b) provides a zoomed-in view of these filament structures. It can be seen that they consist of randomly-orriented polycrystalline grains. With selected area diffraction (SAD) patterning (inlet of Figure 3.7(a)), the reflections of various Si orientations were found: Si (111), Si (022), Si (113), Si

(004) and Si (133). The distribution of the orientations throughout the filaments is random. As in bright field (BF) and high resolution (HR) TEM images, it shows strong diffraction contrast between the crystalline and amorphous materials, therefore only crystalline material can be easily distinguished.

To map the whole Si filament distribution more effectively, energyfiltered (EF) TEM imaging was carried out. Figure 3.8(a) shows an EF

TEM image of the cross-section of the device grade material. It can be seen that silicon is organized in filaments in a tree-like structure starting from the substrate and surrounded by a-SiO

X

:H tissue. These silicon

41

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.9: Planar view BF TEM image of the device grade p-doped nc-SiO

X

:H. Dark spots indicate Si nanocrystals. These nanocrystals can only be found surrounded by a-Si:H, visible as gray clouds.

filaments consist of a mixture of a-Si:H and crystalline Si grains, seen in figure 3.8(b). The filaments have an average width of 3.3 nm, the length is about 20-30 nm.

The TEM analysis shows that the p-doped nc-SiO

X

:H grows more amorphous in the first few nanometers (Fig. 3.8(b)). This initial amorphous incubation layer was confirmed with spectroscopic ellipsometry. When the initial growth phase is examined more closely in Figure 3.8(a), the initially grown filaments are parallel to each other and with increasing crystallinity they branch out. Here, a hypothesis is proposed as to why the filaments are parallel at first: since near the substrate the material is almost completely amorphous, the a-Si:H and a-SiO

X

:H regions grow perpendicular to the substrate as in any PECVD process, not being influenced in any other direction because of their amorphous nature.

Later in the growth, after plasma stabilization, crystal grains with random orientations start to nucleate. Because of these orientations, the filaments start to grow in the preferential growth direction of the crystalline structure, causing them to branch out.

The studied nc-SiO

X

:H samples have high values of lateral conductivity. The TEM images provide clues to why that is the case. Two possibilities are explored. In Figure 3.8(a) it is suggested that for layers thicker than 20 nm these filaments branch out to the extent that they make contact with each other. These interconnected branches facilitate charge carrier transport in the lateral direction, giving high values of lateral conductivity, especially for thick samples (~100 nm). This suggests that

42

3.4 NANOSTRUCTURE ANALYSIS at the top of the nc-SiO

X

:H film the lateral conductivity is larger than at the bottom (substrate side). It has to be noted that the co-planar contacts are evaporated on the top of the nc-SiO

X

:H film. Therefore these terminals are in direct contact with the most conductive part of the film.

The second option is that the significant values of lateral conductivity could be attributed to a certain quality of the a-SiO

X

:H matrix. Figure

3.9 shows another BF TEM image, this time from the planar view. Dark spots indicate Si nanocrystals. Interesting to note is that the nanocrystals appear only in the gray cloud-like areas representing a-Si:H tissue.

This shows that all crystalline Si is encapsulated in a-Si:H. This a-Si:H as well as the a-SiO

X

3.4.3).

:H has been studied in more detail by FTIR (section

It is assumed that the transversal conductivity is similar or higher than the measured lateral conductivity. This is especially true for layers thinner than 20 nm. Using these typical thicknesses in devices, the shunting of tandem cells due to high lateral conductivity of the ZnO:Al IR cannot be observed. The transversal conductivity lacks a reliable direct measurement method. An indirect way of determining the transversal conductivity is by integrating the IR into a complete solar cell and measuring the series resistance ( R

R

S

S

). This R

S

of a reference cell without the IR. If

R

S

is then compared to the

does not increase, it can be stated that the transversal conductivity is 10

-5

S/cm or higher, as this is the value of transversal conductivity of the absorber layers.

Transmission Electron Microscopy (TEM)

Transmission electron microscopy (TEM) is a well-established technique for investigating the structure of materials on an atomic scale. In nanocrystalline materials, TEM can reveal the phases that constitute the film. It is especially useful in recognizing the size, orientation, and distribution of crystalline silicon grains embedded in amorphous tissue. Electrons coming from an illumination source are concentrated onto the specimen via a condenser lenses (Fig.

3.c). The transmitted electrons are then guided through another series of lenses and projected onto a fluorescent screen where they make up the image by contrast monitoring. Cross sections of samples on glass were prepared by gluing a protective glass on top of the deposited sample. In the next step, 2 mm thin slices were cut from the stack, followed by mechanical polishing and ion milling them up to about 1 μ m thickness (electron transparency).

43

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

TEM can operate in several modes. The most common mode of operation is the bright field (BF) imaging mode. In BF mode, an aperture is placed in the back focal plane of the objective lens which allows only the direct beam to pass. The image results from a weaker direct beam due to its interaction with the sample. Therefore, massthickness and diffraction contrast contribute to image formation.

In BF mode, thick areas, areas with heavy atoms, and crystalline areas appear with dark contrast. In the high resolution (HR) TEM mode, images are formed from multiple diffracted beams, allowing to construct an image of a crystal lattice. Energy filtered (EF) TEM mode uses a filter to form images from inelastically-scattered electrons. By imaging using electrons over selected energy ranges above and below ionization edges, elemental maps can be made.

Finally, selected-area diffraction (SAD) TEM imaging patterns are a projection of the reciprocal crystal lattice, with lattice reflections showing as sharp diffraction spots. A diffraction pattern is made under broad, parallel electron illumination with an aperture in the image plane. It is used to select the diffracted region of the specimen, giving site-selective diffraction analysis. SAD patterns can be used to identify crystal structures and measure lattice parameters like crystal orientation. SAD of nanocrystals gives ring patterns and can be used to discriminate nanocrystalline from amorphous phases.

TEM images of the n-nc-SiO

X

:H were taken with a JEM-

2100F setup operating at 200 kV. The p-doped nc-SiO

X

:H were taken in a FEI-Tecnai F30 with incident electron energy of 300 kV. The EF TEM imaging was acquired by a Tridiem Gatan energy filter with a slit width corresponding to 4 eV.

Figure 3.c: Schematic depiction of a TEM setup.

44

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.10: The lateral conductivity vs. the difference in the I

521

/I

480

I

521

/I

480

ratio of n- (a) and p-doped (b) nc-SiO

X

:H. A big

ratio starting above 0.2 for n-doped and above 0.1 for p-doped material can be observed for similar values of conductivity.

3.4.2 Raman spectroscopy analysis

The nc-SiO

X

:H samples have been characterized by Raman spectroscopy to determine the size and abundance of the crystalline silicon grains. Raman spectroscopy is sensitive only to bonding configurations with a center of symmetry, meaning it detects only Si-Si bonds. In pure

Si:H alloys, the crystallinity can be determined from Raman spectroscopy measurements using the ratio between the amorphous 480 cm -1 peak and the crystalline 521 cm -1 peak [65]. However, that is not possible in nc-

SiO

X

:H as it contains a substantial amount of oxygen and small amounts of phosphorus/boron and carbon (more details from XPS measurements in section 3.4.4). Therefore, as a measure of the increasing crystalline phase, the ratio of the integrated area of the fitted Gaussian peaks at

521 cm

-1

(crystalline silicon) and 480 cm

-1

(amorphous silicon) from the

Raman spectrum is used ( I

521

/I

480

).

Conductivity was found to be closely linked with crystallinity (Fig.

3.10). The conductivity changes significantly with the varying

I

521 ratio. Values of over 1 S/cm have been measured for nc-SiO

X

It has been observed that I

521

/I

480

/I

480

:H films.

>20% and >10% is crucial for good conductivity for the n- and p-doped material respectively. However, the device grade materials were above this threshold, with 37% and 22% for the n- and p-doped nc-SiO

X

:H respectively. It is not easy to compare the two materials as their nanostructure is not uniform with thickness. Each series had differences in thicknesses and eventually, the offsets might correspond to the thickness. The p-doped nc-SiO

X

:H with its relatively low crystallinity has the advantage of a relatively high deposition rate

(~2.75 nm/min) while still being highly conductive. This indicates that the amorphous phase is mainly responsible for good conductivity. The

45

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.11: The exact Raman crystalline peak position (ω

521

(b) nc-SiO

X

) vs. the I

521

/I

480

ratio in n- (a) and p-doped

:H. The x-axis indicates the approximation of crystalline Si content.

nanostructure of this amorphous silicon oxide matrix will be studied in the next part (part 3.3.3).

The Raman peak positions were studied in more detail. The typical

Raman peak positioned at 521 cm

-1

in bulk c-Si, reflecting a crystalline

Si matrix, is shifted to lower wavenumbers in nc-SiO

X ratio. In crystalline nc-SiO

X

:H material. Figure

3.11 depicts the crystalline peak position with respect to the I

521

/I

480

:H material, the crystalline peak is shifted from 520 cm

-1

down to 516 cm

-1

for p-doped nc-SiO

X

:H films but only

:H (Fig. 3.11). For the from 520 cm

-1

to 518 cm

-1

in n-doped nc-SiO

X device grade n-doped nc-SiO

X

:H material the Raman crystalline peak was found between 519 and 519,4 cm -1 and around 519 cm -1 for the p-doped nc-SiO

X

:H. From all the series studied, the only clear trend was found in n-doped nc-SiO

X

:H samples deposited in the pressure and CO

2 flow series. For these samples the peak shifts from 519 cm -1 down to

518,3 cm -1 and from 519,7 cm -1 to 519,2 cm -1 respectively (Fig. 3.12).

There are two possible explanations as to why this shift occurs:

1. Longer Si-Si bond lengths in reference to c-Si indicate the presence of tensile stress in the material.

2. It is a signature of quantum confinement effects in silicon crystal grains, which start to play a role when the grain size is below ~7 nm.

The crystalline grain size is decreasing as the peak shifts to lower wavenumbers [67].

The first option can occurs when the material is substantially heated.

This is not the case as the power of the laser used in the Raman setup was set to only 5% of its maximum and the thermal conductivity is too high to locally induce a significant temperature increase. Thus the peak

46

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.12: Raman silicon crystalline peak position overview of the pressure (black squares) and

CO

2

(red circles) deposition series of n-doped nc-SiO wavenumbers with decreasing deposition parameter. The material is more crystalline with a lower value of each deposition parameter. The 0.5 mbar condition is an amorphous sample and therefore is not shown.

X

:H. In both series the peak is shifting to higher

Figure 3.13: Lateral conductivity vs. ω

521 crystalline grain size.

in p-doped nc-SiO

X

:H. The x-axis can be interpreted as the shift is attributed to the quantum confinement effects coming from a crystalline grain size of about 7 nm. This conclusion is also supported by the TEM images in part 3.3.1. With the increase in the I

521

/I

480

ratio, the Si crystalline peak becomes smaller (Fig. 3.11). This is valid for

47

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION both n- and p-doped nc-SiO

X

:H. Therefore, it can be stated that with a larger crystal grain size comes a higher I

521

/I

480

ratio and vice-versa.

There is no clear conclusion on the relation between the crystalline peak shift (ω

521

) and the lateral conductivity. As high values of conductivity are observed in nc-SiO

X

:H films with a wide range of peak shifts, the grain size does not seem to play a crucial role in the conductivity of the nc-SiO

X

:H material (Fig. 3.13). This further supports the importance of the amorphous tissue in determining conductive properties in nc-SiO

X

Several p-doped nc-SiO

X

:H.

:H films have also been characterized by

X-ray diffraction (XRD) spectroscopy. XRD is able to detect crystalline yield any signature of crystal grain orientation and the XRD spectra resembled that of a standard silicon-oxide based glass. It appears that

XRD is not sensitive to reveal orientations of the crystalline grains in most nc-SiO

X

:H samples. The most likely reason for this is that the density of crystalline grains is low and they are very small, as demonstrated by the significant shifts of the 521 cm -1 peak observed in the Raman spectra and from the TEM images (Fig. 3.6 and 3.9). Therefore TEM imaging is the preferred method to study crystal grain orientation in thin nc-SiO

X films.

:H

3.4.3 Fourier transform infrared spectroscopy analysis

As suggested by the results of the TEM imaging and Raman spectroscopy measurements, the a-SiO

X in the charge carrier diffusion of nc-SiO

X silicon grains, the a-SiO

X

:H tissue plays an important role

:H. In contrast to the crystalline

:H tissue has many silicon-oxide and siliconhydride bonds, making Fourier transform infrared (FTIR) spectroscopy an ideal tool to study its nanostructure.

From the measured FTIR spectra, the focus was on extracting oxygen related modes and fitting them with Gaussians. Using this routine, relative concentrations of various configurations of oxygen incorporation in the material could be assessed. The integrated absorption of these Gaussian modes is calculated by equation:

(3.4) where I is the integrated absorption, α is the absorption coefficient, and

ω is the wavenumber. In the amorphous phase of the material, three absorption regions can be distinguished: 600-900 cm -1 , due to Si-H

X wagging and bending, 1000-1200 cm

-1

, due to the stretching of oxygen

48

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.14: Gaussian fits (grey thin lines) of FTIR scans (thick red line) for two n-doped nc-SiO materials (a) and the device grade p-doped nc-SiO deposited with a H

2

:SiH

4

X

X

:H (b). The n-doped nc-SiO nanocrystalline sample deposited with a pressure of 2 mbar ((a) top) and an amorphous sample

ratio of 0 ((a) bottom).

X

:H

:H materials are a double bonded by silicon atoms (Si-O-Si, Fig. 3.14 left), and 2000-

2300 cm

-1

, due to stretching of Si-H bonds with and without oxygen back bonded (O

X

Si-H, x=0-3, Fig. 3.14 right). Several of the oxygen related modes have been identified and assigned to specific bonding configurations in various reports [68]–[70], but the origin of all of them has not been unambiguously revealed. The bending and wagging modes (600-900 cm -1 ) will not be mentioned further as they do not give detailed information on the configuration of oxygen incorporation in the material. The partial contributions X i

of the oxygen related stretching modes from each region of the measured spectra are extracted. The partial contributions of the Si-O-Si stretching modes found at 1050 and

1135 cm

-1

are defined by:

(3.5)

49

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

(3.6)

The partial contributions of the Si-H stretching modes found at 2100,

2140, 2180, and 2250 cm

-1

are defined by:

(3.7) for i=2100, 2140, 2180, 2250. It has to be added that the measured spectra were fitted with a minimum number of Gaussians. An overview of the current status of insights in a selection of silicon-oxide and siliconhydrides modes (but not complete) is presented in Table 2.1.

Before analyzing each deposition series, it should be noted that variation of the PH

3

flow had an insignificant effect on the incorporation of oxygen in the material as reflected by the measured FTIR spectra.

The PH

3

flow only affects the doping. It also has to be noted that the presence of carbon related stretching modes at 740-820 cm -1 shows that some carbon is incorporated in the nc-SiO

X

:H. However, the variations in these modes are insignificantly small. This makes it difficult to make any statements on the dependence of the carbon incorporation on processing conditions.

Some general correlations were observed between the various modes.

Films showing large contributions to the (Si-O-Si)O

X

O

3

Si:H (2250 cm

-1

(1135 cm

-1

) and

) stretching modes (SM) are expected to have more oxygen content and therefore a higher bandgap. It was observed that when

(Fig. 3.14(a)). On the other hand, when X

1050 over

X

X

1135

1135

>

X

1050

the

X

2250

was more dominant over

X then the fractions X

2100

and X

2250

2100

,

X

2140

, and

X

2180

is found to be dominant

are in the same order (Fig.

3.14(b)). The examples in Figures 3.14(a) and (b) were taken from the n-doped and p-doped nc-SiO

X between X

1050

and X

1135

:H respectively, but the general relations have been observed for all samples. From this it is

reflects the silicon-oxygen-silicon stretching mode concluded that X

1050 in a silicon rich environment (Si-O-Si)Si

X

and X

1135 reflects the siliconoxygen-silicon stretching mode in an oxide rich environment (Si-O-Si)

O

X

. This conclusion is in line with the conclusion based on various silicon oxide phases observed in reference [68].

The nanostructure of the a-SiO

X

:H tissue is of great interest. It is found that the partial contribution

X i

of each SM in the FTIR spectra is strongly dependent on the deposition conditions and that the relation

50

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.15: Changes in the partial contributions of different stretching modes in n-doped nc-SiO

X as a function of pressure. The first sample at 0.5 mbar is completely amorphous.

:H between nanostructure and deposition conditions is complex. From all the deposited series, the above mentioned three series of the n-doped nc-SiO

X

:H are presented as they lead to the clearest trends. The dependence on the processing pressure is analyzed first. In Figure 3.15, the partial contributions of various modes are shown as a function of pressure. The sample deposited at the lowest pressure of 0.5 mbar is amorphous, whereas the samples deposited at pressures >0.5 mbar exhibit a crystalline Si peak at around 517 cm

-1

in the Raman spectra.

It can be seen that X

2100

increases with pressure. The 2100 cm -1 mode corresponds to the hydrides residing at the surfaces of nano-sized voids in pure a-Si:H tissue (Fig. 3.12) [71]. Consequently, an increase in X

2100 indicates a relative increase in a-Si:H-like tissue exhibiting nanometersized voids [71]. This trend is accompanied by the reduced contribution of the 2140 cm -1 mode. The origin of this mode is still under discussion.

It is believed that in the far majority of the samples studied in this thesis, the 2140 cm

-1

corresponds to the (OSi

2 contribution to the Si-H

3

)Si-H mode. However, the

mode cannot be completely ruled out [72].

Note that the oxygen only resides in the a-SiO

X crystalline silicon grains.

:H tissue and not in the

From the measured modes a slight difference can be seen between the nanostructure of the a-SiO

X

:H matrix in the pure amorphous films processed at 0.5 mbar and the a-SiO

X in the device grade nc-SiO

X

:H tissue between the c-Si grains

:H (1.25 mbar). The amorphous tissue in samples with pressures rising above 1.25 mbar has less oxygen

51

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

H H

O

1000-1200 cm -1

O

Si

H

Si

Si

Si

Si

H

Si

H

Si

(void)

O

H

Si

H

Si

O

O

2250 cm

H

O

Si

2140 cm

-1

O

Si

Si

2180 cm

-1

-1

Si

2100 cm

-1

Si

Figure 3.16: Schematic of hydrogen stretching modes with and without back-bonded oxygen at a void surface.

incorporated which can be seen on the reduced values of X

1135 and X

2250

(Fig. 3.14(a) top). For pure a-Si:H material it is well known that a low

2100 cm

-1

mode is desired as it indicates higher quality and denser a-Si:H material with less nano-sized voids. These surfaces at nanosized voids are the locations of many electronic active defects [73]. If this trend is extrapolated to the case of the a-SiO

X

:H tissue, material with fewer voids would have less defects leading to an improved conductivity.

The results shown in Figure 3.15 therefore suggest that the conductivity is determined by a competition between the quality of the a-SiO

X tissue and the incorporation of crystalline grains. If the nc-SiO

X

:H

:H is processed at higher pressures, there is less oxygen in the matrix. The amorphous phase is less dense due to the high content of porous a-Si:H tissue represented by X

2100

. Therefore, an as low as possible pressure is desired for high quality a-SiO

X

:H tissue but still high enough to guarantee crystalline grain formation. The pressure for the best performing nc-

SiO

X

:H material has been found to be at 1.25-1.5 mbar, conditions at which the material exhibits crystalline grains.

According to the FTIR analysis a significant amount of oxygen is incorporated in the a-SiO

X

:H tissue of the device grade material. The incorporated oxygen guarantees good optical properties as a reflective layer due to the lower refractive index of the nc-SiO

X

:H. Good electrical properties are guaranteed with a minimal contribution of X

2100

meaning as small as possible incorporation of voids into the tissue. By increasing the CO

2

:SiH

4

ratio from 1 to 1.6 the variation in

X

2180 small (Fig. 3.17). However, X

2140

and

X

2250

is very

increases, suggesting an increase in

52

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.17: Changes in the partial contributions of different stretching modes in n-doped nc-SiO as a function of the CO

2

:SiH

4

ratio. Device grade material (CO

2 contribution of the hydrogen SMs (2100-2250 cm

-1

).

:SiH

4

X

:H

= 1.6) has a comparable partial

OSi-H bonds in the material. From CO an increase in X

1135 and X

2250

2

:SiH

4

ratios of 1.6-2.2 there is

, further supporting larger oxygen contents

(Fig. 3.17). Again it shows that a trade off exists between the optical and electrical properties and its relation to the a-Si:H fraction.

The hydrogen dilution ratio H

2

:SiH

:SiH

4

has also been found to strongly influence the oxygen content (Fig. 3.19). The sample deposited without H

2

(H

2 4

=0) resulted in a fully amorphous material with very high oxygen content.

X

1135

are significantly larger than in any of the other and

X

2250 processed films and the material is very close to a-SiO

2 bottom, Fig. 3.1(a), Fig. 3.19). With increasing the H

2

(Fig. 3.14(a)

ratio above 50, the material is starting to become nanocrystalline (Fig. 3.1(b),(d), Fig. 3.19).

With the H

2

ratio above 150, the material again becomes amorphous but with a high contribution of pure a-Si:H tissue (Fig. 3.1(c)), indicated by a high X

2100

X

1050

and X

(Fig. 3.19). For the a-SiO

X

2100

:H tissue between the c-Si grains,

in the infrared spectra become more dominant (Fig. 3.18).

This shows a reduction in oxygen and higher contribution of pure a-Si:H tissue. For flow ratios H

2

:SiH

4

>50 the oxygen content in the material does not show any significant variation, the samples have the same contributions of the corresponding oxygen modes (Fig. 3.18). The device grade material deposited at H

2

:SiH

4

=100 is comparable. This suggests that the even distribution of OSi-H, O

2

Si-H, and O

3

Si-H bonds in the amorphous tissue can be used as a signature of optimum nanostructure for the n-doped nc-SiO

X

:H.

53

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.18: Changes in the partial contributions of different stretching modes in n-doped nc-SiO as a function of the H

(H

2

:SiH

4

2

:SiH

4

ratio. The first sample at H

2

:SiH

4

=100) has a comparable partial contribution of the hydrogen SMs (2100-2250 cm

-1

).

X

:H

=0 is completely amorphous. Letters

(a), (b), (c), and (d) correspond to the nanostructure cases in Figure 3.1(a-d). Device grade material

To get more insight into the relation between nanostructure and optical and electrical properties, four measured parameters were chosen to be compared with the

X i

values. These are two optical and two electrical parameters which most significantly characterize the nc-SiO

X

:H for its use as a reflective layer: the bandgap, refractive index, activation energy and conductivity. Only the p-doped nc-SiO

X

:H samples are used because more samples were deposited than in case of the n-doped material. This results in acceptable statistics. Measured data is from different materials deposited under the various conditions presented in Table 3.1. It should be addressed that the plots show a wide sample range, however, still some clear trends between

X i

and the bandgap, refractive index, conductivity and activation energy can be observed.

The bandgap is mainly determined by the ratio between silicon and oxygen in the material. First, the SM dependence on the bandgap is discussed. In Figure 3.19, the partial contributions of various modes are shown as a function of the bandgap. It can be seen that the contribution of X

2100

decreases with the increase in bandgap. X

2100

corresponds to the hydrides residing at the surfaces of nano-sized voids in pure a-Si:H tissue [71]. Consequently, a decrease in

X

2100

indicates a relative decrease in a-Si:H content. This implies a more dominant contribution of the higher bandgap a-SiO

X

:H phase. This trend is accompanied by an

54

3.4 NANOSTRUCTURE ANALYSIS increase in X

1050

, X

1135

, and X

2250

while X

2140

and X

2180

remain unchanged.

This indicates that there is more oxygen present in the bulky form of

Si-O-Si bonds in a Si/O matrix than being a back-bonded O atom to a silicon hydride (Si-H) (Fig. 3.12). In a-Si:H material, the presence of voids increases the bandgap. However, in this case, the oxygen atoms play a greater role in the amorphous silicon oxide matrix in increasing the bandgap than the porous a-Si:H.

X

1106

(O

2i

is attributed to interstitial oxygen

) between the Si-O-Si rings and its presence is detrimental for a high bandgap. Clear trends of this mode were not observed.

In case of the relation between refractive index and the SMs (Fig.

3.20), most

X i

Most X i

do not show a clear correlation with the refractive index.

seem to be mostly independent on the measured parameter, except X

1135

, which is most strongly related to oxygen content. This confirms the trends observed in case of the bandgap, as a lower refractive index is connected with a more transparent material (higher bandgap).

The silicon related 2100 cm

-1

mode is increasing sharply with higher refractive index as expected, further confirming the role of oxygen in the optical properties. It is worth to note that n, measured at 600 nm, for device grade material is among the higher values ( n=2.7). It can be stated that the amorphous phase itself is heterogeneous and looks more like the nanostructure illustrated in Figure 3.1(c) than Figure 3.1(a).

Conductivity shows interesting correlations with X i

(Fig. 3.21). All the SM related peaks are decreasing with increasing conductivity except

X

2100

which is increasing. X

1050 dominant. High X

2100

is decreasing similarly as X

1135

, but stays

contributions mean high a-Si:H content. Therefore high lateral conductivity is found in samples with high a-Si:H content.

This confirms the hypothesis based on the TEM analysis (Fig. 3.8, 3.9), as the a-Si:H tissue serves as an encapsulant for the crystalline grains and fills the majority of the transversal Si filaments (Fig. 3.1(d)). The decreasing oxygen modes with increasing conductivity indicate that the a-SiO

X

:H tissue has no influence on lateral conductivity; only more a-Si:H tissue shows a positive influence.

The partial contributions of the SMs are related to activation energy

(Fig. 3.22) as well. In case that activation energy is increasing, all X i are rising except

X

2140

. This can be interpreted as low oxygen content is beneficial for low activation energy. The a-Si:H related 2100 cm

-1

mode is increasing with increasing E a

X

2100

as well. However, in the total contribution

still remains the highest, meaning that low activation energies are still dominated by the a-Si:H tissue. The

X

1106

contribution in both

E a

and conductivity is small (Figure 3.21, 3.22). From the results it is visible that the 1106 cm

-1

mode does not play a crucial role in good optical nor electrical performance of the material.

55

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.19: Partial contributions of fitted Gaussian peaks to each p-doped nc-SiO

X spectrum vs. their bandgap. The green zone represents device grade material.

:H sample’s scanned

56

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.20: Partial contributions of fitted Gaussian peaks to each p-doped nc-SiO

X

:H sample’s scanned spectrum vs. their refractive index. The green zone represents device grade material.

57

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.21: Partial contributions of fitted Gaussian peaks to each p-doped nc-SiO

X spectrum vs. their conductivity. The green zone represents device grade material.

:H sample’s scanned

58

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.22: Partial contributions of fitted Gaussian peaks to each p-doped nc-SiO

X

:H sample’s scanned spectrum vs. their activation energy. The green zone represents device grade material.

59

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

X-ray photoelectron spectroscopy (XPS)

X-ray photoelectron spectroscopy (XPS) is a technique used to determine the elemental composition of a material. The principle of the method is based on irradiating the material with x-rays and detecting the out-coming electrons (Fig. 3.d). The incoming photons ionize the target atoms: m + hν → m

+*

+ e

-

(3.8) where m is an atom in the material. An incident highly energetic x-ray photon causes the ejection of an electron with a discrete kinetic energy, which is measured in XPS. The kinetic energy of the ejected electron is essentially the difference between the energy of the incident photon and the binding energy of the electron to the atom. Atoms of different elements have different characteristic electron-binding energies. This means that the ejected electron is a direct indicator of the presence of a certain element.

The penetration depth of the x-ray beam is usually a few nm.

XPS cannot detect hydrogen or helium because of their very small photoelectron cross-section, being too small for photoemission.

The detection precision is from 1% (light element) to 0.1% (heavy element) of the surface layer. The measurement method requires ultra-high vacuum conditions. XPS is usually combined with ion sputtering to remove the impure (oxidized) initial surface layer to guarantee the probing of the real bulk material of interest.

Energy Analyzer

Collector

Slit

Spectrophotometer

Electron Multiplier

Photoelectrons

Sample

Figure 3.d: Schematic depiction of a XPS setup.

60

Lens System

Source Slit

X-ray Source

3.4 NANOSTRUCTURE ANALYSIS

Table 3.5: Partial contributions of each element to the composition of films from different series of n-doped nc-SiO

X

:H given by XPS measurements.

Sample

0.5 mbar

2.75 mbar

H

2

:SiH

4

=0

H

2

:SiH

4

=150

CO

2

:SiH

4

=1

CO

2

:SiH

4

=1.6

CO

2

:SiH

4

=2.2

SiH

4

:PH

3

=25

SiH

4

:PH

3

=125

Si

43.6%

66.0%

32.0%

51.6%

61.5%

55.9%

50.6%

51.0%

54.6%

Elements

O

53.9%

C

2.4%

29.9%

66.0%

46.1%

36.4%

41.4%

47.1%

46.2%

43.1%

1.8%

1.8%

2.0%

1.5%

1.8%

1.7%

2.0%

2.2%

P

0.1%

2.3%

0.2%

0.3%

0.7%

0.9%

0.6%

0.8%

0.1%

Table 3.6: Partial contributions of each element to the composition of films from different series of p-doped nc-SiO

X

:H given by XPS measurements. The boron content was insignificantly small, therefore is neglected form this table. * 130 H

2

flow

Sample

60 W*

CO

2

:SiH

4

=1.25*

CO

2

:SiH

4

=1.75*

CO

2

:SiH

4

=2.25*

CO

2

:SiH

4

=1.17

CO

2

:SiH

4

=3.5

H

2

:SiH

4

=0

H

2

:SiH

4

=213

H

2

:SiH

4

=250

SiH

4

:B

2

H

6

=700

SiH

4

:B

2

H

6

=250

1.4 mbar

3 mbar

Si

69.1%

70.2%

69.8%

73.7%

66.9%

69.9%

68.4%

77.3%

78.5%

71.6%

65.4%

77.6%

62.3%

70.6%

70.9%

Elements

O

24.0%

21.0%

21.6%

18.9%

21.9%

19.5%

24.1%

15.9%

11.2%

20.2%

27.3%

12.8%

32.9%

21.3%

20.8%

C

3.9%

2.6%

3.1%

3.1%

5.6%

4.2%

2.7%

2.2%

2.5%

3.0%

3.9%

2.7%

2.8%

3.7%

3.1%

61

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

3.4.4 X-ray photoelectron spectroscopy comparison

FTIR can give absolute concentrations of elements when the Si alloy contains only two elements like in the case of a-Si:H. However, for complex alloys like nc-SiO

X

:H it is not possible to measure the absolute elemental concentration using FTIR. For this reason, several nc-SiO

X

:H films have also been characterized by XPS. XPS is able to determine the relative content of elements making up the film. For this experiment, samples with the most extreme nanostructure properties of each series were chosen (Table 3.1, 3.2). The results are shown in Tables 3.5 and

3.6. It can be observed that there is high oxygen content for both n- and p-doped nc-SiO

X

:H with a high CO

2

flow and low pressure. The n-type material in general has higher oxygen content, reaching the stoichiometric value for the material processed without hydrogen dilution. The variation of PH

3

flow has the smallest influence on the oxygen content. As for the doping atoms, only phosphorus is shown as its concentration is high enough to be detected using XPS. In contrast, the percentage of boron content was always lower than the detection limit of XPS.

The sample deposited at conditions that lack hydrogen dilution has the largest oxygen content and is completely amorphous. The FTIR scan of this sample is shown in Figure 3.14(a bottom). The material has a high X

1135

, therefore is dominated by (Si-O-Si). Most of the hydrogen contributing to a SM is bonded to a Si atom with three back-bonded O atoms (2250 cm -1 mode).

The effects of the deposition conditions on the oxygen content are summarized. In contrast to the n-type material, the hydrogen flow does not have such a big influence in the p-type material. As for the dopants, in the p-type material, increased B

2

H

6

flow is reducing oxygen content while in the n-type material, increased PH

3

is increasing oxygen content.

Variation of the temperature has been investigated only for the p-type material and shows increasing oxygen content with lower temperature.

Variation of the CO

2

flow makes a bigger difference in the oxygen content in the p-doped material than in the n-doped.

The CO

2

:SiH

4

dilution can be controlled either by changing the CO

2 flow or the silane flow. It has to be noted that the absolute extremes of the CO

2

series (CO

2

:SiH

4

=1.17 and 3.5) were deposited by varying the silane flow, not the CO

2

flow. Since the CO

2

flow controller has a

flow has been varied to larger values of limited operating range, the SiH

4

CO

2

:SiH

4

dilution. The flows of the other gases were kept constant, which could have led to a slight increase in errors in the results, as the ratios of these gases were also varying with the silane flow.

Another interesting observation is that the carbon content is higher in

62

3.4 NANOSTRUCTURE ANALYSIS the p-doped material than in the n-doped. The origin of this difference in carbon incorporation is for the moment not clear, but it can be speculated that it is due to the boron presence. From the XPS measurements the observed phosphorus content is also very low and does not clearly depend on any of the deposition parameters besides pressure. With higher pressure the phosphorus content is more than double that in the device grade material (CO

2

:SiH

4

=1.6 sample, Table 3.5). Therefore at this point it is not possible to draw any conclusions on neither its incorporation mechanism nor the fraction of active p-dopants.

FTIR gives information on oxygen and silicon content, but those are only relative contents. Therefore XPS measurements were conducted as it gives the absolute elemental content of the material. The comparison between FTIR and XPS is presented. The XPS results (Table 3.5, 3.6) correlate with the partial contributions of the stretching modes assigned to silicon and oxygen bonds from the FTIR measurements (Fig. 3.15,

3.17, 3.18). It can be seen that the samples with higher elemental oxygen content have higher partial contributions of oxygen related modes (1050,

1135, 2140, 2180, and 2250 cm

-1

) and samples with higher silicon content have higher partial contributions of silicon related mode (2100 cm -1 ). To have a direct comparison, the integrated areas of the Gaussian peak intensities measured with FTIR for the oxygen (1050 and 1135 cm -

1

) and hydrogen (2100-2250 cm

-1

) stretching modes were compared with the oxygen to silicon content ratio. This relative oxygen content to silicon was obtained through the following equation:

C

O

=

O

X

+

O

X X

Si

(3.9) where

X

O

and

X

Si

are the fractions of oxygen and silicon in the material respectively, taken from XPS measurements (Table 3.5, 3.6). This equation was chosen to eliminate the influence of the carbon and doping atoms in this study. Hydrogen atoms are not considered, as they are not detected by XPS. Figures 3.23 and 3.24 show

X i

versus

C

O

. In the n-doped material (Fig. 3.23), the partial contributions of all the oxygenrelated modes are rising as the oxygen fraction is increasing. On the contrary, X

2100

which is related to a-Si:H tissue, is decreasing. This can as well be said about the p-doped material (Fig. 3.24), except that

X

1106 is falling with higher oxygen content which seems a contradiction as it is attributed to interstitial oxygen. The device grade nc-SiO

X was calculated to have C

O

:H material

= 22.5% (Fig. 3.24). These results validate

FTIR as a reliable and simpler method of obtaining material composition.

63

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

Figure 3.23: The integrated peak intensity from Gaussian fitted FTIR spectra from n-doped nc-SiO oxygen (1050 and 1135 cm

-1

) and hydrogen (2100-2250 cm

-1

) SMs vs. C

X

:H for

O

calculated with Equation 3.9.

64

3.4 NANOSTRUCTURE ANALYSIS

Figure 3.24: The partial contributions from Gaussian fitted FTIR spectra from p-doped nc-SiO

X for oxygen (1050, 1106, and 1135 cm

-1

) and hydrogen (2100-2250 cm

Equation 3.9. The green zone represents device grade material.

-1

) SMs vs. C

O

:H

calculated with

65

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

From the XPS and FTIR results a simple model for predicting the fraction of Si and O bonds is proposed. This model is based on random distribution of Si-Si bonds without oxygen present and Si-O-Si bonds with oxygen present. The total number of bonds is 2 N

Si

. N

Si

and N

O represent the silicon and oxygen bond density respectively. Assuming that only silicon and oxygen make up the material, the oxygen content is defined by:

(3.10)

Rearranging the equation gives the silicon to oxygen density ratio:

(3.11)

(3.12)

Then the fraction of pure Si-Si bonds is calculated

(3.13)

Substituting N

O oxygen content:

and N

Si

gives the Si-Si bond relation to the relative

(3.14)

Pure Si-Si bonds have an inverted dependence with growing oxygen content, reaching 0 at pure stoichiometric material where Si-O-Si bonds dominate the material (Fig. 3.25). Similarly, the fraction of Si-O-Si bonds in a silicon rich environment is found:

(3.15)

From the experimental results, it can be seen that with low relative oxygen content (

C

O

< 0,4), most of the oxygen is found in Si-O-Si bonds in a silicon-rich environment represented by the 1050 cm

-1

mode

66

3.5 DISCUSSION AND CONCLUSION

Figure 3.25: Model based on random distribution of Si-Si bonds without oxygen present (red) and

Si-O-Si bonds with oxygen present (blue circles and green stars, for the 1050 cm

-1

and 1135 cm

-1

SM respectively). The fraction of these bonds and its trend with increasing oxygen content in the material is shown.

(Fig. 3.25). For C

O

> 0,4, the Si-O-Si bonds are more surrounded by oxygen, giving rise to the 1135 cm

-1

mode (O

X

(Si-O-Si)O

Y

). The measured results in first approximation are below the maximum Si-O-Si possible in silicon oxide material (represented by the black line in Figure

3.25). To conclude, the correlation of XPS and FTIR measurements with the model presents FTIR as a quick and reliable method in estimating the oxygen content in nc-SiO

X

:H layers and its usefulness as a tool to engineer device quality material without the need to apply XPS.

3.5 Discussion and Conclusion

In summary, nc-SiO

X

:H layers have been developed and characterized.

The nanostructure of nc-SiO

X

:H films with a wide array of optical and electrical properties has been studied in detail by TEM, Raman, FTIR and XPS. The relations and correlations between these measurement methods and the influence on the deposition and measured parameters were analyzed. The main findings are discussed in detail in this part.

How much is the conductivity influenced by the nanostructure? And what kind of nanostructure is optimal for high conductivity? Conductivity was found to greatly depend on the nanostructure. The purely amorphous

SiO

X

:H films showed poor conductivity (Fig. 3.10). Therefore, it is concluded that a minimal crystalline volume fraction, here expressed in

67

(a)

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION

(b) a-Si :H c-Si a-SiO :H c-Si a-SiO :H

Figure 3.26: Cross-section depiction of the nanostructure of n-doped (a) and p-doped (b) nc-SiO

X

:H. terms of the I

521

/I

480

ratio (above 20% and 10% for n- and p-doped material respectively), is required for reasonable conductivity. This is true even if the crystalline volume fraction is relatively low. However, for films exhibiting a low crystalline volume fraction, conductivity has to rely heavily on the amorphous phase as well (Fig. 3.26). In case of the p-doped nc-SiO

X

:H, the presence of crystalline grains in the a-Si:H filaments is an effect of a-Si:H filament formation, as a-Si:H grows from the substrate since the start of deposition (Fig. 3.8). This makes a-Si:H mainly responsible for conductivity (Fig. 3.26(b), 3.27(a)). It appears that for the crystalline silicon phase to grow it needs to be separated from the silicon oxide phase (Fig. 3.9).

The amorphous phase providing good conductivity must have certain qualities. The pure a-SiO

X

:H samples have very poor conductivity. FTIR measurements in Figure 3.14(a) show the difference between wellperforming a-SiO

X

:H tissue (top) and non-performing tissue (bottom).

The strongly oxygen related modes (1135 and 2250 cm

-1

) appear to be good indicators for the amorphous tissue with poor conductive properties.

The general observation for well-performing device grade material is that

X

1050

>

X

1135 from the Si-O-Si modes and the hydrogen stretching modes between 2100-2250 cm

-1

have a comparable contribution or with a slight X

2100

3.14(b)).

dominance, indicating a higher a-Si:H tissue fraction (Fig.

To provide sufficient conductivity, a need for minimal crystalline content has been established. The Si crystal orientation was found to be random. It was observed that the crystalline peak position is rising to higher wavenumbers with higher crystalline silicon content, indicated by the

I

521

/I

480

ratio (Fig. 3.11). This implies that the more crystal grains

68

3.5 DISCUSSION AND CONCLUSION

Figure 3.27: The conduction mechanism in p-doped (a) and n-doped (b) nc-SiO

X represent the likely paths of charge carriers.

:H. The blue lines incorporated, the bigger they are in size. The crystalline peak is shifted from 520 cm -1 down to 516 cm -1 in p-doped nc-SiO

X

:H films (Fig. 3.11b) but only from 520 cm

-1

to 518 cm

-1

in n-doped nc-SiO

X

:H (Fig. 3.11a).

It can be stated that for the deposition series studied in this thesis, the crystalline grains in the n-type nc-SiO

X

:H are in general bigger than in the p-type. For the device grade n-doped nc-SiO

X

:H material the Raman crystalline peak was found between 519 and 519,4 cm

-1

and around 519 cm

-1

for the p-doped nc-SiO

X

:H. This shift is practically identical for both n- and p-doped material, indicating that there is an optimum size for the crystal grains in terms of conductivity. A shift to 519 cm -1 represents a grain size of approximately 5-8 nm [67]. That is in line with the grain size observed in the TEM images, where the grains appear as quantum dots (Fig. 3.6 and 3.9). The significance of the presence of crystal grains in the n-doped material is most probably not important. Still the majority of the path the charge carriers need to travel is through amorphous tissue

(Fig. 3.27(b)). It can be speculated that as well as in the p-type material, the c-Si grains could be a side effect of an amorphous phase formation with sufficient conductivity.

The influence of the crystalline peak shift on the material properties was analyzed. The peak shift is affected by the deposition parameters. For increasing pressure and CO

2

flow in the n-doped material, the crystalline peak shifts to lower wavenumbers (Fig. 3.12), but the conductivity, refractive index, and bandgap have opposite trends for the deposition parameter increase in these series (Fig. 3.2). Therefore, no conclusions can be drawn from the peak shift in relation to optical and electrical properties. It has been shown with TEM images that the reason for this

69

3. SILICON OXIDE NANOSTRUCTURE AND MATERIAL CHARACTERIZATION shift is because of smaller crystalline Si grain size and not stress in the material. Stress appears when the material is significantly heated. Due to the low power use (5%) of the laser in the Raman setup and the high thermal conductivity of the material, stress can be ruled out.

From the TEM and Raman analysis it has been established that the amorphous phase plays a crucial role in the conductivity of the material.

Therefore the quality of this amorphous phase was studied using FTIR. In all FTIR measured samples, correlations were found between the trends of certain stretching modes, which are summarized here. When

X

1135 decreasing, X

2140

and X

2250

The opposite is valid when X

1135 observed for varying nanostructure (Fig. 3.15, 3.17, 3.18). With H

2 flow variation only

X

2140

also decrease and X

1050

and X

2100 is

increase. is increasing. This correlation has been

shows an opposite trend in reference to

X

1135

(Fig. 3.18). This suggests that more than one type of bond stretching contributes to the 2140 cm

-1

stretching mode.

A-Si:H tissue has been detected in the material but its exact role has not been confirmed, especially in terms of conductivity. From pure a-Si:H material it is well known that a low 2100 cm

-1

mode is desired. A high 2100 cm

-1

mode indicates lower quality a-Si:H material with nanosized voids [71]. These surfaces at nano-sized voids are the locations of many defects. Therefore a dominance of the 2000 cm -1 SM is desired as it is a sign of high quality a-Si:H material [74], [75]. If this trend is extrapolated to the case of the a-SiO

X

:H tissue, material with fewer voids would have less defects leading to an improved conductivity. This would play an important role especially in the p-doped nc-SiO

X

:H, where a-Si:H is most likely mainly responsible for good conductivity in the vertical a-Si:H filaments. As there is practically no 2000 cm

-1

mode detected in the material (Fig. 3.14) but the 2100 cm -1 mode is largely present, it can be concluded that the void-rich a-Si:H provides sufficient conductivity even with the higher density of defects. The results shown in Figure 3.15 suggest that the conductivity is determined by a competition between the quality of the a-SiO

X

:H tissue and the incorporation of crystalline grains.

This is observed by the lowest values of

X

2100

to be near the transition between amorphous and nanocrystalline material.

The silicon filaments are vertical, therefore logically they should provide transversal conductivity. But why is the lateral conductivity high?

For layer thicknesses above 20 nm it was shown (Fig. 3.8(a)) that these filaments branch out and interconnect. This explains the measured high lateral conductivity for samples which had typical thicknesses of ~150 nm. The device grade transversal conductivity in p-doped nc-SiO

X

:H is attributed to the a-Si:H filaments, in which the c-Si grains inside play a minor role. Conductivity of the filaments could simply be based on

70

3.5 DISCUSSION AND CONCLUSION the fact that they are made of oxygen-free silicon alloys. Whether the silicon phase in the filaments is made of c-Si grains and/or a-Si:H tissue is most probably not crucial in achieving the desired conductivities. 3D

TEM holography images of the filaments could give more insight into this question and is recommended for further research.

The deposition parameters greatly affect the contributions of the SMs.

By increasing the CO

2 the variation in X

2180

:SiH

and

4

ratio from 1 to 1.6, E

X

2250 g

increases (Fig. 3.2) but

is very small (Fig. 3.17). However, X

2140 increases, suggesting that it is a signature of the increasing OSi-H bonds in the material. There is an increase in X

1135 larger oxygen contents resulting in higher E g and X

2250

, further supporting

and decrease in n. Varying

ratio has a direct impact on the amount of oxygen in the the CO

2

:SiH

4 plasma and together with the H

2

:SiH

4

ratio severely affects the oxygen content in the film. The lack of hydrogen for instance results in the material being completely amorphous, therefore containing more oxygen bonds instead of silicon crystals. The FTIR scan of this sample can be seen in Figure 3.14(a) bottom. The material is dominated by Si-O-Si bonds (1135 cm

-1

mode) and most of the small amount of hydrogen that is incorporated is bonded to a Si atom with three back-bonded O atoms

(2250 cm -1 mode). On the contrary, the incorporation of phosphorus atoms is more dependent on other deposition parameters, mainly pressure and H

2

. Pressure is a very critical deposition parameter as it not only influences the crystallinity and quality of the a-SiO

X

:H tissue (Fig. 3.15), but also the doping. In the deposition conditions explored in this thesis, the increase in PH

3

flow lowered conductivity (Fig. 3.2).

71

Parallel to the development and optimization of nc-SiO layers, the qualities of nc-SiO

X

X

:H as single

:H films were studied by integrating them in single junction a-Si:H cells. The device grade p-doped nc-SiO

X was tested as the p-layer and the device grade n-doped nc-SiO

X

:H

:H as the back reflector (BR). A BR functions as a reflector of the not absorbed light that passed through the absorber layer. This light is reflected back into the absorber layers for a second chance in being absorbed. For a layer to be a good reflector, it needs a low refractive index n in reference to the refractive index of the a-Si:H layer. If it is integrated in a cell it also needs to be conductive. Nc-SiO

X

:H fulfills these conditions and is a perfect reflector layer (section 3.3) in combination with a-Si:H. In both the p- and n-doped materials, an improvement in

J

SC

and solar cell efficiency was observed. For the n-doped material, which is serving as spectrum as expected, but over the entire spectrum. Therefore a current enhancement analysis is presented as well.

4. SOLAR CELL APPLICATION

4.1 Experimental details

The individual films of the solar cells were deposited using the

AMIGO RF-PECVD cluster tool (section 2.1.1). The a-Si:H cells were deposited onto Asahi-U (n-doped nc-SiO

X

(p-doped nc-SiO

X

:H series) and Asahi-VU

:H series) TCO (consisting of SnO

2

:F) substrates. For all layers, the bias frequency was 13.56 MHz and substrate temperature

C. For the n-doped nc-SiO

X

:H the cell configuration is a 10 nm a-SiC:H p-layer, 5 nm intrinsic a-SiC:H buffer layer, 300 nm non-

H

2

-diluted a-Si:H i-layer, 10 nm a-Si:H n-layer and the nc-SiO

X with a thickness of about 40 nm. For the p-doped nc-SiO

X

:H

:H series the cell configuration is the same except that the p-layer material is varied

(without buffer layer) at a thickness of about 20 nm and after the n-layer there is 40 nm of n-doped nc-SiO

X

:H. Different series of single junction cells were deposited with varying p- and n-doped nc-SiO

X

:H materials in ranges presented in Tables 3.1 and 3.2. Silver back contacts with an area size of 0.16 cm

2

were evaporated onto the cell (section 2.1.3). To guarantee well defined contact areas, the remaining nc-SiO

X

:H around the metal contacts was removed using reactive ion etching (section cells were characterized by J-V measurements (section 2.2.1) where short-circuit current density ( J

SC

) values are calibrated using external quantum efficiency (EQE) measurements without voltage bias (section

2.2.2). The average of the 10 best cells from each deposition was taken as the result. The device grade materials as defined in section 3.2 are performing the best in cells providing the highest efficiency. The device grade material deposition conditions can be found in Tables 3.1 and 3.2.

The reference cell for the n-doped nc-SiO

X

:H is a cell in which the n-layer only consists of n-doped a-Si:H. The reference cell for the p-doped nc-SiO

X

:H is a cell with the p-layer consisting of an a-SiC:H p-doped layer and an intrinsic a-SiC:H buffer layer. Only the device grade materials are presented in this chapter.

4.2 Results

4.2.1 N-doped nc-SiO

X

:H as a back reflector

The developed n-doped nc-SiO

X

:H layers are well suited as a BR with their low refractive index and high conductivity. Therefore they have been integrated as BRs into single junction a-Si:H cells. The performance of the cells was improved, with almost every type of nc-SiO

X

:H material developed and every tested thickness in reference to a cell without it.

74

4.2 RESULTS

Variation in n-doped nc-SiO

X layer thickness

:H

0-100nm

(a) (b)

Figure 4.1: (a) depiction of the BR in a single junction cell. (b) EQE of a-Si:H cells with different thicknesses of n-doped nc-SiO double a-Si:H n-layer thickness.

X

:H. All other layer thicknesses were kept constant. *This cell has a

An enhancement is observed over the entire spectrum (Figure 4.1(b)), a result of the apparent enhancement of electron collection, discussed in detail in the next section. The device grade material was used in a

BR thickness series from 0-100 nm (Figure 4.1(a)) to find the optimal thickness for the highest cell efficiency. A p-i-n junction with solely a

Ag BR is used as a reference solar cell (Figure 4.1(b)).

The best reflectance in the 550-700 nm range was found at thickness of nc-SiO

X

:H around 50 nm for the optimum nc-SiO

X

:H material (Figure

4.1(b)). It has to be noted that the cell without the nc-SiO

X

:H layer had a twice-as-thick standard a-Si n-layer in comparison to the cells with nc-SiO

X

:H. This will result in slightly enhanced parasitic absorption losses. The best cell had an initial efficiency of 9.34% ( V

OC

= 0.89 V,

= 14.17 mA/cm 2 , FF = 74.06%). The best performing 10 cells with J

SC an nc-SiO

X

= 0.883 V,

:H layer of 50 nm have an initial efficiency of 10.63% (

V

OC

J

SC

= 17 mA/cm

2

, FF = 70.85%), an improvement of 13.8% in comparison to the reference cell. These experiments were done on

Asahi U-type substrates. Depositions of solar cells on Asahi VU-type increased the efficiency to 11.0%.

It is worth to note that n, measured at 600 nm, for this material is among the higher values ( n=2.6-3, Figure 3.2). This implies that the best

BR/IR does not have the highest possible refractive index contrast with the a-Si:H silicon, showing that not only optical parameters determine the performance of the BR in a solar cell. The electrical properties such as conductivity and activation energy play an important role in cell performance as well.

75

4. SOLAR CELL APPLICATION

4.2.1.1 Current enhancement analysis

The n-doped nc-SiO

X

:H was applied at the back of p-i-n solar cells as a back reflector. Thicknesses from 0-100 nm in steps of 25 nm were applied. The effect on the EQE can be seen in Fig. 4.1. The best performing cell had a 50 nm thick nc-SiO

X

:H layer. N-doped nc-SiO

X

:H layers from 25-75 nm thick improve the spectral response of the cell over the entire spectrum. The 100 nm thick layer shows only an improvement in the red part of the spectrum ( > 640 nm) when compared to the reference cell. This improvement can be attributed to better reflection of high-wavelength photons back into the i-layer. In n-doped nc-SiO

X

:H layers thicker than 100 nm, parasitic absorption starts to dominate, thus reducing the reflection effect from the combined n-doped nc-SiO

X back reflector.

:H/Ag

While the improvement in the red was expected and can be easily explained, the improved blue-green response (λ < 550 nm) is not so straight forward. Veneri et al. [43] also report an improvement over the entire spectrum and acknowledge the improved blue response to better charge collection without further analysis. A combination of factors could be responsible for this current enhancement. These include:

1. Prevention of a native oxide formation on the standard a-Si:H n-layer by being covered with the n-doped nc-SiO

X

:H layer, providing a better contact interface with silver. This is due to the processing sequence as the PECVD steps are followed by a vacuum break before the silver contact evaporation.

2. The lower activation energy of n-doped nc-SiO

X

:H in comparison with the standard a-Si:H n-layer facilitates the flow of electrons into the silver contact.

3. Changing of the band states due to the larger bandgap of n-doped nc-SiO

X

:H in reference to n-doped a-Si:H.

4. Thinner a-Si:H n-layer as the one in the reference cell is twice as thick.

5. Lower parasitic plasmonic absorption in the silver back contact.

The n-doped nc-SiO

X

:H provides a better contact interface with silver compared to the standard n-doped a-Si:H layer. This is due to the formation of a native oxide of undesirable quality on the n-doped a-Si:H when it is exposed to ambient air before the silver contact is evaporated. N-doped nc-SiO

X

:H suppresses the formation of this oxide

76

4.2 RESULTS

Figure 4.2: EQE of n-i-p solar cells with and without the n-doped nc-SiO configuration there is also an enhancement over the whole spectrum.

X

:H back reflector. In this

Table 4.1: Measured parameters of the best cells with different n-doped nc-SiO

X

:H thickness.

n-doped nc-SiO

X

:H thickness

0 nm

25 nm

50 nm

75 nm

100 nm

V

OC

[mV]

890

893

882

882

878

J

SC

[mA/cm2]

14.17

16.65

17.00

16.15

14.90

FF

[%]

74.61

72.16

72.08

72.91

73.10

η

[%]

9.38

10.73

10.81

10.38

9.57

R

S

[Ω/cm2]

4.82

5.42

5.01

4.45

4.94

R

P

[Ω/cm2]

1419

1550

1527

1384

1567 by protecting the n-doped a-Si:H from air exposure. This can as well be achieved with ZnO:Al. The native oxide forms when water molecules come into contact with the hydrogenated amorphous silicon surface [70].

This mechanism does not apply to n-doped nc-SiO

X

:H nor ZnO:Al, as they already are oxides. Analysis with SE showed a native oxide layer of around 1.5 nm on the ambient-exposed n-doped a-Si:H but not on nc-SiO

X

:H. The effect of the non-present native oxide can be seen in the improved blue response of the cells with n-doped nc-SiO

X

:H (Fig. 4.1).

The native oxide reduces electron collection by acting as a recombination center.

The significantly improved blue-green response is not present in an n-i-p cell with n-doped nc-SiO

X

:H (Fig 4.2). This is due to the deposition sequence: the n-doped a-Si:H is deposited directly on the n-doped nc-

SiO

X

:H (or on the silver back contact for the sample without n-doped nc-SiO

X

:H) and is then followed by the intrinsic layer deposition without vacuum break, thus not allowing a native oxide to form. Similarly, there is no enhanced blue-green response when p-i-n cells are processed with

77

4. SOLAR CELL APPLICATION

Figure 4.3: Band diagram of a single junction amorphous silicon solar cell in the configuration (from left to right) p-i-(n-doped a-Si:H)-(n-doped nc-SiO

X

:H). EC is the conduction band and EV the valance band. This diagram was made with the solar cell simulation software ASA [ASA, page 90].

a vacuum break before the n-doped nc-SiO

X

:H deposition [8].

The lower electron collection due to the native oxide can be seen as well in the red part of the spectrum in the p-i-n cell. When comparing the p-i-n and n-i-p cell EQEs in Figures 4.1 and 4.2, at 700 nm, the p-i-n cell without the nc-SiO

X without the nc-SiO

X

:H has 0.3 EQE while the n-i-p cell

:H has 0.45 EQE. In case of the n-i-p cell the native oxide is missing even in the cell without the n-doped nc-SiO

X

:H due to the deposition sequence, being as well evident from the higher EQE. This way the pure reflectance contribution can be seen in the red part of the spectrum of the n-i-p cell (Fig 4.2).

In the n-i-p cells, the slight improvement in the blue starting at

450 nm is attributed to the enhanced electron collection from the lower activation energy of the n-doped nc-SiO

X

:H (40 meV compared with

240 meV of the n-doped a-Si:H layer) providing a smoother path for electrons to be collected at the back contact (illustrated in Fig 4.3; band diagram was created by a simulation of a solar cell in the Advanced

Semiconductor Analysis (ASA) software). This is not the case if ZnO:Al is used, as its activation energy is above the conduction band. The lower activation energy of n-doped nc-SiO

X applied in a tandem cell.

:H also makes a good TRJ when

The layer interface with Ag is also improved; the n-doped nc-SiO

X

:H acts as a barrier, preventing the silver from diffusing into the n-doped a-Si:H, avoiding the creation of a poorly defined interface. Inserting

78

4.2 RESULTS

ZnO:Al in the position of n-doped nc-SiO

X

:H as well prevents this diffusion and keeps the silver interface sharp [76]. To what extent are these factors facilitating electron collection requires further investigation.

It is necessary to observe this effect in p-i-n cells where the cell with n-doped nc-SiO

X

:H could be compared with a cell with only an n-doped a-Si:H/Ag interface but without vacuum break.

In samples where the thickness of the n-doped nc-SiO

X

:H layer exceeds 100 nm, the increased current collection at the n-doped nc-

SiO

X

:H/Ag interface is counteracted by increased recombination due to defects within the n-doped nc-SiO

X

:H layer. This is shown by the lower blue response in the sample in Figure 4.1. Finally, because of the phosphorus doping present in n-doped nc-SiO

X

:H, the standard n-doped a-Si:H thickness can be reduced, thus reducing its parasitic absorption.

However, the presence of the n-doped a-Si:H is still necessary because of the sharp band offset between the n-doped nc-SiO

X

:H and the a-Si:H thus the lower parasitic absorption due to a thinner a-Si:H n-layer cannot be considered. the n-doped nc-SiO

X

V

OC

has not been significantly affected; however,

:H increases the series resistance, thus negatively affecting the FF (Table 4.1).

Parasitic plasmonic absorption in silver also lowers the current in solar cells [77]–[79]. As only the long-wavelength light reaches the silver, this effect does not influence the blue spectral response.

However, because of the better n-doped nc-SiO

X presence of the n-doped nc-SiO

X

:H/Ag interface, the

:H contributes to an increased red response by shifting the plasmon resonance to lower wavelengths which have already been absorbed. Again, to what extent this contributes to the red response over the reflection of the n-doped nc-SiO

X further investigation.

:H itself needs

4.2.2 P-doped nc-SiO

X

:H as a p-layer

The developed p-doped nc-SiO

X

:H layers have been integrated as p-layers into single junction a-Si:H cells to test their performance. The substrate used was Asahi VU covered with a thin layer (~10 nm) of

ZnO:Al, to protect the SnO

2

:F of the Asahi from the H

2

plasma of the p-layer deposition, which followed afterwards. The used device grade p-layer deposition parameters are described in Table 3.1, except the

CO

2

:SiH

4

ratio was increased to 2.5. The best performing cell with the best nc-SiO

X

V,

J

SC

:H p-layer has an initial efficiency of 11.4% ( V

OC

=0.887

=18.1 mA/cm 2 ,

FF=70.8%). Although the p-doped nc-SiO

X

:H shows worse performance in the blue part of the spectrum, overall it

79

4. SOLAR CELL APPLICATION

Figure 4.4: EQE of the best performing single junction cell with p-nc-SiO

X

:H as the p-layer [80].

slightly enhances the current (Figure 4.4). This lower blue response is attributed to parasitic absorption in the p-doped nc-SiO

X of p-doped nc-SiO

X

:H. The benefits

:H are visible after 400 nm, as reported elsewhere

[48]. Between 400-680 nm there is an increase in EQE due to the antireflectance effect of the p-doped nc-SiO

X

:H. The p-doped nc-SiO

X

:H results in an improved refractive index gradient (from glass up to the a-Si:H absorber layer) in reference to an amorphous SiC p-layer. This results in a reduced reflection at the front side. The bump at 650 nm could also be due to a difference in thickness of the n-doped nc-SiO

X

:H back reflector, as the two cells were deposited in different deposition runs. They come from the lower parasitic absorption losses and beneficial refractive index grading. The overall efficiency gain in comparison to a cell with a p-doped SiC layer and intrinsic SiC buffer layer (section

3.3.2) is 0.3%. The added advantage of using p-doped nc-SiO

X

:H as a p-layer is in simplified processing as only one layer is necessary in comparison to the p-doped SiC and intrinsic SiC buffer layer duo.

4.3 Conclusion

In summary, p- and n-doped nc-SiO

X

:H layers have been incorporated into single junction a-Si:H cells. The device grade nc-SiO

X

:H film is a compromise of optical and electrical properties. The n-doped nc-SiO

X

:H

BR enables to enhance the performance of single junction a-Si:H solar cells in the red, as well as in the blue part of the spectrum. The p-doped nc-SiO

X

:H showed anti-reflective properties by enhancing the current between 400-600 nm. So far, the best single junction a-Si:H cells with a 300 nm non-diluted i-layer with incorporated nc-SiO

X

:H doped layers

80

4.3 CONCLUSION have an initial efficiency of 11.4 %.

The possible causes of current enhancement by using n-doped nc-

SiO

X

:H/Ag back reflectors have been analyzed. The enhancement in the long-wavelength part of the spectrum is due to back reflection of light into the absorber layer. The enhancement in the blue part, but as well over the entire spectrum is due to a combination of factors, the most apparent of which is the prevention of the formation of a native oxide of undesirable quality on the a-Si:H n-layer. This native oxide reduces electron collection, thus lowering the current of the solar cell. The best single junction amorphous silicon p-i-n cells with a 50 nm n-doped nc-SiO

X

:H/Ag back reflector reached an initial efficiency of 10.81% with a short-circuit current improvement of 19.97% in reference to the cell without n-doped nc-SiO

X

:H.

81

As mentioned in the previous chapters, there are strict requirements for an IR. These requirements can be met by only a handful of materials.

From these materials many complicated (multi-) layer structures can be processed with interesting optical properties [56]. Due to lower costs and level of complexity, so far industry has preferred the simplest of IRs.

These are single layer IRs mainly based on zinc oxide and silicon oxide which have been studied and applied extensively [28], [34], [81], [82].

Adding an IR between two junctions may increase reflectance into the top cell, but every extra layer induces unwanted parasitic absorption as well. To date, the best-working IRs are mainly based on silicon oxide [21], [41], [83]. The doped silicon oxide layers are part of the p-i-n junctions and create an internal electric field across the intrinsic absorber layers to collect the charge carriers generated in them. The optical properties of these doped layers were tuned to serve as IRs as well. These doped silicon oxide layers have therefore dual functionality.

Additional layers between the junctions will increase reflectance of light to the top cell but also increase parasitic absorption that deprives the bottom cell of some longer wavelength light. Advanced concepts as multi-layered distributed Bragg reflector (DBR) stacks have the potential to increase the top cell current with a very thin a-Si:H absorber layer.

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

This minimizes the impact of the SWE on the performance of the solar cell. In this chapter, it will be explored if cells with integrated DBRs have the potential to increase the overall conversion efficiency compared to cells without DBRs.

In the second part of this chapter, a variety of processing approaches to make an IR with an asymmetric morphology will be explored. The

Asahi VU substrate is designed for maximum performance of the top cell so its morphology needs to be preserved in the top cell. Its advantages are high transparency and an ideal texture for light scattering for a-Si:H absorber layers [84]. On the other hand, flat substrates guarantee the growth of nanocrystalline solar cells with good electrical properties.

Nanocrystalline absorber layers absorb longer wavelengths of light, which need to be scattered by larger surface features than the ones used to scatter the green-red light for a-Si:H absorber layers. Thus, the main motivation for an IR with asymmetric morphology is to grow higherquality bottom cell absorber layers with optimal light scattering. Various processing approaches for IRs with asymmetric textures based on wet etching and polishing will be demonstrated.

5.1 Bragg stacks as intermediate reflectors

DBRs have been applied as IRs in tandem TF Si solar cells. The design, modeling and omnidirectionality of DBRs is shown in this section.

5.1.1 Choice of materials and ASA simulations

5.1.1.1 Material characterization

As mentioned in box DBRs, to build-up a wide photonic band gap it is necessary to combine two materials (in multiple stacks) with high refractive index contrast. To apply them as IRs, low extinction coefficients and high transversal conductivity are desired as well. For this experiment, 2% Aluminum doped ZnO (ZnO:Al) was chosen as the low refractive index material. It guarantees low parasitic absorption losses due to a low extinction coefficient and has a high conductivity [34][28].

In addition, this material has several advantages such as non-toxicity, low cost, the material is abundantly available, and high stability against hydrogen plasma and heat cycling [85]. Nc-SiO

X

:H was chosen as the high refractive index material because its refractive index can easily be tuned and its extinction coefficient is low and conductivity high [43]–

[45], [86]. More importantly, because of its low activation energy, it is proven to work well as a tunnel recombination junction (TRJ) between the top and bottom cells.

84

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

Bragg reflector theory

A Distributed Bragg Reflector (DBR) consists of a stack of two alternating materials. Those two media have different refractive indexes and thicknesses resulting in prominent reflectance at each interface. A high reflectance at the Bragg wavelength λ

B

is achieved when the wavelength is 4 times the multiple of the refractive index n and thickness d of each material:

(5.1)

In this way, constructive interferences enhance the reflectance, as illustrated in Figure 5.a(a).

(a)

n i n

1 n

2 n s d

1 d

2

Figure 5.a: Constructive interferences in a distributed Bragg reflector (a). Reflectance of a photonic crystal made with five stacks. Δλ

B

is the photonic band gap (b).

As for a single interface, the larger the difference between the two refractive indexes, the higher the total reflectance. Moreover, the number of pairs of layers M also enhances the constructive interferences and thus the reflectance R:

(5.2)

It has to be noticed that the refractive indexes of the incident medium ni and of the substrate ns play a role as well. A DBR provides high reflectance for a given range of wavelengths, a range which is easily scalable by tuning the thickness of the layers. This region is called the photonic band gap and its width is commonly noted by Δλ

B

(Figure 5.a(b)). The properties of DBRs are of high interest in the field of photovoltaic solar cells since they offer an alternative approach for light management. They are suitable for use as back reflectors in single junction [57] and intermediate reflectors in tandem cells [55].

85

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Table 5.1: Optical and electrical characteristics of materials used as intermediate reflectors.

Name n

1 n

2 p

1 p

2

ZnO:Al

Material

n-nc-SiO

X

:H n-nc-SiO

X

:H p-nc-SiO

X

:H p-nc-SiO

X

:H

ZnO:Al

Conductivity

[S.cm-2]

4,04

1,05 10

-2

1,46 10

-5

1,45

88,40

Extinction coefficient

(at

600

nm)

1,70 10

-2

5,70 10

-2

1,46 10

-3

4,8 10 -2

2,9 10 -3

Activation

Energy

[meV]

32,8

142,6

288,6

35,9

11,9

Refractive

Index

(at

600

nm)

3,06

3,46

2,75

3,96

1,95

ZnO:Al was deposited using sputtering (section 2.1.2) while nc-

SiO

X

:H was deposited using PECVD (section 2.1.1). Four types of nc-

SiO

X

:H have been deposited. Two types of n-doped, referred to as n

1 and n

2

and two types of p-doped, referred to as p

1

and p

2 five materials have different optical and electrical characteristics and are

. All these presented in Table 5.1. The details of the used processing conditions are in Appendix A. All cells with experimental IRs were compared to a reference p-i-n-p-i-n tandem cell with two p-doped and two n-doped nc-SiO

X

:H layers. The processing conditions for all the layers of this cell can be found in Appendix A as well. The experimental IRs were then integrated between the top cell n- and bottom cell p-layer.

5.1.1.2 Bragg reflector design

As explained in section 1.3.4, a DBR consists of stacks of two materials with different refractive indexes. In this thesis, four types of stacks are studied, each of them implementing a ZnO:Al layer and a nc-SiO as illustrated in Figure 5.1. Depending on the type of nc-SiO

X involved (n or p

X

:H layer

:H being n

2

1

, n

,“stack p

1

2

, p

1 2

), the stack is referred to as“stack n

, or“stack p

2

1

,“stack or of a double stack (two stacks on top of each other).

It has to be mentioned that not much space will be devoted to the double stacks in this thesis. The cells with the double stacks (Fig.

5.1(b)) performed badly compared to the ones with a single stack. In addition, the top cell current was not measurable due to shunting issues.

Therefore, these cells are not extensively discussed. The only measured value from these cells is the bottom cell current, which was low (~4 mA/cm

2

), indicating that much light is reflected and/or absorbed. If the shunting issues can be resolved and the properties of the double stacks can be further fine-tuned, the double stacks still remain a good candidate as an IR.

86

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

Figura 5.1: Design of the DBR stacks, implemented as single stacks (a) or double stacks (b).

glass

ZnO:Al

Asahi VU

Simula tion air

Top cell

Simulation

}

Optimal IR

Reflectance comparisson

Fit

Bottom cell

Proc essing glass

ZnO:Al nc-SiO :H air

Figura 5.2: Depiction of optimization procedure of DBRs for IR purposes.

The EQE measurements of the reference cell (Fig. 2.6), clearly show that the top cell absorbs photons with wavelengths up to 750 nm.

Consequently the IR is desired to be reflecting light with a wavelength up to 750 nm back into the top cell. The design of such an IR is presented in Figure 5.2. The DBR has been designed in such a way that the Bragg wavelength λ

B

is about 600 nm and the photonic band gap is 300 nm wide, giving high reflectance values between the desired spectral range of 450-

750 nm (see box about Bragg reflector theory, page 85). Theoretically, the optimal thickness of the material in a DBR is derived through equation

(5.1). However, the refractive index is not constant as a function of the wavelength. Measured n, k data of the glass, ZnO:Al, and nc-SiO

X

:H has been used as the input data in the ASA modeling. This makes the layer directly measure the internal reflectance of the DBR integrated inside the cell. Therefore, simulations were carried out with the Advanced

Semiconductor Analysis (ASA) program in order to ideally design a DBR for use as an IR (see box about ASA, page 90). This resulted in the following thicknesses for the four designed stacks as presented in Table

5.2.

87

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Table 5.2: Calculated and deposited layer thicknesses of the DBRs.

ZnO:Al nc-SiO

X

:H

Stack n

1

75 nm

50 nm

Stack n

75 nm

43 nm

2

Stack p

75 nm

55 nm

1

Stack p

2

75 nm

38 nm

5.1.2 Bragg reflector on glass

In this section the optical behavior of IRs was simulated. For this purpose, the optical performance of a DBR was simulated between the intrinsic layers of the top and bottom cell. Using the simulations, the reflectance from the DBR back into the top cell is calculated. A reflectance with a maximum around 600 nm is desired, seen as the green line in

Figure 5.3. In this way the optimal individual layer thicknesses of the

DBR were determined. A second model was created in ASA, where the optimized DBR is processed on a glass substrate and the entire optical structure is embedded in air (Fig. 5.3, blue line). This optimization of the

DBR design has been previously demonstrated [87].

To demonstrate how sensitive the reflectance of the IR is to the layer thickness, thicker layers are simulated as well. This results in a shift of the photonic bandgap to higher wavelengths. This can be seen in Figure

5.3, where the black line is a simulation of the same DBR with 10 nm thicker nc-SiO

X

:H layers. The calculated optimal thicknesses of the stack layers were deposited on a Corning glass substrate. Note that the ZnO:Al is always the first layer deposited. Single stacks n stacks n

1

and n

1

, n

2

, p

1

, p

2

and double

2

were processed and characterized. The reflectance of the DBRs on glass is measured. The reflectance of the ASA simulation and the measured reflectance of the stack show a close match as can be seen in an example in Figure 5.3.

5.1.2.1 Normal incidence

Reflection and transmission (RT) measurements of the DBRs were compared to the simulation of the stack on glass in air. The results in

Figure 5.3 clearly show the influence of the embedding media on the photonic bandgap. The photonic band gap is much broader in the case of the DBR deposited on glass and measured in air than when the same stacks are simulated inside the cell. The smaller the n of the surrounding medium, the larger the photonic bandgap (n air

The

=1 and n

Si n of air is fixed in the simulations at n air

≈4).

=1. Measured n, k data was used for all other layers. Only the layer thicknesses were tuned for maximum reflection around 600 nm inside the cell assuming an Asahi VU

88

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

Figure 5.3: Reflectance of the n

2

double stack DBR when deposited on glass (simulation & measurement) and when integrated in a cell. The expected reflectance is shown when looking from the top surface. The example of the thicker DBR on glass is 10 nm thicker.

textured layer (Fig 5.2). For all the DBRs deposited on glass, experimental

RT measurements fit very well with the ASA simulations on a flat surface.

From the results it can be assumed that the optical behavior of these IRs implemented in the cells is similar to the predictions using ASA.

5.1.2.2 Omni-directionality

In practice, solar cells receive solar radiation under various angles of incidence. The behavior of the integrated IRs with respect to different angles of incidence needs to be examined. Measurements of the spectral reflectance of DBRs were carried out under various angles of incidence for both p- and s-polarization on the double stack n

2

–configuration deposited on glass. The intensity of the spectral reflectance and photonic band gap strongly depends on the angle of incidence (Fig. 5.4). The reflectance decreases with increasing angle of incidence in case of p-polarized light and increases in case of s-polarized light. This is due to the increasing optical thickness as the incident beam under an angle travels a longer distance through each layer. Thus it can be concluded that reflection intensity of the DBRs on a flat substrate cannot be said to be omni-directional.

In contrast to reflection intensity, the Bragg wavelength λ

B

remains at almost the same wavelength for various angles of incidence. This can be seen in Figure 5.4 as the highest points of the fringes are almost at the same wavelength. Thus the design of DBR-based IRs is applicable to a practical range of angles of incidence. Only the intensity of reflection changes, but not the wavelength range. The basic function of the IR to be spectrally selective is kept, which is crucial in light management between

89

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Figure 5.4: Measured spectral reflectance of double stack n with s-polarized light (left) or p-polarized light (right).

2

DBR for different angles of incidence ( θ i

)

Simulation of optical behavior with ASA

The Advanced Semiconductor Analysis (ASA) program has been introduced by the PVMD group of the Delft University of

Technology in 1987 and has since shown accurate results in the optical and electrical simulations of amorphous, crystalline, and tandem solar cells [89]. This software consists of two parts: an optical solver that derives the absorption profile of the device (or stack) using the Maxwell and Fresnel equations, and an electronic model that solves charge transport equations and calculates solar cell characteristics. All calculations presume a 1D environment.

Layer stacks can be simulated as being flat or textured. Details can be found in the ASA manual [90]. An example of an optical ASA simulation can be seen in Figure 5.3.

Simulations

In ASA, the optical properties of multilayer DBR stacks with different refractive indexes were of interest. Refractive indexes of the embedding media are also of interest. Simulations were carried out for two situations:

1. How the DBR stack performs inside the cell

2. How the DBR stack performs between glass and air

90

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

Spectrophotometry - Testing of Bragg reflector reflectance on glass

Normal incidence

The DBRs deposited on glass substrates are studied with

Reflectance and Transmittance measurements. These measurements were taken using the Integrating Sphere of the Perkin Elmer Lambda

950 spectrophotometer. The spectrophotometer has two light sources, a deuterium arc lamp for ultraviolet light and a tungstenbased halogen lamp providing light from 175 – 3300 nm. There are two detectors, a photomultiplier for the ultraviolet and visible spectrum (< 860 nm) and a PbS detector for the near infrared spectrum (> 860 nm). Once light enters the Integrating Sphere, it undergoes multiple internal reflections causing the electromagnetic field to become homogeneous and the total reflectance (specular and/or diffuse) can be measured (Fig. 5.b).

(a) Transmission reference beam

8° sample beam sample transmission port

(b) Reflection reflection port sample beam reference beam

8° sample

Figure 5.b: Depiction of a planar view of the IS when used to measure transmission (a) and reflection (b). To measure total reflection (transmission) the reflection (transmission) port is closed to keep the specular component of light inside the IS. Figure courtesy of K. Jäger [84].

91

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Omni-directionality

Because of scattering effects in the solar cell, light does not always reach the DBR under perpendicular incidence. The optical path through the stack can be lengthened, changing the optical behavior of the DBR in reference to a flat cell or DBR. Therefore reflectance measurements with varying angles of incidence were carried out in the absolute reflectance / transmittance accessory

(ARTA) compartment of the PerkinElmer setup. Inside the ARTA, the sample is mounted on a rotating stage, allowing angles of incidence from 10°to 350°to be measured (Fig. 5.c). It is equipped with a polarizer for individual polarization measurements.

θ = 90° detector

θ = 170°

NOT REA

CHABLE incoming light

θ = -170° sample

r

θ

θ = 0°

θ = -90°

Figure 5.c: Schematic cross section of the ARTA along the measurement plane. This plane is defined by the direction of the incoming light and the possible detector positions. Figure courtesy of K. Jäger [84].

the two junctions of a tandem cell. This result is of elemental importance, as in a solar cell, the light that reaches the IR is always scattered, thus arrives at various angles of incidence.

Now a tandem cell with an integrated DBR is considered. Such a cell is deposited on a textured substrate, in this case Asahi VU. Then the DBR in the middle of the cell experiences a broad range of incidence angles of incoming light. Note that the DBRs were designed on a flat substrate to optimize them for a solar cell deposited on Asahi VU. Therefore the texture of the Asahi VU needs to be taken into account. The only way to do this is to use the Asahi VU substrate itself (with the TCO). Replicated

Asahi VU texture on only glass was not available for this experiment.

Therefore the resulting measurements of such a sample include the TCO layer of SnO:F as part of the DBR as well. This SnO:F layer drastically

92

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

Asahi VU

Top cell

Asahi VU

Top cell

ZnO:Al n-doped nc-SiO :H

Bottom cell p-doped nc-SiO :H

ZnO:Al

Bottom cell

Figure 5.5: Cell integration of the single stack DBR.

influences the optical behavior of the stack. Thus, the resulting reflectance of a DBR on an Asahi VU texture is unknown, as it was not possible to measure the DBR without the Asahi VU TCO.

The n and k values of each material depend on the wavelength of light. The Bragg wavelength was calculated for light of 600 nm, below which most of the light is already absorbed in the top cell. From Equation

5.2 it can be seen that the total reflection depends on the ratio of the refractive indexes of the two materials of the DBR. For designing the

DBR, the refractive indexes of ZnO:Al and nc-SiO

X

:H at 600 nm were used. For wavelengths above 600 nm, the refractive index ratio of these two materials varies minimally. Thus the DBRs provide similar values of reflection at different wavelengths, although slight errors might arise.

5.1.3 Cell integration

n

2

The DBRs were integrated into tandem cells as IRs. Single stacks n

1

, p

1

and p

2

,

were introduced between the top and bottom cell as illustrated in Figure 5.5. External parameters and EQE of the cells with the DBRs were measured. These results were compared to the performances of a simple p-i-n-p-i-n reference cell in Table 5.3 and Figure 5.6. The integration of the DBRs results in working solar cells, however, some external parameters are deteriorated compared to the reference cell.

The cells with the p

1

and p

2

stacks show a clear gain in the top cell current compared with the reference cell, namely 1,5 mA/cm 2 and 1,3 mA/cm 2 for the stacks p

1

and p

2

respectively. Furthermore, the EQE plot

(Figure 5.6) confirms that this increase comes from a better spectral response in the 500 – 700 nm range. Regarding the stack n

1

, no gain in top cell current is observed. Only in the range from 600 – 700 nm a marginal gain of 0,02 mA/cm 2 is observed in comparison with the reference cell. For the cells with the DBRs, all bottom cells receive less light than the reference cell, resulting in lower J

SC

(Table 5.3). This could be due to additional parasitic absorption of the DBR or higher reflection

93

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Table 5.3: External parameters of tandem cells with integrated single stacks n

1 values are the average of 5 best cells.

, n

2

, p

1

, p

2

as IRs. All

Ref.

p

1 p

2 n

1 n

2

Recipe

V

OC

[V]

1,33

1,29

1,09

1,23

1,22

Top

11,77

11,79

11,60

13,27

13,07

J

SC

[mA

/cm2]

Bottom Total

11,61

η FF

23,38 9,69% 63,5%

6,90

6,74

9,12

8,07

R

S

[Ω.cm2]

9,92

18,89 6,13% 69,6% 10,69

18,34 3,98% 56,5% 63,45

22,39 7,17% 65,1% 10,89

21,14 5,74% 63,6% 12,16

R p

[Ω.cm2]

687

1106

760

1047

1067 back to the top cell. The 1-R measurements (Fig. 5.6(b)) show that the reflection is approximately the same for all stacks. Even though stack p

1

is thicker (Table 5.2), it still has a lower parasitic absorption loss compared to n

1 measured for p

1 the n

1

. This is in agreement with the low extinction coefficient

(Table 5.1). This indicates that parasitic absorption in

stack is lowering the bottom cell current. Stack n

2

performed poorly in all measured parameters, most likely due to its high silicon content.

The measured external parameters in Table 5.3 show that, the single stack n

2

is not working well electrically. This is shown by the drastic increase in series resistance while the shunt resistance drops in reference to the other stacks. Furthermore, the

V

OC

is low. It implies that both part of the cell is shunted and the tunnel recombination junction is not working well. The latter is most likely caused by a bad ZnO:Al – n

2 interface. This is supported by the high activation energy measured for the n

2

stack (Table 5.2).

Furthermore, the n

1

, p

1

, and p

2

cells show a gain over the reference cell of more than 45% in the parallel resistance R

P

, possibly due to less

and shunts. The series resistance R

S

increases by only about 10% for n

1 p

1

. Due to the improvements in

R

P

the

FF of these two cells rises by

6,1% and 1,6% respectively. However, the V

OC

of all the cells is lower compared to the reference cell. This loss is most likely coming from the

TRJ that introduces a barrier potential in the opposite direction of the

V

OC

potential. This is definitely the case for material p

2

, as the

E a

of this material is low. Therefore the V

OC

loss for the cells with DBRs is not originating from the band offset of the n- and p-doped materials forming the TRJ.

94

5.1 BRAGG STACKS AS INTERMEDIATE REFLECTORS

Figure 5.6: (a) EQE of tandem cells with integrated single stacks n

1 current utilization of the same cells and their 1-R measurements.

, n

2

, p

1

, and p

2

as IRs. (b) Total

5.1.3.1 Current matching

A current matching experiment has been carried out for single stacks n

1

and n

2

to compensate for the enhanced light incoupling to the top cell by the DBR and the parasitic absorption of the DBR. A 2,7 μ m-thick bottom cell instead of 1,8 μ m has been deposited. The results of these cells are presented in Table 5.4 and in Figure 5.7.

Table 5.4: External parameters of tandem cells with single stack n

1

and n

2 and a 2,7 μm-thick bottom cell. The reference cell is a tandem cell with a 1,8 μm-thick bottom cell.

Recipe

Ref.

n

1 n

2

V

OC

[V]

1,34

1,21

1,17

J

SC

[mA

/cm2]

Top Bottom Total

11,82 12,17

η FF

R

S

[Ω.cm2]

23,99 10,43% 65,4% 11,46

13,49 10,43

12,86 10,11

23,92

22,97

8,31% 67,9% 10,79

7,65% 63,5% 9,71

R p

[Ω.cm2]

915

736

541

95

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Figure 5.7: (a) EQE of tandem cells with single stack n

1 cell. (b) Total current utilization of the same cells.

and n

2

integrated and a 2,7 μm-thick bottom

The current matching of these cells with a thick bottom cell is much better in reference to the previous series of cells with a thinner bottom cell.

However, the bottom cell is still the current limiting one. The wavelength range where the EQEs of the top and bottom junctions overlap – between

500 and 800 nm – has been shifted towards longer wavelengths and the overlapping fraction is slightly smaller. The IR operates well in distributing the photons better between the top and bottom cells. The spectral utilization is improved as more long-wavelength photons are absorbed in the top cell. Furthermore, one could notice that the slope of the top cell EQE around 700 nm is now much steeper. This is a step towards the welcome segmentation of the solar spectrum between the top and bottom cells, allowing better spectrum utilization by a smaller top/bottom cell current overlap. A drawback of these IRs is the lower total current, mostly due to parasitic absorption of the stack. This can be seen in the 1-R measurement in Figure 5.7(b). Although the reflection

96

5.2 CONTROL OF INTERFACE TEXTURING

Figure 5.8: SEM image of the cross-section of a tandem cell deposited on Asahi VU substrate. The red-circled zones show defect rich filaments, the growth of which originates in sharp valleys from the

Asahi VU substrate.

is smaller for the DBR cells, the total current output is not higher but smaller.

The tandem cell with a single stack IR from the previous series with a 1,8 μ m-thick bottom cell are compared to this series with a 2,7 μ mthick bottom cell. The values for

R

P

have dropped drastically compared to solar cells from the previous run. The lower R

P

values have their origin in the double integration of ZnO:Al layers. The high lateral conductivity of ZnO:Al enhances the probability that shunts in the top junction are electrically connected with the bottom junction. However, the

FF stays relatively high due to a lower R

S

and greater current mismatch compared to the reference cell. Each ZnO:Al deposition step was followed by a hydrogen plasma step to eliminate potential spikes, therefore they can be excluded from causing shunting issues.

5.2 Control of interface texturing

5.2.1 Influence of front TCO texture

A typical Asahi VU texture provides very good light scattering in the top cell. However, it deteriorates the electric quality of the bottom cell due to the incorporation of defect-rich filaments induced by the texture, seen in Figure 5.8. In addition the Asahi VU texture is not very suitable for scattering long wavelength light that is to be absorbed in the bottom cell. Therefore, the ideal structure of a tandem cell would be a top cell having Asahi VU morphology while the bottom cell would have a

97

(a)

Asahi VU

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

(b)

Asahi VU

Top cell

(c)

ZnO:Al

Wet etched

Figure 5.9: Depiction and SEM images of front Asahi VU (a), the back of a cell deposited on Asahi VU substrate (b), and front wet-etched ZnO:Al (c). All scales are 1 µm. The TF Si layers deposited on top of the textured Asahi substrate adopt its texture. The surface features at the back of the cell have the same size, however, the shape has been smoothened out.

morphology with larger (1,2-1,6 μ m) surface features. This change in morphology needs to be achieved in the IR between the two junctions.

Here, approaches that smoothen the texture of the IR are presented. For this purpose ZnO:Al was chosen as an IR. The advantage of ZnO:Al is that the texture can easily be modified by wet etching.

As a reference, the wet-etched ZnO:Al was used as a front TCO to see the optical effect and adoption of its texture throughout the whole solar cell. This solar cell was then compared to a reference solar cell, which is a tandem cell deposited on Asahi VU type. Light scattering from this substrate behaves like Rayleigh scattering and is beneficial for short wavelengths of light (<500 nm) with high values for the haze in transmission [84]. The resulting light trapping enables to keep the top cell as thin as possible (175 nm). This is an advantage in minimizing the effects of light induced degradation on the performance of the solar cells

[24]. After cell deposition on an Asahi VU substrate, the surface features have the same size as the features on the Asahi substrate, however, they appear to be smoother (Fig. 5.9(b)). The sharp peaks are eliminated, but the sharp valleys stay.

The layers of the cell deposited on top of Asahi VU substrates adopt its of long wavelength light suitable for absorption in the bottom cell, thus lowering its potential J

SC

bottom. The surface may appear smoother, but it still contains sharp valleys which can initiate the growth of defective

98

5.2 CONTROL OF INTERFACE TEXTURING

Table 5.5: External parameters of a tandem cell deposited on Asahi VU and on textured ZnO:Al.

Recipe

V

OC

[V]

1352

J

SC

[mA

/cm2]

Top Bottom Total

10,91 12,45 23,36

η FF

R

S

[Ω.cm2]

8.32% 56.9% 34.04

R p

[Ω.cm2]

769

Asahi

substrate

ZnO:Al

substrate

1400 10,10 13,93 24,03 8.34% 59.1% 29.06

976 material appearing as vertical filaments [41], [59], [60] (Fig. 5.8).

These defect rich filaments reduce

V

OC

and

FF in the nc-Si:H junction.

The Asahi VU front substrate texture negatively influences all external parameters of the bottom cell. Therefore modifying the texture before bottom cell deposition is an interesting approach to improve overall cell performance.

5.2.1.1 Asahi VU and etched ZnO:Al comparison

In this section, the effects of different TCO materials with different surface morphologies on solar cells will be studied. First, the effects of an

Asahi VU and etched ZnO:Al substrates on a tandem cell are considered.

One μ m of 2% ZnO:Al was deposited by sputtering (section 2.1.2). The texturing on the ZnO:Al was applied by a wet-etch dip in 0,5% diluted

HCl for 40 seconds. The scattering effect of Asahi VU can be seen by the reflectance measurements in Figure 5.10. Above 700 nm, the light reflected back from the solar cell shows some interference fringes.

Such fringes can only appear if a significant fraction of the light is still coherent (and specular) while traveling through the cell. For wavelengths above 700 nm, no significant scattering occurs and the substrate can be considered optically flat for this wavelength range. However, below 700 nm, the Asahi substrate scatters short wavelength light very efficiently.

The ZnO:Al substrate exhibits large surface features, with sizes of around 1 μ m, beneficial for long-wavelength scattering [84]. It can be seen in Figure 5.10 that the light trapping is extended to longer wavelengths, in reference to the cell on Asahi VU, by a gain in bottom cell current above 700 nm and in the 1-R measurement. Two effects are responsible for this. First, it is an effect of scattering from the larger features of the etched ZnO:Al. Secondly, the ZnO:Al also exhibits less free-carrier (parasitic) absorption above 700 nm of wavelength in comparison with fluorine doped tin-oxide (Asahi VU) [84]. However, as it was not possible to decouple these two effects, the total contribution to the red response due to lower free-carrier absorption remains unclear.

99

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Figure 5.10: EQE of the same tandem cell deposited on Asahi VU and on textured ZnO:Al. The dashed lines represent the amount of light that is trapped in the cell (1 - Reflectance).

Between 300-350 nm, the tin oxide-based Asahi substrate results in higher current production due to its higher band gap. However, the 1-R measurements in Figure 5.10 show that the Asahi VU is reflecting more light above 700 nm compared to the ZnO:Al.

The quality of the nanocrystalline Si is very sensitive to the substrate it is deposited on. A texture with too many sharp features and deep valleys, such as the texture of Asahi VU, could induce defect-rich filaments in the material. Figure 5.8 shows the presence of such regions in the bottom cell of a tandem deposited on Asahi VU. To make such SEM cross-sectional images, the samples need to be broken. During breaking, the material splits in vertical slabs in the defective regions located above the sharps valleys of the substrate. This suggests that the mechanical properties of the material above the valleys are different compared to the dense nc-Si:H. The distribution of stress through the matrix appears to determine the preferential location where the material breaks.

The presence of defect-rich filaments is shown as well by the performance comparison ( FF, V ) of a cell deposited on Asahi VU

OC and on the much smoother ZnO:Al substrate (see Figure 5.9(c)). The cell deposited on ZnO:Al has a better FF (due to a higher parallel resistance and a smaller series resistance) and a higher

V

OC

(Table 5.5).

EQE measurements under reverse bias voltage in reference to no bias confirm the loss of light excited charge carriers in the bottom cell due to recombination (Figure 5.11). The gain in bottom cell current under a reverse bias voltage is greater in the case of a tandem cell deposited on Asahi VU than when deposited on a ZnO:Al substrate. This implies

100

5.2 CONTROL OF INTERFACE TEXTURING

Figure 5.11: EQE of a tandem cell under reverse bias voltage, deposited on Asahi VU (left) and on front wet-etched ZnO:Al (right).

that some of the light excited charge carriers of the Asahi sample are recombined in the defective material.

These results show that different types of textured substrates are optimal for a-Si:H compared to nc-Si:H cells in order to work at their full optical and electrical potential. However, they also demonstrate the strong influence of the front texturing on the interfaces throughout the whole device. It is a challenge to optimize light management and growth morphology for both the top and bottom cells of a tandem with solely using the texture of the front substrate. A front ZnO-based substrate with modulated surface texture morphology has demonstrated excellent light trapping for both sub-cells, however, it has more short-wavelength absorption due to the smaller bandgap of ZnO:Al compared to SnO:F

(Asahi VU) [41], [88]. As can be seen in Figure 5.10, the Asahi VU substrate starts to transmit light from 280 nm, while ZnO:Al transmits from 350 nm. This gives a 0,8 mA/cm 2 difference in absorption in the i-layer, showing that Asahi VU is superior to ZnO as the substrate for the top cell. The need for state-of-the-art light trapping in the top cell in combination with the processing of a high quality nc-Si:H bottom junction shows the necessity for an IR whose texture can be controlled.

5.2.2 Etched zinc oxide as textured intermediate reflector

As seen in Figure 5.10, it is beneficial for the top cell to have the

Asahi UV substrate for both its high transmittance and good scattering properties. However, it is not beneficial for light scattering into the bottom cell, as Asahi VU can be considered optically flat for higher wavelengths.

Therefore, a proposed solution would be to integrate different lightscattering features after the top cell is deposited. In this section such an approach will be explored.

The approach developed in this thesis is illustrated in Figure 5.12.

101

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

(a)

Asahi VU

Top cell

ZnO:Al

(b)

Asahi VU

Top cell

(c)

Top cell

Bottom cell

Wet etched

Figure 5.12: Texturing of Intermediate ZnO:Al. First step: Deposition of a thick layer of ZnO:Al (~475 nm) underneath the top cell (a). Second step: Etching of the ZnO:Al layer and creation of large and smooth features (b). Third step: Deposition of the bottom cell on a smoother substrate (c).

After the deposition of the top cell on an Asahi VU substrate, a thick layer of ZnO:Al (2% doped) is sputtered on the sample. This ZnO:Al is then textured using direct wet-etching in order to create relatively smooth features similar to the ones observed on front wet-etched ZnO:Al (as in Figure 5.9). These features are smoother and larger without sharp valleys as compared to the Asahi VU ones. Finally, the bottom cell is deposited on top of the intermediate ZnO:Al layer, in order to complete the tandem cell.

5.2.2.1 Etching with 0,5% HCl dilution

The ZnO:Al IRs were wet-etched in a 0,5% HCl solution. The wetetching process not only creates features of various different sizes, it also reduces the overall thickness of the ZnO:Al layer. Based on the calculations in section 5.1.1.2, the ZnO:Al layer should be 75 nm thick to achieve good reflective properties. Therefore the etching rate was determined from a single ZnO:Al layer deposited on glass. A top cell with a 475 nm intermediate ZnO:Al layer was deposited on an Asahi

VU substrate (Figure 5.13(a)). The sample was then etched in a 0,5% weight diluted HCl solution for 25, and 33 seconds. SEM images of the samples were taken to see the feature sizes created on the ZnO:Al

(Figure 5.13(b),(c)). They appear sharp and porous with sizes above

1 μ m in both samples and 2 μ m in the 33 s etched sample. Finally, the bottom cell was deposited on top of the ZnO:Al, completing the tandem cell (Figure 5.12(c)). 1x1 cm

2

silver back contacts were evaporated on the sample area.

The characterization of these solar cells showed that they are shunted.

The solar cell with the non-etched ZnO:Al IR was shunted because of the very thick laterally-conducting ZnO:Al layer which electrically connects the shunts in the top and bottom cell. In the cells in which the intermediate ZnO:Al has been etched, the features of the etched ZnO:Al

102

5.2 CONTROL OF INTERFACE TEXTURING

(a) 0 s

(b) 25 s

(c) 33 s

Figure 5.13: SEM images of the ZnO:Al deposited on the top cell. (a) Without etching (0 s), (b) after

25 s etching, and (c) after 33 s etching in a 0,5% weight diluted HCl solution. The ZnO:Al appears bright while the silicon layers of the top cell appear dark. All scales are 2 µm. appear to be too sharp to be applied as an IR, leading to defective material growth and/or non-ideal crystallinity values. The series resistance has risen drastically, leading to very low values for the FF. Together with a decrease in V

OC

, these results show that tandem cells with integrated

ZnO:Al IRs are performing far from optimally. Unfortunately, no signal could be measured using the EQE setup, due to the electrically connected shunts in the top and bottom cells. Therefore EQE measurements could not be used to evaluate top and bottom cell current densities and no conclusions can be drawn concerning the optical behavior of these cells.

103

(a) 150 s

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

(b) 180 s

(c) 210 s

(d) 240 s

Figure 5.14: SEM images of ZnO:Al deposited on the top cell and wet-etched in a 0,1% HCl solution.

(a) depicts the ZnO:Al experiencing 150 s etching time, (b) 180 s etching time, (c) 210 s etching time,

(d) 240 s etching time. (a), (b) and (c) are views under a 45° angle while (d) is a flat top view. All scales are 2 µm. In (d) there is a negligible amount of ZnO:Al found in the valleys.

Instead of smoothening the tips of the Asahi-induced features, the acid solution was most efficient in etching the valleys, expanding them all the way to the top cell surface. The ZnO:Al was aggressively etched due to the high concentration of the acid solution. After etching, the surface features were still too sharp due to the short etching time. For this reason a lower concentration of HCl and longer etching times were applied.

5.2.2.2 Etching with 0,1% HCl dilution

A lower HCl concentration was used to prevent the creation of texture with sharp peaks on the ZnO:Al during etching. An experiment was carried out, implementing a 0,1% weight diluted HCl solution and a

960 nm-thick sputtered ZnO:Al. Four etching steps were studied: 150,

180, 210, and 240 s. The SEM images of these sample after intermediate

ZnO:Al etching are presented in Figure 5.14.

The porosity of the ZnO:Al film is increasing which is indicated by the holes which appear after 210 seconds of etching. After 240 seconds practically all the ZnO:Al is etched off. The 240 s sample still has a negligible amount of ZnO:Al, however, it is most likely not influencing the cell performance. The texture is again dominated by the Asahi VU substrate with etching for 150 and 180 s. In these samples the texture appears smoother with less sharp peaks than in the etching experiment with 0,5% weight diluted HCl (Fig. 5.14). However, the surface feature size is still quite small and similar to Asahi VU. The same conclusions

104

5.2 CONTROL OF INTERFACE TEXTURING

Figure 5.15: EQE and 1-R of tandem cells with a ZnO:Al intermediate reflector wet-etched for different times in a 0,1% weight diluted HCl solution.

can be derived from EQE measurements on these devices (Figure 5.15).

The spectral responses of the top and bottom cells for the porous and etched off ZnO:Al cells (210 and 240 s) are similar to the reference cell. The sample with the ZnO:Al etched for 240 s would be expected to behave the same as the sample with no ZnO:Al. The measurements of the samples show small differences between the reference cell and the one with the ZnO:Al etched for 240 s. It can be concluded that still some

ZnO:Al is left between the junctions of the cell with the 240 s etched IR which is the cause of the differences. These remnants of the ZnO:Al are inducing the opposite of what is desired from the IR – the top cell spectral response is weaker from 450 – 600 nm and the bottom cell stronger from 600 – 700 nm. This could be due to the etching process damaging the top cell layers. On the other hand, for 150 and 180 s etching times, the intermediate ZnO:Al plays the role of an IR. The top cell spectral response slightly increases between 600 and 800 nm and the bottom cell current density decreases in comparison to the reference cell.

The effect of the newly-induced texture on light trapping performance can be analyzed from 1-R measurements presented in Figure 5.15.

By including the ZnO:Al IR in the tandem cell the reflection above 630 nm is smaller than for the reference cell. However, the J

SC

of the cells etched for 210 and 240 s is comparable to that of the reference cell. This suggests that the ZnO:Al IR contributes to parasitic absorption losses. In case of the reference cell there are even more reflective losses above

630 nm. This makes the nc-SiO

X

:H the most effective reflector from the compared samples for the wavelengths above 630 nm. The ZnO:Al is scattering longer wavelengths more effectively. This is confirmed by

105

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Table 5.6: External parameters of the tandem cells with a ZnO:Al intermediate reflector wet-etched for different times in a 0,1% weight diluted HCl solution.

Recipe

Ref.

240 s

210 s

180 s

150 s

V

OC

[V]

1,33

1,31

1,32

1,27

1,21

J

SC

[mA

/cm2]

Bottom Total

η FF

R

S

[Ω.cm2]

Top

10,98 13,18 24,16 9.59% 66.1% 12.27

10,44 13,00 23,44 8.32% 60.0% 24.65

10,79 13,22 24,01 7.87% 60.4% 23.35

11,01 11,91 22,92 8.85% 63.1% 10.18

11,09 11,87 22,96 8.60% 63.9% 9.47

R p

[Ω.cm2]

1225

1052

991

567

518 the light becoming more incoherent (interference fringes progressively disappearing) with increasing ZnO:Al thickness. The thickest ZnO:Al layer (150 s) is the best scaterrer as indicated by the smaller amplitude of the interference fringes. In addition this sample shows the smallest reflection.

In Table 5.6 the external parameters of the tandem cells with ZnO:Al

IRs wet-etched in the weaker 0,1% HCl solution are shown. The approach did not yet result in improved performance of the solar cells compared to the reference cell. The main reason for this is the increased free-carrier absorption in the ZnO:Al. From the SEM as well as the

EQE measurements (Figures 5.14 and 5.15), it can be stated that the IR surface morphology of the 240 s-etched cell is similar to the reference cell. Due to the acid treatment there are more defects on the adjacent doped layers, which could explain the higher R

S

for the 210 and 240 s etched samples. Under this assumption, with greater intermediate ZnO:Al thickness (shorter etching times) the FF of the cell is enhanced, with an absolute gain of 3,5% in reference to the cells with the thinner ZnO:Al IR

(longer etching times) (Table 5.6). Current mismatch does not greatly contribute to the

FF as the cells were almost current matched. On the other hand, the

V

OC

loses up to 0,1 V. This can be due to either Shockley

Read Hall (SRH) recombination and/or a barrier forming on the Si/ZnO:Al interface. No conclusion regarding the quality of the bottom cell material can be drawn here. A recommendation for future work is to make the

ZnO:Al layer thinner and/or use a material with lower Al doping.

5.2.3 Mechanical polishing of the intermediate reflector

Mechanical polishing is very appealing to use in TF solar cell technology, as it can smoothen and flatten a textured layer. This method is of interest for the IR, which has a texture inherited from an Asahi VU

106

5.2 CONTROL OF INTERFACE TEXTURING

(a)

Asahi VU

Top cell

N-doped nc-SiO x

:H

(b)

Asahi VU

Top cell

Polished

(c)

Asahi VU

Top cell

Bottom cell

Figure 5.16: Tandem cell preparation flowchart with mechanical polishing of the IR. A thick nc-SiO deposited on the flat IR surface (c).

X

:H

IR is deposited (a), it is polished down to the tips of the top cell absorber layer (b), the bottom cell is

(a)

Asahi VU

Top cell

ZnO:Al

(b)

Asahi VU

Top cell

(c)

Asahi VU

Top cell

(d)

Asahi VU

Top cell

Bottom cell

Polished Wet etched

Figure 5.17: Tandem cell preparation flowchart with mechanical polishing and etching of the IR. A thick ZnO:Al IR is deposited (a), it is polished to remove the Asahi VU-induced texture (b), the sample is dipped into a HCl solution to induce larger surface features and remove unneeded IR thickness (c), the bottom cell is deposited on the new IR surface with larger features (d).

substrate (Fig. 5.16(a)). Polishing the IR surface would result in a flat interface and would guarantee growth of higher quality nc-Si:H (Fig.

5.16(b)). Mechanical polishing has been attempted by Boccard [61], however, only on silicon oxide. In case of the ZnO:Al IR, after making its surface flat, a wet etching step can be applied to induce the welcome large surface features, beneficial for long-wavelength scattering (Fig.

5.17(c)). In section 5.2.2, texturing the ZnO:Al IR was already studied.

It was found that the etching of the HCl solution is most efficient in the valleys of the substrate texture. As a result the valleys are deepened which results in sharper peaks. The motivation for the polishing approach is to provide a flat surface initial to the start of the etching process (Fig.

5.17(b)). This approach of polishing the ZnO:Al IR was proven in case of the ZnO:Al substrate use. The ZnO:Al is relatively flat when sputtered onto glass. This guaranteed good electrical properties. A wet etching step gives it large surface features for long-wavelength light scattering

(Fig. 5.9(c)). It was shown that these large surface features enhance bottom cell current (Fig. 5.10).

107

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Mechanical polishing

Mechanical polishing is a method in which a deposited layer is made thinner and its surface is smoothened. Removing the layer material is based on friction and pressure between the sample, a polishing material, and a slurry.

The polishing process begins by mounting the sample onto a holder where it is held in place by vacuum contact at the back

(Fig. 5.d). Rotating, the holder is then pressed at a certain pressure against a rotating disk with a fabric surface. The holder and disk rotate in opposite directions. The rotating holder arm sweeps back and forth on the counter-rotating disk. Water and ammonia slurry are poured onto the fabric-coated disk during the polishing process. The polishing time, set pressure, and rotation and sweep speed all influence the total removed thickness. It is possible to remove material with a precision of 30-50 nm. The polishing setup is a Presi Mecapol E 460.

Figure 5.d: The mechanical polishing setup.

108

5.2 CONTROL OF INTERFACE TEXTURING

Figure 5.18: SEM image of polished IR surface. The sample was deposited on an Asahi VU substrate.

The peaks are smoothened out while the valleys can still be seen due to insufficient polishing time.

Two types of IR materials were polished, n-doped nc-SiO

X

:H and

ZnO:Al. The initial layer before polishing had to be thicker than the final desired thickness to accommodate the thickness loss during polishing

(Fig. 5.16(a)). The two materials have different polishing rates under the applied conditions. ZnO:Al was removed at an average rate of 2.48 nm/s and n-doped nc-SiO

X

:H was removed at an average rate of 4.66 nm/s. ZnO:Al proved to be harder than silicon oxide. In practice, the effective polishing rate is not constant in time. The initial polishing rate, determined solely by the peaks of the texture, is faster. Therefore it is not straightforward to determine the exact moment when to stop the polishing procedure. To make the IR flat, the nc-SiO

X to be polished for at least 40 and 70 s respectively.

:H and ZnO:Al had

The solar cell samples with the nc-SiO

X

:H as the IR were polished to remove the Asahi-induced texture. The duration of the polishing process was adjusted to leave the desired nc-SiO

X

:H thickness. The removed thickness depends on the polishing time, set pressure, and rotation speed of the holder and disk. Many samples have to be used to optimize these three processing parameters to achieve the desired results. Often, the

IR was completely polished off or other times the IR was not polished enough, indicated by the presence of valleys (Fig. 5.18). Figure 5.19 shows the first results of a polished IR in a tandem cell.

109

5. INTERMEDIATE REFLECTORS IN TANDEM SOLAR CELLS

Figure 5.19: SEM image of a tandem cell with a polished nc-SiOx:H IR.

In Figure 5.19, it can be seen that the interface between the top and bottom cells is truly flat. However, the deposition of nc-Si:H on top of this surface creates textures again. This can be seen by the texture on the metal contact. Yet an increase in V

OC

and FF was not observed.

This could be due to an IR which is still too thick due to insufficient removal of material during the polishing step. Such an IR is subject to increased parasitic absorption and can also serve as a recombination trap due to lower electrical quality. No defect-rich filaments have been observed, showing the homogeneous cross-sectional surface as a result of breaking of the sample (Figure 5.19). Near the interface of the top and bottom cells, the material has a better homogeneity in the density of the nc-Si:H bottom cell in comparison with the nc-Si:H in Figure 5.8. The nc-Si:H in the bottom cell closer to the metal contact is more divided into vertical slabs. This suggests a higher crystallinity closer to the metal back contact.

5.3 Conclusions

Four configurations of DBR stacks were developed and two were tested with a double stack configuration. They all showed promising optical

110

5.3 CONCLUSIONS properties, with strong reflectance at the Bragg wavelength and reasonable omni-directionality (see section 5.1.2.2). The optimal thicknesses of these stacks were engineered by ASA simulations of light propagation through semiconductor materials. These stacks were confirmed to work inside solar cells as IRs by comparing the measurements of these stacks deposited on glass with their corresponding ASA simulations on glass.

The integration of the DBR stacks in tandem solar cells as IRs strongly enhances the top cell J

SC

, reaching up to 13,5 mA/cm 2 in a 175 nm-thick a-Si:H layer. An additional advantage of DBRs is the increase of FF of the cells due to higher values of R

P

. However, the integration of stacks in the solar cells had a negative impact on V

OC

. Thus there is a tradeoff between lower V

OC

and higher J

SC

(seen in stack n

2

, section 5.1.1.3).

Generally, the DBR stack integration does not improve the overall cell performance. With better current matching overall cell performance can theoretically be raised.

In section 5.2.1, an approach to modify the texture of ZnO:Al serving as an asymmetric IR was developed. No improvement of cell performance was demonstrated yet using this approach. However, a strong need for better control of the development of the interface texture throughout the tandem device was demonstrated. This texture control is important in order to provide good light scattering for both the top and bottom cell and to provide a good substrate for the growth of a defect-free nanocrystalline absorber layer. Because of the excellent performance of the top amorphous cell deposited on Asahi VU, it is beneficial to keep this substrate texture for the top cell and integrate different textures in the layers processed after the top cell. For this purpose, an

IR with an asymmetric texture was integrated in the tandem solar cell.

This IR consisted of sputtered ZnO:Al which was then textured via a wet-etching process. The IR interface facing the top cell has a typical

Asahi VU texture. The IR interface facing the bottom cell has a larger surface features beneficial for long-wavelength scattering. This IR was enhancing the FF of the cell by 3,5% (see section 5.2.2.2) by offering a better substrate for the bottom cell growth. However, lower

V

OC remain an issue.

values

Mechanical polishing was applied to nc-SiO

X

:H and ZnO:Al IRs.

Flattening of the IR surface texture was demonstrated, however, no improvements in cell external parameters were recorded yet. SEM images reveal a nc-Si:H absorber layer with less defect-rich zones when deposited onto a polished IR. Further optimization of the process is recommended.

111

In this section the main findings of this thesis are summarized. This thesis describes in detail doped silicon oxide layers and IR concepts.

Various nc-SiO

X

:H layers with a wide range of optical and electrical properties have been developed, characterized, and applied in solar cells.

The device grade p- and n-doped material, performing best in thin-film silicon solar cells was found and studied extensively. The nanostructure of nc-SiO

X

:H was studied in great detail to identify the crucial properties that determine the optical and electrical properties needed for an IR.

The performance of solar cells with nc-SiO

X

:H layers is compared to ones without. Finally, various IRs are tested to see their effect on cell performance.

6.1 Nanostructure

The nanostructure of nc-SiO

X

:H films with a wide array of optical and electrical properties has been studied in detail by TEM, Raman, FTIR and XPS. Silicon oxide was found to be a very heterogeneous material with complex nanostructure. The amorphous and crystalline phases of nc-SiO

X

:H have been studied in detail. Differences were found between the p- and n-doped materials. Both materials contain crystalline silicon grains and an amorphous silicon oxide matrix.

6. CONCLUSIONS

It is found that the n-doped material has a nanostructure of silicon crystal grains embedded in an amorphous silicon oxide matrix (Fig.

3.25(a)). The p-doped material, however, contains silicon filaments in an amorphous silicon oxide matrix (Fig. 3.25(b)), a finding that has also been reported by other groups [47], [64]. A minimum crystalline content in n-doped and p-doped nc-SiO

X

:H is required to have sufficient conductivity. However, they differ in their amorphous phases. It is noted that the Si crystal orientation was found to be random.

6.1.1 Crystalline phase

It has been shown that crystalline silicon grains are present in nc-

SiO

X

:H. It was found that a minimal fraction of crystalline Si is necessary for sufficient material conductivity. This minimum, expressed in terms of the I

521

/I

480

ratio, is ~20% for the n-doped nc-SiO p-doped nc-SiO

X

X

:H and ~10% for the

:H. However, the relatively small crystalline volume fraction cannot solely provide the conductive properties, the amorphous phase plays an important role as well.

Information on the crystalline silicon grains in nc-SiO

X

:H was obtained by studying the Raman peak at 521 cm -1 corresponding to the crystalline silicon matrix. The amorphous peak is found at 480 cm -1 . It was observed that the crystalline peak position is shifting to higher wavenumbers with higher crystalline silicon content, indicated by the I

521

/I

480

ratio (Fig.

3.11). This implies that the more crystal grains incorporated, the bigger they are in size. It can be stated that for the deposition series studied in this thesis, the crystalline grains in the n-type nc-SiO

X

:H are in general bigger than in the p-type. The crystalline peak shift is practically identical for both device grade n- and p-doped material, suggesting that there exists an optimum size for the crystal grains. The crystalline peak of both device grade materials was found at 519 cm

-1

, representing a crystal grain size of approximately 5-8 nm [67]. That is in line with the grain size observed in the TEM images, where the grains appear as quantum dots in an amorphous tissue (Fig. 3.6 and 3.9). The presence of crystal grains in the n-doped material is most probably not crucial as the majority of the path that the charge carriers need to travel is in amorphous tissue (Fig. 3.26(b)).

The influence of the crystalline peak shift on the material properties was analyzed. The peak shift is affected by the deposition parameters. For increasing pressure and CO

2

flow in the n-doped material, the crystalline peak shifts to lower wavenumbers (Fig. 3.12), but the conductivity, refractive index, and bandgap have opposite trends with the deposition parameters (Fig. 3.2). Therefore, no conclusions can be drawn from

114

6.1 NANOSTRUCTURE the peak shift in relation to optical and electrical properties. It has been shown with TEM images that the crystalline peak shift is due to small crystalline Si grain size and not stress in the material.

6.1.2 Amorphous phase

The amorphous phase in nc-SiO

X

:H is heterogeneous, consisting of amorphous silicon and amorphous silicon oxide. Clear evidence of this comes from the FTIR measurements which show different detected stretching modes. Various oxygen-related modes are detected, giving evidence of oxygen incorporation in the material. The mode with the strongest absorption signal is the 1050 cm -1 mode, representing Si-

O-Si stretching. The 1135 cm -1 mode also has a strong contribution.

Hydrogen stretching modes with back-bonded oxygen atoms between

2140-2250 cm

-1

are detected as well. An even distribution of these modes with a slight dominance of the 2250 cm -1 mode is desired for device grade material. An a-Si:H mode at 2100 cm -1 is detected as well, a sign of the presence of pure a-Si:H. Pure a-Si:H is found to be crucial for sufficient conductivity of the material. Correlations between the trends of certain stretching modes in all FTIR measured samples and XPS elemental composition measurements were found, showing that

FTIR is a reliable method of element content analysis in silicon oxide

The quality of the amorphous phase was related to the optical and electrical properties. Some crucial properties of the amorphous tissue were identified to guarantee good conductivity. The pure a-SiO

X

:H samples have very poor conductivity. The strongly oxygen related modes

(1135 and 2250 cm

-1

) appear to be good indicators for the amorphous tissue with poor conductive properties. The electrical properties, namely activation energy and conductivity were decreasing and increasing respectively with the increasing 2100 cm

-1

mode. However, the optical properties suffered. Therefore it can be stated that nc-SiO

X

:H is a compromise between achieving good optical and electrical properties.

The deposition parameters greatly affect the contributions of the

SMs. Varying the CO

2

:SiH

4

ratio has a direct impact on the amount of oxygen in the plasma and together with the H

2

:SiH

4

ratio severely affects the oxygen content in the film. The lack of hydrogen precursor in the plasma for instance results in the material being completely amorphous.

The incorporation of phosphorus atoms is more dependent on other deposition parameters, mainly pressure and H

2

. Pressure is a very critical deposition parameter as it not only influences the crystallinity and quality of the a-SiO

X

:H tissue (Fig. 3.15), but also the doping.

The silicon filaments in p-doped nc-SiO

X

:H are vertical, therefore

115

6. CONCLUSIONS logically they should provide transversal conductivity. However their lateral conductivity is high as well. For layer thicknesses above 20 nm it was shown (Fig. 3.8(a)) that these filaments branch out and interconnect. This explains the measured high lateral conductivity for samples which had typical thicknesses of ~150 nm. The device grade transversal conductivity in p-doped nc-SiO

X

:H is attributed to the a-Si:H filaments, in which the c-Si grains inside play a minor role. Conductivity of the filaments could simply be based on the fact that they are made of oxygen-free silicon alloys. Whether the silicon phase in the filaments is made of c-Si grains and/or a-Si:H tissue is most probably not crucial in achieving the desired conductivities.

6.2 Cell integration

The doped nc-SiO

X

:H layers have been integrated into single junction thin-film silicon solar cells. The p-doped nc-SiO

X

:H replaced the previously used p-doped silicon carbide as the p-layer. The n-doped nc-SiO

X

:H was applied on top of the n-doped amorphous silicon layer.

This n-doped amorphous silicon layer could be made thinner than in cells without the n-doped nc-SiO

X

:H. In both the p- and n-doped nc-SiO

X

:H application an improvement in cell performance was observed. The p-doped nc-SiO

X

:H showed anti-reflective properties by enhancing the current between 400-600 nm. The best single junction a-Si:H cells with a 300 nm non-diluted i-layer with incorporated nc-SiO

X have an initial efficiency of 11.4 %.

:H doped layers

Possible causes of current enhancement by using n-doped nc-

SiO

X

:H/Ag back reflectors have been analyzed. The enhancement in the long-wavelength part of the spectrum is due to back reflection of light into the absorber layer. The enhancement in the blue part, but as well over the entire spectrum is due to a combination of factors. The most apparent of which is the prevention of the formation of a native oxide on the a-Si:H n-layer. This native oxide reduces electron collection, thus lowering the current of the solar cell.

6.3 Intermediate reflector concepts

6.3.1 Distributed Bragg Reflectors

The use of nc-SiO

X

:H only as doped layers in tandem cells already qualify it as an IR. Reflection in solar cells can be increased by taking advantage of constructive interference when multiple layers with contrasting refractive indexes, also called Distributed Bragg Reflectors

116

6.4 RECOMMENDATIONS

(DBR), are used. Applying multiple layers increases parasitic absorption as well. In this thesis, four DBR stacks were developed and two were tested with a double stack configuration. They all showed promising optical properties, with strong reflectance at the Bragg wavelength and reasonable omni-directionality (section 5.1.2.2). The integration of the DBR stacks in tandem solar cells as IRs strongly enhances the J

SC top. DBRs also increased the FF of the cells due to higher values of R

P

.

. It can be stated that However, all stacks had a negative impact on the V

OC there is a trade-off between lower

V

OC

and higher

J

SC

. Generally, the DBR stack integration does not improve the overall cell performance. With better current matching the overall cell performance can theoretically be raised.

6.3.2 Intermediate reflector texturing

A wet-etching approach to modify the texture of ZnO:Al serving as an IR was developed. After applying this approach, no improvement in cell performance was observed. However, a strong need for better control of the interface texture throughout the tandem device was demonstrated.

An enhancement of the FF of the cell by 3,5% was demonstrated as well (see section 5.2.2.2) by offering a better substrate for the bottom cell growth. However,

V

OC reference cell.

values are lower when compared with the

Mechanical polishing was applied to nc-SiO

X

:H and ZnO:Al IRs in order to eliminate the inherited top cell texture. Flattening of the IR surface texture was demonstrated, however, no improvements in cell external parameters were recorded. SEM images reveal a nc-Si:H absorber layer with less defect-rich zones when deposited onto a polished IR.

6.4 Recommendations

For further study of nc-SiO

X summarized here:

:H, a set of recommendations is

• 3D TEM holography images of silicon filaments in p-doped nc-

SiO

X

:H can give more insight into the significance of the crystal grains for conductivity

• Optimizing the polishing of nc-SiO

X

:H and ZnO:Al until a desired thickness of these layers is achieved and their surface is completely flat

117

6. CONCLUSIONS

• More TEM images of different nc-SiO

X

:H materials for better comparison of measured parameters with nanostructural features

• Study of nc-SiO

X

:H nanostructure and optical and electrical characteristics in relation to light-induced degradation and stable cell efficiency.

118

[1]

[2]

[3]

[5]

[6]

[7]

[8]

[4]

[9]

[10]

[11]

[12]

[13]

[14]

Renewable energy policy network for the 21st century, “Renewables 2012

Global Status Report,” 2012.

M. T. Klare, “The Geopolitics of Oil : Old and New,” 2012.

Google Public Data, “World Electrical Power Consumption,” 2013. [Online].

Available: https://www.google.nl/publicdata/explore?ds=d5bncppjof8f9_&met_ y=eg_use_pcap_kg_oe&tdim=true&dl=en&hl=en&q=world energy consumption#!ctype=l&strail=false&bcs=d&nselm=h&met_y=eg_use_elec_ kh&scale_y=lin&ind_y=false&rdim=region&ifdim=region&tdim=true&hl=en_

US&dl=en&ind=false. [Accessed: 08-Sep-2013].

The World Bank, “Access to electricity (% of population),” 2013. [Online].

Available: http://data.worldbank.org/indicator/EG.ELC.ACCS.ZS. [Accessed:

09-Sep-2013].

J. Tollefson, “Islands of Light,” Nature, vol. 507, pp. 154–156, 2014.

N. Rosen,

Off the Grid: Inside the Movement for More Space, Less Government, and True Independence in Modern America. Penguin Books, 2010.

A. Jaeger-Waldau, “PV Status Report 2012,” 2012.

The Energy Collective, “Solar Energy: Grid Parity In India, Italy, and More to Come in 2014,” 2013. [Online]. Available: http://theenergycollective.com/ josephromm/194506/solar-energy-grid-parity. [Accessed: 09-Sep-2013].

U.S. Geological Survey, “Rare Earth Elements—Critical Resources for High

Technology,” 2002.

US Energy Information Administration of the U.S. Department of Energy,

“Levelized Cost of New Generation Resources in the Annual Energy Outlook

2013,” 2013.

W. Shockley and H. J. Queisser, “Detailed Balance Limit of Efficiency of p-n

Junction Solar Cells,” J. Appl. Phys., vol. 32, no. 3, p. 510, 1961.

A. Richter, M. Hermle, and S. W. Glunz, “Reassessment of the Limiting

Efficiency for Crystalline Silicon Solar Cells,” IEEE J. Photovoltaics, vol. 3, pp.

1184–1191, 2013.

Panasonic, “Panasonic HIT Solar Cell Achieves World’s Highest Energy

Conversion Efficiency of 25.6% at Research Level”, 2014.

J. Zhao, A. Wang, M. a Green, and F. Ferrazza, “19.8% Efficient ‘Honeycomb’

Textured Multicrystalline and 24.4% Monocrystalline Silicon Solar Cells,” Appl.

Phys. Lett., vol. 73, p. 1991, 1998.

BIBLIOGRAPHY

[23]

[24]

[25]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

[26]

[27]

[28]

[29]

Greentech Media, “Solar Balance-of-System Costs Account for 68% of PV

System Pricing: New GTM Report,” 2012. [Online]. Available: http://www.

greentechmedia.com/articles/read/Solar-Balance-of-System-Accounts-for-

68-of-PV-System-Pricing-New-GTM-Repo. [Accessed: 09-Sep-2013].

“Navigant Consulting,” 2013.

D. Levitan, “Gallium Arsenide Solar Panel Breaks Efficiency Record,” IEEE

Spectrum, 2012. [Online]. Available: http://spectrum.ieee.org/energywise/ green-tech/solar/gallium-arsenide-solar-panel-breaks-efficiency-record.

First Solar, “Innovative Cadmium Telluride Technology,” 2013. [Online].

Available: http://www.firstsolar.com/Innovation/CdTe-Technology.

Solar facts and advice, “Sale of CIGS Solar Cell Panels Expected to Reach $1

Billion by 2013.” [Online]. Available: http://www.solar-facts-and-advice.

com/CIGS-solar-cell.html.

Spectrolab, “Company History.” [Online]. Available: http://www.spectrolab.

com/history.htm.

B. Yan, G. Yue, L. Sivec, J. Yang, S. Guha, and C.-S. Jiang, “Innovative dual function nc-SiOx:H layer leading to a >16% efficient multi-junction thin-film silicon solar cell,”

Appl. Phys. Lett., vol. 99, no. 11, p. 113512, 2011.

S. Kim, J.-W. Chung, H. Lee, J. Park, Y. Heo, and H.-M. Lee, “Remarkable progress in thin-film silicon solar cells using high-efficiency triple-junction technology,” Sol. Energy Mater. Sol. Cells, vol. 119, pp. 26–35, Dec. 2013.

R. Schropp and M. Zeman, Amorphous and Microcrystalline Silicon Solar Cells.

Kluyer Academic Publishers, 1998.

D. L. Staebler and C. R. Wronski, “Reversible conductivity changes in dischargeproduced amorphous Si,” Appl. Phys. Lett., vol. 31, no. 4, p. 292, 1977.

M. Stuckelberger, M. Despeisse, G. Bugnon, J.-W. Schuttauf, F.-J. Haug, and

C. Ballif, “Comparison of amorphous silicon absorber materials: Light-induced degradation and solar cell efficiency,” J. Appl. Phys., vol. 114, no. 15, p. 154509,

2013.

J. Meier, R. Fluckiger, H. Keppner, and A. Shah, “Complete microcrystalline p-i-n solar cell—Crystalline or amorphous cell behavior?,” Appl. Phys. Lett., vol. 65, p. 860, 1994.

F. Finger, P. Hapke, M. Luysberg, R. Carius, H. Wagner, and M. Scheib,

“Improvement of grain size and deposition rate of microcrystalline silicon by use of very high frequency glow discharge,” Appl. Phys. Lett., vol. 65, no. 20, p. 2588, 1994.

D. Fischer, S. Dubail, J. A. Anna Selvan, N. Pellaton Vaucher, R. Platz, Ch. Hof,

U. Kroll, J. Meier, P. Torres, H. Keppner, N. Wyrsch, M. Goetz, A. Shah, and K.

Ufert, “The ‘Micromorph’ solar cell: extending a-Si:H technology towards thin film crystalline silicon,” in IEEE, 1996, pp. 4–7.

J. Krc, F. Smole, and M. Topic, “Advanced optical design of tandem micromorph silicon solar cells,” J. Non. Cryst. Solids, vol. 352, no. 9–20, pp. 1892–1895,

Jun. 2006.

120

BIBLIOGRAPHY

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

O. Isabella, M. El-Shinawy, S. Solntsev, and M. Zeman, “3-D Modeling of

Triple Junction Solar Cells on 2-D Gratings with Optimized Intermediate and

Back Reflectors,” Renew. Energy Environ. Opt. Photonics Congr., p. JM5A.11,

2012.

B. Vet, B. Grancic, O. Isabella, S. Solntsev, and M. Zeman, “Optical and Electrical

Simulations of Advanced Silicon Based Solar Cell Devices,” in 2

4th European

Photovoltaic Solar Energy Conference, 21-25 September 2009, Hamburg,

Germany, 2009.

Y. G. Xiao, K. Uehara, M. Lestrade, Z. Q. Li, Z. M. S. Li, and H. Drive, “Modeling of Si-based Thin Film Triple-Junction Solar Cells,” in 34th IEEE PVSC, 2009.

A. Limmanee, K. Sriprapha, J. Sritharathikhun, and C. Piromjit, “ZnO interface layer and CO2 plasma treatment for improving efficiency of micromorph silicon solar cells,” Sol. Energy Mater. Sol. Cells, vol. 95, no. 1, pp. 146–149, Jan.

2011.

T. Soderstrom, F.-J. Haug, X. Niquille, V. Terrazzoni, and C. Ballif, “Asymmetric intermediate reflector for tandem micromorph thin film silicon solar cells,” Appl.

Phys. Lett., vol. 94, no. 6, p. 063501, 2009.

P. Sichanugrist, T. Sasaki, a. Asano, Y. Ichikawa, and H. Sakai, “Amorphous silicon oxide and its application to metal/n-i-p/ITO type a-Si solar cells,” Sol.

Energy Mater. Sol. Cells, vol. 34, no. 1–4, pp. 415–422, Sep. 1994.

T. Suzuki, A. Nakajima, T. Meguro, M. Yoshimi, M. Ichikawa, S. Fukuda, K.

Yamamoto, Y. Koi, T. Suezaki, Y. Tawada, T. Sawada, K. Hayashi, and H. Takata,

“Large area thin film Si module,”

Solar Energy Materials and Solar Cells, vol.

74. pp. 449–455, 2002.

S. Kirner, S. Calnan, O. Gabriel, S. Neubert, M. Zelt, B. Stannowski, B. Rech, and

R. Schlatmann, “An improved silicon-oxide-based intermediate-reflector for micromorph solar cells,” Phys. status solidi, vol. 9, no. 10–11, pp. 2145–2148,

Oct. 2012.

C. Das, A. Lambertz, J. Huepkes, W. Reetz, and F. Finger, “A constructive combination of antireflection and intermediate-reflector layers for a-Si/μc-Si thin film solar cells,”

Appl. Phys. Lett., vol. 92, no. 5, p. 053509, 2008.

C.-N. Li, H.-Y. Fang, Y.-H. Chen, C.-M. Yeh, C.-F. Huang, Y.-C. Wang,

K.-H. Hung, Y.-Y. Pan, C.-L. Wu, Y.-R. Chen, J.-C. Liu, and C.-H. Chen,

“Microcrystalline-Silicon-Oxide-Based N-Type Reflector Structure in

Micromorph Tandem Solar Cells,” Int. J. Photoenergy, vol. 2012, pp. 1–5, 2012.

M. Boccard, C. Battaglia, S. Haenni, K. Soderstrom, J. Escarre, S. Nicolay,

F. Meillaud, M. Despeisse, and C. Ballif, “Multiscale transparent electrode architecture for efficient light management and carrier collection in solar cells.,”

Nano Lett., vol. 12, no. 3, pp. 1344–8, Mar. 2012.

H. Tan, E. Psomadaki, O. Isabella, M. Fischer, P. Babal, R. Vasudevan, M. Zeman, and A. H. M. Smets, “Micro-textures for efficient light trapping and improved electrical performance in thin-film nanocrystalline silicon solar cells,”

Appl.

Phys. Lett., vol. 173905, no. 103, 2013.

121

BIBLIOGRAPHY

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

A. Lambertz, V. Smirnov, T. Merdzhanova, K. Ding, S. Haas, G. Jost, R. E. I.

Schropp, F. Finger, and U. Rau, “Microcrystalline silicon–oxygen alloys for application in silicon solar cells and modules,”

Sol. Energy Mater. Sol. Cells, vol.

119, pp. 134–143, Dec. 2013.

P. Delli Veneri, L. V. Mercaldo, and I. Usatii, “Silicon oxide based n-doped layer for improved performance of thin film silicon solar cells,” Appl. Phys. Lett., vol.

97, no. 2, p. 023512, 2010.

A. Lambertz, T. Grundler, and F. Finger, “Hydrogenated amorphous silicon oxide containing a microcrystalline silicon phase and usage as an intermediate reflector in thin-film silicon solar cells,” J. Appl. Phys., vol. 109, no. 11, p.

113109, 2011.

P. Buehlmann, J. Bailat, D. Domine, a. Billet, F. Meillaud, a. Feltrin, and C.

Ballif, “In situ silicon oxide based intermediate reflector for thin-film silicon micromorph solar cells,” Appl. Phys. Lett., vol. 91, no. 14, p. 143505, 2007.

P. Cuony, M. Marending, D. T. L. Alexander, M. Boccard, G. Bugnon, M.

Despeisse, and C. Ballif, “Mixed-phase p-type silicon oxide containing silicon nanocrystals and its role in thin-film silicon solar cells,” Appl. Phys. Lett., vol.

97, no. 21, p. 213502, 2010.

P. Cuony, D. T. L. Alexander, I. Perez-Wurfl, M. Despeisse, G. Bugnon, M.

Boccard, T. Soderstrom, A. Hessler-Wyser, C. Hebert, and C. Ballif, “Silicon filaments in silicon oxide for next-generation photovoltaics.,”

Adv. Mater., vol.

24, no. 9, pp. 1182–6, Mar. 2012.

P. Babal, J. Blanker, M. Workum, H. J. van Veen, R. Vasudevan, A. H. M. Smets, and M. Zeman, “Doped silicon oxide layers for thin film silicon solar cells,” in

Optical Society of America, 2012.

K. Schwanitz, S. Klein, T. Stolley, M. Rohde, D. Severin, and R. Trassl, “Antireflective microcrystalline silicon oxide p-layer for thin-film silicon solar cells on ZnO,” Sol. Energy Mater. Sol. Cells, vol. 105, pp. 187–191, Oct. 2012.

R. Biron, C. Pahud, F.-J. Haug, J. Escarre, K. Soderstrom, and C. Ballif,

“Window layer with p doped silicon oxide for high Voc thin-film silicon n-i-p solar cells,”

J. Appl. Phys., vol. 110, no. 12, p. 124511, 2011.

B. Vet, “Improvement of a-Si:H devices by analysis, simulations and experiment,” Delft University of Technology, 2014.

V. Demontis, C. Sanna, J. Melskens, R. Santbergen, a. H. M. Smets, a. Damiano, and M. Zeman, “The role of oxide interlayers in back reflector configurations for amorphous silicon solar cells,” J. Appl. Phys., vol. 113, no. 6, p. 064508,

2013.

J. M. Gee, “Optically enhanced absorption in thin silicon layers using photonic crystals,” in Proceedings of 29th IEEE Photovoltaic Specialists Conference

(New Orleans, LA), 2002, pp. 150–153.

O. Isabella, J. Krc, and M. Zeman, “Application of Photonic Crystals as Back

Reflectors in Thin-Film Silicon Solar Cells,” in 24th European Photovoltaic

Solar Energy Conference, 21-25 September 2009, Hamburg, Germany, 2009, pp. 2304–2309.

122

BIBLIOGRAPHY

[61]

[62]

[63]

[55]

[56]

[57]

[58]

[59]

[60]

[64]

[65]

[66]

[67]

[68]

P. G. O’Brien, a Chutinan, K. Leong, N. P. Kherani, G. a Ozin, and S. Zukotynski,

“Photonic crystal intermediate reflectors for micromorph solar cells: a comparative study.,” Opt. Express, vol. 18, no. 5, pp. 4478–90, Mar. 2010.

A. Bielawny, C. Rockstuhl, F. Lederer, and R. B. Wehrspohn, “Intermediate reflectors for enhanced top cell performance in photovoltaic thin-film tandem cells.,” Opt. Express, vol. 17, no. 10, pp. 8439–46, May 2009.

J. Krc, M. Zeman, S. L. Luxembourg, and M. Topic, “Modulated photoniccrystal structures as broadband back reflectors in thin-film solar cells,” Appl.

Phys. Lett., vol. 94, no. 15, p. 153501, 2009.

P. Obermeyer, C. Haase, and H. Stiebig, “Advanced light trapping management by diffractive interlayer for thin-film silicon solar cells,” Appl. Phys. Lett., vol.

92, no. 18, p. 181102, 2008.

M. Python, E. Vallat-Sauvain, J. Bailat, D. Domine, L. Fesquet, A. Shah, and C.

Ballif, “Relation between substrate surface morphology and microcrystalline silicon solar cell performance,” J. Non. Cryst. Solids, vol. 354, no. 19–25, pp.

2258–2262, May 2008.

Y. Nasuno, M. Kondo, and A. Matsuda, “Effects of Substrate Surface Morphology on Microcrystalline Silicon Solar Cells,” Jpn. J. Appl. Phys., vol. 40, no. Part 2,

No. 4A, pp. L303–L305, Apr. 2001.

M. Boccard, “Novel Micromorph Solar Cell Structures for Efficient Light

Trapping and High-Quality Absorber Layers,” EPFL, 2012.

Thermo Nicolet Corporation, Introduction to fourier transform infrared spectrometry. 2001.

K. Yamamoto, A. Nakajima, M. Yoshimi, T. Sawada, S. Fukuda, T. Suezaki, M.

Ichikawa, Y. Koi, M. Goto, T. Meguro, T. Matsuda, T. Sasaki, and Y. Tawada,

“High Efficiency Thin Film Silicon Hybrid Cell and Module with Newly Developed

Innovative Interlayer,” 2006 IEEE 4th World Conf. Photovolt. Energy Conf., vol.

1–2, pp. 1489–1492, 2006.

S. Kirner, “Development of Wide Band Gap Materials for Thin Film Silicon Solar

Cells,” Berlin University of Technology, 2013.

C. Smit, R. A. C. M. M. Van Swaaij, H. Donker, A. M. H. N. Petit, W. M. M.

Kessels, and M. C. M. Van De Sanden, “Determining the material structure of microcrystalline silicon from Raman spectra,” J. Appl. Phys., vol. 94, no. 5, p.

3582, 2003.

A. Sarker, C. Banerjee, and A. K. Barua, “Preparation and characterization of n-type microcrystalline hydrogenated silicon oxide,” vol. 1205, 2002.

G. Faraci, S. Gibilisco, P. Russo, A. Pennisi, and S. La Rosa, “Modified Raman confinement model for Si nanocrystals,” Phys. Rev. B, vol. 73, no. 3, p. 033307,

Jan. 2006.

C. T. Kirk, “Quantitative analysis of the effect of disorder-induced mode coupling on infrared absorption in silica,”

Physical Review B, vol. 38. pp. 1255–

1273, 1988.

123

BIBLIOGRAPHY

[69]

[70]

[71]

[72]

[73]

[74]

[75]

[76]

[77]

[78]

[79]

[80]

[81]

[82]

L. Murin, T. Hallberg, V. Markevich, and J. Lindstrom, “Experimental Evidence of the Oxygen Dimer in Silicon,” Phys. Rev. Lett., vol. 80, no. 1, pp. 93–96,

Jan. 1998.

A. C. Bronneberg, A. H. M. Smets, M. Creatore, and M. C. M. van de Sanden,

“On the oxidation mechanism of microcrystalline silicon thin films studied by

Fourier transform infrared spectroscopy,” J. Non. Cryst. Solids, vol. 357, no. 3, pp. 884–887, Feb. 2011.

A. H. M. Smets, W. M. M. Kessels, and M. C. M. Van De Sanden, “Vacancies and voids in hydrogenated amorphous silicon,” Appl. Phys. Lett., vol. 82, p. 1547,

2003.

V. A. Burrows and S. B. C. Y.J. Chabal, G.S. Higashi, K. Raghavachari, “Ideal hydrogen termination of the Si (111) surface,” Appl. Phys. Lett., vol. 53, pp.

998–1000, 1988.

M. Fehr, a. Schnegg, B. Rech, O. Astakhov, F. Finger, R. Bittl, C. Teutloff, and

K. Lips, “Metastable Defect Formation at Microvoids Identified as a Source of

Light-Induced Degradation”in Phys. Rev. Lett., vol. 112, no. 6, p. 066403, Feb.

2014.

Y. Toyoshima, K. Arai, A. Matsuda, and K. Tanaka, “Real time in situ observation of the film growth of hydrogenated amorphous silicon by infrared reflection absorption spectroscopy,” Appl. Phys. Lett., vol. 56, no. 16, p. 1540, 1990.

N. Blayo and B. Drevillon, “In situ study of the hydrogen rich overlayer at the a-Si:H surface by infrared ellipsometry,” J. Non. Cryst. Solids, vol. 138, pp.

771–774, 1991.

F.-J. Haug, T. Soderstrom, O. Cubero, V. Terrazzoni-Daudrix, and C. Ballif,

“Influence of the ZnO buffer on the guided mode structure in Si/ZnO/Ag multilayers,” J. Appl. Phys., vol. 106, no. 4, p. 044502, 2009.

J. Springer, “Absorption loss at nanorough silver back reflector of thin-film silicon solar cells,” J. Appl. Phys., vol. 95, no. 3, p. 1427, 2004.

F.-J. Haug, T. Soderstrom, O. Cubero, V. Terrazzoni-Daudrix, and C. Ballif,

“Plasmonic absorption in textured silver back reflectors of thin film solar cells,” J. Appl. Phys., vol. 104, no. 6, p. 064509, 2008.

H. Sai, H. Jia, and M. Kondo, “Impact of front and rear texture of thin-film microcrystalline silicon solar cells on their light trapping properties,” J. Appl.

Phys., vol. 108, no. 4, p. 044505, 2010.

“Measurement courtesy of Hairen Tan.” .

G. Bugnon, T. Soderstrom, S. Nicolay, L. Ding, M. Despeisse, A. Hedler, J.

Eberhardt, C. Wachtendorf, and C. Ballif, “LPCVD ZnO-based intermediate reflector for micromorph tandem solar cells,” Sol. Energy Mater. Sol. Cells, vol.

95, pp. 2161–2166, 2011.

R. Biron, S. Haenni, M. Boccard, C. Pahud, G. Bugnon, L. Ding, S. Nicolay, G.

Parascandolo, F. Meillaud, M. Despeisse, F. Haug, and C. Ballif, “Optimization of the Asymmetric Intermediate Reflector Morphology for High Stabilized

Efficiency Thin n-i-p Micromorph Solar Cells,” IEEE J. Photovoltaics, vol. 3, no. 1, pp. 41–45, Jan. 2013.

124

1.2. CHAPTER SUBTITTLE

[83]

[84]

[85]

[86]

[87]

[88]

[89]

[90]

[91]

V. Smirnov, a. Lambertz, B. Grootoonk, R. Carius, and F. Finger, “Microcrystalline silicon oxide (μc-SiOx:H) alloys: A versatile material for application in thin film silicon single and tandem junction solar cells,”

J. Non. Cryst. Solids, vol. 358, no. 17, pp. 1954–1957, Sep. 2012.

K. Jaeger, “On the Scalar Scattering Theory for Thin-film Solar Cells,” Delft

University of Technology, 2012.

D. Song, A. G. Aberle, and J. Xia, “Optimisation of ZnO:Al films by change of sputter gas pressure for solar cell application,” Appl. Surf. Sci., vol. 195, pp.

291–296, 2002.

S. Y. Myong and L. S. Jeon, “Improved light trapping in thin-film silicon solar cells via alternated n-type silicon oxide reflectors,” Sol. Energy Mater. Sol.

Cells, vol. 119, pp. 77–83, 2013.

M. Zeman, O. Isabella, P. Babal, A. H. M. Smets, and J. Krc, “Design and

Integration of 1-D Photonic Crystals in Thin-Film Silicon Solar Cells,” 2011.

M. Boccard, T. Soderstrom, P. Cuony, C. Battaglia, S. Haenni, S. Nicolay, L. Ding,

M. Benkhaira, G. Bugnon, A. Billet, M. Charriere, F. Meillaud, M. Despeisse, and C.

Ballif, “Optimization of ZnO Front Electrodes for High-Efficiency Micromorph

Thin-Film Si Solar Cells,” IEEE J. Photovoltaics, vol. 2, no. 3, pp. 229–235,

Jul. 2012.

B. Pieters, J. Krc, and M. Zeman, “Advanced Numerical Simulation Tool for

Solar Cells - ASA5,” in 2006 IEEE 4th World Conference on Photovoltaic

Energy Conference, 2006, vol. 2, pp. 1513–1516.

“ASA manual.” [Online]. Available: http://www.ewi.tudelft.nl/overde-faculteit/afdelingen/electrical-sustainable-energy/photovoltaischematerialen-en-devices/asa-software/.

J. Sap, “Modelling optical properties of layers for thin-film silicon solar cells,”

Delft University of Technology, 2010.

125

SUMMARY

In summary, this thesis shows the development and nanostructure analysis of doped silicon oxide layers. These layers are applied in thinfilm silicon single and double junction solar cells. Concepts of intermediate reflectors (IR), consisting of silicon and/or zinc oxide, are applied in tandem cells. Multi-stack Bragg reflector IRs are tested in tandem cells, increasing the top cell current output. Finally, mechanical polishing is applied on intermediate reflectors, creating asymmetrically textured IRs.

Doped silicon oxide layers have proven their versatility as multipurpose layers in thin-film silicon solar cells. In chapter 3, the search for device grade n- and p-doped silicon oxide material is described. The nanostructure of silicon oxide films with a wide array of optical and electrical properties is studied in detail by TEM, Raman,

FTIR and XPS. Silicon oxide is found to be a very heterogeneous material with complex nanostructure. Both the amorphous and crystalline phases of silicon oxide are studied in detail. Differences are found between the p- and n-doped materials. It is found that the n-doped material has a nanostructure of silicon crystal grains embedded in an amorphous silicon oxide matrix. The p-doped material, however, contains silicon filaments in an amorphous silicon oxide matrix. These filaments are of intrinsic amorphous silicon with crystalline silicon grains. Intrinsic amorphous silicon is mainly responsible for good conductivity in both n-doped and p-doped silicon oxide, however, minimum crystalline content is also required. Finally, the relations between each phase and element content is related to optical and electrical properties.

N-doped silicon oxide used as a back reflector in single junction solar cells reflects unabsorbed light back into the absorber layer, increasing its current output. The blue part of the spectrum is absorbed in one pass, therefore the response in the red part of the spectrum is expected to increase. However, an increase in the blue part of the spectrum is observed as well and is the topic of chapter 4. This increase is attributed to a combination of factors, but mostly to the prevention of a native oxide formation on the standard a-Si:H n-layer. The standard n-layer is covered with the n-doped silicon oxide layer which prevents the standard layer from oxidizing in ambient air. The silicon oxide also provides a better contact interface with silver. Other factors increasing the blue

127

SUMMARY response include: 1. The lower activation energy of n-doped silicon oxide in comparison with the standard a-Si:H n-layer. 2. The changing of the band states due to the larger bandgap of n-doped silicon oxide in reference to n-doped a-Si:H. 3. The thinner a-Si:H n-layer as the one in the reference cell is twice as thick. 4. The lower parasitic plasmonic absorption in the silver back contact due to the common interface with silicon oxide. P-doped silicon oxide exhibits anti-reflective properties, increasing cell current output in the blue part of the spectrum as well. An initial efficiency of 11.4% is achieved with the application of both p- and n-doped silicon oxide layers in a single junction a-Si:H solar cell.

Intermediate reflector concepts are explored in chapter 5. Distributed

Bragg Reflectors (DBR) have tunable reflective properties and are an interesting candidate for intermediate reflectors in tandem cells. They exhibit nearly the same reflectance range under various angles of incidence. DBRs can be easily designed with the help of optical simulation software such as ASA. The design sequence is as follows: 1. The desired reflectance range inside a solar cell is simulated by varying the thickness of each material. 2. This stack is then simulated in a glass – air environment. 3. The stack is deposited on a glass substrate. 4. The measured reflectance is compared with the air – glass simulation. If a good fit is achieved, the DBR will give the desired simulated reflectance inside the cell. DBRs greatly enhance the top cell current in a tandem cell, reaching up to 13,5 mA/cm

2

in a 175 nm-thick a-Si:H layer. Current matching and lowering of V

OC

remain issues.

Texture control in the IR is important in order to provide good light scattering for both the top and bottom cells of a tandem and to provide a good substrate for the growth of a defect-free nanocrystalline absorber layer. An approach to modify the texture of ZnO serving as an asymmetric IR in a tandem cell is developed. Because of the excellent performance of the top amorphous silicon cell deposited on an Asahi

VU substrate, it is beneficial to keep this substrate texture for the top cell and integrate different textures (with larger surface features) in the layers processed after the top cell. Two approaches to create an asymmetrically-textured IR are chosen: wet etching and mechanical polishing. The wet etching approach is done with two dilution levels of

HCl. Then the IR interface facing the top cell has a typical Asahi VU texture while the IR interface facing the bottom cell has larger surface features beneficial for long-wavelength scattering. The second approach is about applying mechanical polishing to silicon oxide and ZnO IRs. This approach successfully flattened the Asahi-induced texture, leaving it in the IR interface facing the top cell and on the other flat side allowing higher-quality nc-Si:H growth.

128

SAMENVATTING

Dit proefschrift gaat over de ontwikkeling van gedoteerde siliciumoxide

(SiO

X

) lagen en de nanostructurele analyse daarvan. Deze lagen worden in single- en multi-junction dunnefilm silicium zonnecellen toegepast.

Verschillende concepten van tussenliggende reflectoren (IR van Engels intermediate reflectors), die uit silicium- en of zinkoxide (ZnO) bestaan, worden in tandemcellen toegepast. Multilagen-Bragg-reflector IRs worden in tandemcellen getest, waar ze de stroom in the topcellen verhogen. Uiteindelijk wordt mechanisch polijsten van de IRs toegepast, wat tot een asymmetrische texturisatie van de IRs leidt.

Het is vaak aangetoond dat gedoteerde siliciumoxide lagen in dunnefilm silicium zonnecellen op zeer verschillende manieren toegepast kunnen worden. In hoofdstuk 3, wordt de zoektocht naar n- en p-gedoteerd

SiO

X

van device-grade kwaliteit voor zonnecellen beschreven. De nanostructuur van SiO

X

lagen met verschillende optische en elektrische eigenschappen wordt gedetailleerd onderzocht met TEM, Raman, FTIR en XPS. Het wordt ontdekt dat siliciumoxide een zeer heterogeen materiaal met een complexe nanostructuur is. Zowel de amorfe alsook de kristallijne fase worden in detail onderzocht. Verschillen tussen n- en p-gedoteerd materiaal worden gevonden: n-gedoteerd SiO

X

heeft een nanostructuur van kristallijne korrels die in een matrix van amorf SiO

X

in zijn gebed. In tegenstelling heeft het p-gedoteerde materiaal filamenten van silicium in een matrix van amorf SiO

X

. Intrinsiek amorf silicium

(a-Si:H) is hoofdverantwoordelijk voor de goede geleiding zowel in n- alsook in p-gedoteerd materiaal. Uiteindelijk worden de relaties tussen de fases en de elementaire concentraties in verband gebracht met de optische en de elektrische eigenschappen.

n-gedoteerd SiO

X

dat als reflector aan de achterkant van singlejunction zonnecellen wordt gebruikt, reflecteert niet-geabsorbeerd licht terug naar de absorberlaag, waardoor de stroom van de zonnecel wordt verhoogt. Het blauwe gedeelte van het zonnespectrum wordt al in de eerste doorgang door de absorber volledig geabsorbeert, dus is de verwachting dat het effect van de achterreflector vooral in het rode gedeelte van het spectrum te zien is. Hoe dan ook wordt ook in het blauwe gedeelte een toename gezien, wat het onderwerp van hoofdstuk

4 is. Deze toename wordt aan verschillende factoren toegeschreven; het

129

SAMENVATTING hoofdeffect is dat door de SiO

X

-laag geen native oxidelaag aan de a-Si:H n-laag kan ontstaan. Deze n-laag wordt bedekt met een n-gedoteerde

SiO

X

laag, waardoor voorkomen wordt dat de a-Si:H n-laag in de omgevingslucht kan oxideren. Verder is het contact tussen siliciumoxide en zilver beter dan het standaard contact tussen amorf silicium en zilver.

Ander factoren die aan het verbeteren van de zonnecelstroom in het blauw bijdragen zijn: 1. De lagere activatieenergie van n-type SiO

X ten opzicht van het standaard n-type a-Si:H. 2. Het veranderen van de bandtoestanden vanwege de hogere bandkloof van n-type SiO

X

ten opzicht van een n-gedoteerde a-Si:H laag. 3. De dunnere a-Si:H n-laag

(in de referentiecel is deze laag twee keer zo dik). 4. De verlaagde parasitaire plasmonische absorptie in de zilveren achtercontact vanwege de gezamenlijke tussenlaag. p-gedoteerd SiO

X

toont anti-reflective eigenschappen, waardoor ook de stroom in het blauw wordt verhoogt.

Door het toepassen van zowel p- alsook n-gedoteerd SiO

X

wordt een initieel rendement van 11.4% voor een single-junction a-Si:H zonnecel behaalt.

Verschillende concepten voor tussenliggende reflectoren worden in hoofdstuk 5 onderzocht. Verdeelde Bragg reflectoren (DBR van Engels

Distributed Bragg Reflectors) hebben goed controleerbare reflectieeigenschappen en zijn dus een veelbelovende kandidaat voor IRs in tandemcellen. DBRs hebben bijna dezelfde reflectiviteit voor verschillende invalshoeken van het licht. Middels optische simulatiessoftware zoals ASA kunnen DBRs eenvoudig worden ontworpen. Het ontwerp wordt in deze volgorde gemaakt: 1. Het gewenste reflectiebereik binnen de zonnecel wordt door het varieren van de dikte van elk materiaal gesimuleerd. 2. Dit multilagen system wordt dan in een glas-lucht omgeving gesimuleerd.

3. Het multilagen system wordt op glas gedeponeerd. 4. De gemeten reflectiviteit wordt met de gesimuleerde warden vergeleken. Indien een goede overeenstemming wordt gevonden, zal de DBR de gewenste reflectiviteit binnen de cel geven. DBRs kunnen de stroom in de topcell aanzienlijk verhogen, tot waardes van 13.5 mA/cm 2 in een 175 nm dikke a-Si:H laag. Maar het evenaren van de stromen van de top en de bottomcel en de verlaagde openklemspanning blijven problemen.

Het controleren van de texturisatie van de IRs is belangrijk voor het waarborgen van goede verstrooiing van het licht in zowel de top- en de bottomcel van een tandemcel. Verder is texturisatie belangrijk voor de defectvrije groei van een nanokristallijne absorberlaag. Een aanpak om de texturisatie van zinkoxide aan te passen wordt ontwikkelt; op deze manier kan het zinkoxide als asymmetrische IR in een tandemcel dienen.

De topcel wordt op Asahi VU gedeponeerd. Omdat de textuur van dit materiaal tot excellente prestaties van de topcel leidt, is het voordelig

130

SAMENVATTING om deze textuur te behouden en verschillende texturen (met grotere karakteristieken) in de lagen achter de top cel te introduceren. Twee manieren om asymmetrische IRs te maken worden gekozen: nat-chemisch etsen en mechanisch polijsten. De nat-chemische proces wordt met zoutzuur met twee verdunningen gedaan. Daarna heeft de IR tussenlaag aan de topcel-kant de typische Asahi VU textuur terwijl de tussenlaag aan de bottomcel-kant grotere kenmerken heeft, die geschikter zijn voor verstrooiing bij grote golflengtes. In de tweede aanpak worden de siliciumoxide en de zincoxide IRs mechanisch gepolijst. Met deze aanpak kan de door Asahi VU geintroduceerde textuur succesvol afgevlakt worden, zodat op de vlakke tussenlaag van de IR aan de bottom-cel kant nc-Si:H van hoge kwaliteit gedeponeerd kan worden.

131

LIST OF PUBLICATIONS

A. H. M. Smets, R. Vasudevan,

P. Babal

, J. Melskens, M. Fischer, and M. Zeman,“Recent progress in a-Si:H solar cells using approaches based on nanostructure engineering and

M. Zeman, O. Isabella, K. Jaeger,

P. Babal

, S. Solntsev, and R. Santbergen,“Modeling of Advanced Light Trapping Approaches in Thin-Film Silicon Solar Cells,”

MRS Proc., vol. 1321, pp. a23–05, 2011

P. Babal

, J. Blanker, R. Vasudevan, A. H. M. Smets, and M. Zeman, “Microstructure analysis of n-doped μc-SiOx:H reflector layers and their implementation in stable a-Si:H p-i-n junctions,”* in Conference Record of the IEEE Photovoltaic Specialists

Conference, 2012, pp. 321–326

* IEEE best poster award received for this work

P. Babal

, H. J. van Veen, M. Workum, A. Smets, and M. Zeman, “Doped silicon oxide layers for tandem silicon solar cells,”in Renewable Energy and the Environment Optics and Photonics Congress, p. PW1B.4, 2012

P. Babal

, H. Lopez, L. Xie, B. van Veen, M. van Sebille, H. Tan, M. Zeman, and A.

Proceedings - 28th European Photovoltaic Solar Energy Conference and Exhibition (pp.

2580-2587), 2013

H. Tan, E. Psomadaki, O. Isabella, M. Fischer, P. Babal, R. Vasudevan, M. Zeman, and

A. H. M. Smets, “Micro-textures for efficient light trapping and improved electrical performance in thin-film nanocrystalline silicon solar cells,” Appl. Phys. Lett., vol.

173905, no. 103, 2013

D. Y. Kim,

P. Babal

, R. A. C. M. M. van Swaaij, and M. Zeman, “a-SiOx:H single junction solar cell improved by optimizing n-layers and its application to multi-junction solar cells,”Photovoltaic Science and Engineering Conference (PVSEC)-23, 2013

H. Tan,

P. Babal

, M. Zeman, and A. H. M. Smets, “ Wide bandgap p-type nanocrystalline silicon oxide as window layer for high performance thin-film silicon multi-junction solar cells,” Solar Energy Materials and Solar Cells, 2014, submitted

133

ACKNOWLEDGEMENTS

Doing a PhD and finalizing it with a thesis is an excruciating task that cannot be done without the help and support of many special people. I would therefore like to start this list by thanking Professor Miro Zeman for giving me the opportunity to do this fouryear project at a world-renowned institution as the Delft University of Technology is. He trusted me with the job despite that at our first interview I did not understand the basics of solar cell operation. I owe you a “veľké ďakujem”.

My interest in the nanostructure of semiconductors increased tremendously thanks to the guidance from Arno Smets, my daily supervisor. It was a pleasure working with you! I am grateful for all the clear explanations and problem analyses that you helped me with. I also highly appreciate your bohemic spirit and your never-miss-a-party attitude.

I would also like to thank René van Swaaij for a nice semiconductor physics course and consultancy possibilities. And of course my gratitude is deserved by my defense committee members, prof. dr. J.A. La Poutre, prof. dr. L.D.A. Siebbeles, prof. dr. E. Vlieg, dr. W.

Soppe, and dr. A. Gordijn, for taking the time to criticize my work.

In the beginning of my project, as a new member of the PVMD group, I was introduced to the equipment and the way the group works by fellow PhD students

Michael, Solomon, Bas, Dong, Guangtao, Joke, Klaus (cheers for translating the summary of this thesis), Marinus, and Olindo. Thank you for all your patience, from explaining deposition and measurement equipment to physics and solar cell operation. My gratitude goes as well to newer PhD students in the group, namely Jimmy (cheers for translating the propositions), Lihao (thank you for helping me arrange TEM images), Martijn and Mark

(helped arrange XPS measurements), Hairen (for many discussions and an unforgettable trip in the US after the IEEE). I also shouldn’t forget Dimitris, Andrea, Miriam, Fai Tong, and Wendelin, I acknowledge you all!

I would also like to thank the post-docs for all their help, support, and guidance, namely Serge, Sergiy, Karol, Rudi, Doyun, and Tristan. A special word goes to Tristan; it was great to have you as a colleague, thanks for your humor, from all the countless beers sessions to burning pancakes. A special thank you also goes to the PVMD group technicians Stefaan, Martijn, Remko, and Jan Chris. Guys, thanks to your availability and willingness to help could I have done my experiments. You helped me tame the AMIGO, that says it all. I am also very happy that I have had the opportunity to collaborate with industry and to see my know-how applied in the “real world”. For this I thank Edward hammers from HyET solar who gave me the opportunity to see silicon oxide applied on a m

2

scale instead of mm

2

.

Another special place on this page goes to my students: Ravi, Diego, Johan, Bert,

135

ACKNOWLEDGEMENTS

Hector, Boris, Carla. Whether doing a MSc thesis or just a SIP II project, your hard work substantially contributed to this thesis becoming reality. I also want to thank Albert for

XRD measurements, Valeria, and all other master students in the PVMD group. I had the pleasure to work with Paula, Aditya, David, Shin, Nishant at the PV LAB. Thank you for making it such a nice learning experience.

Certain experiments conducted for the purpose of this thesis were carried out in

DIMES. Therefore I owe my gratitude to Chuck for the safety course, Johan for teaching me how to operate the wafer polisher, Mario for explaining the SEM, and Cassan for general support. My thank you for great administrative support goes to the department secretaries Laura, Rinske, Marian, Weby, and Sharmilla. Thank you very much for all your help in navigating through bureaucracy and administrative procedures, processes sometimes so complicated, that I almost made a proposition out of them.

During my PhD I got to know great friends who helped me reset my mind when I needed it the most. Yes, I’m talking about you: Rodrigo, Francesca, Gis, Claudia, Maria,

Mária, Richie, Paola, Camilo, Amanda, Mónica, Pepijn, Caesar, and Asti. You have made sure that I don’t forget that life is also about having fun. Plus, all those parties benefited my productivity the days after, thanx guys! Greetings go as well to my family and friends in

SK, who if I should name them all would make another book - ďakujem za vašu podporu.

A špeciálne ďakujem patrí tým ktorí mi spríjemnili môj pobyt v Delfte svojou návštevou:

Alby, Baška, Andra, Tomáš, Adam, Mižu, Dano, Muko, a Nika. A samozrejme Marek

(thanks for the cover photo!) a Sisa. And to my family in Colombia, thank you for your big support from afar – muchas gracias por todo!

Now a special word to the people closest to my heart. To my parents: Od malička ste ma viedli k pracovitosti a poznaniu. Nie vždy to vychádzalo podľa plánov, no dúfam že môžete byť na mňa hrdí. Srdečne vám ďakujem za všetku vašu lásku a podporu. To

Claudia: My love, you understood when I had to work instead of spending time together, what more, you helped me format the final design of this thesis. Especially during the pregnancy with Katerina you were very patient and forgiving when I could not give you the attention you deserved. Thank you for being my joy and inspiration. Clau, muchas gracias por tu amor y soporte, no podria hacerlo sin ti.

136

ABOUT THE AUTHOR

Pavel Babál was born in Bratislava, Slovakia in 1986.

He received his mechanical engineering degree (Ing.) from the Slovak University of Technology in 2009. In that same year he joined the Photovoltaic Materials and Devices group of Prof. dr. M. Zeman at the Delft University of

Technology to pursue his PhD. He carried out research on doped silicon oxide alloys and intermediate reflectors for thin-film silicon solar cells under the guidance of dr. A.

Smets. Since April 2014 he works as a product engineer on the Brewer spectrophotometer at Kipp & Zonen in Delft, the Netherlands.

137

Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement