Document [original]

Development of heterocontacts for high

efficiency silicon solar cells

vorgelegt von

M. Sc.

Nathan Nicholson

ORCID: 0009-0007-6761-7792

an der Fakultät IV – Elektrotechnik und Informatik

der Technischen Universität Berlin

zur Erlangung des akademischen Grades

Doktor der Ingenieurwissenschaften

- Dr.-Ing. -

Promotionsausschuss:

Vorsitzender: Prof. Dr. Bernd Szyszka

Gutachter: Prof. Dr. Bernd Rech

Gutachter: Prof. Dr. Steve Albrecht

Gutachter: Prof. Dr. Olindo Isabella

Tag der wissenschaftlichen Aussprache: 04. Oktober 2023

Berlin 2024

Abstract

Silicon solar cells using passivating contacts consisting of an ultrathin silicon oxide (SiOx) layer

and a doped polycrystalline silicon layer (poly-Si) are assumed to become the next generation

of industrial solar cells. By realising these passivating contacts on the rear of silicon solar cells,

high efficiencies above 25% are achievable. In the last years, literature regarding these pas-

sivating SiOx/poly-Si contacts focused on the techniques one should employ especially for in-

dustrial applications.

This thesis focuses on passivating contacts that were grown using wet-chemical ozone oxida-

tion for the SiOx layer and electron beam evaporation for the initially amorphous silicon layer.

Fabricated solar cells using amorphous silicon heterojunction (SHJ) structures on the front and

a passivating SiOx/poly-Si contact on the rear of the device show efficiencies of ~20% with

symmetrical SiOx/poly-Si lifetime structures achieving implied open-circuit voltages of up to

735 mV. Cell efficiencies were partly limited by the processing scheme chosen, affecting

mostly the surface passivation quality at the front of the device and the short-circuit current

density resulting in a loss of ~3 mA/cm2 as shown via GenPro4 simulations. Based on simula-

tions conducted, an efficiency of ~22% can be expected for cells using passivating contacts

similar to the efficiency of SHJ devices processed in this thesis.

Investigations relating to the wet-chemical ozone-based silicon oxide layer focused mostly on

structural changes that occur during the fabrication of the passivating contact. Initially, ozone-

based silicon oxide layers show no significant surface passivation with minimal defect densities

of ~1013 cm-2eV-1. Structural changes are expected to occur during a high-temperature anneal-

ing step that is performed for the passivating contact. X-ray photoelectron spectroscopy (XPS)

measurements conducted on oxidized <111> silicon wafers showed that the annealing step

reduces the overall full width at half maximum (FWHM) of the Si3+ oxidation state of the Si 2p

orbital from ~1.35 eV to ~1.02-1.07 eV. The reduction in FHWM is assumed to be linked to a

reduction in defect concentrations present at the interface between silicon oxide and substrate.

No significant change could be observed for the other sub-oxide states, besides a growth in

the bulk silicon dioxide layer.

Lastly, the process used to provide doping for the silicon layers used in this thesis, plasma

doping, was investigated using secondary ion mass spectroscopy (SIMS) and electrochemical

capacitance voltage (ECV) measurements. Total phosphorous concentrations of up to 2x1020

atoms per cm3 were reached for 300 nm thick poly-Si layers using annealing temperatures

above 860°C. Higher temperatures resulted in layers with a more even distribution in Phos-

phorous. Concentrations in electrically active Phosphorous reached similar levels, however,

showed a higher temperature-dependance resulting in larger doping gradients throughout lay-

ers where lower annealing temperatures were used. The shape of doping profiles appeared

similar to ones achieved with more commonly employed doping techniques such as POCL

diffusion.

Zusammenfassung

Derzeit wird angenommen, dass die nächste Generation von Siliziumsolarzellen passivierende

Kontakte verwenden wird, die aus einer ultradünnen Siliziumoxidschicht (SiOx) und einer do-

tierten polykristallinen Siliziumschicht (poly-Si) bestehen. Solarzellen, die den Kontakt auf der

Rückseite verwenden, zeigten Wirkungsgrade über 25%. In den letzten Jahren wurden vor

allem die Verwendung verschiedener Prozesse für die Herstellung von SiOx/poly-Si Kontakten

untersucht auch unter Berücksichtigung einer industriellen Anwendung.

In dieser Dissertation wurde die Verwendung von nasschemischer Ozonoxidation und Elekt-

ronenstrahlverdampfung für die Siliziumschicht untersucht. Resultierende Solarzellen, die

amorphe Siliziumheterokontakte (SHJ) auf der Vorderseite nutzen, zeigten Wirkungsgrade von

~20% mit einer implizierten Leerlaufspannung von bis zu 735 mV für symmetrische SiOx/poly-

Si Lebensdauerproben. Der Wirkungsgrad von Zellen wurde durch die verwendete Prozessab-

folge begrenzt, welche vor allem Einfluss auf die Passivierung der Vorderseite und die Kurz-

schlussstromdichte hatte. Simulationen, die in GenPro4 durchgeführt wurden, zeigten einen

Verlust von ~3 mA/cm2. Laut Simulationen können Wirkungsgrade von bis zu 22% für die un-

tersuchten Strukturen erwartet werden sofern die zuvor genannten Probleme behoben werden

können, wodurch der Wirkungsgrad vergleichbar ist mit SHJ Zellen, die für diese Arbeit her-

gestellt wurden.

Siliziumoxidschichten wurden mittels Röntgenphotoelektronenspektroskopie (XPS) untersucht

um strukturelle Änderung im Siliziumoxid zu bestimmen, die sich während der Herstellung des

SiOx/poly-Si Kontaktes ergeben. Im frisch gewachsenen Zustand zeigen oxidierte <111> Sili-

ziumwafer keine signifikante Oberflächenpassivierung mit einer minimale Defektdichte von

~1013 cm-2eV-1. Änderung in der Struktur der Oxide sind vor allem nach dem Hochtemperatur-

schritt zu erwarten, der für die Herstellung der Kontakte notwendig ist. XPS-Messungen zeig-

ten, dass die volle Halbwertsbreite des Si3+ Oxidationszustands im Si 2p Signal von ~1,35 eV

auf 1,02-1-07 eV sinkt. Es wird angenommen, dass diese Änderung eine Reduktion an Defek-

ten darstellt, die an der Grenzfläche zwischen Siliziumsubstrat und Siliziumoxid liegen. Neben

einem Wachstum in der Siliziumdioxidschicht konnten keine weiteren markanten Änderungen

in der Struktur gefunden werden.

Zuletzt wurde der Plasmadotierprozess, der in dieser Arbeit für die Dotierung der poly-Si-

Schicht genutzt wurde, mittels Sekundärionen-Massenspektrometrie (SIMS) und elektroche-

mischer Kapazität-Spannungs-Messungen (ECV) untersucht. Phosphorkonzentrationen von

bis zu 2x1020 Atome pro cm3 konnten in 300 nm dicken poly-Si Schichten erzielt werden, die

bei mind. 860°C getempert wurden. Durch Erhöhung der Temperatur konnten homogenerer

Phosphorverteilungen in der poly-Si Schicht erzielt werden. Für die elektrisch aktive Phosphor-

konzentration war dieser Zusammenhang stärker ausgeprägt mit höheren Konzentrationsgra-

dienten bei 860°C. Insgesamt waren die resultierenden Dotierprofile qualitativ vergleichbar mit

Dotierprofilen, die mit typischen Methoden z.B. POCL-Diffusion erstellt werden.

Table of content

1 Introduction .................................................................................................................... 1

2 Fundamentals ................................................................................................................ 5

2.1 Working principle of a solar cell ............................................................................... 5

2.1.1 Solar cell under operation conditions ................................................................ 6

2.2 Recombination mechanisms ................................................................................... 9

2.3 Doping of silicon .....................................................................................................11

2.4 Solar cells with passivated surfaces .......................................................................13

2.4.1 Poly-silicon based passivating contacts...........................................................14

2.4.2 a-Si:H/c-Si solar cells ......................................................................................18

2.5 Oxidation of silicon .................................................................................................19

2.5.1 Wet-chemical oxidation of silicon.....................................................................22

3 Deposition and characterisation methods ......................................................................27

3.1 Deposition methods ................................................................................................27

3.1.1 Wet-chemical ozone oxidation .........................................................................27

3.1.2 Electron beam evaporation..............................................................................27

3.1.3 Plasma enhanced chemical vapour deposition ................................................28

3.1.4 Plasma doping ................................................................................................29

3.2 Characterisation .....................................................................................................30

3.2.1 Photoconductance decay ................................................................................30

3.2.2 Photoluminescence .........................................................................................33

3.2.3 JV characteristics ............................................................................................33

3.2.4 External quantum efficiency ............................................................................34

3.2.5 Transfer Length method ..................................................................................36

3.2.6 Ellipsometry ....................................................................................................37

3.2.7 Optical simulations in GenPro4 .......................................................................38

3.2.8 Transmission electron spectroscopy ...............................................................39

3.2.9 X-ray photoelectron spectroscopy ...................................................................39

3.2.10 Modelling of X-ray photoelectron spectroscopy data .......................................42

3.2.11 Secondary ion mass spectrometry ..................................................................43

3.2.12 Electrochemical capacitance voltage profiling .................................................45

4 Wet-chemical ozone-based silicon oxide layers for passivating contacts .......................47

4.1 XPS analysis of wet-chemical ozone-based oxides ................................................47

4.2 XPS sample preparation ........................................................................................48

4.3 Determining the thickness of ultra-thin silicon oxides ..............................................50

4.4 Bulk silicon oxide properties ...................................................................................53

4.5 Evolution of silicon oxide structure and composition during annealing ....................55

4.6 Summary of chapter ...............................................................................................59

5 Plasma doping for passivating contacts .........................................................................61

5.1 SIMS profiles for passivating contacts ....................................................................61

5.2 Initial diffusion of phosphorous ...............................................................................63

5.3 Diffusion and activation of phosphorous during the high-temperature annealing step

5.4 Comparison of plasma doping to other doping techniques for passivating contacts68

5.5 Results and discussion ...........................................................................................70

6 Electron-beam evaporation and wet-chemical ozone-based passivating contacts .........73

6.1 Determining process conditions for passivating contacts ........................................73

6.2 Passivating qualities of passivating contacts ..........................................................75

6.3 Summary ................................................................................................................79

7 Integration of passivating contacts into solar cells .........................................................81

7.1 Solar cell structure and processing scheme ...........................................................81

7.2 Results of measurements .......................................................................................83

7.3 Investigation of different contacting schemes for passivating contacts ...................88

7.4 Efficiency potential of investigated cell concepts ....................................................89

7.5 Summary and conclusions .....................................................................................91

8 Prospects of technologies investigated ..........................................................................93

8.1 Electron-beam evaporated layers and wet-chemical ozone oxides for passivating

contacts ............................................................................................................................93

8.2 TOPCon-like passivating contacts for silicon/perovskite tandem solar cells ...........96

9 Conclusions and outlook ...............................................................................................99

9.1 Outlook ................................................................................................................. 100

Acknowledgements ............................................................................................................ 116

Publications ........................................................................................................................ 117

List of figures

Figure 2.1 Schematic band diagram for a PERC silicon solar cell with aluminium-diffused back

surface field. Figure taken from [26] and edited. ............................................................................. 6

Figure 2.2 I-V characteristics of a solar cell under illumination. Voltage-dependent power as

well as point of maximum power are shown in blue. Recreated from [28]. ................................. 7

Figure 2.3 1 and 2 diode electrical circuit models according to eq. 5 and 6. Figures taken from

[27]. ......................................................................................................................................................... 7

Figure 2.4 Generation and recombination of charge carriers within a semiconductor. Each

recombination mechanism is shown separately. Figure was recreated from [31]. ..................... 9

Figure 2.5 Simplified two-dimensional atomic structure of intrinsic and doped mono-crystalline

silicon. Dopants are built into the crystalline structure of silicon and replace silicon atoms. Free

charge carriers result from a lack of bonding partners. Figure taken from [41]. ....................... 12

Figure 2.6 Doping profile for a TOPCon contact generated using electrochemical capacitance

voltage measurements (ECV). A significant drop in doping concentration can be observed at

the interfacial region between poly-Si and c-Si. b) shows a schematic of the interfacial region

of TOPCon contacts. Included in the schematic are pinholes and areas with reduced oxide

thickness that enable dopants to more easily travel across the interfacial region. Figure is taken

from [45]. .............................................................................................................................................. 13

Figure 2.7 Schematic of a n-type TOPCon solar cell using a passivating SiOx/poly-Si contact

on the rear of the device. The front is passivated using AlOx and combined with an antireflection

coating to improve the optical properties of the device. The figure was taken from [63]. ....... 15

Figure 2.8 Schematic energy band diagram for n-type TOPCon passivating contacts. Holes

are less likely to tunnel through the oxide due to the higher energy barrier that silicon oxide

presents for holes compared to electrons. Figure recreated from [70]. ..................................... 17

Figure 2.9 Energy band diagram and schematic of the structure of an a-Si:H/c-Si silicon

heterojunction solar cell. Energy barriers caused by a mismatch between the bandgap of a-

Si:H and c-Si are shown on the left and right side of the substrate. Figure taken from [80]. .. 19

Figure 2.10 Schematic describing the oxidation of silicon including the separate zones that

are formed during the process. Figure taken from [97]. ............................................................... 21

Figure 2.11 Thickness of silicon oxide layers grown in solutions containing 20 ppm ozone as

calculated from spectral ellipsometry data for different oxidation durations. Data is compared

with two different models that describe the oxidation of silicon. .................................................. 22

Figure 3.1 Results from photoconductance decay measurements showing a) the relation

between implied open-circuit voltage and illumination intensity and b) the effective lifetime of

minority charge carriers for different minority charge carrier concentrations. ........................... 32

Figure 3.2 External quantum efficiency of a solar cell. Deviations between the ideal and

measured quantum efficiency can be connected to losses occurring at the front or rear of the

cell as well as optical properties. ...................................................................................................... 35

Figure 3.3 Schematic of structures used for TLM measurements (a). Method to determine the

contact resistance and transfer length from TLM measurements (b). Figures recreated from

[121] and taken from [122]. ............................................................................................................... 36

Figure 3.4 Deconvolution of the Si 2p spectrum for a silicon wafer oxidized in an ozone

containing solution measured by XPS using an excitation energy of 170 eV. The background

caused by inelastically scattered electrons has already been subtracted from this plot. ........ 41

Figure 3.5 Schematic of a tool used for time-of-flight secondary ion mass spectrometry. Figure

taken from [133]. ................................................................................................................................. 44

Figure 3.6 Schematic of electrochemical cell used for ECV measurements. Samples are

attached to the plunger on the right side. Figure taken from [135]. ............................................ 46

Figure 4.1 XPS surveys for oxidised silicon samples measured using synchrotron radiation

(hν = 1keV). ......................................................................................................................................... 48

Figure 4.2 Process scheme used for XPS measurements. Experiments included freshly grown

oxides as well as oxides that were previously part of a passivating contact. ............................ 49

Figure 4.3 SEM images of c-Si wafers with silicon dioxide and doped poly-Si on top before a)

and after b) etching the samples in TMAH solutions. ................................................................... 50

Figure 4.4 Oxide thickness derived from XPS data using the method described by M. P Seah

and S. J. Spencer and ellipsometry data for oxides grown in 30 ppm ozone solution for different

durations. ............................................................................................................................................. 52

Figure 4.5 TEM measurements performed on an oxidised c-Si wafer. Measurements show the

c-Si substrate (bottom left), the ultra-thin silicon oxide layer and glue used for sample

preparation. .......................................................................................................................................... 52

Figure 4.6 O 1s spectra for non-annealed silicon oxide samples a) that can be adequately

fitted using one peak (top left) or b) require additional peaks (top right). The same behaviour

is shown for samples annealed at 900°C (c and d). Fits were performed assuming one or two

peaks. Samples on the left represent the quality of fits achieved for most samples. Samples

on the right are outliers that showed noticeable asymmetries. ................................................... 54

Figure 4.7 Si 2p spectra for a) freshly grown silicon oxides, b) oxides that were annealed within

a passivating contact at 800°C and c) 900°C after poly-Si removal in TMAH. Shirley

backgrounds were subtracted before normalising the intensity of each spectrum to the intensity

of the Si4+ peak. The effects of charging were compensated by moving each spectrum so that

the Si0 peaks would align. Measurements were performed at 900°C. ....................................... 56

Figure 4.8 Si1+ Sub-oxide fractions within the transitional layer of samples measured with an

excitation energy of 170 eV (left) and 370 eV (right). ................................................................... 58

Figure 4.9 Relative change in equivalent thickness of Si1+ and Si3+ sub-oxides after annealing

at 800°C and 900°C respectively analysed using the approach of M. P. Seah and S.J. Spencer

[137]. ..................................................................................................................................................... 59

Figure 5.1 SIMS profile for a sample with passivating contact that was processed at a pressure

of 0.5 mbar and an annealing temperature of 900°C (a). Signals for Si., O, P, C and H are

shown. The broadening in the O, P and C curves at about 1400 s is assumed to be related to

the surface morphology of the sample at the measured spot. SIMS profile for a sample with

passivating contact that was processed at a pressure of 0.7 mbar and an annealing

temperature of 860°C. ........................................................................................................................ 62

Figure 5.2 Schematic showing the impact of substrate morphology on the result of SIMS

measurements. For planar surfaces (a) the width of the measured signal is representative of

the thickness of the analysed layer. For different surface morphologies, where layers are not

aligned perpendicular to the crater formed during the measurement, signal widths of especially

thin layers are larger compared to the thickness of the layer (b). ............................................... 62

Figure 5.3 Phosphorous concentrations calculated from SIMS data for samples that received

the plasma doping treatment at different pressures without annealing. A silicon reference with

known phosphorous concentration was used as a basis for these calculations....................... 64

Figure 5.4 Phosphorous concentrations calculated from SIMS data for samples that received

the plasma doping treatment at different pressures followed by annealing at 860°C for 30 min.

A silicon reference with known phosphorous concentration was used as a basis for these

calculations. ......................................................................................................................................... 65

Figure 5.5 Phosphorous concentrations calculated from SIMS data for samples that received

the plasma doping treatment followed by annealing at varying temperatures. The first 200 s

are shown at higher resolution in the top left corner A silicon reference with known

phosphorous concentration was used as a basis for these calculations. .................................. 66

Figure 5.6 Results of ECV measurements conducted on the same samples used for SIMS and

PCD measurements. Both pressure during the plasma doping treatment and temperature were

varied here. .......................................................................................................................................... 67

Figure 5.7 Zoom into the data shown in Figure 5.6 showing the indiffusion of phosphorous.

Data was offset horizontally to better compare each sample. ..................................................... 67

Figure 5.8 SIMS data of passivating contacts using different doping processes to realise

doping within the poly-Si layer after the deposition of intrinsic a-Si. Data was taken from [145]

and [142]. ............................................................................................................................................. 69

Figure 5.9 ECV data of passivating contacts using different doping processes to realise doping

within the poly-Si layer after the deposition of intrinsic a-Si. Data was taken from [143] and

[144]. ..................................................................................................................................................... 70

Figure 6.1 Average layer thicknesses measured on different glass substrates that received a-

Si E-Beam layers with a set thickness of 300 nm. ........................................................................ 74

Figure 6.2 Impact of the thickness of a-Si layers on the minority carrier lifetime in symmetrical

lifetime samples using the passivating contact of both sides. Plasma doping was realised at

0.5 mbar followed by an annealing step performed at 900°C. Each group includes 4 samples.

............................................................................................................................................................... 74

Figure 6.3 iVoc values of symmetrical lifetime samples using passivating contacts. Both

pressure during the plasma doping step and annealing temperature were varied. Values were

calculated assuming an optical constant of a) 0.55 and b) 0.95 to reflect the behaviour of the

symmetrical lifetime samples and expected behaviour of cells respectively............................. 77

Figure 6.4 Impact of band bending between (n) c-Si substrate and poly-Si on the iVoc of

symmetrical lifetime samples using passivating contacts. For most samples an almost linear

relation is depicted. Deviations from this behaviour are found for higher annealing

temperatures and are expected to be related to indiffusion. ........................................................ 77

Figure 6.5 Optical microscopy images of pinholes enhanced via TMAH etching for a sample

a) annealed at 860°C and b) 915°C. Field of view: 1237x928 µm2. ........................................... 78

Figure 7.1 Schematic of solar cells investigated and process scheme used. The front side

uses layers and processes that have been previously established. The rear side features the

passivating contact together with a titanium and silver contact. ................................................. 81

Figure 7.2 Flow chart of processes used for the fabrication of cells with passivating contact.

............................................................................................................................................................... 82

Figure 7.3 PL images for the best processed cell before (left) and after (right) edge isolation.

Colour scale for both images is identical. ....................................................................................... 84

Figure 7.4 Trends of JV curves for cells with passivating contact and a silicon heterojunction

reference cell. The figure also includes a JV curve derived from Suns-Voc measurements [150]

for one of the cells with passivating contact. .................................................................................. 85

Figure 7.5 Impact of annealing wafers with a silicon oxide cap before cleaning and deposition

of i/p silicon heterojunction structures. Lifetimes are compared to results achieved for

symmetrical i/p structures after wafer annealing (black), without wafer annealing (red) and for

a sample with poly-Si passivating contact on both sides (green). Lifetime samples of cell-like

structures are also shown (blue), which have the same structure as cells shown in Figure 7.1

excluding the TCO and metal contacts. .......................................................................................... 85

Figure 7.6 Comparison of measured EQE curve without the influence of gridlines and EQE

data extracted from GenPro4 measurements for solar cells with a passivating contact and

Ti/Ag metal contact on the rear. ....................................................................................................... 86

Figure 7.7 Optical simulations of solar cells with passivating contact and Ti/Ag rear contact

performed in Genpro4. Ti on the rear of the cell causes significant parasitic absorption limiting

the magnitude of the current that can be generated by the c-Si substrate. ............................... 87

Figure 7.8 Optical simulations of solar cells with passivating contact and newScot/Ag rear

contact performed in Genpro4. The replacement of Ti with newScot and other TCO leads to a

significant reduction in parasitic absorption on the rear of the device. ....................................... 87

Figure 7.9 Comparison between measured JV data and JV data derived from simulations

based on the 1-diode and 2-diode model. ...................................................................................... 90

Figure 7.10 Results of simulations performed using the 1-diode model and 2-diode model for

solar cells. Potential efficiencies are shown for different dark saturation currents. 2-diode model

simulations were also performed at dark saturation current densities that were obtained for

lifetime samples with i/p a-Si:H layers (symmetrical), poly-Si passivating contacts

(symmetrical) and cell-like samples using PCD. Here, j01 was set to the values obtained from

PCD measurements and j02 was adjusted so that the resulting open-circuit voltage would be

equal to the measured implied open-circuit voltage. ..................................................................... 90

Figure 8.1 Bright (a,c) and dark (b,d,e) TEM images of passivating contacts using emitter

currents of 190 mA (a,b) or 300 mA (c,d) for the growth of the initially amorphous silicon layer.

Crystal sizes vary strongly across the investigated area with singular crystals (e) stretching

across the entire poly-Si layer. ......................................................................................................... 95

Abbreviation and symbols

Area

a-Si

Amorphous silicon

Silver

AlOx

Aluminium oxide

AM 1.5G

Global standard spectrum for an air mass of

1.5

Argon

ARC

Anti-reflective coating

AZO

Aluminium-doped zinc oxide

Deal-Grove model constant

B2H6

Diborane

Bismuth

Speed of light

c-Si

Monocrystalline silicon

Capacitance

C(T)

Temperature-dependant solubility

Solubility reference point

CDTE

Cadmium telluride

CIGS

Copper indium gallium selenide

Zirconium

Caesium

Capacitance-voltage

Czochralski

Deal-Grove model constant

Distance

Distilled water

dox

Oxide thickness

Irradiance

Oxidation potential

Binding energy

E-Beam

Electron beam

ECV

Electrochemical capacitance-voltage

EFi

Fermi energy level

Ekin

Kinetic Energy

EQE

External quantum efficiency

optical constant

Faraday constant

Fill factor

FGA

Forming gas anneal

FWHM

Full-width at half maximum

Float-zone

Gallium

Planck constant

Enthalpy

Molecular hydrogen

HCl

Hydrochloric acid

Hydrofluoric acid

HNO3

Nitric acid

Intrinsic

Current

IMPP

Current at maximum power point

IPCC

Intergovernmental Panel on Climate

Change

ISix+

Suboxide intensity

ITRPV

International Technology Roadmap for

Photovoltaics

iVoc

Implied open-circuit voltage

Flux

Dark saturation current

j01

Dark saturation current (2-diode model)

j02

Dark saturation current (2-diode model)

jMPP

Current density at maximum power point

jph

Photogenerated current

jsc

Short-circuit current density

Boltzmann constant

Length

LPCVD

Low-pressure chemical vapour deposition

LSix+

Attenuation length of silicon oxidation state

LTLM

Transfer line method length

Molecular weight

Electron mass

MOSFET

Metal oxide semiconductor field-effect

transistor

MPP

Maximum power point

Density of electrons

Dopant concentration

Efficiency

Equilibrium density of electrons

N2O

Nitrous oxide

Diode ideality factor

Intrinsic charge carrier concentration

Optical constants

nP(Si)

Solubility of phosphorous in silicon

Oxygen

Ozone

Density of holes

Equilibrium density of holes

PCD

Photoconductance decay

PECVD

Plasma-enhanced chemical vapour

deposition

PERC

Passivated emitter and rear cell

PH3

Phosphine

Photoluminescence

POCl

Phosphorous oxychloride

POLO

Passivating contacts based on

polycrystalline silicon

Poly-Si

Polycrystalline silicon

PTunneling

Tunnelling probability

Elementary charge

QSSPC

Quasi steady-state photoconductance

Reflectance

Ratio between silicon dioxide and silicon

intensity

Contact resistance

Recombination rate of a single

recombination mechanism

Parallel resistance

Parallel polarised reflected light

Series resistance

Perpendicular polarised reflected light

RSix+

Ratio between suboxide intensity and

silicon intensity

Rtotal

Total recombination rate

SC-1

Standard clean 1

SC-2

Standard clean 2

SEM

Scanning electron microscopy

SHJ

Silicon heterojunction

Silicon

SiH4

Silane

SIMS

Secondary ion mass spectroscopy

SiNx

Silicon nitride

SiOx

Silicon oxide

SIPOS

Semi-insulating polycrystalline silicon

Six+

Silicon oxidation state

SPV

Surface photovoltage

Spectral response

SRH

Shockley-Read-Hall

STC

Standard test conditions

Temperature

Time

TCO

Transparent conductive oxide

Titanium

TMAH

Tetramethylammonium hydroxide

TOF

Time-of-flight

TOPCon

Tunnel oxide passivated contact

Electrical field constant

Deal-Grove model constant

VMPP

Voltage at maximum power point

Voc

Open-circuit voltage

Width

Depletion width

Distance

XPS

X-ray photoelectron spectroscopy

Deal-Grove model constant

Dissolution number

Zinc

Phase

Δn

Excess minority carrier concentration

ε0

Vacuum permittivity

εr

Relative permittivity

Wavelength

Frequency

eff

Effective minority charge carrier lifetime

Effective minority charge carrier lifetime for a

single recombination mechanism

Φe

Energy barrier for electrons

ΦFB

Flat band potential

Amplitude

1 Introduction

Through the Paris agreement [1] signed in 2016, multiple nations have declared the necessity

of limiting the effects of global warming to less than 2°C compared to pre-industrial levels by

2050. However, a more recent assessment report by the IPCC [2] showed the need to further

limit the increase in global temperate to 1.5°C by 2045 not only showing the ever-increasing

severity of the issue but also the need for technological advancements to reduce greenhouse

gas emissions. One of the biggest sectors related to the emissions of greenhouse gases is the

energy supply sector, which makes up about ~25% of greenhouse gas emissions based on

data for the European union [3], with many other energy related sectors such as industry and

domestic transport. While scenarios attempting to predict the development of greenhouse gas

emissions and rise in global temperature vary, renewable energy technologies, especially pho-

tovoltaics and wind energy, are generally considered the main path for carbon-neutrality for

energy-based sectors, which is also evident in the development of these sectors over the last

decades.

The photovoltaic industry has shown a significant reduction in levelized cost of electricity from

~300 $ per MWh to less than 100 $ per MWh over the last 10 years [4], which was accompa-

nied by a significant increase in installed photovoltaic power across the globe exceeding 600

GW in 2020 with trends indicating similar growth rates for the following years. Market-relevant

photovoltaic technologies are typically divided into three categories: based on monocrystalline

and multi-crystalline silicon wafers and thin film technologies such as amorphous silicon, CIGS,

CDTE and perovskite. Production for each type of technology has significantly changed over

the last years with monocrystalline silicon taking over 80% of the global market with around

95% [4] of the worldwide market focusing either on multi- or monocrystalline silicon wafer-

based showing not only the relevance of silicon-based photovoltaics but also the importance

of further improving silicon solar cells in particular.

In the last years research for silicon solar cells focused on improving surface passivation es-

pecially on the rear side of silicon wafers leading to the emergence of passivated emitter and

rear contact (PERC) solar cells and similar concepts [5]. The main improvement in device

efficiency was achieved by limiting the area on the rear of the device that is directly connected

to the metal contact, where charge carriers can more easily recombine limiting the efficiency

of solar cells. For PERC solar cells this is typically accomplished by first depositing passivating

structures on the rear of the cell and removing the passivation locally via laser ablation so that

a contact can be formed on the rear. While PERC and similar concepts were able to improve

the surface passivation for silicon solar cells, further gains were expected from fully separating

the metal contact from the substrate. This idea was realised through silicon heterojunction

solar cells employing hydrogenated amorphous silicon layers on both sides of the substrate.

Compared to other passivating structures, amorphous silicon layers are able to conduct charge

carriers so that it is not necessary to locally remove the structure for contacting. In addition,

charge carriers are conducted selectively at each contact further limiting the recombination

rate within the device. Through this concept record efficiencies of more than 26% were

achieved [6], showing the quality of surface passivation provided.

Tunnel oxide passivating contact [7] structures also referred to also TOPCon follow the same

base principle as silicon heterojunction solar cells. Instead of using hydrogenated amorphous

silicon layers, passivation is provided by combining an ultra-thin silicon oxide layer of 1-2 nm

thickness with strongly doped polycrystalline silicon layers. Here, chemical surface passivation

is provided mostly by the silicon oxide layer, while the combination of silicon oxide and doped

polycrystalline silicon enables selective transport of charge carriers towards the metal contact.

Unlike SHJ solar cells, TOPCon structures are typically used only on the rear of the cells due

to parasitic absorption by the poly-Si layer and limited passivation provided on textured sur-

faces [8], [9], [10]. The main advantage of TOPCon compared to SHJ solar cells from an in-

dustrial point of view is that TOPCon structures can be more easily integrated into industrial

grade PERC solar cells, which has already been successfully accomplished by some compa-

nies, who have shifted their focus towards these structures [11].

While TOPCon has already reached the point of market viability, the optimal industrial route

for creating these structures is still discussed in literature. Research tends to focus on the

techniques used for the creation of the oxide and especially the polycrystalline silicon layer.

Typically, the silicon layer is realised by depositing an initially amorphous silicon layer, which

crystallises during a high-temperature annealing step that is always performed for these pas-

sivating contacts. Amorphous silicon can be grown using plasma enhanced chemical vapour

deposition (PECVD), low-pressure chemical vapour deposition (LPCVD) and sputtering among

other techniques [12]. PECVD and LPCVD are the most popular among those techniques, with

PECVD being a method investigated by F. Feldmann et al. [13], who were the first to success-

fully apply TOPCon structures to high-efficiency solar cells. Through PECVD it is possible to

create well passivating structures with layers that can be doped during the initial process. While

record efficiencies for TOPCon solar cells have used PECVD amorphous silicon layers [14],

some properties of these layers make them not as attractive from an industrial point of view.

The initial amorphous silicon layers contain hydrogen, which can lead to the local delamination

and removal of the silicon layer during the high-temperature annealing step also referred to as

blistering [15]. Carbon can be incorporated into these PECVD layers to increase the layer’s

thermal stability [16], although blistering can still occur for thicker layers [15]. As PECVD layers

require low layer thicknesses to ensure stability, the integration of screen-printed and fired

metal contacts, which is the preferred method in the industry to contact solar cells, can be

difficult. Layers from LPCVD have shown similar passivation quality [17], however, single-sided

depositions are not possible necessitating additional processes steps for the removal of wrap-

around [18]. While it is possible to create thick a-Si layers using sputtering, passivation quali-

ties tend to be limited by damage caused to the ultra-thin oxide during the process [19].

Electron-beam evaporation was investigated by the Helmholtz-Zentrum Berlin in cooperation

with the International Solar Energy Research Centre Konstanz to create silicon layers for pas-

sivating contacts, which were combined silicon oxide layers grown in a wet-chemical HNO3

solution [20]. Passivation provided by these structures was comparable to those using LPCVD

amorphous silicon layers. The main advantage of electron-beam evaporation besides the dep-

osition of high-quality silicon layers is that the technique combines many of the properties that

are desirable for passivating contacts with high deposition rates allowing the deposition of thick

and doped layers that initially do not contain any hydrogen with no significant damage caused

to the silicon oxide layer. However, as the application of electron-beam evaporation in photo-

voltaics has been limited, some questions still remain in particular regarding scalability.

Silicon oxide layers for passivating contacts are typically grown wet-chemically or thermally

although reports also include the use of ozone containing gas mixtures [21] and plasma-based

techniques among others [22]. Thermal oxides tend to provide better chemical passivation,

however, require more time to process. Wet-chemical options for silicon oxide layers used in

passivating contacts include HNO3 and solutions containing ozone. Wet-chemical ozone in

particular is rather interesting as it does not require any additional chemicals besides water,

which can be more easily disposed of or treated compared to HNO3, although it should be said

that small amounts of HCl (<1% of the total solution) often need to be used in addition for pH

adjustment [23]. Additionally, wet-chemical ozone oxidation is a technique that has already

been established for the cleaning of silicon wafers.

As both electron-beam evaporation and wet-chemical ozone oxidation show characteristics

that are desirable from an industrial point of view, the present thesis reports investigations

focusing on both techniques separately and in combination for the formation of passivating

contacts. Research regarding electron-beam evaporation focused mostly on the passivating

contact but also includes structural properties of silicon layers and aspects related to the pro-

cess. Passivating contacts were also integrated into solar cells and characterised. Investiga-

tions were supported by simulations conducted in GenPro4. For the wet-chemical ozone-

based silicon oxide, efforts focused mostly on the structural evolution of the oxide within the

passivating contact, a topic which is believed to be rather important for wet-chemical oxides

and a topic that has only received little attention in literature. The creation of passivating con-

tacts involved a doping process, which is referred to as plasma doping. While the process has

been previously used for n-type passivating contacts, it has not been thoroughly examined up

to this point. The profiles of inactive as well as active phosphorous concentrations were deter-

mined for samples that received the plasma doping step as well as samples that were annealed

afterwards.

Chapter 2 focuses on fundamental aspects necessary to understand the behaviour of solar

cells as well as passivating contacts and silicon oxide layers.

Chapter 3 includes techniques used for the creation of passivating contacts and solar cells as

well as techniques used to characterise these structures and layers.

Chapter 4 centres around the silicon oxide layer and discusses structural changes the oxide

undergoes during processing. Discussions include changes in the structure, composition and

growth of the oxide during processes that are typically performed for passivating contacts. A

brief discussion on different approaches for determining the thickness of ultra-thin silicon oxide

layers is also included.

Chapter 5 focuses on plasma doping and the evolution of phosphorous profiles during the

initial plasma doping step as well as the following high-temperature annealing step.

Chapter 6 presents the passivating qualities of passivating contacts processed using electron

beam evaporation and wet-chemical ozone oxides. Results of lifetime measurements are com-

pared with the results of previous chapters to better understand the behaviour of samples in-

vestigated.

In chapter 7 the results of measurements performed on solar cells that use passivating con-

tacts are discussed. Simulations are performed to better understand the behaviour of samples

investigated and existing limitations resulting from the approach chosen.

Prospects regarding the techniques investigated as well as TOPCon-like structures for sili-

con/perovskite tandem solar cells are discussed in chapter 8.

The results of this thesis are summarised in chapter 9. The chapter also includes an outlook

regarding further investigations that could be considered using this thesis as a foundation.

2 Fundamentals

2.1 Working principle of a solar cell

In its most basic form, a solar cell is a diode consisting of a semiconductor with two differently

doped areas (both n- and p-type) with metal contacts. For an ideal solar cell, where loss mech-

anisms are not considered, the electrical output of the device is limited by the extracted current

as well as the implied open-circuit voltage. The current results from an ordered flow of excited

charge carriers. In semiconductors, like silicon, charge carriers can become excited or mobile

if they absorb enough energy to overcome the bandgap of the material. The bandgap is the

difference between conduction band minimum and valence band maximum. The bands repre-

sent energy states that can be occupied by electrons in an excited or dormant state respec-

tively shown in Figure 2.1. The size of the bandgap is also what differentiates semiconductors

from conductive and isolating materials. In conductors no bandgap exists meaning that elec-

trons can move freely within the solid. For insulators the bandgap is significantly higher mean-

ing that more energy is required for excitation to occur. In case of silicon with a bandgap of

1.12 eV photons with a wavelength of ~1100 nm or less can be used for excitation. However,

as silicon is a non-direct semiconductor an additional change in momentum is required, which

is provided by lattice vibrations [24]. Once electrons become excited both the electron and the

electron hole that results from excitation have to be able to reach their respective contact to

contribute towards the current of the device. This is typically achieved by doping the semicon-

ductor so that both n- and p-type areas exist. Doping promotes the conductivity of majority

carriers (holes for p-type and electrons for n-type materials), which in turn increases the col-

lection probability. The presence of both n- and p-type regions also leads to the formation of a

p-n junction. Diffusion of electrons from the n-type and holes from the p-type layer or rather

the potential difference that results from the diffusion leads to the formation of an electrical field

across the p-n junction that limits the direction electrons and holes can flow and, in the end,

determines the direction of the current that flows through the device when no bias is applied

[25]. The presence of the electrical field can also be seen in the band structure shown in Figure

2.1, where the absolute position of the band edges shifts when transitioning from p- to n-type

region. These shifts represent energy barriers that impact the flow of charge carriers across

the region. The alignment is based on the position of the fermi energy, which is the position

where an energy state, if present, has a 50% chance of being occupied by an electron. For

intrinsic materials it is located at the centre of the bandgap, but shifts for doped materials. For

n-type structures the Fermi energy is located closer to the conductive band edge and for holes

closer to the valence band edge. Under illumination the Fermi energy needs to be described

separately for electrons and holes according to eq. 1 and eq. 2 using the Fermi energy level

for an intrinsic semiconductor EFi, the Boltzmann constant kB, temperature T, intrinsic charge

carrier concentration ni and the carrier concentration of electrons n and holes p respectively.

(1)

𝐸𝐹𝑛=𝐸𝐹𝑖+𝑘𝐵𝑇𝑙𝑛𝑛

𝑛𝑖

(1)

𝐸𝐹𝑝=𝐸𝐹𝑖+𝑘𝐵𝑇𝑙𝑛𝑝

𝑛𝑖

(2)

𝑞𝑉𝑜𝑐=𝐸𝐹𝑛−𝐸𝐹𝑝

(3)

The separation of Fermi energy levels under illumination is equivalent to the implied open-

circuit voltage and therefore the maximum voltage that the device can achieve, which shows

that the separation and collection of charge carriers is not only of importance for the current

but also the open-circuit voltage of the device.

Figure 2.1 Schematic band diagram for a PERC silicon solar cell with aluminium-diffused back surface

field. Figure taken from [26] and edited.

2.1.1 Solar cell under operation conditions

Solar cells operate at the point of maximum power (MPP), where the power output of the device

reaches its highest value. At the MMP both current and voltage are slightly below the short-

circuit current of the device, which can be considered the photogenerated current for modern

solar cells. The ratio between maximum power output and the theoretical limit set by the short-

circuit current and open-circuit voltage is called the fill factor (FF) of the cell (eq. 6), which

includes contributions from series and shunt resistances. The series resistance consists of

contributions from the semiconductor as well as the metal contacts including the contact resis-

tivity between metal contact and device, which is typically the reason for high series re-

sistances. The shunt resistance described all alternate paths charge carriers can take within

the device including alternative paths within the device caused by e.g. defects or as a result of

directly contact the p-n junction as well as paths at the edge of the device. The impact of both

shunt and series resistance can be described by the following equations, which are based on

the equivalent circuit model of a solar cell shown in Figure 2.3 [27]:

Figure 2.2 I-V characteristics of a solar cell under illumination. Voltage-dependent power as well as

point of maximum power are shown in blue. Recreated from [28].

Figure 2.3 1 and 2 diode electrical circuit models according to eq. 5 and 6. Figures taken from [27].

rrent o er

r e

he short circ it c rrent

he o en circ it olta e

o er r e

olta e

(1)

𝑗(𝑉)=𝑗0(𝑒𝑥𝑝(𝑞(𝑉−𝑗𝑅𝑠)

𝑛𝐷𝑘𝐵𝑇)−1)+𝑉−𝑗𝑅𝑠

𝑅𝑝−𝑗𝑃ℎ

(4)

(1)

𝑗(𝑉)=𝑗01(𝑒𝑥𝑝(𝑞(𝑉−𝑗𝑅𝑠)

𝑘𝐵𝑇)−1)+𝑗02(𝑒𝑥𝑝(𝑞(𝑉−𝑗𝑅s)

𝑛D𝑘B𝑇)−1)+𝑉−𝑗𝑅s

𝑅p

−𝑗𝑃ℎ

(5)

Here, j0, j01 and j02 are the dark saturation current densities of diodes, nD the diode’s ideality

factor, Rs and Rp the series and shunt resistance respectively and jph the photogenerated cur-

rent density, which is approximately the short-circuit current density for high shunt and low

series resistances. Typically, the model is either built with one or two diodes, where the first

diode is assumed to behave like an ideal diode with nD=1. The main difference between the

one and two diode model is that the two-diode model can more accurately describe the behav-

iour of the device at varying voltages and relates to the different types of recombination mech-

anisms that dominate under these conditions [27], which are discussed in the following chap-

ter.

(1)

𝐹𝐹=𝐼MPP𝑉MPP

𝐼sc𝑉oc =𝑗MPP𝑉MPP

𝑗sc𝑉oc

(6)

ƞ=𝐹𝐹𝑗𝑠𝑐𝑉𝑜𝑐

𝐸

(7)

By measuring the fill factor FF, open-circuit voltage Voc and short-circuit current density jsc, the

efficiency of a solar cell can be calculated using the irradiance E that hits the front surface of

the device. While the efficiency of the device can be calculated rather easily using eq. 7, it

reflects poorly the complex nature and physical limits that exist for solar cells. Some of these

aspects such as the impact of recombination and spectral response are discussed in the fol-

lowing chapters, other aspects such as the impact of the bandgap energy that is more relevant

for semiconductors within tuneable bandgap such as perovskites [29] are not. A more through

discussing on the topic may be found elsewhere [30], [31].

2.2 Recombination mechanisms

Chapter 2.1 described the fundamental principle of a solar cell under the assumption that all

excited charge carriers can be collected. However, interactions between excited charge carri-

ers and other particles, structures, etc. can lead to the recombination of charge carriers, which

happens naturally for all excited charge carriers. Recombination describes electrons falling

from the conduction band to the valence band and the associated release and transfer of en-

ergy [25]. However, the mechanisms as well as the average rate at which recombination oc-

curs vary. Typically, four types of recombination are considered. Assuming a perfect crystal

structure without any defects in the bulk or interface of the material, recombination occurs

either via radiative or Auger recombination [32].

Figure 2.4 Generation and recombination of charge carriers within a semiconductor. Each recombina-

tion mechanism is shown separately. Figure was recreated from [31].

Radiative recombination describes the inverse process to absorption. An electron falls from

the conduction band back to the valence band and releases its energy in the form of a photon.

The energy of the photon is equal to the bandgap energy of the material. For indirect semicon-

ductors, like silicon, the process requires an additional momentum to reach the minimum of

the conduction band, which is provided by lattice vibrations represented by a phonon. As the

process requires the interaction of three different particles electron, hole and phonon, it is not

as likely to occur in silicon leading to a low radiative recombination rate. Radiative recombina-

tion still plays a relevant role in silicon especially in characterisation methods such as photolu-

minescence measurements (PL), which are discussed in chapter 3.

Auger recombination requires the interaction of two excited charge carriers. During the re-

combination of one of the excited charge carriers (either electron with hole or hole with elec-

tron) the resulting energy and momentum is transferred to the second excited charge carrier,

which in turn is exited beyond the band boundary. Excess energy is released via lattice vibra-

tions (phonons), which in the end are converted into thermal energy. As the process requires

3 charge carriers (either two electrons and a hole or two holes and an electron) it becomes

more dominant for high concentrations of excited charge carriers [32].

Shockley-Read-Hall recombination becomes relevant once one considers defects present

within a material. The presence of defects can lead to the creation of energy states within the

bandgap, which increases the likelihood of recombination to occur. The distribution of these

energy states depends on the origin of the defect, e.g. deviations from the crystal structure,

grain boundaries, impurities, etc.

Surface Recombination occurs at the surface of the semiconductor substrate. While the

mechanism is identical to Shockley-Reed-Hall, it is considered separately as it presents a dif-

ferent issue when it comes to the optimisation of semiconductors as well as solar cells. While

the Shockley-Read-Hall recombination rate within the substrate reflects the quality of the sub-

strate and the degradation in quality that may occur during processing e.g. doping processes,

the surface recombination rate reflects the quality of passivating structures applied to the sur-

faces of a semiconductor.

By adding the contributions of all recombination mechanisms, the total recombination rate Rtotal

can be calculated as the sum of individual contributions Ri:

𝑅𝑡𝑜𝑡𝑎𝑙=𝑅𝐴𝑢𝑔𝑒𝑟+𝑅𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑣𝑒+𝑅𝑆ℎ𝑜𝑐𝑘𝑙𝑒𝑦−𝑅𝑒𝑒𝑑−𝐻𝑎𝑙𝑙+𝑅𝑆𝑢𝑟𝑓𝑎𝑐𝑒

(8)

However, it is more common to evaluate the effective lifetime of charge carriers 𝜏𝑒𝑓𝑓:

𝜏eff=1

𝜏Auger+1

𝜏Radiative+1

𝜏Shockley−Reed−Hall+1

𝜏Surface

(9)

The lifetime 𝜏 associated with each separate recombination process is connected to the re-

combination rate via the excess minority carrier density ∆𝑛 [33]:

𝜏=∆𝑛

𝑅

(10)

Minority lifetimes are typically used in literature to evaluate the quality of substrates and sur-

face passivation. They can be measured using photoconductance decay measurements

(PCD), which are discussed in chapter 3.2.1.

2.3 Doping of silicon

Doping describes a controlled introduction of impurities into a semiconductor with the purpose

of changing its electrical, optical or structural properties. For monocrystalline silicon solar cells

doping is mostly used to increase the conductivity of charge carriers within the cell. By intro-

ducing impurities with a larger amount of valence electrons compared to silicon, e.g. phospho-

rous, some valence electrons do not form bonds and thus are mobile within the substrate. The

same can be achieved for electron holes by using elements that have a lower amount of va-

lence electrons compared to silicon, e.g. boron. While the introduction of dopants results in the

formation of bonds between two different elements, the concentration of dopants is kept low

relative to the concentration of silicon atoms so that the overall structure is still the same as for

intrinsic monocrystalline silicon as shown in Figure 2.5 where simplified 2D structures are

shown. However, for other materials the presence of dopants can have a more significant im-

act on the material’s str ct re hich is the case for oly-silicon layers that are created by

crystallising amorphous silicon layers. Here, the concentration of dopants is strongly linked to

shape and size of crystals formed [34]. For solar cell applications doping concentrations of

~1015-1016 atoms per cm3 are used for the substrate. The introduction of mobile electrons or

holes through doping improves the conductivity of majority carriers. Minorities are more likely

to recombine as the recombination rate generally depends on the concentration of both elec-

trons and holes thus limiting the average distance minority carriers can travel before recom-

bining. Changing the doping of a substrate or layer can be achieved using several different

methods, which can be roughly divided into three different groups: they either use a sacrificial

layer that is applied before the doping process e.g. in the form of a doped silicon glass [35],

inject atoms into the substrate, which occupy interstitial spaces, e.g. plasma doping (further

discussed in chapter 5) and implantation [36], or methods where the doping is realised in a

single step e.g. via annealing in phosphorous oxychloride (POCl3) [37] at elevated tempera-

tures. Regardless of the method chosen, for the doping process it is necessary that the follow-

ing two events occur. First, the dopant needs to diffuse into the sample. Diffusion describes

the net movement of atoms driven by a difference in concentration between two points as

sho n by Fick’s la s. Fick’s first la [38] (eq. 11) describes the flux of atoms 𝐽 between two

points with different concentrations dC dependant on how easily one species can diffuse inside

of another solid. While this factor is summarised by the diffusion coefficient D it is important to

note that it strongly depends on several factors, especially temperature. The relation between

diffusion coefficient and temperature can be approximated using the Arrhenius equation [39].

Fick’s first la sho s that the rate at hich diff sion occ rs stron ly de ends on the difference

in concentration. However, this also means that the diffusion flux is not constant as the distri-

bution of atoms during diffusion reduces the concentration gradient and therefore the rate at

which diffusion occurs. The time-de endant diff sion rate is described by Fick’s second la

(eq. 12). Once dopants have diffused into the sample, they can occupy either an interstitial

position between lattice atoms or substitutional positions as shown in Figure 2.5. Only substi-

tutional dopants are electrically active and contribute towards the doping of the semiconductor.

Both the diffusion process and the activation of dopants require energy that is typically provided

thermally, but can also be realised via laser irradiation or a combination of both [40].

Figure 2.5 Simplified two-dimensional atomic structure of intrinsic and doped mono-crystalline silicon.

Dopants are built into the crystalline structure of silicon and replace silicon atoms. Free charge carriers

result from a lack of bonding partners. Figure taken from [41].

𝐽=𝐷𝑑𝐶

𝑑𝑥

(11)

𝑑𝐶

𝑑𝑡=𝐷𝑑2𝐶

𝑑2𝑥

(12)

Beyond the general description given for the doping of monocrystalline silicon, the doping of

more complex structures such as the passivating contact involve additional effects. First, the

diffusion of dopants through the poly-Si layer does not occur homogeneously as the diffusion

coefficient at grain boundaries is typically higher compared to the diffusion through silicon crys-

tals [42]. Secondly, the diffusion coefficient for the diffusion of phosphorous through silicon

oxide is significantly lower compared to the one through silicon [43]. The doping profile, there-

fore, shows a noticeable kink at the poly-Si/SiOx/Si interface as shown in Figure 2.6. Dopants

can still travel to the substrate through pinholes leading to the formation of highly-doped areas

in the close proximity of pinholes. However, quantifying the impact of each effect can be rather

difficult. As passivating contacts are typically annealed at temperatures that allow a homoge-

nous distribution of dopants within the poly-Si layer, the impact of grain boundary diffusion is

not noticeable in the bulk of the poly-Si bulk at the end of the diffusion process. Additionally,

the crystallisation of the Si layer from amorphous to polycrystalline and therefore the formation

of grain boundaries occurs at the same time as the diffusion of dopants, which further compli-

cates the issue. The interaction of dopants and the crystallisation of amorphous silicon also

presents a complex issue as the presence of dopants significantly impacts the shape and

growth of poly-silicon crystals. Pinholes present a similarly complex issue as the formation of

pinholes occurs during the annealing step [44].

Figure 2.6 Doping profile for a TOPCon contact generated using electrochemical capacitance voltage

measurements (ECV). A significant drop in doping concentration can be observed at the interfacial

region between poly-Si and c-Si. b) shows a schematic of the interfacial region of TOPCon contacts.

Included in the schematic are pinholes and areas with reduced oxide thickness that enable dopants to

more easily travel across the interfacial region. Figure is taken from [45].

2.4 Solar cells with passivated surfaces

In the previous chapters the discussion regarding solar cells was mostly limited to p-n homo-

junctions, where both sides of the p-n junction consist of the same base silicon material. For

these types of devices, the substrate has to be directly contacted to the front and rear metal

contacts. Due to the distribution of energy states in metals and the lack of a bandgap, Shock-

ley-Read-Hall recombination is enhanced at the metal-semiconductor interface limiting espe-

cially the open-circuit voltage of the device [46]. Additionally, the diffused area created for the

formation of the p-n junction (either p or n), contains doping concentrations far above the base

doping of the substrate with concentrations above 1x1019 atoms/cm3 [47], which increases the

Auger recombination rate. While the creation of differently doped areas as well as the creation

of further barriers such as the back-surface fields, which in the past were realised via diffusion

of aluminium [48] supports the separation of charge carriers, the ability of these structures to

selectively transport majority carriers is still limited. To reduce recombination at the metal-sem-

iconductor attempts were made to limit the surface of the interface, which led to the creation

of passivated emitter and rear contact solar cell (PERC) [5] and similar designs like the pas-

sivated emitter rear locally diffused (PERL) [49] and passivated emitter rear totally diffused

(PERT) cells [50]. PERC solar cells apply passivating layers to the front and the rear of the

cell. Small openings are created via laser ablation on the rear of the cell so that the semicon-

ductor can be contacted [51]. SiOx, SiNx and AlOx [51] are typically used as passivating layers.

While the introduction of these more advanced contacting schemes led to significant improve-

ments in the properties of cells with reported efficiencies as high as 24% [52], the presence of

the metal-semiconductor junction still significantly affects the recombination rate. Additionally,

the local contacting scheme on the rear of the cell requires the current to travel longer dis-

tances through the cell in addition to potential damage caused to the silicon substrate during

the ablation process [53], [51]. Further improvements were achieved through the separation of

substrate and metal contact made possible by the introduction of heterojunctions. The follow-

ing chapters feature information for two of these structures namely solar cells with passivating

contacts, which are typically found in literature as tunnel oxide passivating contact (TOPCon)

[13] solar cell or rarely as passivating contacts based on polysilicon (POLO) solar cells [54],

and a-Si:H/c-Si solar cells [55].

2.4.1 Poly-silicon based passivating contacts

The creation of poly-silicon based passivating TOPCon, a term coined by F. Feldmann et al.

[13] who were the first group to successfully implement the structure into high-efficiency solar

cell, was inspired by the work of E. Yablonovitch and T. Gmitter [56]. Yablonovitch and Gmitter

showed the successful application of semi-insulating polycrystalline silicon (SIPOS) in solar

cells after the widespread success of using the material in bipolar transistors with implied open-

circuit voltages of ~720 mV. SIPOS is best described as a mixture of microcrystalline silicon

and silicon oxide and in an annealed state provides similar interface defect densities on silicon

surfaces as thermal silicon oxide layers [57]. Due to the success of poly-Si emitters in bipolar

transistors [58], several groups attempted to implement poly-Si layers in solar cells with gen-

erally poor cell properties [59], [60]. Feldmann et al. split up the SIPOS structure into two sep-

arate and defined layers. A layer of silicon oxide chemically passivates the rear surface of the

device and a highly doped poly-Si layer enables high selectivity as well as high conductivity

for majority carriers [13]. The structure of the TOPCon cell shown in Figure 2.7 uses a full area

metal contact on the rear, which does not require laser ablation as the doped poly-Si layer is

highly conductive for majority carriers, as well as a typical approach for industrial solar cells for

the front side consisting of a mostly-passivated surface with antireflection coating. For sym-

metrical lifetime samples, where the passivating contact is created on both sides, iVoc values

of ~715 mV were achieved with cells efficiencies of up to 23.7 % [13]. Further development

and improvements in the TOPCon solar cell led to record efficiencies of 25.8% for n-type sub-

strates [61] and 26.1% for p-type substrates [62], although it should be stated that these cells

were realised under laboratory conditions on small cell areas of ~4 cm2 with processing tech-

niques that are partly limited in their application to research or laboratory based work. The

industrial mass production of TOPCon solar cells is able to achieve average efficiencies of

~23% [52].

Figure 2.7 Schematic of a n-type TOPCon solar cell using a passivating SiOx/poly-Si contact on the rear

of the device. The front is passivated using AlOx and combined with an antireflection coating to improve

the optical properties of the device. The figure was taken from [63].

The creation of the TOPCon or passivating contacts typically involves 4-5 steps. First, the

surface or silicon wafers are oxidised. Then an initially amorphous silicon layer is deposited on

top of the oxide. In most cases the amorphous silicon layer is already doped, although doping

can also be provided ex-situ [20]. Groups including the group of Feldmann et al. reported the

usage of doped amorphous silicon layers containing carbon to improve the thermal stability of

the layer and to prevent issues that may arise especially when dealing with hydrogenated a-Si

layers e.g. the formation of blisters. Hydrogen effuses during the following processing steps,

which can tear the amorphous silicon layer from the substrate leading to the formation of blis-

ters locally delaminating and rupturing the a-Si layer [15]. After the deposition of the a-Si layer

samples are annealed at elevated temperatures between 850°C and 1050°C, with values

around 900°C being more commonly reported. The high temperature annealing step serves

multiple purposes. First, it allows the activation of dopants or the doping of the initially amor-

phous silicon layer for cases where the doping is provided ex-situ. Dopants are also driven

deeper into the sample and, to some degree, diffuse past the silicon oxide layer. The shape of

the in-diffused doping profile tends to affect the passivating quality of the contact. A steep drop

in doping concentration at the SiOx/c-Si interface tends to be associated with higher iVoc values

[12], [20]. Parallel to the formation of the doping profile, the amorphous silicon layer crystallises

to poly-Si. The crystallisation to poly-Si is necessary to achieve high conductivities. Lastly, the

poly-Si/SiOx/c-Si interface undergoes significant structural changes. This includes the restruc-

turing of the silicon oxide layer, which is necessary especially for wet-chemical oxides to pro-

vide good chemical passivation, and the formation of pinholes. Pinhole formation may lead to

the degradation of the chemical passivation provided by the oxide if the concentration is high

enough [64], but can also be necessary for the transport of majority carrier. The flow of charge

carriers across the passivating contact is further discussed in the following chapter. Finally,

after the completion of the annealing step samples undergo hydrogenation to saturate open

bonds especially at the interface of the device to enable low interface defect densities and

therefore high iVoc values. While some amorphous silicon layers may initially contain high con-

centrations of hydrogen, hydrogen (if present) effuses from the sample during the high-tem-

perature annealing step and therefore needs to be replenished. Hydrogenation can be pro-

vided by annealing samples in forming gas, exposing them to a plasma containing hydrogen,

or by depositing and annealing layers containing high concentrations of hydrogen e.g. PECVD-

based SiNx layers [65], [66]. While the overall processing structure to realise passivating con-

tacts has remained the same, the approach can still vary significantly based on the deposition

techniques chosen. In particular, for the amorphous silicon layer multiple approaches have

been suggested with some of the more popular ones found in literature being plasma-en-

hanced chemical vapour deposition (PECVD) [67], [68]] and low-pressure chemical vapour

deposition (LPCVD) [17] although sputtering [19] and the usage of electron-beam evaporation

(E-Beam) [20] have been investigated. For the growth of the silicon oxide either wet-chemical

oxides, thermal oxides, plasma based [12] oxides or UV ozone oxides [21] have been investi-

gated. Thermal oxides typically provide the best chemically passivation, however, low defect

densities have also been reported for wet-chemical approaches especially HNO3- [21], [68]

and O3-based oxides [69].

2.4.1.1 Transport of charge carriers across the c-Si/SiOx/poly-Si interface

While some fundamental properties of the TOPCon structure such as the selective conduction

of majority carriers were previously known, it was, initially, not exactly clear how charge carriers

overcome the SiOx barrier. Tunnelling was assumed to be the main transport mechanism [37].

Tunnelling describes the ability of charge carriers to bypass energetic barriers which they could

not overcome otherwise. The tunnelling probability can be estimated using the Wentzel-Kra-

mers-Brillouin approximation, eq. 13 [45].

𝑃𝑡𝑢𝑛𝑛𝑒𝑙𝑖𝑛𝑔,𝑒≈𝑒𝑥𝑝(−4𝜋𝑑𝑜𝑥

ℎ√2𝑚𝑒𝑞∆𝛷𝑒)

(13)

Figure 2.8 Schematic energy band diagram for n-type TOPCon passivating contacts. Holes are less

likely to tunnel through the oxide due to the higher energy barrier that silicon oxide presents for holes

compared to electrons. Figure recreated from [70].

From this relation it becomes clear that tunnelling is mostly affected by the thickness of the

barrier or in this case oxide layer dox as well as the height of the energy barrier for electrons

𝛷𝑒. To calculate the tunnelling probability for holes the height of the barrier as well as the

effective mass of electrons me need to be replaced by their hole equivalent. Eq. 13 is, however,

a simplification, which assumes flat band edges for the silicon oxide as opposed to the trape-

zoidal shape shown in Figure 2.8. While comparisons between measured and modelled data

assuming only tunnelling as transport mechanisms showed similar results for tunnel oxide

thicknesses of ~1 nm, more noticeable deviations were present for samples with thicker oxide

layers. Peibst et al. [45] suggested an alternate explanation for passivating contacts that used

thicker oxide layers up to 2.4 nm. The model assumes charge carriers either use pinholes

formed during the annealing process which enable the formation of direct localised c-Si/poly-

Si contacts and/or the localised reduction in oxide thickness resulting from the restructuring of

the oxide and a-Si layers during the annealing process as shown in Figure 2.6 in chapter 2.3,

where the doping of silicon is discussed. From this one can infer not only that the thickness of

the silicon oxide layer has an impact on the behaviour of the passivating contact but also the

range of appropriate process conditions for differently designed passivating contacts. Higher

interfacial oxide thicknesses require not only higher annealing temperatures that allow the for-

mation of pinholes and/or the local reduction in oxide thickness, but also silicon oxide layers

that do not fully degrade under these conditions. For oxide thicknesses of up to ~ 1.5 nm lower

temperatures can be used in combination with oxidation methods that produce oxides which

degrade at elevated temperatures above 900°C (typical for wet-chemical oxides) [12], [70].

2.4.2 a-Si:H/c-Si solar cells

Similar to passivating poly-Si contacts, the saturation of defects states at the surface of the

substrate is mostly realised via hydrogenation, which here is provided during the deposition of

the amorphous silicon layers. Unlike passivating contacts where the bandgap of polycrystalline

and monocrystalline silicon is the same, the bandgap of hydrogenated amorphous silicon is

~0.6 eV higher [71] compared to crystalline silicon. The difference in bandgap energy is visible

in the band structure of the material, which causes the formation of energy barriers within the

valence band and to a minor degree in the conduction band. Especially the energy barrier in

the valence band can affect JV characteristics of the device leading to s-shapes in the trend

of the JV curve [72]. The formation of s-shapes can, however, be avoided e.g. by limiting the

thickness of the intrinsic a-Si layer and by using highly-doped a-Si:H (p) layers. Silicon hetero-

junction cells currently hold the record for the highest efficiency achieved for non-concentrator

crystalline cells with 26.7% [73]. However, the main difference between both types of solar

cells is their suitability as an industrial solar cell. For silicon heterojunction solar cells process

temperatures are limited to ~200°C degrees to prevent the degradation of the hydrogenated

amorphous silicon layers [74]. This limits the application of several techniques which have

become a standard for industrial solar cells such as the deposition of some antireflective coat-

ings such as silicon nitride, which typically is performed at ~400°C [75] and the high-tempera-

ture firing of metal contacts at ~800°C [76]. Passivating contacts can provide the thermal sta-

bility necessary to undergo these process steps as shown by the TOPCon solar cell. In addi-

tion, the application of passivating contacts only adds three to four steps to the process

schemes of industrial PERC and similar solar cells, which allows easy integration. However, it

is possible to use some industry-preferred techniques in the creation of silicon heterojunction

cells such as the printing and low-temperature formation of metal contacts [77].

While silicon heterojunction structures were used for the creation of solar cells investigated in

this thesis, no thorough variations and investigations were performed for these structures, as

they were simply used to see how fabricated passivating contacts would perform within a solar

cell. More thorough investigations may be found elsewhere [78], [79].

Figure 2.9 Energy band diagram and schematic of the structure of an a-Si:H/c-Si silicon heterojunction

solar cell. Energy barriers caused by a mismatch between the bandgap of a-Si:H and c-Si are shown

on the left and right side of the substrate. Figure taken from [80].

2.5 Oxidation of silicon

The oxidation of silicon is well researched in particular motivated by the emergence of

MOSFET transistors, and similar technologies, where they are used as insulating layer [81].

While several approaches exist to oxidise silicon such as wet-chemical and thermal methods

as well as oxidation of silicon in ambient atmosphere, the thermal approach is the most studied.

For transistors as well as solar cells, the silicon oxide has to be able to sufficiently passivate

the surface of the silicon substrate. Defects mostly relate to bond-stretching, dangling bonds

and absorption of impurities from the silicon surface. Thermal oxides typically provide lower

min. interface defect densities as shown in tab. 1. Other methods typically provide defect den-

sity distributions with larger minimal defect densities of up to 1013 cm-2eV-1. However, it should

be noted that these values refer to the qualities of freshly grown oxides. This fact is of im-

portance when considering passivating SiOx/poly-Si contacts as the high-temperature anneal-

ing step that is performed for the formation of the structure affects the properties of the oxide

layer. Nevertheless, the values shown in tab. 1 clearly reflect the difference in quality that can

be achieved through thermal oxidation.

Tab. 1 Comparison of minimal interface defect density for oxidised silicon substrates using different

oxidation methods. The values were taken from several different sources.

Oxidation method

Min. interface defect density in cm-2eV-1

Thermal

~1010-7x1011 [10], [82], [83]

Wet-chemical ozone

~1013 [84], [85]

Wet-chemical HNO3

~3x1012 [84]

While several models exist to describe the oxidation process, the Deal-Grove model is typically

accepted as an accurate depiction of the thermal growth of silicon oxide for high temperatures

[86]. The oxide thickness dox can be calculated for a given oxidation time tox by using two

constants B and D that reflect reaction kinetics, which are normally extracted from growth data

by using the following equation:

𝑑𝑜𝑥

2+𝐵𝑑𝑜𝑥=𝐷𝑡𝑜𝑥

(14)

The model assumes that the oxidation process can be divided into two main processes: the

transport of interstitial oxide species towards the silicon interface and the reaction of silicon

and oxygen at the interface. However, when looking at the entire oxidation process, the for-

mation of a transitional area between silicon and oxide has to be considered. The difference

in bonding lengths between Si-Si bonds (~2.35 Å [87]) and Si-O bonds (~1.6 Å [88]) causes a

mismatch between both structures that would require a significant stretching or compressing

of bonds and change in bonding angles. Instead, research by several groups using photoelec-

tron spectroscopy (discussed in chapter 3) and other methods have shown that a transitional

area is formed between SiO2 and Si with sub-stoichiometric oxides (suboxides) consisting of

silicon that forms between 1 and 3 bonds with oxygen atoms smoothening the transition be-

tween silicon substrate and the stoichiometric amorphous SiO2. The exact shape and form of

this transitional area is, however, not quite clear or at least may be strongly dependant on the

oxidation method used. Reports have been made regarding non-ordered [89] and ordered [90],

[91], [92] transitional areas with some reports even indicating the formation of crystallites at

the silicon interface [93], [94], [95], [96]. Considering the basic assumptions made for the Deal-

Grove model as well as the formation of the transitional area, the oxidation of silicon occurs

according to Figure 2.10. As the oxide grows, the transitional area moves together with the

Si/SiOx interface deeper into the sample as more and more silicon atoms are used for the

growth of the silicon dioxide layer.

Figure 2.10 Schematic describing the oxidation of silicon including the separate zones that are formed

during the process. Figure taken from [97].

The transitional oxide, therefore, also exists as an intermediate stage of the oxidation of silicon-

to-silicon dioxide. It is assumed that the shape of the transitional area strongly depends on the

reaction method with the reaction rate playing an important role at least for the thermal oxida-

tion of thick silicon dioxides. From this one can infer that the reaction is limited by the transport

of oxide species towards the interface, which in turn is limited by the thickness of the oxide and

the kinetics of the transport mechanism involved. While the Deal-Grove model accurately de-

picts the thermal oxidation of silicon, other models such as the one by Fehlner and Mott [98]

are required for low-temperature and wet-chemical approaches. One major difference between

these models is the transport mechanism that allows oxygen species to reach the silicon inter-

face. In case of thermal oxidations, the transport happens via diffusion. However, for wet-

chemical oxidation methods it has been shown that instead the transport may be drift assisted

[98]. The presence of negatively charged oxide species at the silicon dioxide surface and pos-

itively charged silicon ions at the silicon interface erect an electrical field across the silicon

oxide that supports the drift of oxygen ions across the silicon oxide. The time dependant thick-

ness of the oxide can be described using eq. 15 together with the fit parameters V and Y, which

similar to B and D from the Deal-Grove model are also a convolution of various properties

relating to the kinetics of the silicon oxide growth:

𝑑ox=𝑉∗ln⁡(1+𝑌𝑡)

(15)

In Figure 2.11 measured thicknesses of oxides grown in solutions containing 30 ppm and 50

ppm ozone are qualitatively compared to the Deal-Grove (eq. 14) and Fehlner-Mott (eq. 15)

model. The data shows a rather immediate formation of the oxide with a significant reduction

in oxidation rate as the oxidation process continues. Both models show different curve trends

with the Deal-Grove model showing a steadier oxidation rate compared to the Fehlner-Mott

model, which similar to the data shown has an initially high oxidation rate which drastically

decreases. The growth of the oxide reduces the effectiveness of the electrical field formed over

the silicon oxide layer, which in turn limits the reaction-rate of the process significantly. For the

Deal-Grove model or thermal oxidation the formed oxide acts as a diffusion barrier, which,

however, can be easily overcome at elevated temperatures.

Figure 2.11 Thickness of silicon oxide layers grown in solutions containing 20 ppm ozone as calculated

from spectral ellipsometry data for different oxidation durations. Data is compared with two different

models that describe the oxidation of silicon.

2.5.1 Wet-chemical oxidation of silicon

The oxidation of silicon is a rather important reaction for the creation of silicon-based devices

even in cases where no oxide is present within the structure. The removal of contaminants

from the silicon surface is crucial for the functionality and quality of silicon devices and is typi-

cally accomplished wet-chemically through methods that involve the growth of an oxide. Re-

search regarding the subject started in the 1950s [99] but has remained as relevant as ever

due to the ever-increasing requirements for silicon devices. For silicon solar cells the main

issue that contaminants present is the introduction and formation of defects, which, as previ-

ously discussed, limit the quality of the device. J. John et al. [100] investigated the impact of

metal contaminants on silicon substrates by submerging wafers into cleaning solutions con-

taining defined concentrations of several different metal species. After a high-temperature an-

nealing step, which allowed metal contaminants to diffuse into the substrate, surfaces were

passivated and compared to samples that did not come in contact with the contaminated so-

lutions. In case of copper, surface concentrations as low as 5*1010 atoms/cm2 would limit the

effective lifetime of the device to ~50% of what could be achieved normally. Zn, Ti and Cr need

around 1012 atoms/cm2 for similar results. Co and Fe completely degrade the passivation within

the investigated ranges. The removal of contaminants both organic and inorganic has in the

past been accomplished using RCA clean [99]. While the method itself presents one of the

most effective ways to remove contaminants from silicon surfaces, other methods including

the usage of piranha solution, the IMEC cleaning procedure [101], and wet-chemical solutions

containing ozone are more typically used in industrial environments [102] partially to reduce

chemical waste and processing time. RCA cleaning consists of three main processing steps

which are interrupted by rinsing in deionised (DI) water. The first step, also referred to as

standard clean 1 (SC-1), consists of an oxidation of the wafer surface in a solution containing

DI water, ammonia water (29 wt%) and hydrogen peroxide (30 wt%) at ~80°C with a ratio of

5:1:1 for 10 min, although the exact conditions may vary. During the process oxide is created

but also dissolved within the alkaline solution. The interplay of oxidation and removal or etching

of silicon oxide enables undercutting below particles, which in turn can be more easily removed

from the wafer surface. After the SC-1 step particles and especially organic contaminants are

either removed from the surfaces or captured within the surface oxide that remains at the end

of the process. The oxide is removed using ~1% HF. As a final step the samples are sub-

merged in a DI:HCl:H202 solution with 6:1:1 (HCl 37 wt%) ratio at similar temperatures and

durations as the SC-1 step. The solution removes remaining metal traces as well as alkaline

traces from the SC-1 step and similar to the SC-1 step includes the formation of an oxide on

the wafer surface. The removal of contaminates is accomplished via sequential desorption and

complexing of metal particles with the solution. SC-2 also oxidises the wafer surface, which

acts as a protective layer that is typically removed before the next process step.

When it comes to the application of silicon oxides for semiconductor devices, oxides resulting

from SC-1 and SC-2 are typically avoided due to the poor surface passivation and oxide quality

that oxides from these methods possess [103]. Wet-chemical oxidation methods that are typi-

cally used for surface passivation in particular for the creation of passivating contacts include

the oxidation of silicon via ozone containing solutions, which is the focus of this thesis, and hot

nitric acid solutions. Wet-chemical solutions containing ozone were initially investigated as an

alternati e to R A cleanin d e to the sol tion’s hi h oxidation otential E0=2.08 V [104], which

surpasses other relevant solutions. Originally, the effectiveness of the method was fairly limited

as reactions between ozone and metal components within the ozone generator as well as the

degradation of pipes and other solution-holding components via reaction of ozone introduced

contaminates into the cleaning solution. However, these issues have been solved through ap-

plication of ozone resistant materials. Ozone for the process is provided by an ozone genera-

tor, which causes oxygen to form ozone via UV excitation. Ozone together with remaining

oxygen and nitrogen (to suppress the decay of ozone [105]) is then directed to a wet bench

where the ozone is absorbed into a wet-chemical solution containing mostly water. Small vol-

umes of other chemicals, typically HCl, are added to the solution to adjust the pH value to

improve absorbability and reduce the degradation rate of ozone, which is not stable within the

aqueous solution. pH values of 2-4 have shown to enable the highest solubility for solutions

using HCl as additive and are required to reach high ozone concentrations [106]. The degra-

dation rate, on the other hand, is more severe at higher ph-Values above 7 and low pH values

above 1.2, which affects the decay of ozone via reaction with chloride ions [106]. However, the

degradation rates are typically low enough that a regulated supply of ozone is able to stabilise

O3 concentrations within the solution. Temperature also needs to be considered as the solu-

bility is strongly connected to the temperature according to the Arrhenius equation [107]:

𝐶(𝑇)=𝐶0∗𝑒𝑥𝑝(𝑑𝐻

𝑘𝐵𝑇)

(16)

Here, C(T) is the temperature dependant solubility, C0 the solubility at reference point and dH

the change in enthalpy caused by absorbing a substance in a solvent. While the reaction of

silicon and ozone is not fully understood, it is believed that the oxidation species responsible

for the growth of the oxide is either O1- [108] or O2- and O3- [109]. Based on the work of F. De

Smedt et al. and a hypothesis by S. L. Nelson et al. [110] the creation of the oxidation species

is realised close to the SiO2/liquid interface with the following reactions for n-type silicon:

O3↔O2+O

(17)

O+𝑒→O−

(18)

Silicon oxide layers grown in ozone-containing solutions are believed to have bulk qualities

similar to thermally grown oxide albeit a severely limited oxide thickness of the order of 1-2

nm. However, passivation qualities of ozone-based oxides are initially poor with measured

minority lifetimes of ~3 µs. The limitation is believed to result from stress formed at the interface

between bulk silicon oxide and silicon substrate during the initial growth of the oxide as dis-

cussed in chapter 4. This behaviour is not detrimental for the application of ozone-based ox-

ides in passivating contacts due to structural changes that occur during annealing. The corre-

lation between oxide properties after initial growth and annealing is not fully understood and

initial growth conditions and growth methods chosen may very well affect the final properties

of the oxide. Although passivation qualities of passivating contacts using ozone- and nitric-

based silicon oxides show comparable minority lifetimes [111].

3 Deposition and characterisation methods

The following chapters explain the processes used for the deposition and growth of materials

investigated as well as the techniques used for characterisation.

3.1 Deposition methods

3.1.1 Wet-chemical ozone oxidation

The aforementioned wet-chemical oxidation of silicon was conducted in cleanroom conditions

using a system consisting of wet benches from the company Arias, circulator pumps, ozone

sensors and a HeliO3 LIQUOZON ozone generator. The ozone generator provides a mixture

of ozone, oxygen, which is not converted during the ozone formation process, and nitrogen,

which is used to suppress the degradation of ozone into oxygen. The gas mixture is then di-

rected to the wet bench consisting of two 12 l basins. One basin is used exclusively to remove

contaminants from the surface of wafers after saw-damage etching and texturing. The other is

used for the oxidation process and final cleaning steps. With this system it is possible to reach

temperatures between 20°C and 65°C with ozone concentrations as high as 50 ppm, although

one has to consider the solubility of ozone in water-based solutions, which strongly decreases

with temperature. For solutions containing 20 ml HCl resulting in a pH value of 2.3, which were

used exclusively for this thesis, ozone concentrations were limited to ~35 ppm at 35°C and

~20 ppm at 50°C. Ozone sensors are used to measure and regulate the concentration of ozone

within the solution. Oxidations were performed 40 min after starting the process to ensure that

the set temperature and ozone concentrations were reached and stable.

3.1.2 Electron beam evaporation

Electron-beam (E-Beam) evaporation was used to deposit silicon layers for passivating con-

tacts. During E-beam evaporation electrons emitted by an electron gun are directed towards a

crucible containing a silicon ingot. The resulting collision causes silicon to melt and sublimate

resulting in a silicon particle cloud, which covers the surfaces within the chamber including the

samples mounted at the top of chamber, which are rotated. Heaters are placed above the

sample holders to heat up samples before and during the deposition. The growth rate of the

silicon layer can be controlled by an emission current, which controls the intensity of the elec-

tron beam, enabling deposition rates above 600 nm/min. The structure of the deposited silicon

layer is affected by the temperature used during the process resulting in either amorphous

structures for temperatures of 540°C and below or partly crystalline structures with higher tem-

peratures resulting in more crystalline structures [20].

E-Beam evaporated a-Si layers were investigated for passivating contacts as layers grown

through this process show characteristics that are considered advantageous for passivating

contacts. This includes a high deposition rate enabling thick layers that are more easily com-

bined with the firing of metal contacts, hydrogen-free layers that prevent the formation of blis-

ters during the high-temperature annealing step, and one-sided depositions. Experiments were

performed at 450°C using an emission current of 190 mA. Depositions times were adjusted

during the process by measuring the deposition rate using a resonator. Deposition rates are

determined at four different times during the process during which the resonator is moved into

the particle beam for less than a minute. The thickness of layers deposited was measured

using a profilometer. Measurements were performed at the corners of 10x10 cm2 glass sub-

strates, which are covered during the deposition by the sample holder. Samples would typically

show a decrease in layer thickness related to the position within the chamber with areas closer

to the centre showing thicker layers. This deviation was shown to be within +/-10% of the

averaged thickness.

3.1.3 Plasma enhanced chemical vapour deposition

Plasma enhanced chemical vapour deposition (PECVD) is a vacuum deposition technique that

allows the growth of thin layers at lower temperatures as energy is partially provided electri-

cally. During the process, gas molecules and atoms from a chosen combination of precursor

gases are ionised via electrical discharge between two electrodes located within the chamber.

Radicals resulting from ionisation react with surfaces found within the chamber leading to the

growth of a layer. The choice of process gases and process conditions depends on the desired

composition and structure of grown layers. For a-Si:H layers, which are used in this thesis for

the front side of solar cells to test the behaviour of passivating contacts and to create reference

cells, silane, hydrogen as well as diborane and phosphine are used. Silane is required for the

growth of the amorphous silicon layer. Hydrogen provides hydrogenation for defects found

within the amorphous structure but mainly the interface between amorphous layer and crystal-

line substrate. The ratio between silane and hydrogen gas flow is also of importance not only

for hydrogenation but also for the morphology of the structure, which can turn microcrystalline

for sufficiently low SiH4/H2 ratios [112]. Diborane and phosphine provide doping for p- and n-

doped layers respectively. Besides the chosen process gases, process conditions also need

to be considered as they significantly influence the properties of the grown layer. Important

ones include, temperature, pressure, gas flow rates and electrical power.

In table 2 process conditions for intrinsic and doped Si:H layers are listed. The layers were

deposited in an AKT1600 PECVD cluster providing electrical power via a radio frequency gen-

erator operating at 13.56 MHz. The tool consists of three PECVD chambers, which are used

to grow the intrinsic, p-type and n-type nc-Si:H layers respectively. The same chamber used

for n-type nc-Si:H layers is also used for silicon oxide layers, which were used as etching

barriers for the creation of one-side textured wafers. Thick silicon oxide layers were also used

to protect the textured front side of cells during the high-temperature annealing step, which

were grown using a PECVD chamber of a CS400PS cluster tool, which uses a 65 MHz gener-

ator.

Table 2 Parameters used for the deposition of PECVD layers

Layer

nc-Si:H (n)

a-Si:H (i)

a-Si:H (p)

SiOx

Tool

AKT

Temp. in °C

185

190

205

185

450

Power density in

mW/cm2

200

150

Duration in s

101

20 (i/n) / 6+30

(i/p)

12+24

135s

(per 100 nm)

Pressure in mbar

2.5-3.0

1.5

0.3

Gas flow rates in sccm

PH3

SiH4

300 (i/n) / 400

+ 60 (i/p)

4.5

B2H6

10+100

N2O

500

100

3000

300 (i/n) /

400+1380 (i/p)

540

3.1.4 Plasma doping

Doping for the E-Beam silicon layers was provided via a PECVD-based doping process re-

ferred to as plasma doping. While the process uses the same tools and base principles as the

previously described PECVD process, no layer is deposited during the process. Instead, a

plasma containing H2 and PH3 is ignited resulting in the creation of phosphorous containing

radicals, which diffuse into the sample. Doping is then finalised during a high-temperature an-

nealing step, which is also used here for the formation of the passivating contact. While previ-

ous experiments performed have shown the successful application of this process [20], no

detailed investigations have been performed up to this. It is assumed that during the initial

plasma doping step phosphorous mostly accumulates close to the surface of the sample. The

majority of phosphorous atoms is assumed to be electrically inactive after the completion of

the plasma doping step based on the process temperature [113] and the high concentration of

phosphorous atoms necessary to achieve high doping concentrations of ~1020 atoms/cm3

throughout the poly-Si layer. During the high-temperature annealing step phosphorous diffuses

deeper into the sample and becomes electrically active leading to the formation of the doping

profile.

To confirm assumptions made about the functionality of the process investigations were per-

formed using secondary ion mass spectroscopy and electrochemical capacitance-voltage

measurements to determine the total phosphorous concentrations after the plasma doping and

annealing steps and to determine the electrically active phosphorous concentrations after the

annealing step. The results of these investigations are further discussed in chapter 5.

Plasma doping was performed using the same 65 MHz PECVD chamber used for the deposi-

tion of silicon oxide capping layers (VA). Processes were performed at 450°C and 250W using

100 sccm H2 and 20 sccm PH3 (2% diluted in H2) for a duration of 3 min. Pressure during the

process was varied from 0.3 up to 1.5 mbar, although most of the discussions in the following

chapters focus on processes performed at 0.3, 0.5 and 0.7 mbar as higher doping pressures

of 1.0 mbar and above would lead to a significant reduction in passivation quality for a-Si layer

thicknesses of ~340 nm and annealing temperatures of 885°C and above most-likely linked to

more significant indiffusion.

3.2 Characterisation

3.2.1 Photoconductance decay

Photoconductance decay (PCD) measurements are typically performed to evaluate the quality

of passivation provided by layers and structures realised on the surfaces of silicon wafers as

well as the impact of process steps and treatments on the passivation quality. During the meas-

urement a sample is illuminated for a short duration resulting in the generation of excess

charge carriers, which recombine over time. A change in the concentration of excess charge

carriers also causes the conductance of the sample to change, which is measured by a circuit

that is inductively coupled to the sample. The measured change in conductance is used to

calculate an equivalent excess minority charge concentration and in turn the effective lifetime

of the sample. While the effective lifetime is not only affected by the quality of surface pas-

sivation but also recombination mechanisms present within the bulk of the material as previ-

ously discussed in chapter 2.2, surface recombination generally presents the limiting factor for

the lifetime of devices using monocrystalline silicon wafers.

Photoconductance decay measurements can be performed either under transient or quasi-

steady-state conditions (QSSPC). The main difference between both methods is the duration

of the flash used for excitation. For samples with high lifetimes, where recombination occurs

over a longer duration, a short flash (transient) is sufficient to determine accurate results. For

sample with low lifetimes, however, the intensity of the flash light decays over a longer period

of time (QSSPC). For the transient version of this characterisation method, the effective mi-

nority lifetime is calculated according to eq. 19. For QSSPC eq. 19 has to be changed to con-

sider the generation rate of charge carriers G, resulting in eq. 20 [114].

𝑡𝑒𝑓𝑓,𝑡𝑟𝑎𝑛𝑠𝑖𝑒𝑛𝑡=− 𝑛(𝑡)

𝑑𝑛(𝑡)

𝑑𝑡

(19)

𝑡𝑒𝑓𝑓,𝑄𝑆𝑆𝑃𝐶=− 𝑛(𝑡)

𝐺(𝑡)−𝑑𝑛(𝑡)

𝑑𝑡

(20)

Data derived from PCD can be used to estimate the potential of device structures resulting in

an implied open-circuit voltage (iVoc) according to eq. 21. The equation is based on the ideal

voltage of a solar cell, which is only limited by fermi-level splitting under illumination [115].

i𝑉𝑜𝑐=𝑘𝑇

𝑞𝑙𝑛[(𝑛0+∆𝑛)(𝑝0+∆𝑛)

𝑛𝑖2+1]

(21)

Through the use of a reference cell iVoc values can not only be linked to an excess minority

charge carrier concentration but also an illumination intensity. However, optical properties of

the investigated structure also need to be considered as they affect the number of photons

absorbed and therefore the generation rate of the device. An optical factor f is introduced,

which described the ratio of generation rates between sample and reference cell. Current den-

sities generated by the device derived from optical simulations can also be used here to esti-

mate the optical constant f according to the following equation:

𝑓≈𝑗sc,sample

𝑗sc,ref

(22)

For symmetrical lifetime samples with the passivation contact on both sides and a poly-Si

thickness of ~300 nm an optical constant of ~0.55 was determined. While this value is accurate

regarding the investigated structure, it poorly reflects the potential of the device as poly-Si

layers are not typically employed for the front side of TOPCon-like cells especially with such a

high thickness. An optical constant of 0.95 was, therefore, used to estimate the potential of the

contact for solar cell applications. The value was chosen based on optical simulations and JV

measurements performed on cells using the passivating contact on the rear of the device.

However, one could also consider optical constants above 1 based on the results reported for

TOPCon cells [116].

Figure 3.1 Results from photoconductance decay measurements showing a) the relation between

implied open-circuit voltage and illumination intensity and b) the effective lifetime of minority charge

carriers for different minority charge carrier concentrations.

Measurements were conducted using a WCT – 120 Sinton lifetime tester. The heater of the

device was switched on 15 min before the start of measurements to ensure that a temperature

of 25°C was reached. For the measurements a resistivity of 3 Ωcm and a thickness of 135 µm

was assumed for all samples, based on previous measurements performed. Minor deviations

can be expected in particular for the base resistivity of silicon wafers used as the n-type base

doping for wafers is provided via diffusion processes resulting in a doping gradient between

the centre and edge of the wafer.

3.2.2 Photoluminescence

Similar to PCD measurements, the results of photoluminescence measurements (PL) are

strongly tied to the effective lifetime of minority charge carriers within the sample. While it is

possible to gain quantitate information about lifetimes using PL [117], the technique is typically

used to qualitatively compare samples. The main advantage compared to PCD is that meas-

urements show spatial information rather than an averaged value. PL requires an illumination

source typically in the form of LEDs, a detector and filters. Samples are illuminated so that

charge carriers can be excited and recombine. Some charge carriers recombine via radiative

recombination, as discussed in chapter 2.2, resulting in the emission of photons with a fixed

wavelength according to the s bstrate’s band a . Optical filters are designed to let these pho-

tons pass and reach a detector, where each photon is counted. For samples with well passiv-

ated surfaces or low SRH recombination, more charge carriers recombine via radiative recom-

bination compared to samples with poor surface passivation linking passivation quality of struc-

tures used with radiative recombination and, therefore, the intensity of the measured signal.

Ho e er the sam le’s o tical ro erties also need to be considered as they limit the number

of photons that can be absorbed by the sample. Samples with the same or similar passivating

structures but with additional layers e.g. antireflective coatings, show significantly different re-

sults despite similar minority carrier lifetimes. For this reason, PL measurements are typically

performed in combination with PCD measurements.

PL measurements were performed using a custom tool consisting of LED lamps, a filter block-

ing wavelengths below 800 nm, a camera that is sensitive in the near-IR wavelength range

and a charge-coupled device sensor, which only detects wavelengths around ~1100 nm.

Measurements were performed on samples with SDE surfaces unless stated otherwise. Qual-

itive comparisons were made between samples with the same structure and to confirm the

homogeneity of applied surface passivating structures.

3.2.3 JV characteristics

Solar cell parameters discussed in chapter 2.1.1 are obtained through JV measurements.

Here, an external voltage is applied to the sample and the resulting current density is meas-

ured. By varying the voltage and measuring the resulting current a JV curve can be created

from which solar cell parameters can be extracted. As the performance of solar cells is tied to

the property of light used for illumination in addition to the temperature of the device, meas-

urements are performed under so-called standard test conditions (STC). These consist of an

irradiance of 1000 W/m2, a spectrum close to the AM 1.5G spectrum [118] and a temperature

of 25°C. Spectral conditions are realised by using a combination of lamps or LEDs. Measure-

ments are also performed in the dark to extract other cell parameters such as series resistance,

dark saturation current, etc. For the series resistance in particular several methods are sug-

gested, which involve the evaluation of the resistance at different points of a single or multiple

JV curves measured at different illumination conditions. The use of both dark and light JV

measurements to calculate the series resistance is generally considered a more accurate

method [119]. Here, the dark JV curve is shifted so that the short-circuit current for the dark

measurement jsc,dark is equal to the measured jsc of the cell. The series resistance in then cal-

culated from the difference in voltage at the MPP for the light and shifted dark JV curve. A

more accurate version of this method also considers the dark series resistance according to

eq. 24, which is used here.

𝑅𝑠=𝑉𝑑𝑎𝑟𝑘,𝑚𝑝𝑝−𝑉𝑙𝑖𝑔ℎ𝑡,𝑚𝑝𝑝−(|𝐽𝑠𝑐|−|𝐽𝑚𝑝𝑝|)𝑅𝑠,𝑑𝑎𝑟𝑘

|𝐽𝑀𝑃𝑃|

(23)

𝑅𝑠,𝑑𝑎𝑟𝑘=𝑉𝑑𝑎𝑟𝑘,𝑗𝑠𝑐−𝑉𝑜𝑐

|𝐽𝑠𝑐|

(24)

Measurements were performed using a Wavelabs LED sun simulator. Samples were meas-

ured after adjusting the irradiance, which was tested and confirmed using a reference cell.

Samples are placed on a water-cooled chuck, to connect the rear side of the cell and to control

the temperature of the device. Cell areas are defined via masks. Measurements were per-

formed using a voltage range of -0.2 – 0.9 V with a step size of 0.01 V. Dark and light JV

measurements were performed in sequence for each sample.

3.2.4 External quantum efficiency

While parameters extracted from JV measurements are typically used to evaluate solar cells,

they give little information about the origin of losses limiting the performance of the device.

External quantum efficiency (EQE) defines the percentage of incident photons of a specific

wavelength, which are used to generate collected minority charge carriers. In an ideal case,

the EQE would be 100% for photons with energies above the bandgap and would drop to 0%

for photon energies below the bandgap. However, the EQE is limited by multiple factors as

shown in Figure 3.2. Losses related to recombination at the front surface decrease the re-

sponse for lower wavelengths. Similarly, recombination losses at the rear surface affect high

wavelengths close to the bandgap. As the optical properties of the device affect the entire EQE

curve by limiting the number of photons absorbed via reflection, data from EQE measurements

is typically compared to reflectance (R) measurements. This is either done by comparing the

EQE curve with a 100%-R curve, which represents the max. achievable EQE, or an internal

quantum efficiency is calculated using reflectance data.

Figure 3.2 External quantum efficiency of a solar cell. Deviations between the ideal and measured

quantum efficiency can be connected to losses occurring at the front or rear of the cell as well as optical

properties.

EQE is determined from spectral response (SR) measurements, which describes the ratio be-

tween photogenerated current and power of the incident irradiation [120], according to eq. 25

sin lanck’s constant h the s eed of li ht c, the elementary charge q and the wavelength λ.

𝐸𝑄𝐸(𝜆)=𝑆𝑅(𝜆)ℎ𝑐

𝑞𝜆

(25)

Measurements were performed using two custom setups. The first setup was used to measure

the EQE across the entire width of the cell including the influence of the front fingers. Here, a

400 W halogen lamp is used as illumination source with LEDs to realise bias illumination. The

second setup allows measurements to be performed on smaller areas, which was used to

measure the EQE without the influence of the front grid used as a basis for simulation per-

formed in GenPro4. The second setup uses a combination of 400 W halogen and 75 W xenon

lamps together with LEDs or an additional halogen lamp for bias illumination. For both setups

measurements were conducted from 300 nm up to 1200 nm with a step size of 10 nm.

3.2.5 Transfer Length method

Issues relating to the series resistance of solar cells are typically connected to the formation

of the metal contact and the associated contact resistivity, which can be calculated using the

transfer length method (TLM). The method requires contact stripes with equal width and length

to be placed on the sample with increasing distance between each contact stripe. I-V

characteristics are then measured by contacting a pair of stripes, typically those placed next

to each other. By plotting the distance between contacts vs. the measured resistance, the

contact resistance of the contact can be determined by extrapolation the curve to the y-Axis

as shown in Figure 3.3b. The contact resistivity is then calculated using the contact resistance

as well as the geometry of the contact stripes and transfer length, shown in Figure 3.3a

according to eq. 26.

Figure 3.3 Schematic of structures used for TLM measurements (a). Method to determine the contact

resistance and transfer length from TLM measurements (b). Figures recreated from [121] and taken

from [122].

𝜌𝑐=𝑅𝑐𝐿𝑇𝐿𝑀𝑊

𝑐𝑜𝑡ℎ( 𝐿

𝐿𝑇𝐿𝑀)

(26)

TLM measurements were used to investigate alternative contacting schemes for the passivat-

ing contact relying on the use of Ag and TCOs. As the method necessitates the formation of

ohmic contacts as well as defined current paths, 100 nm thick silicon oxide layers were depos-

ited on top of c-Si substrates before the deposition and formation of the passivating contact so

prevent current from flowing through the c-Si substrate. Patterning for the contact was realised

using photolithography according to [79]. While it was possible to create TLM structures using

this process, the metal contact would frequently lift-off during the etching process, which is

assumed to be the result of under-etching. Therefore, an alternative processing scheme was

also investigated. Here, the Ag layer was grown using TLM masks resulting in thicker contact-

ing stripes compared to the ones resulting from the previous method. After coating the Ag

stripes in an etch-resistant coating, the TCO was patterned using the previously used method.

Geometries of TLM structures formed using both methods are shown in table 3.

Contact resistivities resulting from these experiments do not represent the contact formed be-

tween two layers but instead the sum of the Ag/TCO and TCO/poly-Si contact. While a more

thorough analysis is required to determine the contribution of each contact formed, measured

contact resistivities discussed in chapter 7.3. were low enough that they are not considered a

limiting factor for the device.

Table 3 Dimensions of TLM structures investigated

Type

Photolithography [79]

Adjusted scheme

Width W in mm

Length L in mm

0.5

1.1

Distance d1 in mm

0.05

0.35

Distance d2 in mm

0.1

0.65

Distance d3 in mm

0.2

0.85

Distance d4 in mm

0.4

1.35

Distance d5 in mm

0.6

1.85

Distance d6 in mm

2.85

3.2.6 Ellipsometry

Ellipsometry is a method to determine the optical properties and thickness of layers. Mono-

chromatic light with known polarisation is directed towards a sample, where part of the beam

is reflected towards a sensor measuring the intensity and polarity of the reflected beam. From

the intensities of the parallel and perpendicular polarised reflected light rp and rs one can then

derive the change in amplitude ψ and phase difference Δ according to the following equation:

𝑟p

𝑟s=𝑡𝑎𝑛(𝜓)𝑒𝑖∆

(27)

Optical properties as well as the thickness of layers cannot be directly extracted from ellipsom-

etry data but have to be determined via data fitting using an optical model that describes the

sample. For the poly-Si layer refractive data from Jellison et al. was used [123], which was

able to accurately model ~325 nm thick doped poly-Si layers, which were grown on c-Si sub-

strates with an ~300 nm thick silicon dioxide layer, with some deviations for wavelengths below

500 nm, which can be explained through the roughness of the sample as well as the presence

of silicon oxide on top of the poly-Si layer strongly affected the properties of the sample. Data

for the poly-Si layer was mostly used for optical simulations performed in GenPro4, which are

further discussed in chapter 3.2.7 and 7.3. The model for the ultra-thin silicon oxide layer was

created based on investigations performed by A. Moldovan [124]. Here, the transitional area

was realised by mixing the properties of the bulk silicon dioxide layer, which was modelled

using a Cauchy equation, and the silicon substrate. The thickness of the interfacial layer was

assumed to be 0.2 nm for all samples. While this approach considers the presence of the

interfacial layer, properties of the interfacial layer could not be determined via parameter fitting

as the thickness would either reach values beyond 1 nm, which significantly exceed the ex-

pected value of ~0.2 nm [125], or would reach limits set for the thickness of the interfacial layer.

It should, however, be said that the quality of fits did not change significantly with the thickness

of the interfacial layer.

Measurements were performed using a SE850 ellipsometer by SENTECH. Samples were

characterised within a wavelength range of 190-850 nm (ultra-thin silicon oxide) and 190-1200

nm (poly-Si). For the ultra-thin silicon oxide layer measurements were only performed at high

angles (70° and 80°), which feature higher optical light paths.

3.2.7 Optical simulations in GenPro4

To better understand the behaviour and limitations of cells processed, simulations were

performed using GenPro4 [126]. GenPro4 is a simulation program used to determine losses

within a device related to optical effects such as reflection and parasitic absorption, which is

determined for each layer, reflection and transmission. The basis of the model is formed by

data sets for the refractive index (nk) of each layer, which are used to construct the device

within the program. Simulations are then performed using ray-tracing as a basis.

Simulations were performed here to determine and quantify limitations regarding the short-

circuit current density of cells processed. Tab 4. summarises thicknesses used. Thicknesses

for the a-Si:H layers were not measured but chosen based on the results of previous

ellipsometry measurements performed. For the silicon oxide layer the thickness measured

before the deposition of the a-Si layer and formation of the passivating contact was chosen,

although the impact that the silicon oxide has on the results of simulations performed is

negligible. Thicknesses for the metal layers on the rear of the device were not directly

measured, however, are expected to be close to the assumed values as the growth rate of

both layers is monitored and controlled during the deposition.

Table 4 Device structure and layer thicknesses used for optical simulations performed in GenPro4

Layer

Thickness in nm

newScot (front)

a-Si:H (n)

a-Si:H (i)

c-Si

135000

SiO2

1.5

poly-Si (n)

300

AZO

newScot (front)

1000

3.2.8 Transmission electron spectroscopy

Transmission electron microscopy (TEM) measurements were performed externally at the

HZB (Wannsee) for oxidised wafers to determine the thickness of the oxide as a point of com-

parison to values extracted via XPS and ellipsometry and to determine the structure for poly-

crystalline layers. Oxidised samples were prepared by cleaving the sample and gluing two

pieces together (oxide pointing towards the centre) and reducing the thickness of the sample

via ion milling. Samples with poly-Si were prepared via slicing with a focused ion beam. Oxide

thickness was determined from high-resolution measurements. For the poly-Si layers bright

and dark field measurements were performed to differentiate between amorphous and crystal-

line structures.

3.2.9 X-ray photoelectron spectroscopy

X-ray photoelectron spectroscopy (XPS) is an analytical technique used for the characteriza-

tion of surfaces and interfaces, and is based on the photoelectric effect. Typically, XPS is used

to determine the chemical composition of solids in a volume close to the surface. Other prop-

erties may also be determined such as the thickness of ultra-thin films, changes in the structure

of the material and the valence band edge using ultraviolet radiation instead of X-rays (UPS).

For XPS, X-rays are generated in an ultra-high vacuum and the associated photons are ab-

sorbed within a sample. The energy of x-ray photons is high enough to excite electrons from

core levels to an energy state that allows electrons to be released into the vacuum. Released

electrons can be analyzed to determine the number of electrons in respect to their kinetic en-

ergy Ekin. By knowing the work function of the analyzer WF and the energy of photons from the

excitation source ℎ𝑣, the binding energy EB of the emitted electron can be calculated according

to the following equation:

𝐸B=ℎ𝑣−𝐸Kin−𝑊𝐹

(28)

Based on the position of peaks found within the spectrum one can then determine elements

that the sample consists of including the presence of contaminants. While the binding energy

of core levels is mostly related to the element and orbital it originates from, smaller changes,

also referred to as chemical shifts, may occur. For silicon oxide a chemical shift in the binding

energy of the Si 2p orbital can be observed as electrons used to form covalent bonds tend to

be attracted by highly electronegative oxygen atoms. These electrons are no longer able to

shield other electrons from the nucleus meaning that the remaining electrons experience

higher binding energies, which grow with the amount of bonds formed between silicon and

oxygen atoms. This means that up to 5 so-called oxidation states may be observed for Si 2p

in silicon oxide films depending on how many covalent bonds silicon forms with oxygen (0 to

4). By deconvoluting the Si 2p spectrum, the composition of silicon oxide layers may be deter-

mined as shown in Figure 3.4.

Another important aspect, which may give rise to a shift in measured peak position, is the

b ild of char e on the sam le’s s rface. his can be obser ed for ins latin [127] and low-

conductivity materials, as the buildup of positive charge due to emission of electrons cannot

be compensated as additional electrons cannot flow through the sample or are transferred at

a slow rate. Positively charged surfaces attract emitted electrons resulting in a decrease in

kinetic energy / rise in effective binding energy.

‘For electrons originating from the silicon dioxide layer a further shift in binding energy with

layer thickness has been reported for ultra-thin layers up to ~ 2 nm [128]. This additional thick-

ness dependent shift in binding energy is assumed to be the result of interactions of released

electrons and core hole image charges formed in the silicon substrate close to the interface.

Once an electron is released, Coulomb interactions between electron and remaining core hole

(attraction) are eakened by the core’s ima e char e (re lsion) leadin to an increase in

kinetic energy that expresses itself through a decrease in binding energy. For thicker layers

the average distance between the interface and excited electrons from the silicon dioxide layer

increases limiting the interactions with the image charge [128].

Figure 3.4 Deconvolution of the Si 2p spectrum for a silicon wafer oxidized in an ozone containing

solution measured by XPS using an excitation energy of 170 eV. The background caused by inelastically

scattered electrons has already been subtracted from this plot.

While absorption lengths of photons from x-ray sources commonly used for XPS, such as

magnesium and aluminum anodes with photon energies of ~1254 eV and ~1487 eV respec-

tively, may reach depths of 1-10 µm, XPS measurements are limited to depths of about 1-10

nm by the mean free path of electrons, which describes the average distance electrons can

travel before interaction with other electrons, atoms, etc. [129]. The information depth can be

adjusted by changing the excitation energy either by using a different x-ray source or by using

synchrotron radiation. For this thesis experiments were mostly conducted using synchrotron

radiation. Measurements were performed at an excitation photon energy of 170 eV and 370

eV to investigate the Si 2p spectrum, 600 eV for the O1s spectrum and 1000 eV for survey

measurements to determine the presence of contaminants and other unexpected elements.

Excitation energies for measurements (excluding surveys) were kept low to decrease the ine-

lastic mean free path of electrons, which grows with the kinetic energy of electrons for kinetic

energies above ~100eV [129]. For the Si 2p signal, the use of higher excitation energies would

lead to enhanced photoelectron excitation in the silicon bulk, resulting in an increase in Si0

intensity, which would dominate the spectrum.

Other effects, which impact the shape of the XPS signal, include interactions between excited

electrons and their orbitals also referred to as spin-orbit splitting [130]. Spin-orbit splitting in

particular has a very noticeable impact on the shape of the XPS spectra for all orbital levels

except for s-orbitals (e.g. O 1s) and leads to the appearance of two peaks for each chemical

state as shown in Figure 3.4. A more thorough discussion of spin-orbit splitting and other ef-

fects that affect the shape of the XPS spectrum may be found elsewhere [130].

3.2.10 Modelling of X-ray photoelectron spectroscopy data

XPS data fitting was performed using the software fityk [131]. Spin-orbit-splitting was imple-

mented for each oxidation state belonging to the Si 2p spectra with an area ratio of 1:2 and a

difference in binding energy position of 0.6 eV. Each curve was created using an approximation

of the voigt function, which is a convolution of the Gaussian and Lorentzian distribution func-

tions and is used to describe the shape of XPS signals resulting from a single state, that uses

the area of the curve as fitting parameter [132]. It was assumed that the Si 2p spectra consisted

of 5 oxidation states. Besides the peak belonging to elemental Silicon Si0, oxidation states for

the interfacial oxides Si1+, Si2+ and Si3+ as well as the oxidation state for pure silicon dioxide,

Si4+, were considered. Peak positions of the interfacial oxidation states Si1+ (+0.95 eV), Si2+

(+1.75 eV) and Si3+ (+2.48 eV) were set relative to the binding energy of the Si0 3/2 peak

according to the research of Himpsel et al. [125]. For ultra-thin silicon oxide layers the binding

energy of Si4+ shifts with the thickness of the oxide layer as previously mentioned. Therefore,

the position for the Si4+ 3/2 peak was set so that the peak would always have a binding energy

higher than the Si3+ peak. A Shirley background was automatically calculated by fityk and re-

moved before curve fitting was performed. To compensate charging, which occurred during

the measurements, peak positions were shifted so that all Si0+ 2p 3/2 peaks were at the same

position as the 3/2 peak of a silicon reference sample, where silicon oxide on the surface of

the sample was removed with 1% HF. Samples that received the same treatment after oxida-

tion were fitted at the same time. The constraints for each group included that the ratio between

Lorentzian and Gaussian width or shape of each oxidation state has to be the same and that

the Gaussian width of the curve for each separate oxidation state needs to be the same across

all samples within the group. A summary of constraints and starting values used can be found

in table 5.

Table 5 Model parameters and restrains used for fitting data from measured Si 2p spectra

Model parameters

Oxidation state

Si0

Si1+

Si2+

Si3+

Si4+

3/2 peak position in

Si0+0.95

Si0+1.76

Si0+2.48

Si3+< Si4+

Gaussian width in eV

0-1 eV (Same for each spin-orbit pair)

Lorentz/Gauss ratio

(shape)

Same as

Si0

Same as

Si0

Same as

Si0

Same as

Si0

Starting values

(width in eV, shape,

area in eV)

0.2

Area

No negative values

3.2.11 Secondary ion mass spectrometry

Secondary ion mass spectrometry (SIMS) provides information regarding the composition of

substrates or layers close to the surface of the sample. Here, samples are bombarded by a

beam of ions (primary ions) typically Ga+, O2+ Ar+ or Cs+, which penetrate the surface of the

sample and cause a collision cascade leading to the ejection of particles similar to sputtering.

Ejected particles include ions (secondary ions), which can be collected and analysed in a mass

spectrometer. The analysis depth of this technique is typically limited to 1-2 nm, although it

depends on the kinetic energy of primary ions used. Surface measurements can, therefore, be

rather sensitive to the presence of contaminants such as carbon. By continuously bombarding

the surface with primary ions resulting in the continuous removal of matter from the sample

and measuring secondary ions, it is possible to create a time-resolved intensity profile, which

correlates to a depth-resolved distribution of elements within a sample. Compared to XPS,

reference samples are required to determine the exact composition of the material. It is not

only important that relevant elements appear in a known ratio but also that the structure of

these references is similar to the samples investigated, which is also known as the matrix effect

[133]. The detection limit of the technique varies depending on the elements investigated.

A variation of the SIMS technique is the so-called time-of-flight (TOF) SIMS. In TOF SIMS ion

lenses that filter ions by their mass-charge ratio as well as the mass analyser are replaced with

a TOF analyser. Here, ions that are able to reach the flight tube are accelerated by an electrical

field. Ions leave the electrical field with the same kinetic energy Ekin but with different velocities

based on their mass according to eq. 29. By substituting the velocity with the ratio of distance

s and time t and rearranging the equation, one can determine the relationship between mass

and the known values for the electrical field constant U, distance d and the time it takes for

ions to reach the TOF analyser t according to eq. 30 [133]:

𝑞𝑈=1

2𝑚𝑣2=𝐸kin

(29)

𝑚=2𝑞𝑈

𝑑2𝑡2

(30)

The technique also allows to decouple the sputtering and analysis phase. Two primary ion

beams are used one for the sputtering process with high doses but low beam energy and one

for the analysis of the material with high beam energy but low doses. This improves the overall

quality of the measurement as low beam energies are necessary for high depth resolutions

whereas high beam energies are required for high mass resolutions.

Figure 3.5 Schematic of a tool used for time-of-flight secondary ion mass spectrometry. Figure taken

from [133].

TOF-SIMS measurements were performed at the Forschungszentrum Jülich to investigate the

plasma doping step discussed in chapter 5. Here, samples were investigated that were pro-

cessed using varying pressures during the plasma doping step as well as annealing tempera-

tures. Sputtering was performed using Cs ions at 1 keV over an area of 250x250 µm2. For the

analysis a Bi1+ ion beam was used at 30 keV over an area of 30x30 µm2. Samples were com-

pared to a doped silicon reference with known maximum doping concentration of 3x1020 atoms

per cm3.

3.2.12 Electrochemical capacitance voltage profiling

Electrochemical capacitance voltage (ECV) is a destructive characterisation method partly

based on the same principles as capacitance voltage (CV) measurements. The main differ-

ence between both techniques is that with ECV the doping of the sample can be determined

for different depths by etching into the sample leading to the creation of depth-dependant dop-

ing profiles. Etching is accomplished by dissolving the semiconductor electrolytically. The tech-

nique is commonly used to determine the distribution of electrically active dopants within silicon

solar cells as well as the passivating contact as opposed to SIMS measurements, which show

the total concentration of dopants.

During the measurement a semiconductor is pressed against a sealing ring to define the active

area of the reaction and analysis, and to prevent the flow of chemicals across the surface of

the sample. The sealing ring is attached to an electrochemical cell, which is filled with an elec-

trolyte that forms a Schottky contact with the semiconductor. By applying a potential, which is

measured with respect to the reference electrode, between a platinum electrode and the sem-

iconductor (shown in Figure 3.6) a depletion region is formed close to the surface of the sample

which has a capacitance C. The concentration of dopants N at the edge of the depletion region

can be calculated from the measured capacitance using the following equation, where 𝜀𝑟 is the

relative permittivity of the sample, 𝜀0 is the vacuum permittivity, A is the active area and dC/dV

is the slope of the CV curve at the edge of the depletion region [134].

𝑁= 1

𝑞𝜀𝑟𝜀0𝐴2𝐶3

𝑑𝐶/𝑑𝑉

(30)

𝑤𝑟=𝑀

𝑍𝐹𝐷𝐴∫𝐼𝑑𝑡

𝑡

(31)

Figure 3.6 Schematic of electrochemical cell used for ECV measurements. Samples are attached to the

plunger on the right side. Figure taken from [135].

The amount of material removed depends on the etch current flowing through the system,

which is controlled through a circuit between sample and counter electrode. From this the

thickness of the removed layer wr can be calc lated accordin to Faraday’s la of electrolysis.

In eq. 31, M is the molecular weight, Z is the dissolution number, F the Faraday constant. It is

important to note that the removed layer is not equal to the thickness of the profiled layer,

which is the sum of the etched layer and depletion width wd. Eq. 32 describes the depletion

width dependant on the flat band potential 𝛷FB and applied voltage.

𝑤𝑑=√2(𝛷𝐹𝐵−𝑉)𝜀𝑟𝜀0

𝑞𝑁

(32)

Etching and CV measurements are alternated so that a depth-dependant dopant concentration

profile can be calculated. While the technique is able to measure doping concentrations as low

as 1012 atoms per cm3 [134], the error in the determined doping concentration tends to be

rather high for doping concentrations close to the ones found in silicon substrates used for

silicon solar cells (~1015-1016 atoms per cm3).

ECV measurements were used to further investigate the plasma doping method. Measure-

ments were conducted externally at the Forschungszentrum Jülich using a wafer profiler

CVP21 from WEP-C. These investigations used the same samples as the ones investigated

via SIMS and PCD measurements.

4 Wet-chemical ozone-based silicon oxide

layers for passivating contacts

Chapter 4 focuses on investigations conducted to determine the structural composition of sili-

con oxides grown in wet-chemical ozone solutions and structural changes these oxides un-

dergo during the processing of passivating contacts. Investigations focus mostly on the results

of XPS measurements that were performed on freshly grown and processed oxides.

4.1 XPS analysis of wet-chemical ozone-based oxides

The following chapter includes the results of XPS measurements that were performed on

<111> silicon wafers oxidised in wet-chemical ozone solutions. The goal of these investiga-

tions is to determine structural changes the oxide undergoes during the annealing step and to

link this behaviour to the generally good surface passivation wet-chemical ozone oxides pro-

vide within passivating contacts. For this purpose, freshly oxidised samples were compared to

oxides used for passivating contacts. Poly-Si from the passivating contact was removed using

TMAH as it is otherwise not possible to characterise the oxide using XPS. Oxide growth con-

ditions were varied within ranges that were possible to reach for the wet-chemical ozone pro-

cess and deemed relevant for passivation contacts based on the thickness of the resulting

oxide layer. The variation in growth conditions or rather the use of oxides with different thick-

nesses also helps to evaluate damage caused to the oxide layer during the TMAH etching

process. Investigations focused mostly on the Si 2p and O 1s spectra, which are visible in the

XPS surveys shown in Figure 4.1. Besides oxygen and silicon, only carbon was present, which

most-likely ori inates from the sam les’ s rfaces res ltin from reactions bet een the sam le

and carbon dioxide in the air. Si 2p spectra were investigated to determine changes in the

structure of the transitional area. O 1s spectra were investigated to locate other bonds besides

Si-O (if present) that could be an indicator for additional structural changes that occur during

the annealing step besides the ones that are assumed to take place at the interface.

Figure 4.1 XPS surveys for oxidised silicon samples measured using synchrotron radiation (hν = 1keV).

4.2 XPS sample preparation

<111> Si FZ wafers were first cut into 5x5 cm2 pieces using a laser scribing system MD-

U1020C from Keyence. Samples were cleaned using a RCA cleaning procedure according to

chapter 2.5.1. After removing the oxide originating from the cleaning procedure using 1% HF

for 2 min, samples were oxidised in wet-chemical ozone solutions. Samples where then divided

into three different groups. The first group consists of freshly grown oxides, which were imme-

diately placed on a sample holder and stored in a vacuum chamber adjacent to the XPS cham-

ber until the measurement occurred. For the other two sample groups amorphous silicon with

a thickness of ~300 nm was deposited directly on top of the oxide. Plasma doping was then

performed at a pressure of 0.5 mbar using the process conditions discussed in chapter 3.1.4.

Samples were then annealed at either 800°C or 900°C for 30 min followed by annealing in

forming gas at 450°C for 30 min to saturate recombination-active dangling bond defects. Poly-

Si was removed using 5% TMAH at 60°C for 1.25 min. Afterwards both the samples annealed

at 800°C and 900°C were stored in vacuum and measured.

Figure 4.2 Process scheme used for XPS measurements. Experiments included freshly grown oxides

as well as oxides that were previously part of a passivating contact.

One issue with this approach is that the oxide could be removed or damaged during the TMAH

etching process, although it should be said that TMAH is rather selective when it comes to the

removal of silicon compared to silicon oxide [136]. To limit the damage caused to the silicon

oxide, process conditions were investigated for the TMAH etching step. Temperature and con-

centration were chosen based on research showing high selectivity for low temperatures and

TMAH concentration. A duration of 1 min was initially chosen and tested based on the visible

development of hydrogen during the process. Samples were prepared using wet-chemical

ozone oxides as well as a thick PECVD oxide layer with doped poly-Si layers on top and in-

vestigated using ellipsometry and scanning electron microscopy (SEM). According to ellipsom-

etry measurements, no poly-Si was visible after 1 min etching in TMAH for both the sample

with ultra-thin and thick oxide. Similarly, SEM measurements performed on samples with thick

oxide shown in Figure 4.3 show no poly-Si layer after the etching step. As the properties of

poly-Si layers grown using E-beam evaporation can vary, the process time was increased to

1.25 min to ensure that the poly-Si is completely removed. Under the assumptions that silicon

oxide reacts with TMAH for 0.25 min or 1.25 min one can expect that on average ~0.025 or

~0.125 nm of the oxide is removed respectively.

Figure 4.3 SEM images of c-Si wafers with silicon dioxide and doped poly-Si on top before a) and

after b) etching the samples in TMAH solutions.

4.3 Determining the thickness of ultra-thin silicon oxides

The thickness of ultra-thin oxides can be determined using XPS data according to a method

proposed by M. P. Seah and S. J. Spencer [137]. Here, contributions from each oxidation state

towards the total thickness of the transitional area are determined in the form of equivalent

layer thicknesses using the following equations:

𝑑Si3+ =𝐿Si3+cos⁡(𝜃)𝑙𝑛[1+( 𝐼Si3+

𝑅Si3+𝐼Si)]

(33)

𝑑Si2+ =𝐿Si2+cos⁡(𝜃)𝑙𝑛[1+( 𝐼Si2+

𝑅Si2+𝐼Si)]

(34)

𝑑Si1+ =𝐿Si1+cos⁡(𝜃)𝑙𝑛[1+( 𝐼Si1+

𝑅Si1+𝐼Si)]

(35)

Besides the intensity for each peak, which can be extracted from XPS data, information about

the attenuation length L as well as the ratio RSix+ between suboxide and silicon intensity for

bulk materials are required. It is, however, impossible to determine these values through meas-

urements for the sub-oxides. M. P. Seah and S. J. Spencer, therefore, suggest a simplified

approach that assumes R and L linearly increases from Si0 to Si4+ [137].

𝑅Six+ =1+0.25(𝑅Si4+−1)∗𝑥

(36)

𝑅Si4+ =𝐼Si4+bulk

𝐼Si0bulk

(37)

𝐿Six+ =𝐿Si0+(𝐿Si4+−𝐿Si0)∗0.25𝑋

(38)

𝑑ox=𝐿Si0𝑐𝑜𝑠𝜃𝑙𝑛[1+ 𝐼Si4+ +0.75𝐼Si3++0.5𝐼Si2++0.25𝐼Si1+

𝑅0(𝐼Si+0.75𝐼Si1++0.5𝐼Si2+ +0.25𝐼Si3+)]

(39)

RSi4+ is derived from bulk measurements of samples with thick silicon dioxide layers and silicon

wafers after the removal of native oxides via HF. The total thickness of the oxide is then cal-

culated according to eq. 39.

Figure 4.4 contains the results of calculations performed for ozone oxides grown in solutions

containing 30 ppm ozone, showing values between 0.85 and 0.9 nm based on the oxidation

duration. R0 (0.397) and LSi0 (0.61 nm) were derived from measurements performed on silicon

wafer after removing the surface oxide in 1% HF and a 4 nm oxide grown using PECVD. These

values are about 30-40% lower compared to values measured using ellipsometry, which are

shown to reach thicknesses as high as 1.5 nm. Deviations in the thickness extracted from both

methods have previously been reported by other groups showing similar results [138]. Devia-

tions could be explained through the presence of carbon- or OH-based contaminants present

on the surface of the sample, however, XPS spectra further discussed in the following chapters

have shown low contributions from other sources besides the silicon substrates and silicon

oxides. Measurements were, therefore, compared to oxide thicknesses extracted from TEM

measurements. TEM measurements performed on the sample shown in Figure 4.5, which was

oxidised for 30 min, indicate a thickness of ~1.4 nm, which is similar to the value extracted

from ellipsometry and similar to reported thicknesses for silicon oxides grown using similar

wet-chemical ozone solutions [111]. The results of ellipsometry measurements are, therefore,

assumed to be more accurate compared to the method proposed by M. P. Seah and S. J.

Spencer. However, the method is still considered useful to investigate changes within the struc-

ture of the transitional area promoted by high-temperature annealing further discussed in chap-

ter 4.5.

Figure 4.4 Oxide thickness derived from XPS data using the method described by M. P Seah and S. J.

Spencer and ellipsometry data for oxides grown in 30 ppm ozone solution for different durations.

Figure 4.5 TEM measurements performed on an oxidised c-Si wafer. Measurements show the c-Si

substrate (bottom left), the ultra-thin silicon oxide layer and glue used for sample preparation.

4.4 Bulk silicon oxide properties

Excluding contributions from silicon oxides and silicon substrates, other bonds may also be

present within the bulk of the silicon oxide layer. As the oxide is grown in a wet-chemical solu-

tion mostly consisting of water, one could assume that O-H bonds may also be present, in

addition to O-H and carbon-based bonds located at the surface of the sample, which also

contribute towards the shape of the O 1s spectrum. As the O 1s spectrum only consists of one

peak for silicon oxide layers, the presence of additional bonds besides Si-O can be more easily

spotted compared to the Si2p spectrum. Initial fitting of O 1s peaks of freshly grown oxides

with just one peak showed mostly symmetrical curves with minor asymmetric characteristics

as shown in Figure 4.6a. Figure 4.6b, however, shows a more obvious asymmetry towards

lower binding energies. The addition of a second peak 1.23 eV below the main silicon oxide

peak led to a slight improvement in the quality of fits as shown in table 6. The additional peak

could, however, not fully replicate the broad asymmetric behaviour of the 20 ppm sample. A

second set of data using samples that were annealed at 900°C was used as a point of com-

parison. For the annealed samples one would assume a more symmetric behaviour as O-H

bonds should break up during the annealing process. Samples once again show mostly minor

asymmetries similar to the one shown in Figure 4.6c with the sample in Figure 4.6d showing a

more obvious asymmetric trend, however, this time with a signal broadening towards higher

binding energies.

Tab. 6 Sum of squared residuals of fitted O 1s spectra. Fits using 1 and 2 peaks were compared for

non-annealed samples as well as samples annealed at 900°C

1 peak

2 peaks

Growth conditions

Not annealed

900°C

Not annealed

900°C

20 ppm 10 min 20°C

0.131

0.153

0.039

0.120

30 ppm 30s 20°C

0.053

0.056

0.054

0.049

30 ppm 10 min 20°C

0.037

0.047

0.024

0.056

30 ppm 30 min 20°C

0.156

0.113

40 ppm 10 min 20°C

0.061

0.047

0.042

0.041

20 ppm 10 min 50°C

0.019

0.025

0.019

0.023

Figure 4.6 O 1s spectra for non-annealed silicon oxide samples a) that can be adequately fitted using

one peak (top left) or b) require additional peaks (top right). The same behaviour is shown for samples

annealed at 900°C (c and d). Fits were performed assuming one or two peaks. Samples on the left

represent the quality of fits achieved for most samples. Samples on the right are outliers that showed

noticeable asymmetries.

The addition of a second peak with the previously determined peak position did not improve

the quality of the fit for the annealed samples. As the behaviour of samples investigated were

rather inconsistent and no temperature-dependant behaviour was observable, other bonds that

contribute towards the O 1s spectrum besides silicon-based ones are assumed to originate

from surface bonds rather than bonds present within the bulk silicon dioxide or transitional

area. The broadening of the O 1s peak may, therefore, only be connected to the time samples

spent in ambient environments.

4.5 Evolution of silicon oxide structure and composition

during annealing

Structural changes in the silicon oxide layer as well as in the transitional area may express

themselves either through a change in composition e.g. through the contribution of the

interfacial suboxide peaks to the total Si 2p spectrum, the ratio between oxidation states, or a

change in structure. While XPS cannot directly give information about the structural changes

within a sample or layer, it is possible to see the impact of such changes through the full width

at half maximum (FWHM) of each peak. Figure 4.7 includes Si 2p spectra for all samples at

an excitation energy of 170 eV. The impact of growth conditions on the oxide thickness can be

seen indirectly through the intensity of the Si0 peak at a binding energy of ~100 eV. A decrease

in Si0 intensity, which can be observed by comparing freshly grown samples with the annealed

ones, corresponds to the growth of the silicon oxide layer. Changes in oxide thickness can also

be overserved through the position of the Si4+ peak position as discussed in chapter 3.2.9.

However, as the shift in peak position is logarithmically linked with the thickness of the oxide,

the deviation is not quite as clear especially for thicker oxides as the ones annealed at 800°C

and 900°C. Besides the growth of the silicon oxide during annealing, other trends can be

observed related to the structure of the oxide. For the freshly oxidised samples a small hill can

be observed between the Si and SiO2 peak, which disappears after annealing, which is related

to the restructuring of the sub-oxides. To determine the origin of this change, the average

FWHM for each oxidation state and group was investigated, which are shown in Tab. 7.

Tab. 7 FWHM values extracted for each sample group and oxidation state

Peak FWHM in eV

Model with Si2+ state

Si0

Si1+

Si2+

Si3+

Si4+

900°C

0.47

0.70

0.52

1.02

1.37

800°C

0.44

0.66

0.42

0.91

1.29

Non-annealed

0.46

0.59

0.62

1.2

1.25

Model without Si2+ state

Si0

Si1+

Si2+

Si3+

Si4+

900°C

0.46

0.69

1.07

1.36

800°C

0.43

0.64

1.02

1.28

Non-annealed

0.45

0.61

1.35

1.25

Figure 4.7 Si 2p spectra for a) freshly grown silicon oxides, b) oxides that were annealed within a

passivating contact at 800°C and c) 900°C after poly-Si removal in TMAH. Shirley backgrounds were

subtracted before normalising the intensity of each spectrum to the intensity of the Si4+ peak. The effects

of charging were compensated by moving each spectrum so that the Si0 peaks would align.

Measurements were performed at 900°C.

Freshly oxidised samples show a FWHM for the Si3+ that is close to the value of the Si4+ state.

For thermally grown silicon oxide layers the FHWM typically increases with the oxidation state

[125]. As the intensity from the Si2+ state is rather low across all samples with contributions

towards the total suboxide intensity never exceeding 10% it is likely that the FWHM values

shown for this state do not represent physical properties but instead are the result of minimising

the error of the model. The fitting procedure was therefore repeated assuming that Si2+ states

do not exist by setting the area of the corresponding peak to 0. The presence of the Si2+ state

seems to mostly affect the fitting of the Si3+ state, whose FWHM even surpasses the Si4+ state’s

when Si2+ is not considered. Regardless of the approach chosen, a significant change in

FHWM can be observed by comparing the freshly grown oxides with those that were annealed

within passivating contacts. Interestingly, the change in FWHM mostly affects the Si3+ peak. A

narrowing of the FWHM could indicate that the sam le’s str ct ral order im ro es and there-

fore the amount of stress and stress-related defects decreases [139]. This could indicate that

the silicon dioxide layer does not undergo significant structural changes when annealed at

800°C and that silicon atoms that form bonds with only one oxygen atom are more closely tied

to the atomic structure of the base silicon substrate and are, therefore, not affected by the

annealing step. In turn, this could mean that the non-existing passivation provided by wet-

chemical ozone-based oxides after oxidation is the result of stress formed during the initial

growth of the oxide for Si3+ bonds. It is therefore assumed that during the annealing step the

continued growth of the oxide layer may result in an order for the interface or specifically the

Si3+ state that is more akin to the ones found for thermally grown oxides, although the growth

of the oxide may not be necessary. As the use of other wet-chemically grown silicon oxide

layers e.g. grown in HCl or RCA solutions still result in poor passivation qualities even after

annealing in passivating contacts, the properties of the initially ozone-based silicon oxide layer

may enable this behaviour such as the stoichiometric bulk silicon dioxide or the exact shape

of the transitional layer after the initial growth.

Further changes in FWHM can be observed for samples annealed at 900°C. Both Si3+ and Si4+

states seem to broaden when annealing samples at 900°C rather than 800°C The increase in

FHWM could be the result of degradation or pinhole formation. For passivating contacts using

wet-chemical silicon oxide layers degradation in the passivation quality has been observed for

temperatures around 900°C and above related to the concentrations of pinholes formed, which

increases with temperature. 900°C with an annealing duration of 30 min may therefore not be

optimal annealing conditions for the oxide. However, the formation of passivating contacts al-

ways presents an optimisation process, where multiple properties such as passivation pro-

vided by the silicon oxide and doping concentration of the poly-Si layer need to be considered.

Therefore, the increase in FWHM at 900°C may not significantly limit the passivation quality of

the passivating contact as shown in the following chapters.

In the next step area fractions of Si1+, Si2+ and Si3+ states from the transitional region were

compared to see if the composition of the transitional sub-oxide layer changes upon annealing.

While some deviation can be observed in the composition of the interfacial layer, average area

fractions are similar across all sample groups without a clear trend. The spread in area fraction

may therefore be considered an indicator for the accuracy of the fits and model used, however,

no similar set of data could be found in literature to compare this data to. It is important to note

that the shown area fraction does not directly reflect the proportions of Si1+, Si2+ and Si3+ bonds

present within the interfacial layer. At low excitation energies, photoionization cross sections

are different for each oxidation state. For excitation energies of ~400 eV and above [125] the

cross section of each oxidation state is the same and the composition of the interfacial region

can be easily determined. Using the data for samples that was obtained at an excitation energy

of 370 eV, Si1+ area fractions were once again calculated as shown in Figure 4.8b. Similar to

the measurements conducted at 170 eV no clear trend can be observed. However, the spread

in data is higher as the impact of the sub-oxide states on the trend of each curve is not as

significant at 370 eV. On average the transitional region between oxide and silicon wafer con-

sists of ~46% Si1+ bonds, ~1.5% Si2+ bonds and ~53% Si3+. By assuming a capture cross

section ratio of 0.725 between Si1+ and Si3+ at 170 eV according to [128], a similar composition

with area fractions of ~45%, ~4% and 52% respectively can be calculated. Overall, the com-

position of the transitional area does not change significantly during the annealing process.

Figure 4.8 Si1+ Sub-oxide fractions within the transitional layer of samples measured with an excitation

energy of 170 eV (left) and 370 eV (right).

Lastly, an attempt was made to quantify the change in thickness of the transitional area through

annealing according to the method described in chapter 4.3. The overall results of these cal-

culations were rather mixed. While some sample such as the oxide grown with 20 ppm ozone

show a clear decrease in oxide thickness with annealing temperature for both Si1+ and Si3+

bonds, other samples are not as consistent. On average the annealing step seems to cause a

decrease in sub-oxide thickness, however, a significant portion of these values also indicate a

growth in sub-oxide thickness. Similarly, there does not seem to be any consistency related to

the initial growth conditions. Therefore, no conclusion can be drawn regarding the impact of

annealing conditions on the thickness of the transitional area between silicon and silicon diox-

ide. However, this also means that it is not clear what enables the further growth of the silicon

dioxide layer during the annealing process. A decrease in sub-oxide thickness would have

implied that bonds are reformed to form silicon dioxide. As there is no clear trend for the change

in sub-oxide thickness one could consider other reasons for the growth of the silicon dioxide

layer or rather the origin of excess oxygen necessary for the growth. According to Polzin et al.

[138] the growth of the silicon oxide during annealing is enabled by interstitial oxygen at the

Si/SiOx interface. Pinhole formation as well the reformation of sub-oxides were considered as

oxygen source but deemed insufficient. SIMS measurements shown in chapter 5.1 show an

oxygen signal throughout the poly-Si layer, which could have contributed to the growth. How-

ever, it is no clear if this oxygen was present at the same or similar magnitude during the

annealing process or if it diffused into the sample during measurements performed at ambient

conditions after the samples were completed. ´

Figure 4.9 Relative change in equivalent thickness of Si1+ and Si3+ sub-oxides after annealing at 800°C

and 900°C respectively analysed using the approach of M. P. Seah and S.J. Spencer [137].

4.6 Summary of chapter

Chapter 4 discussed changes in the structure and composition of wet-chemical ozone oxides

that undergo annealing within a passivating contact. Through XPS measurements it was

possible to identify some factors related to the behaviour that wet-chemical ozone-based

oxides present within passivating contacts. Based on these measurements it is concluded that

significant improvements in the surface passivation provided by ozone-based oxides after

annealing (see chapter 6) are related to a reduction of stress related defects at the

silicon/silicon oxide interface connected mostly to the reordering of Si3+ bonds. In contrast, on

the level of changes detectable with XPS, the composition of the interface does not appear to

be affected by the high-temperature annealing step with Si1+ and Si3+ bonds of annealed and

non-annealed samples appear approximately in a 1:1 ratio as suggested by Himpsel and

McFeely et al. [125] for oxides grown on <111> silicon crystals. It was not possible to obtain a

conclusive result regarding the impact that annealing has on the thickness of the transitional

area. The thickness of the silicon dioxide layer was shown to increase through annealing,

although the origin of oxygen atoms required for the growth could not be determined. Overall,

it can be said that both the Si3+ and Si4+ oxidation states reflect properties of the oxide that are

of importance for the passivation provided by the passivating contact. The presence of Si3+

states can be used as a measure for the structural disorder of the oxide, reflecting the formation

of defects at the interface, while the Si4+ oxidation states ori inate from the oxide’s b lk and

change with the thickness of the oxide, influencing the selectivity of the contact as discussed

in chapter 2.4 through a change in tunnelling probability. Si3+ states are, therefore, assumed

to be linked with the defect density at the c-Si interface affecting the surface recombination

rate.

5 Plasma doping for passivating contacts

The following chapters discuss results of SIMS and ECV measurements conducted at the For-

schungszentrum Jülich. The goal of these experiments was to determine the fundamental prin-

ciples of the plasma doping process and to form a basis for discussions related to the pas-

sivation provided by passivating contacts.

Experiments were performed on <100> n-type CZ silicon wafers with a resistivity of 1-5 Ωcm.

Saw damage was removed from the surface of wafers using KOH solutions at 80°C for 10 min.

Passivating contacts were realised using the same process steps described in chapter 4. Sam-

ples used silicon oxide films that were exclusively grown in solutions containing 30 ppm at a

temperature of 20°C for 30 min. Plasma doping treatments were varied by changing the pres-

sure from 0.3 to 0.7 mbar. Similarly, the annealing step was performed using different peak

temperatures varying from 860°C to 915°C. Samples discussed in chapter 5 were also char-

acterised using PCD measurements and are part of the samples discussed in chapter 6.

5.1 SIMS profiles for passivating contacts

Figure 5.1 shows the results of SIMS measurements performed on one poly-Si contact,

including silicon, oxygen, phosphorous, hydrogen and carbon. Similar signal intensities can be

observed for both the bulk poly-Si layer and the crystalline substrate to the point that the

interface only becomes visible in the silicon signal through a small bump, which correlates with

the maximum of the oxide signal. It is interesting to note that an oxygen signal is present

throughout the entire poly-Si layer most-likely originating from interstitial oxygen atoms. The

peak related to the ultra-thin ozone-based silicon oxide is fairly broad especially considering

the thickness of the silicon oxide layer of ~1.5 nm and the thickness of the film with ~320 nm.

The width of the peak is influenced by the shift from sputtering poly-Si to SiOx. Different

sputtering rates between the poly-Si and SiOx layer as well as the non-homogeneous removal

of atoms from the investigated area during depth-resolved SIMS measurements cause signals

from separate layers to merge at the interface, which is further enhanced by the topography of

the saw-damage etched CZ silicon substrate, especially in Figure 5.1. It is, therefore, not

possible to extract the thickness of the oxide layer or comment on the diffusion of silicon oxide

at the interfacial region using this approach. However, the position of the oxide layer is most-

likely close to the peak in oxygen intensity, which correlates with peaks in the silicon and

phosphorous signal. These peaks do not necessarily indicate an increase in silicon or

phosphorous concentrations but are related to the different structures and sputtering rates of

silicon oxide and silicon. The intensity of the hydrogen signal, while present, is low, limited by

the detection limit for hydrogen.

Figure 5.1 SIMS profile for a sample with passivating contact that was processed at a pressure of 0.5

mbar and an annealing temperature of 900°C (a). Signals for Si., O, P, C and H are shown. The broad-

ening in the O, P and C curves at about 1400 s is assumed to be related to the surface morphology of

the sample at the measured spot. SIMS profile for a sample with passivating contact that was processed

at a pressure of 0.7 mbar and an annealing temperature of 860°C.

Figure 5.2 Schematic showing the impact of substrate morphology on the result of SIMS measurements.

For planar surfaces (a) the width of the measured signal is representative of the thickness of the ana-

lysed layer. For different surface morphologies, where layers are not aligned perpendicular to the crater

formed during the measurement, signal widths of especially thin layers are larger compared to the thick-

ness of the layer (b).

Issues occurred for some samples relating to the morphology of CZ substrates used. The sam-

ple shown in Figure 5.1 shows a broad shoulder in the oxygen profile at around 1500 s, which

did not occur for most samples and seems to also affect the trend of the phosphorous signal.

It is assumed that the qualitatively different signal trends for oxygen are caused by the irregular

surface of the saw-damage etched samples also referred to as crater-edge effect [140]. A

simplified version of these surfaces is shown in Figure 5.2, where the surface is described as

alternating plateaus and valleys. The plateaus, of course, show some roughness (approx. 29

nm based on atomic force microscopy performed on a 5x5 µm2 area), however, the more se-

vere issue is the transition from valley to plateau. For measurements that were performed close

to the edge, signals from multiple layers could overlap for a longer period of time compared to

measurements performed on plateaus. As the surface-related broadening effects did not ap-

pear consistently across all samples, quantitative as well as qualitive discussions of signal

trends are not possible. However, bulk measurements for the poly-Si layer should not be af-

fected. Discussions regarding the distribution of phosphorous at the interface as well as indif-

fussion are, therefore, limited to ECV data.

5.2 Initial diffusion of phosphorous

SIMS measurements were performed after the plasma doping step, where phosphorous dif-

fuses into the sample, as shown in Figure 5.3, to investigate both the initial diffusion of phos-

phorous and the diffusion of phosphorous during the high-temperature annealing step. Pres-

sures during the plasma doping step were also varied, which affect the total number of phos-

phorous atoms present within the sample. During the initial plasma doping step, phosphorous

mostly diffuses into a volume close to the surface of the sample, which is estimated to be ~30

nm thick based on thickness derived from profilometry measurements and the sputter time

necessary for the removal of the poly-Si layer. For the sample where the plasma doping step

was performed at 0.3 mbar a monotone decrease in phosphorous concentration can be ob-

served. By further increasing the pressure, phosphorous concentration of ~1022 cm-3 were de-

tected. For the 0.5 mbar and the 0.7 mbar sample a plateau is formed close to the surface of

the sample where the doping concentration is almost constant especially for the 0.7 mbar sam-

ple. As the difference in phosphorous concentrations between both plateaus does not differ

too much with 0.9x1022 cm-3 and 1.1 x1022 cm-3 for the 0.5 mbar and 0.7 mbar sample respec-

tively, max. concentrations of fully saturated samples may not differ too much especially con-

sidering the temperature-dependant solid solubility limit of phosphorous in silicon for thermal

diffusion processes, which can reach concentrations above 1021 cm-3 [141].

While the majority of phosphorous atoms are located close to the surface, phosphorous was

detected throughout the bulk of the poly-Si material albeit with a rather low signal intensity

close to the detection limit of the tool especially for the sample where the plasma doping step

was performed at 0.3 mbar. Fluctuations in the signal intensity between ~1017-1018 cm-3 and 0

show that the concentration of phosphorous (if present) is too low to be accurately measured

in the bulk of the a-Si layer and the substrate, where the base doping concentration of ~1015

cm-3 is below the detection limit. Although it should be mentioned that the behaviour of the

phosphorous signal in the bulk of the poly-Si layer was similar to the behaviour in the n-type

c-Si substrate. While it is not possible to accurately determine the concentration of phospho-

rous throughout the entire poly-Si layer, it is still assumed that a low concentration of phospho-

rous diffuses deep into the sample.

Figure 5.3 Phosphorous concentrations calculated from SIMS data for samples that received the

plasma doping treatment at different pressures without annealing. A silicon reference with known

phosphorous concentration was used as a basis for these calculations.

All measured surface concentrations exceed the solubility of phosphorous in silicon nP(Si),

which was calculated for a temperature of 450°C according to eq. 40 [141], which is based on

empirical data:

𝑛P(Si)=9.2x1021𝑒𝑥𝑝(−0.33𝑒𝑉

𝑘𝑇 )

(40)

For a process temperature of 450°C a limit of 4.61x1019 cm-3 was calculated. Assuming that

the limit regarding solubility is reached for these non-annealed samples, the majority of phos-

phorous atoms close to the surface would still be inactive. Surface concentrations similar to

the those found here have been reported previously for POCl3 doping [70], [111], [21].

5.3 Diffusion and activation of phosphorous during the

high-temperature annealing step

During the annealing step phosphorous diffuses deeper into the sample as shown in Figure

5.4 and Figure 5.5, where the SIMS profiles of samples are shown that were annealed after

receiving the plasma doping step. Compared to samples, which were not annealed, a high

phosphorous concentration of approximately 1-3x1020 cm-3 can be seen throughout the entire

poly-Si layer with a steep drop at the interfacial region. The difference in peak position between

annealed samples results from the difference in poly-Si layer thickness. Samples annealed at

higher temperatures show an even distribution of phosphorous throughout the bulk of the poly-

Si layer despite an overall higher poly-Si thickness. For the samples annealed at 860°C and

885°C more phosphorous remains close to the surface of the sample with a more noticeable

slope in the phosphorous profile, which is also present in the ECV profile for the same sample

shown in Figure 5.6, although the distribution for the 885°C sample is more even.

Figure 5.4 Phosphorous concentrations calculated from SIMS data for samples that received the

plasma doping treatment at different pressures followed by annealing at 860°C for 30 min. A silicon

reference with known phosphorous concentration was used as a basis for these calculations.

The introduction of larger quantities of phosphorous during the plasma doping step seems to

mostly affect the average concentration of phosphorous found within the sample as the sam-

ples annealed at 860°C present similar slopes regardless of pressure used, most-likely the

result of a temperature-limited diffusion process. Looking at the tail end of the phosphorous

profiles leading into the c-Si substrate, one could assume a more or less temperature inde-

pendent indiffusion of phosphorous through the SiOx layer as the 860°C samples show a far

more significant indiffusion compared to the 885°C sample. As previously discussed, concen-

trations determined close to the silicon oxide interface are severely affected by the exact shape

of the s bstrate’s s rface. D rin E meas rements are erformed o er a si nificantly

larger surface area with a lower depth resolution compared to SIMS meaning that position

de endant chan es in the s bstrate’s to o ra hy do not si nificantly im act the meas red

doping profiles.

Figure 5.5 Phosphorous concentrations calculated from SIMS data for samples that received the

plasma doping treatment followed by annealing at varying temperatures. The first 200 s are shown at

higher resolution in the top left corner A silicon reference with known phosphorous concentration was

used as a basis for these calculations.

The previously discussed temperature and pressure dependant distribution of phosphorous in

the poly-Si layer can also be observed in the ECV profile shown in Figure 5.6. Once again

samples annealed at 860°C show a visible slope in the poly-Si layer meaning that the limited

diffusion of phosphorous affects both the total and active phosphorous concentration within

the poly-Si layer. The magnitude of the slope also exceeds the error in the measurement of

active phosphorous in the poly-Si layer, which is less than 1%. Comparing the measured

concentrations of phosphorous from both SIMS (Figure 5.4 and Figure 5.5) and ECV (Figure

5.6), one can see that most of the phosphorous becomes electrically active during the

annealing process for the samples annealed at 860°C and 885°C. For higher temperatures the

difference appears more significant despite the fact that the limit for activation of phosphorous

of ~3.5x1020 cm-3 has not been reached.

Figure 5.6 Results of ECV measurements conducted on the same samples used for SIMS and PCD

measurements. Both pressure during the plasma doping treatment and temperature were varied here.

Figure 5.7 Zoom into the data shown in Figure 5.6 showing the indiffusion of phosphorous. Data was

offset horizontally to better compare each sample.

The degree of indiffusion appears to be mostly related to the annealing temperature used as

indicated by Figure 5.7, which shows a smaller section of the data shown in Figure 5.6. Data

points were moved horizontally so that data points close to the end of the poly-Si layer / the

beginning of the c-Si substrate would appear at 0 µm. While the exact position of the silicon

oxide layer could not be derived from the limited resolution of ECV data provided, indiffusion

appears to occur more strongly at higher temperatures resulting in not only an increase in the

average phosphorous concentration close to the surface of the c-Si substrate but also an in-

crease in the depth of the induffused area ranging from ~50 nm to ~200 nm. Some outliers

exist like the samples processed at 0.5 mbar that were annealed at 885°C and 900°C, which

show very similar trends as well as the samples annealed at 860°C processed at 0.3 and 0.5

mbar, although some deviations have to be expected as the depth resolution of ECV meas-

urements shown was not good enough to gain detailed information about the interfacial region.

The impact that these annealing conditions may have on the passivation quality of passivating

contacts created using the same process conditions is discussed in chapter 6.2.

5.4 Comparison of plasma doping to other doping

techniques for passivating contacts

Figure 5.8 and Figure 5.9 include the results of SIMS and ECV measurements performed by

other groups investigating passivation contacts using ion implantation [142], [143] and POCl3

to provide doping for the poly-Si layer [144], [145]. SIMS data shows that each method is able

to produce homogeneous distributions of phosphorous within the poly-Si layer, although the

plasma doping profile appear less homogeneous by comparison. Doping concentrations re-

ported vary for both methods based on the ion implantation doses and POCl3 gas flow rates

as well as annealing conditions used. Values reported for POCl3 diffusions vary between

8x1019 cm-3 and 5x1020 cm-3- for annealing temperatures of 950°C and layer thicknesses of

~300 nm. Phosphorous distributions resulting from POCl3 diffusion display similar characteris-

tics as the ones achieved via plasma doping for passivating contacts including the elevated

level of hos horo s at the sam le’s s rface despite the different nature of both doping pro-

cesses as the POCl3 diffusion process uses an infinite diffusion source compared to the limited

diffusion source of the plasma doping process. Results reported by U. Römer et al. [142] for

ion implanted samples showed similar profiles as the ones achieved for POCl3 and plasma-

doped samples. The main difference here is a drop in phosphorous concentration at the sam-

le’s s rface. nitial hos horo s rofiles created ia ion im lantation are realised ithin the

sample rather than the surface of the sample with the depth of the profile depending on the

doses and excitation energy used, unlike POCl3 and plasma doping processes, where phos-

phorous diffuses into the sample from the surface.

ECV profiles in Figure 5.8 once again show similar results compared to the plasma doping

process. Interestingly, profiles reported by S. Yuan et al. [144], which used an annealing tem-

perature of 860°C, show a slope in the ECV profile similar to the ones measured for the plasma

doped samples albeit with a more severe decline in dopant concentration. Using the same

annealing temperature but different gas flow rates for the POCl3 diffusion it was still possible

to achieve a homogeneous distribution of phosphorous within the poly-Si layer. From this one

could also infer that homogeneous dopant concentrations are also possible for plasma doped

samples annealed at 860°C if the conditions used for the plasma doping are adjusted e.g. by

further increasing the pressure beyond 0.7 mbar.

Figure 5.8 SIMS data of passivating contacts using different doping processes to realise doping within

the poly-Si layer after the deposition of intrinsic a-Si. Data was taken from [145] and [142].

Overall, both SIMS and ECV profiles have shown similar results as other doping processes, at

least for parameters ranges that were deemed appropriate for passivating contacts. Neither

the type of diffusion source (infinite vs finite) nor the method used to transport phosphorous

into the sample (diffusion vs acceleration) seemed to have a significant impact here. This

shows that the plasma doping process is not only suitable for the doping of passivating con-

tacts but also comparable to other doping processes. Although more significant deviations may

become apparent for higher doping concentrations, which may be limited by the finite nature

of plasma doping and ion implantation, or in cases where the silicon oxide diffusion barrier is

not present e.g. emitter formation. Additionally, comparisons between methods were limited

by data available in literature as annealing conditions would deviate significantly from the pro-

cess ranges investigated within this thesis.

Figure 5.9 ECV data of passivating contacts using different doping processes to realise doping within

the poly-Si layer after the deposition of intrinsic a-Si. Data was taken from [143] and [144].

5.5 Results and discussion

This chapter focused on the plasma doping step used to provide n-type doping for poly-Si

layers used for passivating contacts. Symmetrical lifetime samples with the passivating contact

on both sides were realised so that the same samples investigated here could also be used

for lifetime measurements using PCD. SIMS measurements were performed to determine the

total phosphorous concentration within the sample after the initial plasma doping step and to

see how the profile evolves during the high-temperature annealing step that is performed for

passivating contacts. Measurements were compared to the results of ECV measurements,

which showed electrical active phosphorous concentrations. For samples that just received the

plasma doping step, high phosphorous concentrations of up to ~1x1022 cm-3 were observed

close to the surface of the poly-Si layer for pressures of 0.7 mbar. Based on the solubility of

phosphorous in silicon at 450°C, it is assumed that most if not all of the phosphorous diffused

into the sample during the plasma doping step is electrically inactive. The shape of the phos-

phorous profile and magnitude of phosphorous surface concentration is clearly linked to pres-

sure with higher pressures leading to higher surface concentrations (~1022 cm-3 at 0.7 mbar

and ~1021 cm-3 at 0.3 mbar) and larger numbers of phosphorous atoms within the sample.

Phosphorous was mostly measured within a depth of ~30 nm. During the annealing process

phosphorous diffuses deeper into the poly-Si layer resulting in mostly homogeneous phospho-

rous profiles within the poly-Si layer with concentrations of ~1x1020, although larger phospho-

rous concentrations where still visible at the surface of the sample. For the sample annealed

at 860°C that received the plasma doping step at 0.3 mbar, the distribution of phosphorous

was not quite as homogeneous, as observed through both SIMS and ECV measurements

showing a limited diffusion of phosphorous under those conditions, at least in regards to 300

nm thick poly-Si layers. ECV-measurements showed slightly lower but still similar active phos-

phorous concentrations compared to the total phosphorous concentrations obtained from

SIMS. ECV measurements also showed that the degree of indiffusion can be linked to the

process conditions investigated, with both higher annealing temperatures and higher pres-

sures showing more significant indiffusion. However, the accuracy of values obtained close to

the interfacial region was limited. For SIMS measurements, effects assumed to be related to

the morphology of saw-damage etched silicon wafers have limited the accuracy of information

gained at the interfacial region. While interfaces present a general issue for TOF SIMS meas-

urements, as signals from different layers convolute at the interface, improvements in the qual-

ity of SIMS measurements performed at the interface are assumed to improve by using planar

substrates. However, by using the same samples for SIMS, ECV and PCD measurements a

better basis for comparison is formed.

Comparisons between plasma doping and other doping processes have shown that phospho-

rous distributions resulting from plasma doping can be obtained that are similar to other tech-

niques applied to SiOx/poly-Si passivating contacts. SIMS measurements indicated that phos-

phorous distributions resulting from plasma doping appear not quite as homogenous compared

to POCl3, however, with only minor deviations for temperatures above 885°C. The quality of

the comparison was, however, limited by the variance in process conditions investigated in

literature. Further deviations are expected to occur between plasma doping and especially

POCl3 diffusion when applied to other structures e.g. for the formation of emitters in silicon,

although further investigations are required here.

6 Electron-beam evaporation and wet-

chemical ozone-based passivating

contacts

The following chapters feature the results of investigations carried out to create and evaluate

passivating contacts that were created for this thesis using E-beam evaporation and ozone

oxidation. Sample preparation was previously described in chapter 5, where the same samples

were used to investigate the plasma doping step. Additional samples were created to qualita-

tively discuss the temperature-dependant degradation of the silicon oxide layer. The results of

PCD/QSSPC measurements are compared to the results of experiments previously discussed

to better understand the contributing factors that affect the quality of passivation. Investigations

were also conducted into the consistency of the thickness of a-Si layers deposited using E-

beam evaporation, which influences the doping of the sample as thicker layers present lower

average doping concentrations and therefore lower doping concentrations at the poly-Si/SiOx

interface, which is known to affect the selectivity of the passivating contact [146].

6.1 Determining process conditions for passivating

contacts

Initial experiments performed focused on determining appropriate process conditions for each

process step involved in the creation of passivating contacts. Passivating contacts were

realised on both sides of saw-damage-etched CZ wafers as described in chapter 4 and 5.

Samples were processed using different process conditions and characterised using PCD and

QSSPC. Silicon oxides were grown at 30 ppm for 30 min at room temperature resulting in a

thickness of ~1.5 nm according to ellipsometry and TEM measurements previously discussed

in chapter 4.3. For passivating contacts that are annealed at temperatures of ~900°C or less

an oxide thickness of 1.5 nm has shown optimal passivation properties for n-type passivating

contacts [64]. Consequently, oxide growth conditions were not varied for experiments involving

passivating contacts. Annealing was investigated for temperatures between 860°C and 915°C

resulting in good passivating properties depending on the plasma doping conditions used as

shown in Figure 6.3. Hydrogenation was realised using forming gas containing 5% H2 at 450°C

for 30 min. Neither an increase in annealing time from 30 to 60 min nor a change in annealing

temperature from 450°C to 400°C had a noticeable impact on the passivation quality of

passivating contacts.

Figure 6.1 Average layer thicknesses measured on different glass substrates that received a-Si E-Beam

layers with a set thickness of 300 nm.

Figure 6.2 Impact of the thickness of a-Si layers on the minority carrier lifetime in symmetrical lifetime

samples using the passivating contact of both sides. Plasma doping was realised at 0.5 mbar followed

by an annealing step performed at 900°C. Each group includes 4 samples.

The deposition of a-Si layers via E-Beam evaporation presented more significant issues

related to the layer thickness. As the evaporation process is strongly tied to the condition of

the silicon ingot, which changes with each deposition, deviations in the thickness of layers

deposited at the same set conditions have to be expected. As the doping for the a-Si layer is

provided ex-situ, the thickness of the layer affects the formation of the doping profile including

doping within the bulk of the silicon layer, doping concentration at the silicon oxide interface

and the degree of indiffusion. For larger deviations in the a-Si layer thickness the impact can

be severe as shown in Figure 6.2, where 3 different sample groups are shown, which were

processed using the same process conditions except for the thickness of the silicon layer which

was varied. Using the samples with a thickness of ~330 nm as a point of reference, both a

decrease and increase in layer thickness can cause a significant drop in the passivation quality

of the passivating contact. For lower film thicknesses the issue is assumed to be related to

more significant indiff sion accordin to Frick’s la (eq. 11 and eq. 12), whereas a significant

increase in film thickness reduces the average dopant concentration within the film as well as

the dopant concentration at the silicon oxide interface affecting the selective transport of

charge carriers across the silicon oxide layer. While the examples shown may appear extreme,

consecutive depositions of a-Si layers have shown deviations in the average thickness of

layers as high as +/-10% as shown in Figure 6.1. Additionally, one has to consider the

inhomogeneity of the layers, which are shown in Figure 6.2. While attempts were made to

further increase the reproducibility of the E-Beam process by increasing the amount of

measurements performed during the deposition to determine the deposition rate and adjust

the duration of the process accordingly, significant deviations still remain. However, the impact

of the thickness of the a-Si layer on the lifetime of the passivating contact shown here assumes

identical doping conditions. Deviations in the thickness of the a-Si layer can be compensated

by adjusting the doping conditions accordingly. This can be accomplished e.g. by adjusting the

pressure during the plasma doping step. An increase in pressure increases the number of

phosphorous atoms that diffuse into the sample leading to higher doping concentrations as

previously shown in chapter 5. Experiments were therefore conducted with variations in the

plasma doping conditions not only to investigate the plasma doping step discussed in chapter

5. but to also limit the impact of thickness variations.

6.2 Passivating qualities of passivating contacts

Figure 6.3 shows the result of PCD and QSSPC performed on symmetrical lifetime samples.

QSSPC was only performed for samples annealed at 915°C where the plasma doping step

was performed at 0.7 mbar. A wide spread in iVoc values were obtained reaching from 662 mV

up to 734 mV depending on the process conditions used. Results are comparable to some of

the better iVoc values found in literature for passivating contacts that use wet-chemical ozone-

based oxides [124], [70]. Samples show clear trends according to the process conditions used.

For samples annealed at 860°C a monotone increase in iVoc can be observed with an increase

in pressure. Samples annealed at 885°C and 900°C show similar trends, where an initial in-

crease in pressure from 0.3 mbar to 0.5 mbar led to an increase in iVoc followed by decrease

when further increasing the pressure to 0.7 mbar. For an annealing temperature of 885°C the

initial jump in passivation quality is far more significant compared to the 900°C, whereas the

decrease in passivation provided when going from 0.5 to 0.7 is less severe compared to 900°C

samples. Finally, samples annealed at 915°C show average to low iVoc values with an obvious

degradation in passivation quality when high temperatures are combined with high pressures.

The trends that are observed here are influenced significantly by the annealing step, which in

turn affects the shape of doping profiles as previously shown in chapter 5. The quality of pas-

sivation provided by TOPCon-like passivating contacts has been linked to both the doping

concentration close to the poly-Si/SiOx interface and the degree of indiffusion. Higher doping

concentrations within the poly-Si layer typically led to an increase in selectivity, which in turn

affects the effective lifetime of minority carriers and therefore iVoc. However, for higher doping

concentrations trends are not quite as clear as indiffusion reduces the selectivity of the contact

by reducing the height of the barrier affecting the tunnelling probability of both electrons and

holes as described in eq. 13. While the exact impact of indiffusion can vary, more significant

indiffusion is typically linked to a decrease in passivation provided by the passivating contact.

For samples annealed at 860°C trends observed can be clearly linked to the doping profiles

shown in Figure 5.6 where the doping concentration at the poly-Si/SiO2 interface increases

from ~ 2 to 7x1019cm-3 with the doping pressure used. For samples annealed at 900°C a similar

increase can be observed, however, less significant. While the width of the diffused dopant

profile close to the surface of the substrate appears to be similar to samples annealed at

860°C, doping concentrations are generally higher. Additionally, for the 900°C sample which

received the plasma doping treatment at a pressure of 0.7 mbar the depth of the indiffused

profile seems to occur over a larger depth compared to other samples within the same group.

Finally, Samples annealed at 915°C shown both the highest doping concentration but also the

most significant indiffusion. The drop in iVoc at 0.7 mbar visible for all samples except the ones

annealed at 860°C is, therefore, strongly linked to indiffusion that takes places during the pro-

cess. The impact of indiffusion can be seen more clearly in Figure 6.4 where iVoc values are

compared to band bending between the poly-Si layer close to the poly-Si/SiO2 interface and

the bulk of the base substrate. At 860°C, an almost linear increase between iVoc and band

bending can be observed with the 885°C sample and the 900°C samples processed at 0.3 and

0.5 mbar following that trend. Only the 900°C 0.7 mbar and the 915°C samples shown deviate

from this trend, where, as previously mentioned, more significant indiffusion was observable.

Besides the doping of the poly-Si layer the chemical passivating provided by the oxide needs

to be considered.

Figure 6.3 iVoc values of symmetrical lifetime samples using passivating contacts. Both pressure during

the plasma doping step and annealing temperature were varied. Values were calculated assuming an

optical constant of a) 0.55 and b) 0.95 to reflect the behaviour of the symmetrical lifetime samples and

expected behaviour of cells respectively.

Figure 6.4 Impact of band bending between (n) c-Si substrate and poly-Si on the iVoc of symmetrical

lifetime samples using passivating contacts. For most samples an almost linear relation is depicted.

Deviations from this behaviour are found for higher annealing temperatures and are expected to be

related to indiffusion.

Attempts were made at determining the minimum defect density at the c-Si interface through

which it is possible to quantify chemical passivation provided by the oxide. A simulation tool

previously developed by C. Leendertz [147], which was created to analyse lifetime data for

silicon heterojunction solar cells was considered as most of the assumptions made for the

model are also valid for c-Si/SiOx interfaces. However, results from these simulations did not

present clear results as the interface defect density and density of states were strongly linked

resulting in inconsistent values. Surface photovoltage (SPV) measurements performed on ox-

idised wafers showed low reproducibility with only comparisons between annealed and non-

annealed samples showing clear trends with annealed samples showing a minimum interface

defect density of ~1012 cm-2eV-1 compared to ~1013 cm-2eV-1 for non-annealed samples. In-

stead, the impact of annealing conditions on the silicon oxide and possibly the chemical pas-

sivation provided was investigated qualitatively by estimating the number of pinholes formed

during the annealing process according to a process used by [148]. TMAH is used to etch

through the pinhole formed during the annealing process into the wafer leading to the formation

of etch pits, which become visible using optical microscopy. Samples were prepared in the

same way as lifetime samples (excluding KOH etching), except that the passivating contact

was only realised on one side using 0.5 mbar for the plasma doping step on polished <100>

1-5 Ωcm n-type FZ wafers with a thickness of 280 µm. Poly-Si was removed using the same

TMAH etching step as the one used for XPS samples except that the step was performed for

5 min. Pinhole concentrations were estimated based on five measurements performed for each

sample p on randomly chosen areas.

Figure 6.5 Optical microscopy images of pinholes enhanced via TMAH etching for a sample a) an-

nealed at 860°C and b) 915°C. Field of view: 1237x928 µm2.

As shown in Figure 6.5, pinholes appear clearly after etching in TMAH. For the 915°C sample

a larger number of pinholes appear compared to the sample annealed at 860°C. Pinhole con-

centrations were estimated to be ~1000 per cm2 for the 860°C sample, ~1300 per cm2 for the

885°C, ~1700 per cm2 for the 900°C and ~2000 per cm2 for the 915°C. While a general in-

crease in pinhole concentration was observed related to an increase in annealing temperature

for all samples investigated, estimated pinhole concentrations were significantly lower than

reported pinhole concentrations that were connected with the degradation or even breakdown

of the oxide layer [61]. While it was possible to see a difference in the concentration of pinholes,

it is assumed that the behaviour of iVoc is more strongly connected with the doping profile as

previously discussed.

6.3 Summary

Chapter 6 focuses on the application of E-beam evaporation, plasma doping and ozone oxi-

dation for the formation of passivating contacts. Limitations in the reproducibility of the thick-

ness of E-Beam layers were discussed as a point of consideration for the samples processed

for this thesis and the approach used. Using these techniques, it was possible to realise well-

passivating structures enabling high iVoc values on symmetrical lifetime samples of up to 734

mV, comparable to values found in literature published by groups using wet-chemical ozone

oxides. A strong dependence of doping conditions and passivation quality was shown. Com-

parisons between data discussed in chapter 5 and PCD / QSSPC measurements showed an

almost linear trend between iVoc and band-bending between c-Si base and poly-Si layer. Only

samples processed at high temperatures thus with significant indiffusion deviated strongly from

this trend. The degradation of the oxide layer was also considered especially for the samples

annealed at 915°C. However, determined pinholes concentrations are not considered signifi-

cant enough to noticeably impact the iVoc of samples investigated. While other approaches

were considered to quantify the degradation of the silicon oxide layer, however, further inves-

tigations are required to determine an appropriate method.

7 Integration of passivating contacts into

solar cells

In the next step passivating contacts were realised on the rear side of solar cells and tested.

A p-type silicon heterojunction structure was used for the front side. This approach is unusual

as most TOPCon-like solar cells use boron-based diffusion techniques to create a p-type area

for n-type silicon substrates with aluminium oxide and silicon nitride layers or layers with similar

properties on the front to provide surface passivation and anti-reflective properties. However,

this would be out of the scope of this thesis. Silicon heterojunction solar cells have been de-

veloped and optimised at the HZB and were therefore deemed the best approach to test the

passivating contacts investigated in this thesis. GenPro4 simulations were also performed in

combination with TLM measurements to determine and evaluate alternative contacting

schemes for the poly-Si layer as the presence of Ti, discussed below, caused significant par-

asitic absorption. Finally, the potential of the investigated device structure was investigated by

performing simulations based on the 1-diode and 2-diode model.

Figure 7.1 Schematic of solar cells investigated and process scheme used. The front side uses layers

and processes that have been previously established. The rear side features the passivating contact

together with a titanium and silver contact.

7.1 Solar cell structure and processing scheme

First, <100> n-type CZ silicon 6’’ afers ith a thickness of 150 µm and a resisti ity of 1-5 Ωcm

were etched in KOH at 80°C for 10 min to remove saw damage (~7.5 µm per side) followed by

a O3 cleaning process at 20°C using a solution that contained 20 ppm ozone and 20 ml HCl

for 2 min. A 200 nm silicon oxide capping layer was deposited on the rear side of wafers using

PECVD so that only the front side of wafers would be textured. Texturing was performed at

80°C using KOH together with an alcohol mixture followed by the same O3 cleaning procedure.

The capping layer was then removed using 5% HF. A Keyence laser was used to divide each

wafer into 6 5x5 cm2 samples. Samples were then cleaned using RCA. The growth of the

silicon oxide layer used for passivating contact was performed in a solution containing 30 ppm

ozone and 20 ml HCl at 20°C for 30 min. After depositing ~300 nm a-Si on the rear side using

E-beam evaporation and performing the plasma doping step at 0.5 mbar, a 300 nm thick silicon

oxide layer was once again deposited on top of the textured side of the wafer. This was done

to protect the front side during the high-temperature annealing step, which was performed at

900°C for 30 min. After removing the oxide cap in 5% HF, samples were once again cleaned

using RCA. Intrinsic a-Si:H and p-type a-Si:H layers were then deposited on the front using

PECVD as described in chapter 3.1.3. The rear contact was then realised using thermal evap-

oration after remo in nati e silicon oxide from the sam le’s s rface in 1% HF. 20 nm i and

1 µm Ag were evaporated on the rear of the cell. This method was chosen as the contact

formation between metal and the passivating contact had not been investigated yet and ther-

mally evaporated Ti/Ag contacts were considered the safest approach based on past experi-

ments, where low contact resistivities were achieved using this contact for SHJ solar cells.

Figure 7.2 Flow chart of processes used for the fabrication of cells with passivating contact.

An indium-oxide-based TCO (newScot) was deposited on the front side via sputtering through

masks, which defined a 2x2 cm2 cell area in the centre of the sample. The silver grid on the

front of the cell was realised via screen-printing and drying on a hotplate at 210°C for 10 min.

As a final step light-soaking was performed to further improve the performance of samples.

Besides the cells shown in Figure 7.1. reference samples were also processed in parallel in-

cluding symmetrical lifetime samples with the passivating contact on both sides, lifetime sam-

ples with cell-like passivating structures (front p-type silicon heterojunction, rear passivating

contact) and a silicon heterojunction solar cell as a reference process. Lifetime samples were

investigated via PCD and PL measurements. For the cells, PL, JV and EQE measurements

were performed. Attributes of solar cells processed were also compared to simulations per-

formed in GenPro4 to further analyse the results of EQE measurements.

7.2 Results of measurements

The results of JV measurements are shown in Figure 7.4. Three solar cells with the design

shown in Figure 7.4 are compared to a reference silicon heterojunction solar cell, where the p-

type a-Si:H layer was realised on the front of the cell and the nc-Si layer on the rear so that the

reference would be comparable with the investigated solar cell structure. The highest efficiency

reached for the cells with poly-Si passivating contact was 19.8%, with the other two cells reach-

ing 17.7% and 19.6% respectively. All samples with passivating contacts shared very similar

Jsc values of ~35 mA/cm2. Fluctuations in the Voc and FF were more apparent with values

reaching 696-711 mV and from 70-78% respectively, although one of the three cells performed

significantly worse. However, more significant differences can be seen when comparing cells

to the silicon heterojunction reference. The Voc, in particular, is significantly higher with 737

mV, which in combination with a Jsc of 37.6 mA/cm2 and FF of 79% leads to an efficiency of

21.9%. The difference in Voc seems unusual as i/n a-Si:H lifetime samples shown in Figure 7.5,

which were realised on the same substrate, reached similar lifetime values as those reached

by the passivating contacts discussed in previous chapters. It was, therefore, assumed that

the front surface passivation was affected by the processing scheme used for solar cells with

passivating contact. Noticeably, the evaporation step as well as the high-temperature anneal-

ing step are exclusive to the cells with passivating contact. During the evaporation of titanium

and silver the front side has to be in close contact to the sample holder so that one could

assume that the surface might get damaged when placing and removing samples. PL meas-

urements of the cell with the highest efficiency shown in Figure 7.3. did, however, not show

scratches or similar str ct res that o ld indicate dama e to the sam le’s s rface. h s the

impact of the annealing step was investigated.

Figure 7.3 PL images for the best processed cell before (left) and after (right) edge isolation. Colour

scale for both images is identical.

Figure 7.5 shows the lifetimes of symmetrical p/i passivated substrates that underwent the

same process conditions before a-Si:H deposition as the cells and cleaned samples that were

just etched in 1% HF and passivated. Annealed samples clearly show a significant reduction

in lifetime compared to the non-annealed samples. This would imply that the cause for the drop

in lifetime are contaminants that diffuse into the sample during the annealing step. This in turn

would indicate that the 300 nm silicon oxide cap is insufficient as a capping layer and that the

RCA cleaning step is not able to completely remove contaminants. While the capping layer is

rather thick, surface concentrations as low as 1010 atoms per cm2 can lead to lifetime degra-

dation of similar magnitude, at least for some elements [100]. Additionally, groups have re-

ported the formation of micro-cracks [149] during the annealing of PECVD-based silicon oxide

layers alon hich contaminants can more easily mo e to ards the s bstrate’s s rface. Re-

garding the RCA step one has to consider that the cleaning procedure only affects the surface

and the volume oxidised during the process, which is limited to a couple nanometres. Contam-

inants close to the surface but not within reach of the RCA procedure can, therefore, still affect

the sample as long as the concentration is high enough. While the presence of contaminants

would explain the behaviour that annealed samples display, it is not clear what kind of con-

taminants are present and where they originate from, except for particles resulting from friction

between tube and sample holder when placing and removing samples from the tube furnace.

Figure 7.4 Trends of JV curves for cells with passivating contact and a silicon heterojunction reference

cell. The figure also includes a JV curve derived from Suns-Voc measurements [150] for one of the cells

with passivating contact.

Figure 7.5 Impact of annealing wafers with a silicon oxide cap before cleaning and deposition of i/p

silicon heterojunction structures. Lifetimes are compared to results achieved for symmetrical i/p

structures after wafer annealing (black), without wafer annealing (red) and for a sample with poly-Si

passivating contact on both sides (green). Lifetime samples of cell-like structures are also shown (blue),

which have the same structure as cells shown in Figure 7.1 excluding the TCO and metal contacts.

From the PL image shown in Figure 7.3 it was also assumed that shunting may have occurred

as the PL intensity at the edge of the wafer, where some of the TCO is located, is rather low.

Approx. 0.1 nm was removed from each edge of the cell via laser cutting before remeasuring

the sample. A slight increase in PL intensity can be seen after edge isolation was performed.

The increase is, of course, minor as the shunt at the side of the cell is essentially replaced with

a defect rich surface as the passivation was also removed there. While this aspect would have

to be considered for large scale solar cells, the defined cell area in the centre is far enough

away from the edge so that it is unlikely that it is affected.

Besides the higher open-circuit voltage, the reference cell also shows a higher short-circuit

current density. EQE measurements show that the difference in Jsc mostly results from losses

in the IR part of the spectrum between ~950 nm and 1200 nm. Differences between reference

and cell with passivating contact in this wavelength range are connected to the different

structures on the rear side of both cells. To quantify the impact of those layers, simulations

were performed in GenPro4. Results of simulations are first compared to EQE data, which was

measured on the non-metallised part of the cell.

Overall, the model is able to replicate the trend of the EQE data with only some minor

deviations. Deviations are expected as the thicknesses of a-Si:H layers were not measured

but assumed to be close to values that were previously obtained. Additionally, EQE

measurements performed on the part of the cell without grid fingers showed a worse signal to

noise ratio compared to other measurements. Simulations indicate that approx. 0.7 mA/cm2

and 4.6 mA/cm2 are parasitically absorbed in the poly-Si layer and the Ti layer respectively.

This shows clearly that Ti is the limiting factor regarding Jsc. Therefore, other contacting

schemes were tested to find one that is more suitable.

Figure 7.6 Comparison of measured EQE curve without the influence of gridlines and EQE data

extracted from GenPro4 measurements for solar cells with a passivating contact and Ti/Ag metal contact

on the rear.

Figure 7.7 Optical simulations of solar cells with passivating contact and Ti/Ag rear contact performed

in Genpro4. Ti on the rear of the cell causes significant parasitic absorption limiting the magnitude of the

current that can be generated by the c-Si substrate.

Figure 7.8 Optical simulations of solar cells with passivating contact and newScot/Ag rear contact

performed in Genpro4. The replacement of Ti with newScot and other TCO leads to a significant

reduction in parasitic absorption on the rear of the device.

Tab. 8 Current loss analysis for different rear contacting schemes for solar cells shown in generated

with GenPro4.

Rear contact scheme

Ti/Ag

AZO/Ag

newScot/Ag

Reflection in mA/cm2

1.66

2.62

2.45

2.58

Parasitic absorption

(front) in mA/cm2

3.41

3.66

3.60

3.65

Parasitic absorption

(rear) in mA/cm2

5.08

1.91

2.36

2.58

Absorption in mA/cm2

36.31

38.26

38.05

38.29

7.3 Investigation of different contacting schemes for

passivating contacts

As the titanium layer clearly limits the properties of solar cells processed, other contacting

schemes were investigated via GenPro4 simulations and TLM measurements. Tab. 8 com-

pares simulated parasitic losses and absorbed current densities of cells with Ti to cells that

use just Ag, or a TCO (AZO or newScot) with Ag. All of the investigated contacting schemes

show rather similar results regarding currents generated by the substrate with the variants

using AZO and no TCO being slightly higher with ~38.3 mA/cm2. Parasitic absorption caused

by the poly-Si layer does increase once the Ti layer is removed as part of the reflected spec-

trum, which was previously absorbed by the Ti layer, is now absorbed by the poly-Si layer

leading to an increase in parasitic absorption caused by the poly-Si layer of ~1.2 mA/cm2.

As all investigated contacting schemes showed similar photogenerated current densities, it

was attempted to investigate all variants.

TLM samples were prepared on the same substrate and using the same process conditions

as the ones used for solar cells. However, the ultra-thin silicon oxide layer was replaced with

a 100 nm thick silicon oxide layer (PECVD) so that no current would flow through the substrate

and in turn better define the current path for TLM structures. First TLM structures were realised

for the newScot/Ag contact using photolithography based on the work of P. Wager [79] as

discussed in chapter 3.2.5. Using the exact method as proposed by P. Wagner, a contact

resistivity of ~20 mΩcm2 was determined, which could be further improved via annealing per-

formed at 200°C for 10 min to ~3 mΩcm2. These values are considered good not only for the

poly-Si contact [151], [152], but also for solar cells in general as contact resistivities below 5

mΩcm2 contribute little towards the series resistance of the device. What also has to be con-

sidered is that the contact resistivity shown is the sum of the newScot/Ag and poly-Si/newScot

contacts expected to be even lower than the values shown. Similar values were also achieved

for the alternative contacting scheme using thicker contact stripes with ~27 mΩcm2 before and

~3 mΩcm2 after annealing. However, measurements had to be limited to 4 contact stripes here

as the alternative processing scheme did not completely prevent the delamination of the con-

tact stripes during etching leading to the complete or partial removes of stripes. While it was

attempted to perform measurements on structures with partially removed stripes, the meas-

ured resistance deviated from the linear behaviour of the distance vs. resistance curve. Devi-

ations can be explained through the change in contact geometry according to eq. 26 especially

as resistance measured was still ohmic in nature. While the measurements for the second

sample are assumed to be less accurate, values especially after annealing do not differ

strongly from the previously measured value.

For AZO/Ag and pure Ag contacts TLM could not be performed. AZO was shown to be less

resistant against the etching solutions used for the photolithography step leading to the total

removal of the AZO layer. While it was possible to create TLM structures using Ag, measure-

ments indicated the formation of Schottky contacts most-likely caused by the formation of an

oxide layer on top of the poly-Si layer before the deposition of the Ag layer. While the oxide

was initially removed using HF, the time between HF-etch and placement of samples may have

been significant enough.

7.4 Efficiency potential of investigated cell concepts

The results of simulations conducted in GenPro4 and TLM measurements discussed in

chapter 7.2 and 7.3 showed that the current of cells processed could be further improved by

replacing the Ti/Ag contact with a newScot/Ag contact. However, issues relating to the

passivation of the front side of cells could not be addressed, which limits the Voc thus efficiency

potential of cells processed. Therefore, the impact of recombination on device performance

was investigated by modelling JV curves using the 1-diode and 2-diode models described in

chapter 2.1.1 (eq. 4 and eq. 5). Variations in recombination were expressed through the dark

saturation currents j0, j01 and j02. Simulations were performed using the 2/3-Diode Fit program

with the highest efficiency to extract the series resistance, shunt resistance and dark saturation

currents of the device. The short-circuit current density was then increased to 38.05 mA/cm2

according to the results of simulations performed in GenPro4. The series resistance was set

to 0.54 Ωcm2 based on JV measurements conducted on previously processed cells and a

diode ideality factor of 1. However, the substitution of Ti with newScot is expected to lead to a

slight increase in series resistance based on the higher sheet resistance of newScot [154].

Shunt resistances extracted from JV measurements showed values exceeding 50kΩcm2 and

were, therefore, omitted from the simulations by setting the value to 100kΩcm2 as the impact

of high shunt resistances on the shape of the JV curve is negligible. Results of PCD

measurements performed on lifetime samples were also considered as a point of comparison.

While initial 2-diode model simulations assumed a constant j02 of 1.4x108 Acm-2, j02 was varied

here so that the simulated Voc would be equal to measured iVoc values.

Figure 7.9 Comparison between measured JV data and JV data derived from simulations based on the

1-diode and 2-diode model.

Figure 7.10 Results of simulations performed using the 1-diode model and 2-diode model for solar cells.

Potential efficiencies are shown for different dark saturation currents. 2-diode model simulations were

also performed at dark saturation current densities that were obtained for lifetime samples with i/p a-

Si:H layers (symmetrical), poly-Si passivating contacts (symmetrical) and cell-like samples using PCD.

Here, j01 was set to the values obtained from PCD measurements and j02 was adjusted so that the

resulting open-circuit voltage would be equal to the measured implied open-circuit voltage.

The results of simulations in Figure 7.10 show potential efficiencies of solar cells dependant on

the dark saturation current j0/j01. For values close to the ones obtained on symmetrical lifetime

samples for both i/p a-Si:H stacks (non-annealed before deposition) and passivating contacts

of ~10 fA/cm2, efficiencies as high as ~23.5% were calculated for the 1-diode model and

21.25% for the 2-diode model. Efficiencies derived from the 1-diode model are assumed to

overestimate the potential of cells processed as the model is not able to accurately depict the

trend of cells especially around the MPP as shown in Figure 7.9 leading to efficiencies that are

~1% absolute higher than expected. While the accuracy of results obtained from the 1-diode

model can be improved using diode ideality factors above 1, resulting j0 values were

significantly higher than values obtained from PCD measurements and values typically found

in literature for similar Voc and iVoc values (~1x10-12 Acm-2 for n=1.25). For the 2-diode model

issues are related to j02, which affects both iVoc and FF. The impact can be seen especially

through the fill factor, which reaches values below 78% (obtained for processed cells) despite

a significant increase in Voc. A change in passivation quality affects j02 similar to j01 [155],

although j02 values are hard to estimate without available data. 2-diode model simulations

were, therefore, also performed using the results of PCD measurements as a basis with

adjusted j02 values. Potential efficiencies for lifetime samples show both higher efficiencies for

lower dark saturation current densities and lower efficiencies for higher dark saturation current

densities compared to previous calculations using a constant j02 resulting in efficiencies of up

to 22.75%. j02 values obtained from simulations using PCD data ranged from 8x10-9 Acm-2 to

2.4 x10-8 Acm-2 following a similar trend as j01 values. Therefore, one can assume that previous

simulations underestimated the potential of the investigated cell design for cases where good

surface passivating can be achieved resulting in efficiencies of ~22-22.5%. However, the

accuracy of these results depends on the accuracy of j02 values obtained. Alternatively, one

could view 2-diode model simulations using a constant j02 value 1.4x108 Acm-2 as a lower limit

regarding the potential of the investigated device structure. In either case one can expect that

cells using passivating contact on the rear side are able to achieve similar results as the SHJ

reference cell shown in Figure 7.4 showing the potential of passivating contacts processed.

Further limitations that exist for cells processed (both SHJ ref and ones using passivating

contacts) are related to other factors such as the design of the front contact grid and deviations

between optimal and deposited front TCO thicknesses, which were not further investigated.

7.5 Summary and conclusions

In this chapter, the results of solar cells processed and contacting schemes for the poly-Si

layer were discussed. Cells using the SiOx/poly-Si passivating contact on the rear of the cell

were able to reach efficiencies of up to 19.8%. However, comparisons to the silicon hetero-

junction reference indicated issues with the design chosen. Voc values were ~ 20 mV below

the ones of the reference cell, which is assumed to be related to the passivation of the front

side. While it was possible to show that degradation is caused by the high-temperature an-

nealing step by the diffusion of contaminants, it was not possible to locate the exact issue.

Short-circuit current densities of cells processed with the passivating contact also showed

lower values compared to the reference of ~36 mA/cm2. Through EQE measurements together

with GenPro4 simulations it was possible to identify the Ti layer as the cause for parasitic

absorption on the rear of the cell reducing the current of the device by ~2 mA/cm2. Alternative

process schemes were investigated using GenPro4 showing that several contacting schemes

could be considered including the removal of Ti, and the replacement of Ti with a TCO (either

AZO or newScot) with all alternative variants showing similar photogenerated current densities

of ~38 mA/cm2. Contacting structures were tested using TLM measurements. Results from

these measurements showed clearly that the newScot/Ag contacts were not only usable for

contacting poly-Si layers, but also led to low contact resistivities of less than 5 mΩcm2. AZO

layers were shown to be incompatible with the process scheme chosen requiring alternative

methods to realise TLM patterns. Ag contacts showed non-ohmic behaviour, which is assumed

to be the result of the formation of an oxide layer on top of the poly-Si layer. While the issue

regarding the rear contact could be solved, unresolved issues regarding the passivation limited

further investigations. However, simulations performed using the 2-diode model indicated that

efficiencies of ~22% can be achieved for cells with improved front surface passivation, showing

that the cells with passivating contact are able to reach the same efficiency as SHJ reference

cells.

8 Prospects of technologies investigated

8.1 Electron-beam evaporated layers and wet-chemical

ozone oxides for passivating contacts

Through the use of E-beam evaporated a-Si layers it was possible to create passivating con-

tacts with high passivation quality. However, complications may arise from the variance in layer

thicknesses produced from this technique as shown in Figure 6.1. While this does impact the

aptitude of this technique for process schemes that realise the doping after the deposition of

the initially amorphous silicon layer, such as POCl3 diffusion, E-Beam layers may also be

doped during the process using effusion cells through which high doping concentrations can

be achieved or by using doped silicon ingots. Through the use of in-situ doped layers, thickness

related issues can be avoided as long as a minimal thickness is reached necessary for the

contact to provide adequate field-effect passivation [10] and the passivating contact is formed

on the rear of the cell, which is the most commonly used design. Should trends, however, shift

towards designs, where the passivating contact is used on both front and rear side other tech-

niques might have to be considered unless significant improvements can be made in monitor-

ing and controlling the thickness of deposited layers.

Other applications may also be considered for these a-Si layers beyond the creation of pas-

sivating contacts. The successful application of E-Beam silicon layers as absorbers in thin-film

solar cells has been previously shown [156]. The main advantage here is the high deposition

rate achievable with E-Beam evaporation. Additionally, one could consider the use of doped

a-Si layers for the formation of recombination contacts for tandem cell applications, consisting

of n- and p-doped silicon layers processed directly onto each other using E-beam [12]. Beyond

the deposition of silicon layers, E-Beam has also been successfully used for the growth of

metal contacts [157] for silicon solar cells. While several applications can be considered, one

big issue that E-Beam evaporation faces is that the technique has not yet been established in

industrial photovoltaic environments. Further research may, therefore, be required before E-

Beam evaporation can be considered as alternative for other established techniques such as

PECVD and LPCVD.

While experiments performed showed generally good results regarding the passivation pro-

vided by passivating contacts using E-Beam layers, one has to consider deviations between

process conditions used for investigations and preferred process conditions in industrial envi-

ronments. Regarding the a-Si layer, deposition rates were kept rather low at 1-4 A/s to better

control the thickness of the sample resulting in deposition times of ~15 up to 30 min. While the

thickness of a-Si might vary from the range of thicknesses investigated for this thesis, one has

to consider that higher deposition rates, easily achieved by the process, are preferable. To

ensure that the deposition rate for the deposition of a-Si layers does not significantly impact

the structure of resulting poly-Si layers, poly-Si layers were investigated using TEM for 3 dif-

ferent samples grown using an emission current of 190 mA, used in this thesis for the creation

of passivating contacts, as well as 260 mA and 300 mA resulting in deposition rates that are

around five and ten times higher respectively. The samples were annealed at 900°C without

performing a plasma doping step prior to the annealing process. TEM measurements showed

that structures of poly-Si layers were not significantly impacted by the deposition rate. Meas-

urements showed mostly crystalline structures with similar crystal sizes across all samples

ranging from ~200 nm2 to ~20000 nm2. Measurements show the presence of single crystals

within the investigated area stretching from the silicon oxide interface up to the surface of the

sample covering the entire thickness of the poly-Si layer. From these measurements it can be

assumed that the performance of E-Beam a-Si layers should not be impacted by the deposition

rate used for the growth of a-Si layers leading to similar results as the ones shown in chapter

6.2.

While ozone oxides have shown to be able to sufficiently passivate the surfaces of wafers after

annealing and hydrogenation, thermal oxides still provide better surface passivation. Thermal

oxidation of silicon may also seem preferable if the cleaning of wafers is not accomplished via

an ozone cleaning step as the thermal oxide can be grown with a tube furnace, which is re-

quired for the crystallisation of the silicon layer and formation of the passivating contact. Ozone

oxidation may, therefore, be considered if the infrastructure for wet-chemical ozone oxidation

is available and used for other process steps such as the cleaning of wafers. Ozone cleaning

solutions have shown to be as effective at removing most contaminants from wafer surfaces

as other cleaning solutions excluding Cu originating from the saw-cutting process used to di-

vide silicon ingots into wafers [158] while possessing several advantages over other methods

used for wafer cleaning and silicon oxide growth such as low process times, low chemical

consumption and reduced environmental impact. A more meticulous analysis is required to

better assess the cost-performance ratio for ozone-based processes and other methods to

determine an optimal route.

Figure 8.1 Bright (a,c) and dark (b,d,e) TEM images of passivating contacts using emitter currents of

190 mA (a,b) or 300 mA (c,d) for the growth of the initially amorphous silicon layer. Crystal sizes vary

strongly across the investigated area with singular crystals (e) stretching across the entire poly-Si layer.

8.2 TOPCon-like passivating contacts for

silicon/perovskite tandem solar cells

While the application of silicon solar cells with TOPCon-like passivating structures has mostly

been investigated for single junction solar cells, attempts have been made to combine those

cells with perovskites solar cells to form tandem solar cells. The main advantage of tandem or

multi-junction solar cells compared to single-junction devices is that losses resulting from the

difference between photon energy and bandgap energy, as discussed in chapter 3.2.4, can be

reduced by absorbing different parts of the light spectrum using cells with different bandgaps.

Therefore, it is possible to achieve higher efficiencies compared to single-junction devices al-

beit with a more complex device structure. Organic-inorganic metal halide perovskites combine

several qualities, which are preferred for solar cell and tandem applications, including a tune-

able bandgap, high absorption coefficients [159] for light from the visible light spectrum ena-

bling the use of thin layers and easy fabrication. The combination of perovskite and silicon

heterojunction cells has enabled several record efficiencies for perovskite/silicon tandem de-

vice beyond the theoretical limit of single-junction silicon solar cells of 29.4% [160] with the

most recent record reaching 32.5% recently reported by S. Mariotti and E. Köhnen et al. [161].

While the commercial viability of perovskite solar cells is still limited by the long-term stability

of the perovskite cell [162], silicon heterojunction cells may present an additional limitation as

current and estimated market trends show a preference towards different silicon solar cell de-

signs including PERC and TOPCon-like structures. C. Messmer et al. compared different types

of silicon bottom cells based on several factors including efficiency, cost, feasibility and adapt-

ability [163]. The performance of tandem devices was estimated using optical simulations as-

suming a monolithic device current-matched structure as well as electrical simulations for the

silicon bottom cell. Optimised bottom cells using TOPCon and SHJ structures showed compa-

rable cell parameters with almost identical tandem efficiencies of ~30% with PERC solar cells

performing overall worse compared to both TOPCon and SHJ with around 29% tandem effi-

ciency. Estimated costs for the module are slightly lower for the variant using a TOPCon bottom

cell ith 21.3 €/Wpeak com ared to 22.3 €/Wpeak for the SHJ variant resulting in a slightly lower

levelized cost of electricity (LCOE) for the TOPCon cell. The lower LCOE for the perov-

skite/TOPCon tandem cell together with the estimated growth in the world market share based

on the 2021 ITRPV report [164], would indicate that TOPCon bottom cells are preferable for

an industrial adaptation of perovskite/silicon tandem cells. However, it should be said that re-

ported efficiencies for tandem devices using TOPcon or PERC bottom cells have not yet

reached the same efficiencies as values reported for tandems using SHJ cells. More recent

results include reports by Y. Wu et al. [165] achieving 27.6% using a TOPCon bottom cell and

K. Svenbjörnsson et al. [166] featuring a 28.7% tandem cell. The comparably lower tandem

efficiencies are not surprising as research on perovskite/silicon tandem cells focuses mostly

on devices using SHJ bottom cells. Properties of TOPCon-like bottom cells may also be limited

here as only industrial relevant techniques were used for fabrication.

Overall, TOPCon-like solar cells show potential regarding their use as bottom cells for perov-

skite/silicon cells, but are currently not able to achieve efficiencies comparable to perov-

skite/SHJ solar cells. While further improvements are expected, the necessity or required mag-

nitude of these improvements depends on developments in the photovoltaic market, which are

expected to favour TOPCon solar cells.

9 Conclusions and outlook

The main goal of this thesis was to combine wet-chemical ozone-based oxides with E-beam

evaporated amorphous silicon layers and plasma doping to form passivating contacts. Aspects

relating to the performance of passivating contacts processed with these techniques were also

investigated, which included the structural evolution of wet-chemical ozone oxides as well as

doping profiles realised using plasma doping.

ha ter 4 foc sed mostly on the e ol tion of the oxide’s str ct re ithin the assi atin con-

tact. The high-temperature annealing step mostly affected the FWHM of Si3+ bonds located at

the transitional region. Si1+ bonds showed no significant change in FWHM as their structural

order is most-likely connected to the silicon substrate. Annealing may have affected the quality

of the oxide, which could also be seen through the evolution of pinholes formed at tempera-

tures between 860°C and 915°C. The composition of the transitional area did not change

through annealing as similar ratios between Si1+ and Si3+ states were still present. While the

growth of the oxide during the annealing step using nitrogen as process gas could be seen for

all samples, the origin of oxygen atoms necessary for the growth could not be determined.

Restructuring of the transitional area into Si and SiO2 was considered, however, data showed

non-conclusive results.

Chapter 5 consisted of SIMS and ECV measurements performed to investigate plasma doping

for SiOx/poly-Si passivating contacts. During the initial plasma doping step phosphorous mostly

diffuses into a volume close to the surface of the sample, although data suggested that a lower

concentration of phosphorous can still be seen within the bulk of the poly-Si layer. An increase

in pressure resulted in a clear increase in the amount of phosphorous present within the sam-

ple. Through annealing phosphorous diffuses deeper into the sample leading to a mostly ho-

mogenous distribution of phosphorous throughout the poly-Si layer, which correlates with a

homogeneous distribution of electrically active phosphorous determined via ECV. For an an-

nealing temperature of 860°C both ECV and SIMS showed a gradient within the phosphorous

profile indicating a limited diffusion of phosphorous. Information gained regarding indiffusion

was limited by the resol tion of E and to some de ree the sam le’s s rface mor holo y,

however, clear trends between the degree of indiffusion and annealing temperature and doping

pressure were shown. Phosphorous profiles, both active and total, resulting from plasma dop-

ing showed characteristics similar to those achieved via ion implantation and POCl3 diffusion

when applied to passivating contacts.

Passivation qualities of the passivating contacts developed in this thesis were evaluated in

chapter 6. iVoc values as high as 734 mV were reached. Trends of lifetimes samples could be

linked to the results of previous chapters including the impact of doping conditions chosen.

While XPS and optical microscopy measurements pointed towards a potential degradation of

100

the oxide layer at an annealing temperature of 915°C, the actual effect is assumed to be minor

and not significantly impact the results of QSSPC / PCD measurements shown.

Finally, results of solar cells using SiOx/poly-Si contacts on the rear of the device are shown in

chapter 7. Cells with efficiencies of up to 19.9% were processed indicating a successful inte-

gration of the passivating contact. However, results showed clear limitations that arose from

the chosen approach. Titanium, which was used on the rear of the cell to for contacting the

passivating contact, severely limited the properties of cells through parasitic optical absorption.

Alternative contacting schemes were investigated via GenPro4 and tested using TLM. newS-

cot/Ag contacts were used to contact the poly-Si layer leading to contact resistivities below 5

mΩcm2. Other contacting schemes using AZO/Ag or Ag could not be successfully applied.

Besides issues relating to the titanium layer, Voc values were also limited. Issues are assumed

to be caused by contaminants present at the front surface of the cell, although it was not pos-

sible to find the root cause.

9.1 Outlook

Developments in the photovoltaic market over the last decades have clearly shown a prefer-

ence towards wafer-based silicon devices with cell designs focusing on improvements in the

quality of surface passivating structures. Passivating contacts consisting of doped poly-Si and

ultra-thin silicon oxide layers have shown to be a suitable candidate for achieving that goal

resulting in record efficiencies. Solar cell manufacturers have already adopted these structures

for their own devices showing that passivating contacts have already achieved market viability

most-likely leading to growth in the market in years to come. The form that the photovoltaic

market may take is likely to be shaped by cells with passivating contact structures, although

competition especially from PERC+ cells need to be considered. Growth in the market share

of solar cells with passivating contacts may also assist in establishing a market for perov-

skite/silicon tandem solar cells once perovskite have reached market viability.

While TOPCon solar modules use PECVD silicon layers, further growth in the market could

result in the use of different techniques as the question regarding the optimal approach for

creating passivating contacts has still not been answered. The E-Beam silicon layers investi-

gated in this thesis have shown to combine multiple desirable aspects for the application of

passivating contacts on the rear of the cell enabling high passivating qualities when used for

passivating contacts. The application of E-Beam evaporation for photovoltaics has, however,

been limited requiring further research to assess the potential of the technique especially re-

garding scalability and throughput. Ozone-based silicon oxides have shown to be more than

suitable for the chemical passivating of silicon wafers offering an alternative to the thermal

101

oxidation process with reduced processing time. Wet-chemical ozone oxidation may also pro-

mote the adaptation of ozone cleaning processes reducing the amount of hazardous waste

created during the manufacturing process while offering similar cleaning quality. Plasma dop-

ing is unlikely to find use in industrial environments for passivating contacts as PECVD cham-

bers in which the plasma doping process is conducted may as well be used to immediately

deposit doped silicon layers. The technique is, however, useful in research environments as

the impact of process conditions can be assessed more easily for decoupled process steps.

Investigations regarding structural changes in wet-chemical ozone oxides showed potential

reasons for the good performance of wet-chemical ozone oxides in passivating contacts. How-

ever, further investigations are required in particular to determine changes in the width of the

transitional area between c-Si substrate and SiO2 bulk, which were non-conclusive. Addition-

ally, the scope of experiments should be expanded to directly compare structures character-

ised via XPS with passivation quality of passivating contacts using the same oxides.

102

Bibliography

[1]

C. A. Horowitz, "Paris agreement," International Legal Materials, vol. 55, pp. 740-755, 2016.

[2]

H.-O. Pörtner, D. C. Roberts, H. Adams, C. Adler, P. Aldunce, E. Ali, R. A. Begum, R. Betts, R. B.

Kerr, R. Biesbroek and others, "Climate change 2022: Impacts, adaptation and vulnerability,"

IPCC Sixth Assessment Report, 2022.

[3]

European Environment Agency, "Sectoral greenhouse gas emissions by IPCC sector," 21 06

2016. [Online]. Available: https://www.eea.europa.eu/en/analysis/maps-and-charts/change-

of-co2-eq-emissions-2#tab-dashboard-01. [Accessed 17 11 2022].

[4]

Fraunhofer Institute for Solar Energy Systems, "PHOTOVOLTAICS REPORT 2023," 2023.

[5]

A. Blakers, "Development of the PERC solar cell," IEEE Journal of Photovoltaics, vol. 9, pp. 629-

635, 2019.

[6]

J. Haschke, O. Dupré, M. Boccard and C. Ballif, "Silicon heterojunction solar cells: Recent

technological development and practical aspects - from lab to industry," Solar Energy Materials

and Solar Cells, vol. 187, pp. 140-153, 2018.

[7]

B. Kafle, B. S. Goraya, S. Mack, F. Feldmann, S. Nold and J. Rentsch, "TOPCon–Technology

options for cost efficient industrial manufacturing," Solar Energy Materials and Solar Cells, vol.

227, p. 111100, 2021.

[8]

M. Lozac'h, S. Nunomura and K. Matsubara, "Double-sided TOPCon solar cells on textured wafer

with ALD SiOx layer," Solar Energy Materials and Solar Cells, vol. 207, p. 110357, 2020.

[9]

J.-I. Polzin, F. Feldmann, B. Steinhauser, M. Hermle and S. Glunz, "Realization of TOPCon using

industrial scale PECVD equipment," AIP Conference Proceedings, vol. 1999, p. 040018, August

2018.

[10]

Z. P. Ling, Z. Xin, P. Wang, R. Sridharan, C. Ke and R. Stangl, "Double-sided passivated contacts

for solar cell applications: An industrially viable approach toward 24% efficient large area silicon

solar cells," in Silicon Materials, IntechOpen, 2019, pp. 1-44.

[11]

Jinko Solar, "Tiger Neo - N-Type TOPCon," [Online]. Available:

https://www.jinkosolar.com/en/site/tigerneo#s1. [Accessed 02 11 2023].

[12]

S. W. Glunz, B. Steinhauser, J.-I. Polzin, C. Luderer, B. Grübel, T. Niewelt, A. M. O. M. Okasha, M.

Bories, H. Nagel, K. Krieg, F. Feldmann, A. Richter, M. Bivour and M. Hermle, "Silicon-based

passivating contacts: The TOPCon route," Progress in Photovoltaics: Research and Applications,

vol. 31, no. 4, pp. 341-359, 2023.

103

[13]

F. Feldmann, M. Bivour, C. Reichel, M. Hermle and S. W. Glunz, "Passivated rear contacts for

high-efficiency n-type Si solar cells providing high interface passivation quality and excellent

transport characteristics," Solar Energy Materials and Solar Cells, vol. 120, pp. 270-274, 2014.

[14]

A. Richter, R. Müller, J. Benick, F. Feldmann, B. Steinhauser, C. Reichel, A. Fell, M. Bivour, M.

Hermle and S. W. Glunz, "Design rules for high-efficiency both-sides-contacted silicon solar cells

with balanced charge carrier transport and recombination losses," Nature Energy, vol. 6, pp.

429-438, 2021.

[15]

S. Choi, O. Kwon, K. H. Min, M. S. Jeong, K. T. Jeong, M. G. Kang, S. Park, K. K. Hong, H.-e. Song

and K.-H. Kim, "Formation and suppression of hydrogen blisters in tunnelling oxide passivating

contact for crystalline silicon solar cells," Scientific Reports, vol. 10, pp. 1-14, 2020.

[16]

A. Ingenito, G. Nogay, Q. Jeangros, E. Rucavado, C. Allebé, S. Eswara, N. Valle, T. Wirtz, J. Horzel,

T. Koida and others, "A passivating contact for silicon solar cells formed during a single firing

thermal annealing," Nature Energy, vol. 3, pp. 800-808, 2018.

[17]

M. Fırat, H. S. Radhakrishnan, M. R. Payo, P. Choulat, H. Badran, A. van der Heide, J. Govaerts,

F. Duerinckx, L. Tous, A. Hajjiah and J. Poortmans, "Large-area bifacial n-TOPCon solar cells with

in situ phosphorus-doped LPCVD poly-Si passivating contacts," Solar Energy Materials and Solar

Cells, vol. 236, p. 111544, 2022.

[18]

Y. Zhou, K. Tao, A. Liu, R. Jia, J. Bao, Y. Sun, S. Yang, Q. Wang, Q. Zhang, S. Yang, Y. Cao and H.

Qu, "The impacts of LPCVD wrap-around on the performance of n-type tunnel oxide passivated

contact c-Si solar cell," Current Applied Physics, vol. 20, pp. 911-916, 2020.

[19]

J. Linke, J. Rinder, G. Hahn and B. Terheiden, "Compensation of the sputter damage during a-Si

deposition for poly-Si/SiOx passivating contacts by ex-situ p-doping," AIP Conference

Proceedings, vol. 2147, p. 040018, 2019.

[20]

J. Lossen, J. Hoß, S. Eisert, D. Amkreutz, M. Muske, J. Plentz and G. Andrä, "Electron beam

evaporation of silicon for poly-silicon/SiO2 passivated contacts," in 35th European Photovoltaic

Solar Energy Conference and Exhibition, 2018, pp. 418-421.

[21]

A. Moldovan, F. Feldmann, G. Krugel, M. Zimmer, J. Rentsch, M. Hermle, A. Roth-Fölsch, K.

Kaufmann and C. Hagendorf, "Simple cleaning and conditioning of silicon surfaces with

UV/ozone sources," Energy Procedia, vol. 55, pp. 834-844, 2014.

[22]

Z. P. Ling, Z. Xin, C. Ke, K. J. Buatis, S. Duttagupta, J. S. Lee, A. Lai, A. Hsu, J. Rostan and R. Stangl,

"Comparison and characterization of different tunnel layers, suitable for passivated contact

formation," The Japan Society of Applied Physics, vol. 56, p. 08MA01, 2017.

[23]

G. Chen, "The applications of DI-O3 water on wafer surface preparation," in International

Conference on Wafer Rinsing, Water Reclamation and Environmental Technology for

Semiconductor Manufacturing, Industrial Technology Research Institute, Hsinchu, Taiwan,

Hsinchu, Taiwan, 1999, pp. 100-123.

104

[24]

M. Fukuda, "Optical semiconductor devices," vol. 46, John Wiley & Sons, 1998, p. 7.

[25]

M. Fukuda, "Optical semiconductor devices," vol. 46, John Wiley & Sons, 1998, pp. 56-59.

[26]

J. Linke, "Dependencies Between poly-Si Composition and Solar Cell Performance of poly-

Si/SiOx Passivating Contacts," Ph.D. dissertation, Dept. of Physics, University of Konstanz,

Konstanz, Germany, 2020.

[27]

M. Farrag, D. Mathew, R. Chinnappa Naidu, R. Muthu, Y. Wang and K. Busawon, "An improved

optimization technique for estimation of solar photovoltaic parameters," Solar Energy, vol. 157,

pp. 116-124, November 2017.

[28]

C. Honsberg and S. Bowden, "PVEDUCATION," [Online]. Available:

https://www.pveducation.org/pvcdrom/solar-cell-operation/iv-curve. [Accessed 2 Oktober

2022].

[29]

S. Gholipour and M. Saliba, "Chapter 1 - Bandgap tuning and compositional exchange for lead

halide perovskite materials," in Characterization Techniques for Perovskite Solar Cell Materials,

M. Pazoki, A. Hagfeldt and T. Edvinsson, Eds., Elsevier, 2020, pp. 1-22.

[30]

N. J. Ekins-Daukes and M. Guenette, "Photovoltaic Device Operation at Low Temperature," in

2006 IEEE 4th World Conference on Photovoltaic Energy Conference, vol. 1, 2006, pp. 146-149.

[31]

P. Szaniawski, "From Light to Dark : Electrical Phenomena in Cu(In,Ga)Se2 Solar Cells," Ph.D.

dissertation, Faculty of Science and Technology., Uppsala University, Uppsala, Sweden, 2017.

[32]

P. Würfel, "Physics of Solar cells," Wiley, 2005, pp. 65-76.

[33]

P. Würfel, "Physics of Solar cells," Wiley, 2005, p. 86.

[34]

S. Kalainathan, R. Dhanasekaran and P. Ramasamy, "Grain size and size distribution in heavily

phosphorus doped polycrystalline silicon," Journal of crystal growth, vol. 104, pp. 250-256,

1990.

[35]

M. L. Polignano, P. Picco and G. F. Cerofolini, "Phosphorus silica glass as dopant source," Journal

of the Electrochemical Society, vol. 127, p. 2734, 1980.

[36]

M. I. Current, "Ion implantation of advanced silicon devices: Past, present and future," Materials

Science in Semiconductor Processing, vol. 62, pp. 13-22, 2017.

[37]

Y. Tao, V. Upadhyaya, K. Jones and A. Rohatgi, "Tunnel oxide passivated rear contact for large

area n-type front junction silicon solar cells providing excellent carrier selectivity," AIMS

Materials Science, vol. 3, pp. 180-189, January 2016.

[38]

A. Fick, "Ueber Diffusion," Annalen der Physik, vol. 170, pp. 59-86, 1855.

105

[39]

S. R. Logan, "The origin and status of the Arrhenius equation," Journal of Chemical Education,

vol. 59, p. 279, 1982.

[40]

C. Duran, S. J. Eisele, T. Buck, R. Kopecek, J. R. Köhler and J. H. Werner, "Bifacial solar cells with

selective B-BSF by laser doping," in 24th European Photovoltaic Solar Energy Conference, 2009,

pp. 1-4.

[41]

J. A. Röhr, "Electron Transport in Solution Processed Antimony Sulphide Thin Films made from

a Xanthate Precursor," Ph.D. dissertation, Dept. Phy., Imperial College London, London,

England, 2014.

[42]

P. H. Holloway, "Grain boundary diffusion of phosphorus in polycrystalline silicon," Journal of

Vacuum Science and Technology, vol. 21, pp. 19-22, 1982.

[43]

R. N. Ghoshtagore, "Phosphorus diffusion processes in SiO2 films," Thin Solid Films, vol. 25, pp.

501-513, 1975.

[44]

D. Tetzlaff, J. Krügener, Y. Larionova, S. Reiter, M. Turcu, R. Peibst, U. Höhne, J.-D. Kähler and T.

Wietler, "Evolution of oxide disruptions: The (W) hole story about poly-Si/c-Si passivating

contacts," in 2016 IEEE 43rd Photovoltaic Specialists Conference (PVSC), 2016, pp. 0221-0224.

[45]

R. Peibst, U. Römer, Y. Larionova, M. Rienäcker, A. Merkle, N. Folchert, S. Reiter, M. Turcu, B.

Min, J. Krügener and others, "Working principle of carrier selective poly-Si/c-Si junctions: Is

tunnelling the whole story?," Solar Energy Materials and Solar Cells, vol. 158, pp. 60-67, 2016.

[46]

J. Melskens, B. W. H. van de Loo, B. Macco, L. E. Black, S. Smit and W. M. M. Kessels, "Passivating

contacts for crystalline silicon solar cells: From concepts and materials to prospects," IEEE

Journal of Photovoltaics, vol. 8, pp. 373-388, 2018.

[47]

F. Ye, W. Deng, D. Chen, Y. Chen, J. Dong, J. Ding and others, "The Influence of a Low Doping

Concentration Emitter on the Performance of Selective Emitter Silicon Solar Cells," in 29th

EUPVSEC Conference Proceedings, Amsterdam, 2014.

[48]

S. Narasimha, A. Rohatgi and A. W. Weeber, "An optimized rapid aluminum back surface field

technique for silicon solar cells," IEEE Transactions on Electron Devices, vol. 46, pp. 1363-1370,

1999.

[49]

J. Benick, B. Hoex, M. C. M. Van De Sanden, W. M. M. Kessels, O. Schultz and S. W. Glunz, "High

efficiency n-type Si solar cells on Al 2 O 3-passivated boron emitters," Applied Physics Letters,

vol. 92, p. 253504, 2008.

[50]

J. Zhao, A. Wang, X. Dai, M. A. Green and S. R. Wenham, "Improvements in silicon solar cell

performance," in The Conference Record of the Twenty-Second IEEE Photovoltaic Specialists

Conference-1991, 1991, pp. 399-402.

106

[51]

P. Saint-Cast, S. Werner, J. Greulich, U. Jäger, E. Lohmüller, H. Höffler and R. Preu, "Analysis of

the losses of industrial-type PERC solar cells," physica status solidi (a), vol. 214, p. 1600708,

2017.

[52]

R. Chen, H. Tong, H. Zhu, C. Ding, H. Li, D. Chen, B. Hallam, C. M. Chong, S. Wenham and A.

Ciesla, "23.83% efficient mono-PERC incorporating advanced hydrogenation," Progress in

Photovoltaics: Research and Applications, vol. 28, pp. 1239-1247, 2020.

[53]

T. Dullweber and J. Schmidt, "Industrial silicon solar cells applying the passivated emitter and

rear cell (PERC) concept—A review," IEEE journal of photovoltaics, vol. 6, pp. 1366-1381, 2016.

[54]

M. Rienäcker, M. Bossmeyer, A. Merkle, U. Römer, F. Haase, J. Krügener, R. Brendel and R.

Peibst, "Junction resistivity of carrier-selective polysilicon on oxide junctions and its impact on

solar cell performance," IEEE Journal of Photovoltaics, vol. 7, pp. 11-18, 2016.

[55]

W. Fuhs, K. Niemann and J. Stuke, "Heterojunctions of Amorphous Silicon and Silicon Single

Crystals," AIP Conference Proceedings, vol. 20, pp. 345-350, 1974.

[56]

E. Yablonovitch, T. Gmitter, R. M. Swanson and Y. H. Kwark, "A 720 mV open circuit voltage SiO

x: c-Si: SiO x double heterostructure solar cell," Applied Physics Letters, vol. 47, pp. 1211-1213,

1985.

[57]

T. Aoki, T. Matsushita, H. Yamoto, H. Hayashi, M. Okayama and Y. Kawana, "Oxygen-doped

polycrystalline silicon films applied to surface passivation," in Journal of the Electrochemical

Society, 1975.

[58]

I. R. C. Post, P. Ashburn and G. R. Wolstenholme, "Polysilicon emitters for bipolar transistors: a

review and re-evaluation of theory and experiment," IEEE Transactions on Electron Devices, vol.

39, pp. 1717-1731, 1992.

[59]

F. A. Lindholm, A. Neugroschel, M. Arienzo and P. A. Iles, "Heavily doped polysilicon-contact

solar cells," IEEE electron device letters, vol. 6, pp. 363-365, 1985.

[60]

N. G. Tarr, "A polysilicon emitter solar cell," IEEE electron device letters, vol. 6, pp. 655-658,

1985.

[61]

S. Richter, Y. Larionova, S. Großer, M. Menzel, H. Schulte-Huxel, R. Peibst, R. Brendel and C.

Hagendorf, "Evaluation of localized vertical current formation in carrier selective passivation

layers of silicon solar cells by conductive AFM," in AIP Conference Proceedings, vol. 2147, 2019,

p. 040017.

[62]

F. Haase, C. Hollemann, S. Schäfer, A. Merkle, M. Rienäcker, J. Krügener, R. Brendel and R.

Peibst, "Laser contact openings for local poly-Si-metal contacts enabling 26.1%-efficient POLO-

IBC solar cells," Solar Energy Materials and Solar Cells, vol. 186, pp. 184-193, 2018.

107

[63]

A. Richter, J. Benick, F. Feldmann, A. Fell, M. Hermle and S. W. Glunz, "n-Type Si solar cells with

passivating electron contact: Identifying sources for efficiency limitations by wafer thickness

and resistivity variation," Solar Energy Materials and Solar Cells, vol. 173, pp. 96-105, 2017.

[64]

A. Kale, W. Nemeth, S. Nanayakkara, H. Guthrey, M. Page, M. Al-Jassim, S. Agarwal and P.

Stradins, "Tunneling or Pinholes: Understanding the Transport Mechanisms in SiOx Based

Passivated Contacts for High-Efficiency Silicon Solar Cells," in IEEE 7th World Conference on

Photovoltaic Energy Conversion, 2018, pp. 3473-3476.

[65]

T. N. Truong, D. Yan, W. Chen, M. Tebyetekerwa, M. Young, M. Al-Jassim, A. Cuevas, D.

Macdonald and H. T. Nguyen, "Hydrogenation Mechanisms of Poly-Si/SiOx Passivating Contacts

by Different Capping Layers," Solar RRL, vol. 4, p. 1900476, 2020.

[66]

B. W. H. van de Loo, B. Macco, M. Schnabel, M. K. Stodolny, A. A. Mewe, D. L. Young, W. Nemeth,

P. Stradins and W. M. M. Kessels, "On the hydrogenation of Poly-Si passivating contacts by

Al2O3 and SiNx thin films," Solar Energy Materials and Solar Cells, vol. 215, p. 110592, 2020.

[67]

F. Feldmann, T. Fellmeth, B. Steinhauser, H. Nagel, D. Ourinson, S. Mack, E. Lohmüller, J. I.

Polzin, J. Benick, A. Richter and others, "Large area TOPCon cells realized by a PECVD tube

process," in 36th European Photovoltaic Solar Energy Conference and Exhibition, 2019, pp. 304-

308.

[68]

F. Feldmann, G. Nogay, J.-I. Polzin, B. Steinhauser, A. Richter, A. Fell, C. Schmiga, M. Hermle and

S. W. Glunz, "A study on the charge carrier transport of passivating contacts," IEEE Journal of

Photovoltaics, vol. 8, pp. 1503-1509, 2018.

[69]

H. Yousuf, M. Q. Khokhar, M. A. Zahid, M. Rabelo, S. Kim, D. P. Pham, Y. Kim and J. Yi, "Tunnel

Oxide Deposition Techniques and Their Parametric Influence on Nano-Scaled SiOx Layer of

TOPCon Solar Cell: A Review," Energies, vol. 15, p. 5753, 2022.

[70]

S. Glunz, M. Bivour, C. Messmer, F. Feldmann, R. Müller, C. Reichel, A. Richter, F. Schindler, J.

Benick and M. Hermle, "Passivating and Carrier-selective Contacts - Basic Requirements and

Implementation," in 2017 IEEE 44th Photovoltaic Specialist Conference (PVSC), 2017, pp. 2064-

2069.

[71]

W. R. Fahrner, "Amorphous Silicon / Crystalline Silicon Heterojunction Solar Cells," in

Amorphous Silicon / Crystalline Silicon Heterojunction Solar Cells, Berlin, Springer, 2013, p. 11.

[72]

L. Sims, U. Hörmann, R. Hanfland, R. C. I. MacKenzie, F. R. Kogler, R. Steim, W. Brütting and P.

Schilinsky, "Investigation of the s-shape caused by the hole selective layer in bulk heterojunction

solar cells," Organic Electronics, vol. 15, pp. 2862-2867, 2014.

[73]

K. Yoshikawa, H. Kawasaki, W. Yoshida, T. Irie, K. Konishi, K. Nakano, T. Uto, D. Adachi, M.

Kanematsu, H. Uzu and K. Yamamoto, "Silicon heterojunction solar cell with interdigitated back

contacts for a photoconversion efficiency over 26%," Nature Energy, vol. 2, p. 17032, 2017.

108

[74]

D. L. Bätzner, Y. Andrault, L. Andreetta, A. Büchel, W. Frammelsberger, C. Guerin, N. Holm, D.

Lachenal, J. Meixenberger, P. Papet, B. Rau, B. Strahm, G. Wahli and F. Wünsch, "Properties of

high efficiency silicon heterojunction cells," Energy Procedia, vol. 8, pp. 153-159, 2011.

[75]

M. Raval and S. Madugula, Industrial Silicon Solar Cells, IntechOpen, 2019.

[76]

D. M. Huljic, D. Biro, R. Preu, C. C. Castillo and R. Ludemann, "Rapid thermal firing of screen

printed contacts for large area crystalline silicon solar cells," in Conference Record of the

Twenty-Eighth IEEE Photovoltaic Specialists Conference, 2000, pp. 379-382.

[77]

A. Khanna, L. I. N. G. Zhi Peng, V. Shanmugam, M. Boreland, I. Hayashi, D. Kirk, H. Akimoto, A.

ABERLE and T. Mueller, "Screen printed metallisation for silicon heterojunction solar cells," in

28th European Photovoltaic Solar Energy Conference and Exhibition, 2013, pp. 1336-1339.

[78]

A. B. Cruz, "Transparent conductive oxides for silicon heterojunction solar cells: interaction

between materials and device," Ph.D. dissertation, Dept. Elect. Eng., Technische Universität

Berlin, Berlin, Germany, 2021.

[79]

P. Wagner, "Interdigitated back-contact silicon heterojunction solar cells: development of

patterning techniques and applications in tandem devices," Ph.D. dissertation, Dept. Elect. Eng.,

Technische Universität Berlin, Berlin, Germany, 2021.

[80]

A. Descoeudres, Z. C. Holman, L. Barraud, S. Morel, S. D. Wolf and C. Ballif, ">21% Efficient

Silicon Heterojunction Solar Cells on n- and p-Type Wafers Compared," IEEE Journal of

Photovoltaics, vol. 3, pp. 83-89, 2013.

[81]

J. Korec, "Low Voltage Power MOSFETs: Design, Performance and Applications," Springer

Science & Business Media, 2011, p. 38.

[82]

O. V. Aleksandrov and A. I. Dus’, "A model of formation of fixed charge in thermal silicon

dioxide," Semiconductors, vol. 45, pp. 467-473, 2011.

[83]

A. G. Aberle, S. Glunz and W. Warta, "Impact of illumination level and oxide parameters on

Shockley–Read–Hall recombination at the Si-SiO2 interface," Journal of Applied Physics, vol. 71,

pp. 4422-4431, 1992.

[84]

H. Angermann, "Conditioning of Si-interfaces by wet-chemical oxidation: Electronic interface

properties study by surface photovoltage measurements," Applied Surface Science, vol. 312, pp.

3-16, 2014.

[85]

K. Wolke, C. Gottschalk, J. Rentsch and H. Angermann, "Ozone based chemical oxide growth for

crystalline solar cell production," in Solid State Phenomena, vol. 187, 2012, pp. 321-324.

[86]

B. E. Deal and A. S. Grove, "General relationship for the thermal oxidation of silicon," Journal of

applied physics, vol. 36, pp. 3770-3778, 1965.

109

[87]

Z. Rappoport and Y. Apeloig, "The Structural Chemistry of Organosilicon Compounds," in

Chemistry of organic silicon compounds, John Wiley & Sons, 1998, pp. 181-265.

[88]

W. H. Baur, "The prediction of bond length variations in silicon-oxygen bonds," American

Mineralogist: Journal of Earth and Planetary Materials, vol. 56, pp. 1573-1599, 1971.

[89]

D. Rouchon, N. Rochat, F. Gustavo, A. Chabli, O. Renault and P. Besson, "Study of ultrathin silicon

oxide films by FTIR-ATR and ARXPS after wet chemical cleaning processes," Surface and

Interface Analysis: An International Journal devoted to the development and application of

techniques for the analysis of surfaces, interfaces and thin films, vol. 34, pp. 445-450, 2002.

[90]

N. Herbots, J. Shaw, Q. B. Hurst, M. P. Grams, R. Culbertson, D. Smith, V. Atluri, P. Zimmerman

and K. T. Queeney, "The formation of ordered, ultrathin SiO2/Si(100) interfaces grown on (1×1)

Si(100)," Materials Science and Engineering: B, vol. 87, pp. 303-316, December 2001.

[91]

I. Ohdomari, H. Akatsu, Y. Yamakoshi and K. Kishimoto, "Study of the interfacial structure

between Si (100) and thermally grown SiO2 using a ball‐and‐spoke model," Journal of Applied

Physics, vol. 62, pp. 3751-3754, 1987.

[92]

G. Kovačević and B. Pivac, "Structure, defects, and strain in silicon-silicon oxide interfaces,"

Journal of Applied Physics, vol. 115, p. 043531, December 2013.

[93]

A. Munkholm, S. Brennan, F. Comin and L. Ortega, "Observation of a Distributed Epitaxial Oxide

in Thermally Grown Si O 2 on Si (001)," Physical review letters, vol. 75, p. 4254, 1995.

[94]

A. Ourmazd, D. W. Taylor, J. A. Rentschler and J. Bevk, "Si -> SiO2 transformation: Interfacial

structure and mechanism," Phys. Rev. Lett., vol. 59, no. 2, pp. 213-216, July 1987.

[95]

I. K. Robinson, W. K. Waskiewicz, R. T. Tung and J. Bohr, "Ordering at Si (111) a- Si and Si (111)/Si

O 2 Interfaces," Physical review letters, vol. 57, p. 2714, 1986.

[96]

I. Takahashi, T. Shimura and J. Harada, "X-ray diffraction evidence for epitaxial microcrystallinity

in thermally oxidized SiO2 thin films on the Si (001) surface," Journal of Physics: Condensed

Matter, vol. 5, p. 6525, 1993.

[97]

N. F. Mott, S. Rigo, F. Rochet and A. M. Stoneham, "Oxidation of silicon," Philosophical Magazine

B, vol. 60, pp. 189-212, 1989.

[98]

F. P. Fehlner and N. F. Mott, "Low-temperature oxidation," Oxidation of metals, vol. 2, pp. 59-

99, 1970.

[99]

W. Kern and others, "Handbook of semiconductor wafer cleaning technology," New Jersey:

Noyes Publication, pp. 189-212, 1993.

[100]

J. John, A. Hajjiah, M. Haslinger, M. Soha, A. Urueña, E. Cornagliotti, L. Tous, P. Mertens and J.

Poortmans, "Deposition behaviour of metal impurities in acidic cleaning solutions and their

110

impact on effective minority carrier lifetime in n-type silicon solar cells," Solar Energy Materials

and Solar Cells, vol. 194, pp. 83-88, 2019.

[101]

M. Meuris, P. W. Mertens, A. Opdebeeck, H. F. Schmidt, M. Depas, G. Vereecke, M. M. Heyns

and A. Philipossian, "The IMEC clean: A new concept for particle and metal removal on Si

surfaces," Solid State Technology, vol. 38, pp. 109-113, 1995.

[102]

J. R. Vig, "Ultraviolet-ozone cleaning of semiconductor surfaces," ARMY RESEARCH LAB FORT

MONMOUTH NJ ELECTRONICS AND POWER SOURCES DIRECTORATE, Fort Monmouth, USA,

1992.

[103]

H. Angermann, K. Wolke, C. Gottschalk, A. Moldovan, M. Roczen, J. Fittkau, M. Zimmer and J.

Rentsch, "Electronic interface properties of silicon substrates after ozone based wet-chemical

oxidation studied by SPV measurements," Applied surface science, vol. 258, pp. 8387-8396,

2012.

[104]

Z. Ying, Y. Yecheskel, M. Huo, U. Hübner and I. Zucker, "Ozone Consumption by Soils: A Critical

Factor in In Situ Ozonation Processes," ACS ES&T Water, vol. 1, pp. 2403-2411, 2021.

[105]

N. Osawa, T. Tsuji, R. Ogiso and Y. Yoshioka, "Effect of nitrogen addition to ozone generation

characteristics by diffuse and filamentary dielectric barrier discharges at atmospheric pressure,"

The European Physical Journal Applied Physics, vol. 78, p. 20804, 2017.

[106]

F. De Smedt, S. De Gendt, M. Claes, M. M. Heyns, H. Vankerckhoven and C. Vinckier, "The

increasing importance of the use of ozone in the microelectronics industry," Ozone: science &

engineering, vol. 24, pp. 379-390, 2002.

[107]

K. J. Laidler, "The development of the Arrhenius equation," Journal of chemical Education, vol.

61, p. 494, 1984.

[108]

F. De Smedt, C. Vinckier, I. Cornelissen, S. De Gendt and M. Heyns, "A detailed study on the

growth of thin oxide layers on silicon using ozonated solutions," Journal of the Electrochemical

Society, vol. 147, p. 1124, 2000.

[109]

C. Gondek, R. Hanich, F. Honeit, A. Lißner, A. Stapf and E. Kroke, "Etching Silicon with Aqueous

Acidic Ozone Solutions: Reactivity Studies and Surface Investigations," The Journal of Physical

Chemistry C, vol. 120, pp. 22349-22357, 2016.

[110]

S. L. Nelson, R. T. Fayfield, K. K. Christenson and B. E. Deal, "The mechanism of growth of thin

oxide film on silicon using ozonated water," Proc. UCPSS, pp. 287-290, 1996.

[111]

A. Moldovan, F. Feldmann, K. Kaufmann, S. Richter, M. Werner, C. Hagendorf, M. Zimmer, J.

Rentsch and M. Hermle, "Tunnel oxide passivated carrier-selective contacts based on ultra-thin

SiO2 layers grown by photo-oxidation or wet-chemical oxidation in ozonized water," in 2015

IEEE 42nd Photovoltaic Specialist Conference (PVSC), 2015, pp. 1-6.

111

[112]

N. El Arbi, R. Jemai, K. Khirouni and H. Khemakhem, "The Variation of Crystalline Structure

Induced by Gas Dilution and Thermal Annealing in Silicon Layers Deposited by PECVD

Technique," Silicon, vol. 11, pp. 2075-2086, 2019.

[113]

H. Wagner, A. Dastgheib-Shirazi, B. Min, A. E. Morishige, M. Steyer, G. Hahn, C. del Cañizo, T.

Buonassisi and P. P. Altermatt, "Optimizing phosphorus diffusion for photovoltaic applications:

Peak doping, inactive phosphorus, gettering, and contact formation," Journal of Applied Physics,

vol. 119, p. 185704, 2016.

[114]

H. Nagel, C. Berge and A. G. Aberle, "Generalized analysis of quasi-steady-state and quasi-

transient measurements of carrier lifetimes in semiconductors," Journal of Applied Physics, vol.

86, pp. 6218-6221, 1999.

[115]

R. A. Sinton and A. Cuevas, "Contactless determination of current–voltage characteristics and

minority-carrier lifetimes in semiconductors from quasi-steady-state photoconductance data,"

Applied Physics Letters, vol. 69, pp. 2510-2512, 1996.

[116]

Y. Chen, D. Chen, C. Liu, Z. Wang, Y. Zou, Y. He, Y. Wang, L. Yuan, J. Gong, W. Lin and others,

"Mass production of industrial tunnel oxide passivated contacts (i-TOPCon) silicon solar cells

with average efficiency over 23% and modules over 345 W," Progress in Photovoltaics: Research

and Applications, vol. 27, pp. 827-834, 2019.

[117]

D. Skorka, A. Zuschlag and G. Hahn, "Injection resolved spatial lifetime mapping using

photoluminescence," in AIP Conference Proceedings, vol. 2147, 2019, p. 020015.

[118]

National Renewable Energy Laboratory, "www.nrel.gov," [Online]. Available:

https://www.nrel.gov/grid/solar-resource/spectra-am1.5.html. [Accessed 15 October 2022].

[119]

D. Pysch, A. Mette and S. W. Glunz, "A review and comparison of different methods to

determine the series resistance of solar cells," Solar Energy Materials and Solar Cells, vol. 91,

pp. 1698-1706, 2007.

[120]

E. F. Lisbona, "Chapter IIA-2—Calibration, Testing, and Monitoring of Space Solar Cells," Solar

Cells, 2nd ed.; McEvoy, A., Castañer, L., Markvart, T., Eds, pp. 501-529, 2013.

[121]

T. Abbas and L. Slewa, "Transmission line method (TLM) measurement of (metal/ZnS) contact

resistance," Int. J. Nanoelectronics and Materials, vol. 8, pp. 111-120, 2015.

[122]

S. Grover, "Effect of transmission line measurement (TLM) geometry on specific contact

resistivity determination," Theory of Computing Systems \/ Mathematical Systems Theory, p.

35, 2016.

[123]

G. E. Jellison Jr, M. F. Chisholm and S. M. Gorbatkin, "Optical functions of chemical vapor

deposited thin-film silicon determined by spectroscopic ellipsometry," Applied physics letters,

vol. 62, pp. 3348-3350, 1993.

112

[124]

A. Moldovan, Ozonbasierte Reinigungs-und Konditionierungsverfahren für die Herstellung

hocheffizienter Silizium Solarzellen, Fraunhofer Verlag Stuttgart, 2016.

[125]

F. J. Himpsel, F. R. McFeely, A. Taleb-Ibrahimi, J. Yarmoff and G. Hollinger, "Microscopic

structure of the SiO 2/Si interface," Physical review B, vol. 38, p. 6084, 1988.

[126]

R. Santbergen, T. Meguro, T. Suezaki, G. Koizumi, K. Yamamoto and M. Zeman, "GenPro4 optical

model for solar cell simulation and its application to multijunction solar cells," IEEE journal of

photovoltaics, vol. 7, pp. 919-926, 2017.

[127]

D. R. Baer, K. Artyushkova, H. Cohen, C. D. Easton, M. Engelhard, T. R. Gengenbach, G.

Greczynski, P. Mack, D. J. Morgan and A. Roberts, "XPS guide: Charge neutralization and binding

energy referencing for insulating samples," Journal of Vacuum Science & Technology A: Vacuum,

Surfaces, and Films, vol. 38, p. 031204, 2020.

[128]

J. W. Keister, J. E. Rowe, J. J. Kolodziej, H. Niimi, H.-S. Tao, T. E. Madey and G. Lucovsky,

"Structure of ultrathin SiO 2/Si (111) interfaces studied by photoelectron spectroscopy," Journal

of Vacuum Science & Technology A: Vacuum, Surfaces, and Films, vol. 17, pp. 1250-1257, 1999.

[129]

A. Jablonski and C. J. Powell, "Relationships between electron inelastic mean free paths,

effective attenuation lengths, and mean escape depths," Journal of Electron Spectroscopy and

Related Phenomena, vol. 100, pp. 137-160, 1999.

[130]

J. Chastain and R. C. King Jr, "Handbook of X-ray photoelectron spectroscopy," Perkin-Elmer

Corporation, vol. 40, p. 221, 1992.

[131]

M. Wojdyr, "Fityk: a general-purpose peak fitting program," Journal of Applied Crystallography,

vol. 43, pp. 1126-1128, 2010.

[132]

J. J. Olivero and R. L. Longbothum, "Empirical fits to the Voigt line width: A brief review," Journal

of Quantitative Spectroscopy and Radiative Transfer, vol. 17, pp. 233-236, 1977.

[133]

L. Veith, "Optimization and application of Time-of-Flight Secondary Ion Mass Spectrometry

(ToF-SIMS) for the detection of nanomaterials in tissue thin sections," Ph.D.,

Naturwissenschaftlich-Technische Fakultät, Universität Siegen, Siegen, 2018.

[134]

E. Peiner, A. Schlachetzki and D. Krüger, "Doping profile analysis in Si by electrochemical

capacitance-voltage measurements," Journal of the Electrochemical Society, vol. 142, p. 576,

1995.

[135]

Probion Analysis, "Probion.fr," [Online]. Available: https://probion.fr/ecv-en/. [Accessed 05 10

2022].

[136]

MicroChemicals, "www.microchemicals.com," [Online]. [Accessed 02 09 2021].

[137]

M. P. Seah and S. J. Spencer, "Ultrathin SiO2 on Si IV. Intensity measurement in XPS and deduced

thickness linearity," Surface and Interface Analysis: An International Journal devoted to the

113

development and application of techniques for the analysis of surfaces, interfaces and thin films,

vol. 35, pp. 515-524, 2003.

[138]

J.-I. Polzin, S. Lange, S. Richter, A. Moldovan, M. Bivour, C. Hagendorf, M. Hermle, S. W. Glunz

and F. Feldmann, "Temperature-induced stoichiometric changes in thermally grown interfacial

oxide in tunnel-oxide passivating contacts," Solar Energy Materials and Solar Cells, vol. 218, p.

110713, 2020.

[139]

F. Reale, P. Palczynski, I. Amit, G. F. Jones, J. D. Mehew, A. Bacon, N. Ni, P. C. Sherrell, S. Agnoli,

M. F. Craciun and others, "High-mobility and high-optical quality atomically thin WS 2," Scientific

Reports, vol. 7, p. 14911, 2017.

[140]

S. Seki, H. Tamura, Y. Wada and K. Tsutsui, "Depth profiling in secondary ion mass spectrometry

for ultra-thin layer with nanometer order thickness by mesa-structure fabrication," Surface and

interface analysis, vol. 44, pp. 614-617, 2012.

[141]

S. Solmi, A. Parisini, R. Angelucci, A. Armigliato, D. Nobili and L. Moro, "Dopant and carrier

concentration in Si in equilibrium with monoclinic SiP precipitates," Physical Review B, vol. 53,

p. 7836, 1996.

[142]

U. Römer, R. Peibst, T. Ohrdes, B. Lim, J. Krügener, T. Wietler and R. Brendel, "Ion implantation

for poly-Si passivated back-junction back-contacted solar cells," IEEE Journal of Photovoltaics,

vol. 5, pp. 507-514, 2015.

[143]

D. L. Young, W. Nemeth, V. LaSalvia, R. Reedy, N. Bateman and P. Stradins, "Ion implanted

passivated contacts for interdigitated back contacted solar cells," in 2015 IEEE 42nd

Photovoltaic Specialist Conference (PVSC), 2015, pp. 1-5.

[144]

S. Yuan, Y. Cui, Y. Zhuang, P. Chen, Y. Hu, B. Yang, Y. Yu, Y. Ren, W. Wang, W. Chen and others,

"Passivated emitter and rear cell silicon solar cells with a front polysilicon passivating contacted

selective emitter," physica status solidi (RRL)–Rapid Research Letters, vol. 15, p. 2100057, 2021.

[145]

H. Park, S. Bae, S. J. Park, J. Y. Hyun, C. H. Lee, D. Choi, D. Kang, H. Han, Y. Kang, H.-S. Lee and

others, "Role of polysilicon in poly-Si/SiO x passivating contacts for high-efficiency silicon solar

cells," RSC advances, vol. 9, pp. 23261-23266, 2019.

[146]

R. Brendel and R. Peibst, "Contact selectivity and efficiency in crystalline silicon photovoltaics,"

IEEE Journal of photovoltaics, vol. 6, pp. 1413-1420, 2016.

[147]

C. Leendertz, "Effizienzlimitierende Rekombinationsprozesse in amorph/kristallinen und

polykristallinen Siliziumsolarzellen," Ph.D. dissertation, Dept. Elect. Eng., Technische Universität

Berlin, Berlin, Germany, 2012.

[148]

T. F. Wietler, D. Tetzlaff, J. Krügener, M. Rienäcker, F. Haase, Y. Larionova, R. Brendel and R.

Peibst, "Pinhole density and contact resistivity of carrier selective junctions with polycrystalline

silicon on oxide," Applied Physics Letters, vol. 110, p. 253902, 2017.

114

[149]

X. Zhang, K.-S. Chen, R. Ghodssi, A. A. Ayon and S. M. Spearing, "Residual stress and fracture in

thick tetraethylorthosilicate (TEOS) and silane-based PECVD oxide films," Sensors and Actuators

A: Physical, vol. 91, pp. 373-380, 2001.

[150]

R. A. Sinton, A. Cuevas and others, "A quasi-steady-state open-circuit voltage method for solar

cell characterization," in Proceedings of the 16th European photovoltaic solar energy

conference, vol. 1152, 2000, p. 4.

[151]

L. Tutsch, F. Feldmann, J. Polzin, C. Luderer, M. Bivour, A. Moldovan, J. Rentsch and M. Hermle,

"Implementing transparent conducting oxides by DC sputtering on ultrathin SiOx/poly-Si

passivating contacts," Solar Energy Materials and Solar Cells, vol. 200, p. 109960, 2019.

[152]

B. Steinhauser, F. Feldmann, D. Ourinson, H. Nagel, T. Fellmeth and M. Hermle, "On the

influence of the sinx composition on the firing stability of poly-si/sinx stacks," physica status

solidi (a), vol. 217, p. 2000333, 2020.

[153]

S. Suckow, T. M. Pletzer and H. Kurz, "Fast and reliable calculation of the two-diode model

without simplifications," Progress in Photovoltaics: Research and Applications, vol. 22, pp. 494-

501, 2014.

[154]

A. Cruz, D. Erfurt, P. Wagner, A. B. Morales, F. Ruske, R. Schlatmann and B. Stannowski,

"Optoelectrical analysis of TCO+Silicon oxide double layers at the front and rear side of silicon

heterojunction solar cells," Solar Energy Materials and Solar Cells, vol. 236, p. 111493,

December 2021.

[155]

O. Breitenstein, "Understanding the current-voltage characteristics of industrial crystalline

silicon solar cells by considering inhomogeneous current distributions," Opto-Electronics

Review, vol. 21, pp. 259-282, June 2013.

[156]

S. Garud, Laser Processes for Silicon Solar Cells: As Investigated on Polycrystalline Silicon on

Glass, Ph.D. dissertation, Dept. Elect. Eng., Technische Universität Berlin, Berlin, Germany, 2021.

[157]

T. Wahl, J. Hanisch and E. Ahlswede, "Comparison of the Al back contact deposited by

sputtering, e-beam, or thermal evaporation for inverted perovskite solar cells," Journal of

Physics D: Applied Physics, vol. 51, p. 135502, 2018.

[158]

S. Choi, B. Jang, J. Kim, H. Song and M. Han, "Cu-contamination of single crystalline silicon wafers

with thickness of 100 μm during multi-wire sawing process," Solar Energy, vol. 125, p. 198–206,

2016.

[159]

C. T. Crespo, "Absorption coefficients data of lead iodine perovskites using 14 different organic

cations," Data in brief, vol. 27, p. 104636, 2019.

[160]

W. Shockley and H. J. Queisser, "Detailed balance limit of efficiency of p-n junction solar cells,"

Journal of applied physics, vol. 32, pp. 510-519, 1961.

115

[161]

Helmholtz-Zentrum Berlin, "www.helmholtz-berlin.de," 19 12 2022. [Online]. Available:

https://www.helmholtz-berlin.de/pubbin/news_seite?nid=24348;sprache=en. [Accessed 01 01

2023].

[162]

T. A. Chowdhury, M. A. B. Zafar, M. S.-U. Islam, M. Shahinuzzaman, M. A. Islam and M. U.

Khandaker, "Stability of perovskite solar cells: issues and prospects," RSC Advances, vol. 13, pp.

1787-1810, 2023.

[163]

C. Messmer, B. S. Goraya, S. Nold, P. S. C. Schulze, V. Sittinger, J. Schön, J. C. Goldschmidt, M.

Bivour, S. W. Glunz and M. Hermle, "The race for the best silicon bottom cell: Efficiency and cost

evaluation of perovskite–silicon tandem solar cells," Progress in Photovoltaics: Research and

Applications, vol. 29, pp. 744-759, 2021.

[164]

International Technology Roadmap for Photovoltaics, "Crystalline SIlicon Technology - Current

Status and Outlook," Siemens, 2021.

[165]

Y. Wu, P. Zheng, J. Peng, M. Xu, Y. Chen, S. Surve, T. Lu, A. D. Bui, N. Li, W. Liang and others,

"27.6% Perovskite/c-Si Tandem Solar Cells Using Industrial Fabricated TOPCon Device,"

Advanced Energy Materials, vol. 12, p. 2200821, 2022.

[166]

K. Sveinbjörnsson, B. Li, S. Mariotti, E. Jarzembowski, L. Kegelmann, A. Wirtz, F. Frühauf, A.

Weihrauch, R. Niemann, L. Korte and others, "Monolithic Perovskite/Silicon Tandem Solar Cell

with 28.7% Efficiency Using Industrial Silicon Bottom Cells," ACS Energy Letters, vol. 7, pp. 2654-

2656, 2022.

116

Acknowledgements

First, I would like to thank Prof. Dr. Bernd Rech for the opportunity and resources necessary

for conducting research for my PhD. In that regard I would also like to express my gratitude

towards my supervisors Dr. Lars Korte and Dr. Daniel Amkreutz for supporting me throughout

my entire time at the HZB.

I would also like to thank Prof. Dr. Steve Albrecht and Prof. Dr. Olindo Isabella together with

Prof. Dr. Brend Rech for being part of the doctoral committee and for evaluating my work.

For proofreading my thesis, I would like to thank Dr. Lars Korte, Dr. Daniel Amkreutz, Dr.

Philipp Wagner and Dr. Dorothee Menzel.

A special thanks goes out to my former and current roommates Dr. Phillip Wagner, Dr. Alvaro

Tejada Esteves and Dr. Ganna Chistiakova, who helped me in endeavours related to PECVD,

ellipsometry and XPS respectively. I would also like to thank them for the discussions we had

throughout the years, both scientific and non-scientific.

o ld like to thank the members of E L’s X team incl din rof. Dr. arc s Bär Dr.

Regan Wilks and especially Dr. Johannes Frisch for discussions related to XPS and XPS

measurements performed.

My gratitude goes towards Kerstin Jacob and Mona Witting for RCA cleaning and texturing,

Tobias Henschel and Franziska Biegalke for texturing, Martin Muske for E-Beam depositions

and PECVD processes, Dr. Philipp Wagner, Tobias Henschel and Dr. Cham Trinh for PECVD

depositions, Katja Meyer-Stillrich, Denise Debrassine and Manuel Hartig for sputtering

processes, Jannik Kleesiek and Stefan Janke for screen printing Carola Klimm for SEM

measurements and Dr. Markus Wollgarten, Holger Kropf and Ulrike Bloeck for TEM

measurements.

I would like to thank Dr. Maurice Nuys, Dr. Uwe Breuer and Mohammed Jaber from the

Forschungszentrum Jülich for performing SIMS and ECV measurements, which formed the

basis for discussions related to the plasma doping process.

A special thank goes to Martin Muske who kept the VA cluster running and fixed any issue that

occurred.

I would also like to thank Dr. Cham Trinh, Dr. Florian, Ruske and Dr. Johannes Frisch for

detailed discussions on several topics.

Finally, I would like to thank Dr. Lars Korte and Prof. Dr. Steve Albrecht and Prof. Dr. Bernd

Stannowski as well as every member of the former silicon group, NPET group and PVcomB

for providing a pleasant work environment.

117

Publications

Articles

• W. Duan, G. Mains, H. T. Gebrewold, K. Bittkau, A. Lambertz, B. Xu, V. Lauterbach, A.

Eberst, N. Nicholson, L. Korte, M. A. Yaqin, K. Zhang, Q. Yang, U. Rau and K. Ding,

“Enhancing the Selectivity and Transparency of the Electron Contact in Silicon Hetero-

junction Solar Cells by Phosphorus Catalytic Doping “ Adv. Funct. Mater., vol. 34, p.

2310552, 2024.

Presentations

• N. Nicholson, J. Frisch, M. Nuys, U. Breuer, M. Jaber, J. Kleesiek, D. Amkreutz, L.

Korte and B. Rech, “Wet-chemical Ozone-Based Oxide Layers and Electron-Beam

E a orated ilicon Layers for assi atin ontacts “ ral resentation (N. Nicholson)

8th World Conference on Photovoltaic Energy Conversion, Milan, Spain, Sep. 2022.