Spectroscopic Ellipsometry for the In-situ Investigation of Atomic Layer Depositions

Inhaltsverzeichnis

CONTENTS

Lists of Abbreviations and Symbols

Lists of Figures and Tables

1 Introduction

I Theoretical Part

2 Alumina in electronic industry

3 Atomic Layer Deposition
3.1 History
3.2 Process definition
3.3 Benefits and limitations
3.4 ALD growth mechanism of Aluminium oxide from TMA/O
3.5 Growth kinetics
3.6 Comparison of TMA/O3 and TMA/H2O - A literature survey

4 Spectroscopic Ellipsometry
4.1 Introduction
4.2 Measuring Principle
4.3 Fitting and models
4.4 Advantages and limitations

5 X-Ray Photoelectron Spectroscopy
5.1 Introduction
5.2 XPS mechanism
5.3 XPS analysis
5.4 Advantages and limitations

6 Atomic Force Microscopy

II Experimental Part

7 Methodologies
7.1 Experimental setup
7.2 ALD process
7.3 Experiment design
7.4 Spectroscopic Ellipsometry
7.4.1 Tool and software
7.4.2 Data acquisition
7.4.3 Data evaluation
7.4.4 Post processing of data
7.4.5 Sources of errors in SE

8 Results and discussion
8.1 Introduction
8.2 Kinetic ALD characteristic curves
8.2.1 TMA exposure
8.2.2 Argon purging after TMA exposure
8.2.3 Ozone exposure
8.2.4 Argon purging after ozone exposure
8.3 Impact of process parameters on characteristic ALD growth attributes and film properties
8.3.1 Total process pressure
8.3.2 Ozone flow
8.3.3 Deposition temperature
8.4 Reproducibility

9 Conclusions and outlook

Contents

References

III Appendix

A Reference temperatures and ozone flow

B Process parameters

TASK FORMULATION FOR A STUDENT RESEARCH PROJECT

Student’s name: Varun S H A R M A

Subject:

»Spectroscopic ellipsometry for the in-situ investigation of atomic layer depositions« Spectroscopic ellipsometry (SE) is a non-invasive measurement technique, which is fast, accurate, self-referencing, and highly sensitive to the smallest change on a solid-state surface. Thus, SE is ideally suited for the in-situ characterization of a material’s deposition, modifying an investigated sample from its initial stage towards a film covered surface.

Atomic layer deposition (ALD) is a special kind of chemical vapor deposition, which is mainly characterized by a self-terminating surface coverage in the progress of the two distinctive reaction steps. As a result, ALD enables a highly conformal coating of complex three- dimensional-shaped surfaces including a unique film thickness control in the sub-nanometer range. The ALD of aluminum oxide (Al2O3) can be regarded as a standard model process.

Objectives: As a result of the student research project, the capability of in situ SE is to be demonstrated with respect to a detailed characterization of the Al2O3 ALD model process, comprising a comprehensive study of various process parameter interdependencies. Consequently, a reliable process for the Al2O3 ALD has to be developed from trimethylaluminum (TMA) and ozone (O3). The ALD characteristics of the TMA/O3 process should be compared, applying two different gas distribution plates in the gas shower head.

Sub-tasks:

1. Surveying the literature in comparison of the TMA/water with the TMA/O3 process
2. Planning of experiments
3. Familiarizing with the FHR ALD 300 tool, including the in situ ellipsometer as well as a selection of direct surface analysis techniques for referencing purposes
4. Conducting and minuting the planned experiments; evaluating the acquired ellipsometric spectra as well as referencing measurements
5. Interpreting, formulating and discussing the received results in writing Handed over on: October 16, 2012 Supervisors Responsible professor

SELF DECLARATION

I herby declare, that I have completed and submitted my student research project on

’Spectroscopic Ellipsometry for the in-situ investigation of Atomic Layer Depositions’ on this day. I also declare that this project report was independently written by me. And I have used no resources and quotations other than those specified in this report.

Dresden, April 30, 2014

Varun Sharma

LIST OF ABBREVIATIONS

illustration not visible in this excerpt

LIST OF FIGURES

3.4.1Schematic illustration of an ALD cycle

3.5.1complete ALD cycle averaged over 10 ALD cycles

4.2.1Schematic illustration of experimental setups of spectroscopic ellipsome- try. Linearly polarized light is incident to a sample substrate with incident angle of θ. The ellipsometry parameters Ψ and Δ, which represent the ratio of amplitudes of s- and p-polarized components of the incident light and the phase difference between them, respectively[ Fujiwara 2007, p.82 ]

4.3.1Schematic representation of a simple optical layered model with two films, parameterized by thickness (t1,t2) and optical constants(n,k) [ adapted from J.A. Woollam Co. 2011, p.3-43 ]

4.3.2characterization of physical properties by spectroscopic ellipsometry [ adapted from Fujiwara 2007, p.4 ]

5.2.1Diagram of photoelectric effect in XPS

7.1.1 Comparison of Type I (a) and Type II (b) shower head designs

7.1.2 Schematic diagram of ALD-300 chamber from FHR Analgenbau Gmbh with surface characterization and analysis system from Omicron Nan- otechnology Gmbh

7.4.1 course of the optical layer thickness as revealed by iSE time-discrete data acquisition over the Al2O3 ALD cycle number

7.4.2 Time-discrete and time-continuous iSE measurements for shower head design of type II

7.4.3 optical stack layer model for Al2O3 on Si substrate with native oxide

7.4.4 actual temperature versus target substrate temperature

7.4.5 Schematic representation of 10 ALD cycles

7.4.6 signal intensity of 250nm wavelength (red) and optical layer thickness (black) versus ALD process time

8.2.1kinetic ALD characteristic curves for shower head design type I and II at deposition temperature of 150°C and total process pressure of 150 Pa

8.2.2variation in TMA exposure time

8.2.3variation in Ar purging time after TMA exposure

8.2.4variation in O3 exposure for both type of shower head designs - a time- continuous irtSE measurement

8.2.5variation in Ar purging time after ozone exposure

8.3.1variation in total process pressure

8.3.2variation in O3 flow

8.3.3variation in deposition temperature

8.3.4XPS analysis at different deposition temperature

8.3.5effect of temperature on surface roughness

8.3.6AFM topography after 100 Al2O3 ALD cycles at different temperatures

8.4.1reproduced ALD cycle curves

LIST OF TABLES

A.1 actual substrate temperatures corresponding to set-point temperatures

A.2 O2 and O3 gas flows with their corresponding needle valve positions

B.1 process parameter variations

B.2 variations in TMA exposure time

B.3 variations in Ar purge time after TMA exposure

B.4 variations in O3 exposure time

B.5 variations in Ar purge time after O3 exposure

1 INTRODUCTION

Atomic Layer Deposition(ALD) is a special type of Chemical Vapor Deposition(CVD) technique based on self-terminating sequential gas reactions for a conformal and pre- cise growth down to few nanometers range. Ideally due to the self-terminating reac- tions, ALD is a surface-controlled process, where process parameters other than the choice of precursors, substrates, and deposition temperature have little or no influence [Puurunen 2005 P. 121301-22]. In spite of the numerous applications of growth by ALD, many chemical and physical processes that control ALD growth are not yet sufficiently understood [Jaebum 2002 P.733].

Aim of this student research project is to develop an Aluminium Oxide (Al2O3) ALD pro- cess from trimethylaluminum (TMA) and Ozone in comparison of two shower head de- signs. Then studying the detailed characteristics of Al2O3 ALD process using various measurement techniques such as Spectroscopic Ellipsometry (SE), x-ray photoelec- tron spectroscopy (XPS), atomic force microscopy (AFM). The real-time ALD growth was studied by in-situ SE. In-situ SE is very promising technique that allows the time- continuous as well as time-discrete measurement of the actual growth over an ALD process time. The following ALD process parameters were varied and their inter-dependencies were studied in detail: exposure times of precursor and co-reactant as well as Argon purge times, the deposition temperature, total process pressure, flow dynamics of two different shower head designs. The effect of varying these ALD process param- eters was studied by looking upon ALD cycle attributes. Various ALD cycle attributes are: TMA molecule adsorption (Mads), Ligand removal (Lrem), growth kinetics (KO3) and growth per cycle (GPC).

2 ALUMINA IN ELECTRONIC INDUSTRY

Al2O3 is the most common form of Aluminum oxides that occur naturally in its crys- talline polymorphic phase α-Al2O3 known as corundum. Al2O3 has relatively high melt- ing of 2,345 K. It is resistant to attack by many chemicals and gases. It is also a good electrical insulator but has a very good thermal conductivity[ Patnaik 2003, p.11 ]. Al2O3 is widely accepted in many industries like Power Electronics, Optics, Semiconductor, Mechanical, Memories and Computing. In thin film technology Al2O3 is deposited in its amorphous phase.

Al2O3 is a very important material in high performance electronics[ Perrella 2004, p.2 ]. Al2O3 is used in various fields of microelectronic industry. It is used as a gate-oxide in transistors, dielectric in dynamic Random Access memory (DRAM), fast computing applications and in making passive components.

Moreover, at 1 nm thick SiO2 the leakage current becomes unacceptable therefore there was a strict need to substitute SiO2 by high-k and metal oxides according to ITRS road map to follow the current moore’s law [ Manchanda u. a. 2001, p.351 ]. Al2O3 comes under high-k category which makes it suitable for gate dielectric in metal oxide semiconductor field effect transistors (MOSFETs).

The development of dynamic Random Access memory (DRAM) capacitor dielectrics has been similar to that of gate dielectrics. The idea behind this is to use Al2O3 in elec- tronic industry as semiconductor-insulator-semiconductor (SIS) and metal-insulator- metal (MIM) capacitances. Capacitance can be defined as C=ϵrϵ0 A/d and this can be increased by using high-k values, rest are constants for particular geometry[ Yeong Kwan Kim u. a., p.370 ]. The ϵr value for Al2O3 is 10 (for vacuum ϵr=1) which is approxi- mately 2.5 times higher than conventional Silicon oxide [ Jakschik u. a. 2003, p.353 ].

It is also used as an insulator in magnetic tunnel junctions (MTJs) used for magnetic random access memory (MRAM) which may allow the fabrication of solid state Q-bits based on josephson junctions (JJs)[ Mizuguchi u. a. 2005, P.1 ][ Zhu u. Park 2006, P.36 ]. To achieve high performance computing, silicon oxide was replaced by many high-k di- electrics like Al2O3[ Manchanda u. a. 2001, p.351 ]. According to the property of Joseph- son junctions to perform switching at very high speed of few THz which can be used in high performance computing [ Hasuo 1992, p.21 ] and fast switching memories[ H. Suzuki u. a. 1989, p.783 ].

Al2O3 is used as a passivation layer on silicon, the ceramic form is often used as a carrier for thick and thin film passive components and for other small printed circuit boards (PCB). It has relatively low production cost that is combined with its high thermal conductivity[ Bernardi u. a. 2004, p.475 ]

3 ATOMIC LAYER DEPOSITION

3.1 HISTORY

ALD principle was first published under name ”Molecular Layering” in the early 1960s by Prof. S.I. Koltsov from Leningrad Technological Institute. The concept of the ALD process was first proposed by Prof. V.B. Aleskovskii in his Ph.D. thesis published in 1952 [ Puurunen 2005 ]. It was the work of Dr. Tuomo Suntola and coworkers in Finland in mid-1970s that made a scientific idea a true thin film deposition technology and later on adopted by an industrial use. Since then it is widely used in manufacturing industries as well as research institute[ Nalwa 2002 ].

3.2 PROCESS DEFINITION

ALD is one of the thin film deposition technique that is based on sequential use of self- terminating surface reactions[ Puurunen 2005, p. 121301 - 3 ]. A material of desired thickness can be grown with ALD by repeating the following characteristic four pulses:

1. Exposure of the first precursor, typically an metal-organic compound (such as TMA).
2. Purge or evacuation of the reaction chamber to remove the non-reacted precursor molecules and the gaseous reaction by-products.
3. Exposure of the second co-reactant(such as Ozone) or another treatment, to acti- vate the surface again for the reaction of the first precursor.
4. Purge or evacuation of the reaction chamber [ Puurunen 2005, p. 121301 - 3 ] Ideally, each reaction cycle adds a monolayer (ML) of material to the surface, referred to as the growth per cycle GPC. But practically it is not the case: Most ALD processes do not fulfill these criteria[ Puurunen 2005, p. 121301 - 38 ]. Figure 3.4.1 shows a pictorial representation of detailed ALD process and how the depositions occur and reactant by products leave the surface.

3.3 BENEFITS AND LIMITATIONS

In ALD, the thickness of grown material can be controlled very precisely because in the ideal case the film thickness depends only on the number of reaction cycles. Deposi- tions done with an ALD has a better step coverage and side wall deposition. ALD tech- nique is capable to grow very thin films and film properties can be reproduced. ALD allows high density, low defect density and conformal coatings. Practically, the physical state of films like amorphous and crystallinity can be controlled by varying the process parameters like substrate temperature [ cambridge nanotech July ].

ALD gives a choice to grow heterogeneous, graded index and doped layers of differ- ent materials. Due to availability of various precursors in the market, it is possible to grow almost all materials used in semiconductor industry [ Nalwa 2002, p. 116 - 119 ] [ Puurunen 2005 ].

Other ALD techniques like spatial ALD, roll-to-roll and batch processing can be incorporated to achieve high throughput of wafers per minute. Thus several wafers can be deposited with very fine thin films of different materials [ Nalwa 2002, p.108 ]. Also, in combination of PE-ALD and Flash-ALD, depositions at lower temperature can be achieved esp. for substrates like polymers. Films grown with the help of ALD can be of low stress due to molecular self-assembly.

The major limitation of ALD is that it is time consuming as compared to other widely used deposition techniques like physical vapor deposition (PVD), CVD and molecular beam epitaxy (MBE). Although we have wide range of materials available that can be grown by ALD, many important materials like Si, Ge, Si3N4, several multi-component ox- ides, certain metals cannot currently be deposited by ALD in a cost-effective way [Ritala 2002 p.108]. ALD is technique based on chemical reactions that occur on the surface of substrate, thus there is always a risk of residues being left from the precursors specially carbon or hydrogen.

3.4 ALD GROWTH MECHANISM OF ALUMINIUM OXIDE FROM TMA/O3

Al2O3 depositions were performed on crystalline silicon (c−Si) as a substrate. In subfig- ure 3.4.1 a, -OH terminated dangling bonds on the surface are shown that allows TMA molecules to form a chemical bond on pulsing of TMA. It was also found that TMA frag- ments only stick permanently to the surface after protons from the surface migrate to TMA and eliminate one or more ligands as CH4. The resulting formation of an Al-O bond is the first step in film growth[ Elliott u. a. 2006 ]. In figure 3.4.1 b, the byproduct was methane molecule also was confirmed in [ Goldstein u. a. 2008, p.19530 ], after reacting with -OH group from the surface. Therefore creating the self terminating re- action until the whole substrate surface was left with no -OH groups; this is the case in ideal conditions and with optimised process parameters. Moreover, excessive TMA molecules were left non-reacted on the surface or on the walls of the chamber. Later that can be removed easily by Ar purge along with methane by products. In figure 3.4.1

Figure 3.4.1 Schematic illustration of an ALD cycle

a pulsing of TMA precursor
b adsorbed TMA molecules on -OH terminated substrate sample
c pulsing of Ozone as a co-reactant
d after reaction of Ozone molecule with CH3 ligands of surface absorbed TMA molecule
e Al2O3 monolayer (ML) formation after one complete ALD cycle
f Al2O3 deposited after several ALD cycles

[ adapted from cambridge nanotech July ]

3.4 ALD growth mechanism of Aluminium oxide from TMA/O3

c, the second reactant ozone was pulsed that acts as oxygen source. Thus after reac- tion with the monolayer of TMA adsorbate, a layer that looks like in figure 3.4.1 d was obtained. This mechanism was proposed by [ Goldstein u. a. 2008, p.19537 ] along with other reaction by products like carbonate, formate complexes. After the accomplish- ment of step shown in figure 3.4.1 d, second Ar gas purging was ensured to remove all by products and to obtain a nice Al2O3 monolayer shown in figure 3.4.1 e. By perform- ing these pulses over and again, a fine Al2O3 thin film of desired thickness that looks similiar to figure 3.4.1 f can be obtained.

3.5 GROWTH KINETICS

illustration not visible in this excerpt

Figure 3.5.1 complete ALD cycle averaged over 10 ALD cycles

The complete ALD process includes many ALD cycles required to deposit the desired amount of Aluminum oxide. Every ALD cycle comprises of four steps as described in section 3.2. Figure 3.5.1 shows the amount of material evolving during one ALD cycle. In this figure every ALD step has been shown with different colors.

Starting with TMA exposure, marked as light green, TMA molecules get adsorbed on the substrate surface until there are no more adsorption sites available. This amount of adsorbed TMA molecules has been marked with an arrow showing ’Mads’, indicating the maximum molecules that can absorb on substrate at particular instance and with unique process parameter conditions. This is the reason for self-limitation in ALD.

Now after this TMA exposure, it is necessary to purge excessive molecules out of the chamber. This step has been shown in figure 3.5.1 with light yellow mark and is called as Ar purging after TMA exposure. As Argon gas was used to purge the chamber because it is inert in nature. It should also be stated here, that the amount of adsorbed TMA was not dropped in the figure because ideally Ar (due to its inert nature) purge cannot break the chemical bond formed during first step.

The ozone exposure has been shown as light blue in the figure 3.5.1. Ozone starts to react with TMA adsorbate and form various complex intermediates on the surface[ Goldstein u. a. 2008, p.19537 ]. This ligand removal from TMA adsorbate on ozone exposure is denoted by ’Lrem’. This curve is assumed to show an exponential decay, which can be approximately defined by a fitting function described in formula 3.3 and will be discussed later in section 7.3.3. Mathematically the growth kinetics described by ’KO3 ’ for TMAOzone reaction, is the decay rate of the exponential function shown in formula 3.3. The physical meaning of ’KO3 ’ is a ligand removal from TMA adsorbate on ozone exposure which have been shown before in figure 3.2.1 d. On completion of this step, all methyl ligands have been removed from the TMA adsorbed surface.

After accomplishment of this step it is now important to remove all the reaction byproducts out of the chamber. Thus the final step of Ar purging after ozone exposure has been followed to clear the chamber. Ar purge causes no damage to the recently formed Al2O3 on surface because Ar gas is inert in nature and Al-O bond is sufficiently strong to be damaged by Ar purge.

After completion of one ALD cycle, the amount of Al2O3 deposited is denoted as growth per cycle(GPC) that can be described as in formula 3.1, where Tthickness and Ncycles are total thickness of material deposited and total number of ALD cycles respectively. Cycle time of a complete ALD cycle is described mathematically in equation 3.2, where tTMA, tAr1, tO3, tAr2 and tcycle are TMA exposure, Ar pulse time after TMA exposure time, ozone exposure time, Ar pulse time after O3 exposure and complete ALD cycle time, respectively. Similarly, the two constants ’KTMA’ and ’KO3’ describe the growth kinetics for TMA absorption and ligand removal. Theoretically, adsorption of TMA molecules can also be described in terms of exponential growth with KTMA as one of its parameters but due to practical reasons and fast adsorption it was complicated to do the fitting pro- cedure. Therefore, in this project the focus was more on studying the growth kinetics for ozone.

illustration not visible in this excerpt

3.6 COMPARISON OF TMA/O3 AND TMA/H2O - A LITERATURE SURVEY

A brief literature review is devoted to this section, that covers only small part of the en- tire work done on Al2O3 from TMA and H2O/O3. According to previous research done one cannot really tells which co-reactant (O3/H2O) is better because it depends on the chamber design, quality of both the precursors etc. But one can obtain the same re- sults with both the precursors by tweaking other parameters like temperature, pressure etc.

Various growth mechanisms for TMA/H2O and TMA/O3 have been proposed by many researchers. In [ Delabie u. a. 2012 ] the growth mechanism for TMA/H2O was proposed. where [ Elliott u. a. 2006 ] [ Goldstein u. a. 2008 ] proposes the mechanism for TMA/O3. In case of TMA and H2O the growth is inhibited from the second reaction cycle and then increases back to the steady value (4 Al atoms/nm² ) after approximately 20 reac- tion cycles[ Delabie u. a. 2012, p.3 ]. It has been also shown in [ Soto 1991, p.2695 ] that there are two phases of growth procedures for TMA/H2O so called interface and bulk formation. On the other hand, for TMA/O3 there is no island formation but layer by layer growth mechanism. It also does not show any inhibited first reaction cycles[ Park u. a. 2003 ].

ALD experiments shown in [ J. Spiegelman u. J. Sundqvis 2011, p.25 ] assure the re- placement of ozone with water vapor is very efficient, self-limiting and increased the film growth rates by 14.5%. Saturation pulse times were similar between the two oxi- dants as this is directly related to gas flow dynamics in the specific ALD process cham- ber. However the company RASIRC equipment was used in this case to generate ultra pure water vapors [ J. Spiegelman u. J. Sundqvis 2011, p.25 ]. High electrochemical po- tential of ozone results in fast reaction rates at relatively low temperatures as compare to water which is less reactive [ Hans sundstrom 2005, p.2 ]. Also a little enhancement in growth rate has been seen after treating the growing surface with excessive water (H2O) (meaning long water pulse time or increased water flow) throughout TMA/H2O process [ Elliott u. a. 2006 ]. However, the use of ozone as a second reactant does not significantly alter the growth rate when compared to water based Al2O3 ALD films [ Kim u. a. 2006 ].

It was also found that substrate temperature plays crucial role rather than a choice of oxygen source while describing Al2O3 film properties like film thickness uniformity, step coverage and growth rate. In case of O3-based Al2O3 films, step coverage was found to increase with increasing substrate temperature unlike H2O-based Al2O3 films, where the growth rate drops beyond 350°C. Moreover in case of TMA/O3, the step coverage was determined nearly 100% for the films deposited at a higher substrate temperature 450°C[ Kim u. a. 2006 ]. On the contrary, for both oxygen precursors, the molar growth rate per cycle decreases with increasing temperature, and at no temperature does the O3 process yield faster growth than the H2O process [Elliott u.a. 2006].

It is also important to do a comparison of both oxygen sources in terms of quality and defects in grown Al2O3 films. Al2O3 films grown from TMA and ozone had lower car- bon and hydrogen impurities as compared to TMA and water [ Goldstein u. a. 2008; Jakschik u. a. 2003; Kim u. a. 2006 ] because of absence of hydrogen in ozone molecule [ Hans sundstrom 2005 ]. It has been demonstrated that the Al2O3 films prepared with O3 have significantly less amount of defect states like Al-Al and -OH bonds compared to those prepared with water[ Goldstein u. a. 2008; Jakschik u. a. 2003; Kim u. a. 2006 ] [ Kim u. a. 2002, p.6742 ]. At low temperature, the interaction of H2O with Si seems to produce a thicker oxygen-rich interfacial layer, which adheres poorly to the Si substrate [ Jakschik u. a. 2003 ]. In contrary to, O3 as the oxygen source yields uniform films than H2O based Al2O3 films[Jakschik u.a. 2003] but lower-quality films than those grown from TMA/H2O; the films are less dense and rougher, especially at low growth temper- atures. However, use of O3 as oxygen source yields thicker films at low temperature, because of their lower density, correlated with the existence of voids in the alumina lay- ers [ Elliott u. a. 2006, p. 3764 - 3767 ]. By contrast, treatment of the growing surface with excess H2O in the TMA/H2O process ensures homogeneous surface -OH groups, a denser and smoother film, regardless of temperature.

Also the Al2O3 films prepared with O3 shows 1-2 order lower leakage current density compared with the Al2O3 film prepared with H2O. The wet etch(due to H2O as co- reactant) rate has been found to have little effect on Al2O3 films prepared using O3 thus they are more stable to environment conditions as compared to those prepared with water [ Kim u. a. 2002 ] [Ritala 2002 p.120].

Last but not least, lower TMA consumption were observed when using ozone than

when using water [ Elliott u. a. 2006 ]. Another reason to prefer ozone as reactant than water is that water molecules stick to the walls of the inside chamber and is hard to purge[ Elliott u. a. 2006, p.3764 ].

4 SPECTROSCOPIC ELLIPSOMETRY

4.1 INTRODUCTION

Ellipsometry was first developed and defined mathematically by Drude in 1887. Later this technology has been exploited very much since mid 19th century and still research is going on to make it better[ Fujiwara 2007, p.8 ]. Spectroscopic Ellipsometry is an op- tical technique used for analysis and metrology of the various samples like dielectrics, semiconductors, metals, superconductors, etc. The measured response depends on optical properties and thickness of individual materials. Thus, ellipsometry is primarily used to determine film thickness and optical constants. It is also applied to characterize composition, crystallinity, roughness, doping concentration, and other material prop- erties associated with a change in optical response. The name ’ellipsometry’ comes from the fact that polarized light upon reflection with sample becomes ’elliptical’ and spectroscopic comes from the measurements that are carried out in the wide range of electro magnetic (e.m.) spectrum (200nm to 1700nm approx.)[ Fujiwara 2007, p.1 ]. It has both capabilities for in-situ as well as ex-situ measurements. In situ is Latin for ”in the place” while ex-situ means ”off-site”. In situ measurements of the surface can mon- itor the layer-by-layer growth that occurs exactly in ALD process. In situ SE monitoring has been applied to many deposition techniques (e.g., MBE, MOCVD, sputtering, ALD etc.)[ Dale E. Morton u. a. 2002, p.2 ].

4.2 MEASURING PRINCIPLE

In Ellipsometry, a beam of linearly polarized light with known wavelength is incident onto a sample of interest as shown in figure 4.2.1. Upon reflection from surface of sam- ple, the polarization state of light is changed to an elliptically polarized light. This change in polarization state of light is measured by two parameters (Ψ, Δ as shown in figure 4.2.1) and are known as ellipsometry parameters or angles. This is the basic principle of Ellipsometry and is described beriefly as below.

The polarization state of light can be described by the coordinates of p-(parallel) and s-(senkrecht from German) polarizations. For convenience, the ratio ρ of the Fresnel amplitude reflection coefficients for p- and s- polarized light can be defined as ρ = rp/rs. Where rp, rs are the Fresnel reflection coefficients for the p- and s- polarized light wave, respectively. Both rp, rs are complex, so is the ratio ρ, which is expressed in terms of two ellipsometry parameters (Ψ, Δ) as follows[ Fujiwara 2007, p.81 - 84 ].

ρ ≡ tan(ψ).e iΔ (4.1)

The raw measurement is determined by two values, amplitude ratio called Psi (Ψ) and phase difference is denoted by Delta (Δ). Figure 4.2.1 illustrates the measurement prin- ciple of ellipsometry. Where linearly polarized incident light beam E falls on the sample at angle θ to the plane of incidence as shown in figure 4.2.1 and contains both parallel (Eip) and perpendicular (Eis) polarized wave components. Where Eis = Eip for linearly po- larized light, since the amplitudes of p- and s- polarizations are the same and the phase difference between the polarizations is zero. This incident light beam is reflected from the sample surface and becomes elliptically polarized with Ers and Erp reflected s- and p-polarized light beam components respectively. And it undergoes a significant change in its amplitude and phase for both p- and s- polarized states[ Fujiwara 2007, p.82 ]. The change in polarization is known as the ellipsometry measurement, commonly written as:

ρ ≡ (Erp/Eip)(Ers/Eis) (4.2)

For accurate Ellipsometry measurements, a linear polarized light wave is irradiated onto a sample at the Brewster angle (as the difference between rp and rs is maximized at brewster angle θB, hence the measurement sensitivity is maximum too).

This is the basic principle of ellipsometry measurement. Then measured (Ψ,Δ) values are used to calculate other optical parameters[ Dale E. Morton u. a. 2002, p.3 ][ Fujiwara 2007, p.40 ]. Particularly, when a sample structure is simple, the amplitude ratio Ψ is characterized by the refractive index (n), while Δ represents light absorption described by the extinction coefficient κ. In this case, the two values (n,κ) can be determined di- rectly from the two ellipsometry parameters Ψ and Δ obtained from a measurement[ Fujiwara 2007, p.84 ].

Figure 4.2.1 Schematic illustration of experimental setups of spectroscopic ellipsometry. Linearly polarized light is incident to a sample

substrate with incident angle of θ. The ellipsometry parameters Ψ

and Δ, which represent the ratio of amplitudes of s- and p-polarized components of the incident light and the phase difference between them, respectively[ Fujiwara 2007, p.82 ]

4.3 FITTING AND MODELS

SE software Models

SE is a model based analysis technique. In general models are classified into two types; optical layer stack (OLS) model and material model. OLS models consist of layers of different materials e.g. single layer, multi layer etc. with different film parameters e.g. as shown in figure 4.3.1 and these layers can be further elaborated with roughness and graded indices. OLS model represents the basic structure of the deposited material on sample substrate[ J.A. Woollam Co. 2011, p.3-42 ]. A specific OLS model depends on best assumption made according to a specific deposition process. For instance, in this research Al2O3 was deposited on silicon substrate (with natOx on it more detail in figure 7.4.3).

illustration not visible in this excerpt

Figure 4.3.1 Schematic representation of a simple optical layered model with two films, parameterized by thickness (t1,t2) and optical constants(n,k) [ adapted from J.A. Woollam Co. 2011, p.3-43 ]

The material model is a model that represents the type of material and its properties e.g. dielectric, semiconductor etc. According to the optical properties of sample an appropriate dielectric function model is selected. The various dielectric function mod- els are Lorentz, Sellmeier, Cauchy, Tauc-Lorentz, and Drude models. Basically, all these models are derived from the Lorentz oscillator model [ Fujiwara 2007, p.160 ].

Fitting procedure

Figure 4.3.2, shows the detailed characterization of physical properties by spectroscopic ellipsometry. It describes Psi and Delta, are the only two values determined directly by SE. In SE, the obtained raw data (Ψ,Δ) is meaningless without performing any mod- elling. The interesting characteristics such as film thickness, surface roughness and op- tical material parameters can be determined only by optically modelling the near-surface

Figure 4.3.2 characterization of physical properties by spectroscopic ellipsometry [ adapted from Fujiwara 2007, p.4 ] region of the sample[ Jellison 1998, p.1 ]. In general, three steps are involved while fitting SE data, to obtain valuable information on various thin film parameters. These steps are as follows:

1. Choosing an optical layer stack (OLS) model (OLS model specific to this research project is shown in figure 7.4.3). After obtaining a specific OLS model, it is re- quired to define the material model for each layer that provides the information on various optical parameters and absorption coefficient (α)[ Fujiwara 2007, p.161 ]. Various material models can be selected from the inbuilt library of CompleteEASE software.

2. After defining the model, fitting of the optical layer parameters like thickness and temperature (also shown in figure 7.4.3) is an important step. The CompleteEASE software runs some algorithms like Levenberg-Marquardt method[ J.A. Wool- lam Co. 2011, p.3-46 ] to improve the agreement between the measured and model generated SE data[ J.A. Woollam Co. 2011, p.3-42 ].

3. If the modelled data and generated data fits well upto specific interest. Then the fitting is acceptable and various information is evaluated. If the fit is not accept- able then it is required to find a suitable model or modify the fit parameters.

MSE definition

During optical modelling (SE data evaluation), the main goal is to obtain the minimum mean squared error(MSE). MSE describes describs how nicely the model-generated data fits the measured Ψ, Δ values. Mathematically, MSE is defined by equation 7.3 [ Himcinschi u. a. 2001, p.808 ].

illustration not visible in this excerpt

Where ’N’ is number of experimental (Ψ, Δ) pairs, ’M’ is number of fit parameters and

σ is the standard deviation on the experimental data points[ Himcinschi u. a. 2001, p.808 ]. The lower the MSE, the better is the fit agreement between generated and modeled data. So for a best fit, the MSE value is the lowest value possible[ J.A. Woollam Co. 2011, p.3-45 ]. For accurate determination of model parameters, MSE should be con- stant over the desired range of wavelength and should be minimum among various al- lowed OLS models for a particular deposition[ J.A. Woollam Co. 2011, p.3-44 ][ Dale E. Morton u. a. 2002, p.1 ].

4.4 ADVANTAGES AND LIMITATIONS

Ellipsometer provides numerous advantages over conventional techniques based on reflection and absorption. First it is ratio based technique that can be measured accurately even in less intense light conditions. Thus it is capable to measure precisely even if there are some intensity fluctuations in the source[ Dale E. Morton u. a. 2002, p.1 ]. Secondly, it is highly precise in measuring thickness with the sensitivity close to 0.1Å. It is fast in measurement and the delay can also be introduced (in case of triggered measurements). Thus it is flexible in nature. Third advantage is its in-situ capabilities and determining the real time measurements [ Fujiwara 2007, p.3 ].

Ellipsometry is used to measure a wide variety of thin films. There is no restriction

on the type of material measured: dielectrics, semiconductors, metals, organics, and composites of materials etc. provided that light reflects from the surface[ Dale E. Morton u. a. 2002 ]. It can measure an isotropic, anisotropic, homogenous and graded film coatings. Spectroscopic ellipsometry can even be applied to multilayer structures, with films of different materials. Moreover, substrate temperature, quality, absorption and desorption of material can also be measured[ B. Johs u. a. 1999 ]

However, SE has few limitations including necessity of a correct optical model for an- alyzing the data, hence data analysis is little bit complicated. In combination with ar- ray scan technique, it is possible to analyze the whole wafer surface area of 300mm or more but at the expense of time. With help of SE, it is difficult to characterize the op- tical constants for material with low absorption coefficients[ Siah u. a. 2013, p.452 ]. If the coatings are too rough, it scatters the probe beam away from the detector, which prevents spectroscopic ellipsometry measurements[ Dale E. Morton u. a. 2002 ]

The upper thickness limit for SE measurement depends on the wavelength used and type of material. For thick transparent films, many data oscillations become difficult to perceive at shorter wavelengths but are more resolvable in longer wavelength range. Thus, by using near infrared or even mid-infrared measurements one can measure films up to 50 microns thick. However, this range is beyond a typical SE measurement. The preferred limit for most visible-to-near infrared measurements is less than 5 microns. Even for films that are 1 to 5 micron thick, it is best to measure with multiple angles of incidence to have a unique thickness solution[ B. Johs u. a. 1999 ]. For metal and absorbing films, the upper detection limit is below 100 nm which is dependent on extinction coefficient(κ).

5 X-RAY PHOTOELECTRON SPECTROSCOPY

5.1 INTRODUCTION

XPS is a kind of electron spectroscopy in which a sample is irradiated with a beam of mostly monochromatic x-rays and then measuring the energies of resulting photoelec- trons. In 1887, the photoelectric effect was discovered by Heinrich Rudolf Hertz and in year 1905 explained by Albert Einstein. In year 1954, a Swedish physicist Kai Sieg- bahn recorded the first high-energy-resolution XPS spectrum of cleaved sodium chlo- ride (NaCl). A few years later in 1981, Siegbahn won Nobel Prize for developing the method of electron spectroscopy for chemical analysis (ESCA), now known as XPS [ Skoog 1985, p.538 ].

5.2 XPS MECHANISM

Surface analysis by XPS is accomplished by irradiating mono-energetic x-rays and analyzing the electrons emitted from the surface. Normally, MgKα and AlKα x-ray sources are used with 1253.6 eV and 1486.6 eV energies, respectively. These x-rays penetrate the sample surface and interact with atoms on the surface by photoelectric effect. The emitted photoelectrons have kinetic energy given by equation (5.1)

Ek=hν − Eb − φ (5.1)

Where, hν is energy of exciting photons and Eb is binding energy of atomic orbital from which electron originates and φ is known as work function [ Wagner u. Muilenberg 1979, p.4 ]. In the XPS detector the (Ek) kinetic energy and number of electrons that escape from the top (1 to 10 nm) layer of material is measured. A schematic shown in figure 5.2.1 depicts the photoelectric effect.

5.3 XPS ANALYSIS

XPS spectrum is a plot of the number of electrons detected (ordinate) versus the bind- ing energy of those detected electrons (abscissa). Each element produces a character- istic set of XPS peaks at characteristic binding energy values that corresponds to each element that exists in or on the surface of the material that is being analyzed. These characteristic peaks correspond to the electron configuration of the electrons within the atoms, e.g., 1s, 2s, 2p, 3s, etc. The number of detected electrons in each of the charac- teristic peaks is directly related to the amount of element within the irradiated area (vol-

illustration not visible in this excerpt

Figure 5.2.1 Diagram of photoelectric effect in XPS

ume). To generate atomic percentage values, each raw XPS signal must be corrected by dividing its signal intensity (number of electrons detected) by a "relative sensitivity factor" (RSF) and normalized over all of the elements detected.

To count the number of electrons at each kinetic energy value, with the minimum error, XPS must be performed under ultra-high vacuum (UHV) conditions because electron counting detectors in XPS instruments are typically one meter away from the material irradiated with X-rays. It is important to note that XPS detects only those electrons that have actually escaped within the vacuum of the instrument. The photo-emitted elec- trons that have escaped into the vacuum of an instrument are those that originated from within the top 10 to 12 nm of the material layer. For most applications, in-vacuo XPS analysis is efficient and a non-destructive technique (no sputtering is required in case of in-vacuo sample transfer right after finishing the process) that measures the surface chemistry of any material.

5.4 ADVANTAGES AND LIMITATIONS

It is quantitative and surface chemical analysis technique that measures the elemen- tal composition, empirical formula, chemical state and electronic state of the elements that exist within a material. It is non destructive surface analysis technique where the sample surface is not altered. The biggest advantage is its surface and element sen- sitivity (parts per 1000)[ Andrade 1985 ]. It is easy to understand while the instrument works with a direct interpretation of its theory. XPS is widely used to analyze inorganic as well organic compounds, semiconductors, polymers, ceramics, bio-materials and many others. XPS is very reliable technique that provides with high information content for samples under test[ Andrade 1985, p.106 ].

5.4 Advantages and limitations

However, there are few limitations. For effective analysis, samples are analyzed in Ultra high vacuum (UHV, P ≤ 10−⁷ Pa). The experimental setup itself is very costly and must be handled with care. The instrument is complex in design. The biggest disadvantage is that we cannot detect hydrogen and helium atoms[ Andrade 1985, p.106 ]. For thorough analysis of some samples it can be very slow. Sometimes the charge on the sample surface can be a problem that may shift the peaks in either directions. This can generate some confusion while analyzing some elements.

6 ATOMIC FORCE MICROSCOPY

AFM is a very high-resolution type of scanning probe microscopy, with demonstrated high resolution on the order of fractions of a nanometer. AFM consists of a cantilever with a sharp tip at its end that is used to scan the test sample surface. The cantilever is typically silicon or silicon nitride with a tip radius of curvature on the order of few Angstroms. When the tip is brought into proximity of a sample surface, forces between the tip and the sample lead to a deflection of the cantilever according to Hooke’s law. Depending on the situation, forces that are measured in AFM include van der Waals forces, electrostatic forces, magnetic forces, Casimir forces etc. Typically, the deflec- tion is measured using a laser spot reflected from the top surface of the cantilever into an array of photodiodes. Other methods that are used include optical interferometry, capacitive sensing or piezoresistive AFM cantilevers[ Binnig u. Quate 1986 ].

AFM gives topographical scan of a sample surface. The micro and nano-scale features of the sample can be observed, portraying the roughness of the material. If the tip was scanned at a constant height, a risk would exist that the tip collides with the surface, causing damage. Hence, in most cases a feedback mechanism is employed to adjust the tip-to-sample distance to maintain a constant force between the tip and the sam- ple. Traditionally, the sample is mounted on a piezoelectric sensor that can move the sample in the z direction for maintaining a constant force, and the x and y directions for scanning the sample.

The AFM can be operated in a number of modes, depending on the application. In gen- eral, possible imaging modes are divided into static (also called contact) modes and a variety of dynamic (non-contact or ”tapping”) modes where the cantilever is vibrated throughout the measurement. In non-contact mode, the tip of the cantilever does not contact the sample surface [ Zhong u. a. 1993, p.L688 ]. The cantilever is instead os- cillated at either its resonant frequency (frequency modulation) or just above (ampli- tude modulation) where the amplitude of oscillation is typically a few nanometers (<10 nm) down to a few picometers [ Gross u. a. 2009 ]. The van der Waals forces, which are strongest from 1 nm to 10 nm above the surface, or any other long range force which extends above the surface acts to decrease the resonance frequency of the can- tilever. This decrease in resonant frequency combined with the feedback loop system maintains a constant oscillation amplitude or frequency by adjusting the average tip-to- sample distance. Measuring the tip-to-sample distance at each (x,y) data point allows the scanning software to construct a topographic image of the sample surface[ Binnig u. Quate 1986 ].

II EXPERIMENTAL PART

7 METHODOLOGIES

7.1 EXPERIMENTAL SETUP

All experiments in this student project were performed on a top injection ALD-300 reac- tor type from the company FHR Anlagenbau Gmbh and are described in detail in chap- ter 8. Figure 7.1.2 shows the ALD-300 tool, that consists of two process chambers, a load lock and a wafer transport system[ more details in Schmidt u. a. 2008, p.528 ]. All chambers are separated from each other with the help of vacuum locks that are con- trolled pneumatically. The ALD chamber is a warm wall reactor with shower head for exposure of gases. Shower head type of exposure is used because it provides with a uniform flux flow-inside the reaction chamber[ Kim u. a. 2003, p.734 ].

Two kinds of shower head designs were used to study the influence of flow dynamics on pulsing and purging times. Better flow dynamics may lead to faster growth with bet- ter conformality, homogeneity, quality and less contaminations[ Kim u. a. 2003, p.734 ]. In general (see figure 7.1.1), both shower head designs implemented two different gas channels to keep the reactants (TMA and ozone) separate and are entwined in a comb like structure. A slight continuous Argon gas flow is used to prevent any kind of deposi- tion occuring inside these channels.

Type I shower head design has less number of pin holes per each gas channel. In figure 7.1.1 a, it can be seen very clearly that all the holes lie in the middle of channels. While on other hand in figure 7.1.1 b, Type II shower head design has holes almost twice in number as compared to Type I. Also the holes are not in the middle of channels but on the side walls, forming like a zigzag pattern. The results for Al2O3 ALD depositions for both shower head designs will be discussed in chapter 8.

a back side of Type I shower head b back side of Type II shower head

design design

illustration not visible in this excerpt

Figure 7.1.1 Comparison of Type I (a) and Type II (b) shower head designs

Various measurement equipments are attached to both ALD chambers. This includes spectroscopic ellipsometer (SE), X-ray photoelectron spectroscope (XPS), UV photo- electron spectroscope (UPS) and an atomic force microscope (AFM). The SE (shown in figure 7.1.2 with colored rectangular blocks; red being the light source and blue as de- tector part) is attached to the ALD metal chamber for in-situ access. Direct surface anal- ysis is done with the help of XPS and AFM after sample transfer within a high vacuum (≈ 10−⁵ Pa).

Figure 7.1.2 Schematic diagram of ALD-300 chamber from FHR Analgenbau

Gmbh with surface characterization and analysis system from Omicron Nanotechnology Gmbh

7.2 ALD PROCESS

During all experiments, a substrate of crystalline silicon (c−Si) with native oxide (na- tOx) was used. The size of substrate samples were approximately 1cm × 1cm. The substrate samples were cut into small coupons from original 8 inch wafer. Al2O3 de- positions were performed by thermally enhanced ALD technique. The surface reac- tions were accomplished by two reactants; TMA and Ozone. In each experiment 100 ALD cycles were performed. Argon gas was used to purge the chamber and as a car- rier gas for TMA precursor. All gases except ozone were controlled by MFCs. However, O3 flow was fixed to a particular value with the help of manually controlled needle valve because over time O3 can damage the built-in MFC. TMA-bubbler temperature (ϑTMA) was kept at 16 °C according to the specifications provided by the vendor. Substrate heating was controlled with the help of resistive heater attached at the bottom of sub- strate holder. To obtain the desired substrate temperature (ϑSubstrate) pre-heating was always done at higher pressure that is 300Pa for 15 mins to fasten the heating process and then stabilizing the temperature by reducing the pressure to process conditions (e.g. 200Pa, 150°C)[ more details in appendix and Junige u. a., p.2 ].

7.3 EXPERIMENT DESIGN

In order to investigate and study the interdepency of ALD process parameters, the fol- lowing process parameters were varied: TMA exposure time, Argon purge time after TMA exposure, ozone exposure time, Argon purge time after ozone exposure, respec- tively. These exposure and purging variations were helpful to study the optimum cycle time for one complete ALD cycle. This optimum cycle time is necessary to ensure the saturated monolayer formation of TMA adsorbate, minimum time required to purge the excessive TMA molecules and reaction by-products from chamber, minimum ozone ex- posure time required to have a complete reaction with TMA adsorbate on the surface and minimum time required to purge the reaction by-products. Varying the exposure and purge times of both reactants provide the kinetic ALD characteristic curves (dis- cussed in 8.2.1 - 8.2.4 subsections), i.e. the saturation of both ALD half-reactions and the purging behaviour, respectively. It also gives the minimum amount of precursor needed. The GPC tends to saturate at a point where no chemisorption is possible any- more. In an ideal case, further extension of exposure or purging times (keeping all other process parameters constant) will not improve the GPC or quality of thin films[ Nalwa 2002, p.140 ].

Other process parameters that were varied are total process pressure, O3 gas flow and deposition temperature. Studying the temperature characteristics plays a crucial role in the development of an ALD process, as it reveals the temperature window in order to achieve a saturated and thus stable GPC. Finally, the effect of varying above mentioned process parameters and their interdepencies were studied by taking various ALD char- acteristic curve attributes into account. The ALD characteristic curve attributes are GPC, growth kinetics for ozone reaction step, ligand removal and TMA adsorption. For in- stance, the effect of varying substrate temperature was noticed by recording the GPC, TMA adsorbtion and growth kinetics for ozone step(see figure 8.3.6). For studying the impact of various process parameters on an ALD process, only one parameter was var- ied at a time by keeping others constant (see Appendix) [ as described in Nalwa 2002, p.139 ]. All other process parameters were kept constant in order to achieve similar pro- cess conditions inside the reaction chamber.

As one of the most important ALD characteristic curve attribute, the growth per cycle (GPC) has been defined as the film thickness increment resulting from the application of one single ALD cycle (i.e. the exposure and purging of precursor and co reactants, respectively) [ Puurunen 2005 ]. However in recent research, commonly a cummulative GPC has been determined by total film thickness over the total number of ALD cycles applied for the deposition of thicker measurable film. Nevertheless in this report, a real GPC measurement was incorporated by studying a film thickness increment after com- pletion of each ALD cycle. This is one big advantage of irtSE measurements reported in this research project.

Temperature was varied from 100°C to 500°C in steps of 100°C. This temperature was a set point temperature but the actual surface temperature (ϑSurface) was obtained by fitting in-situ SE data as shown in table A.1 (in section A of Appendix). The chamber pressure was varied from 100Pa to 500Pa in accordance with table B.1. The ozone flow was varied from 50sccm to 500sccm [as described in table A.2]. Ozone flow was con- trolled with the help of a needle valve by observing the chamber pressure. The effect on chamber pressure was noticed by changing the oxygen flow (in sccm) with the help of a needle valve position. Ozone flow was assumed to show the same effect on cham- ber pressure. Thus Ozone flow was indirectly controlled and set to its corresponding value [as shown in table A.2] for a particular set of process parameters to accomplish one ALD process.

However, in most ALD processes a compromise between growing a desirable amount of material with acceptable quality and given time as well as temperature regime has to be found. As the ALD characteristic curves themseleves do not include any information about the film quality. A further characterisation (for example by XPS, AFM, etc.) of film properties is needed.

7.4 SPECTROSCOPIC ELLIPSOMETRY

7.4.1 Tool and software

A Rotating Compensator Ellipsometer (RCE) of type M-2000 FI from the company J. A. Woollam Co. Inc, was mounted right on the ALD metal chamber. RCE is one of the possible configurations in SE with many advantages over other configurations like Ro- tating Analyser Ellipsometer (RAE). RCE has uniform sensitivity over an entire Ψ(0◦ − 90◦) and Δ(0◦ − 360◦) range[ Aspnes u. Hauge 1976, p. 949 - 953 ] and thus, measures these both ellipsometric angles more accurately. Type M-2000 FI ellipsometer covers 673 wavelengths of spectral range from 190nm to 1690nm[ J.A. Woollam Co. 2011 ].

Figure 7.1.2 shows the red block mounted on metal chamber at 64◦ of angle of inci-

dence (AOI), which is a 75 Watts Xe arc lamp light source (FLS-350) and blue is detector part of SE tool. While measuring in-situ data, the light beam passes through a special pre-strained glass windows with no shutters installed. A continuous Ar gas flow prevents the windows from any undesired depositions.

The CompleteEASE (ver. 4.90) software was used in these experiments to measure and evaluate the in-situ data. The details of the analysis procedure can be found in the literature [ J.A. Woollam Co. 2011; Jakschik u. a. 2003; Kim u. a. 2006 ].

7.4.2 Data acquisition

A CompleteEASE software that runs on a local computer acquires in-situ data from the SE hardware through a TCP/IP link. Before measuring the in-situ data, a system check was performed once. A system check is a most common method for calibrating the el- lipsometer[ J.A. Woollam Co. 2011, p.12-358 ]. A system check is a software automated process that calculates a coarse AOI, includes in- and out-of-plane window effects and write back to hardware settings. This procedure automatically corrects the acquired spectra of Ψ and Δ. SE data acquisition can be classified in to two broad categories as described below.

Time-discrete data acquisition

In time-discrete data acquisition mode a special kind of synchronisation device between the ALD tool and CompleteEASE software is required that allows the triggering of data acquisition at the beginning of each purging step.

illustration not visible in this excerpt

Figure 7.4.1 course of the optical layer thickness as revealed by iSE time-discrete

data acquisition over the Al2O3 ALD cycle number

In chapter 8, two shower head designs are compared in terms of their purge behaviour, by employing time-discrete SE measurements. Figure 7.4.1 plots the evaluated opti- cal layer thickness in the discretized course of the Al2O3 ALD cycle number. Blue and red data points show the measurements that were triggered to the Ar purge right after TMA and Ozone exposures, respectively. Red data points depict the optical layer thick- ness after completion of the O3 exposure (second ALD half-reaction). In other words, red data points show the amount of Al2O3 deposited upon completion of each ALD cy- cle. From figure 7.4.1, the GPC can be easily extracted by just subtracting the two con- secutive red data points.

Time-continuous data acquisition

Evolution of the film thickness for ALD grown Al2O3 over the process time was also monitored with the help of in-situ real-time SE (irtSE) in a time-continuous manner. In time-continuous acquisition, each data point was acquired with a sampling rate of 0.8 s. This sampling rate was enough to measure the whole data set values through out a complete ALD process time. The time-continuous acquisition can also be selected either for fast or high accuracy mode. For high accuracy with standard configuration mode the polarizer moves to two positions and measures raw data accurately by can- celling the alignment issues[ J.A. Woollam Co. 2011, p.11-347 ]. Also it takes twice as much time per measurement but significantly improves the accuracy. SE tool allows to monitor the various film properties like substrate temperature, thickness etc in real time. Figure 7.4.5 a plots the change in optical layer thickness versus the time continu- ous ALD process time for 10 ALD cycles. After acquiring data the next step is to evalu- ate it and is described in section 7.4.3.

7.4 Spectroscopic Ellipsometry

Comparison of time-discrete and time-continuous data acquisition

Figure 7.4.2 exploits time-discrete and time-continuous data acquisition capabilities of iSE. For time-continuous SE measurement, a decremental of optical layer thickness (in- verted blue region in figure 3.5.1) was plotted to compare with GPC saturation curve for O3 exposure time that was obtained from time-discrete SE measurements i.e. for each data point the thickness slope was calculated after applying 10 ALD cycles. Figure shows that growth kinetics can be revealed for both type of measurements. The both Kinetic ALD characteristic curves overlay on each other showing a saturation around 55

seconds of O3 exposure time. However, one big advantage of time-continuous SE measurement is that it has more data points (acquiring each data point almost every one second) in shorter time as compare to time-discrete SE measurement and thus providing more information.

illustration not visible in this excerpt

Figure 7.4.2 Time-discrete and time-continuous iSE measurements for shower head design of type II

7.4.3 Data evaluation

For all experiments done in this project, Al2O3 cody-lorentz oscillator material model for low absorbing and amorphous Al2O3 layer was chosen to describe the optical con- stants of film more precisely[ Fujiwara 2007, p.172 ][ Ferlauto u. a. 2002 ]. Al2O3 layer was assumed to have isotropic, temperature and depolarization independent behaviour. Surface roughness was also neglected according to in-vacuo AFM measurements (see figure 8.3.9). A NTVE_JAW [ from Herzinger u. a. 1998 ] material model for native SiO2 was chosen. Also in order to get the information regarding the actual Si-surface temper- ature, Si_TEMP_JAW(Temp_Library) [proposed by Jellison and modine] material model was selected from the software because it has the temperature as a fit parameter. A relationship between the actual Si-surface and set point temperature has been plotted in figure 7.4.3 [ more details in appendix and Junige u. a., p.2 ]. Figure 7.4.4 shows a difference between a set point temperature and the actual ϑSurface which was obtained by fitting in-situ SE data.

illustration not visible in this excerpt

Figure 7.4.3 optical stack layer model for Al2O3 on Si substrate with native oxide

illustration not visible in this excerpt

Figure 7.4.4 actual temperature versus target substrate temperature

7.4.4 Post processing of data

Figure 7.4.5 a shows the evaluated optical layer thickness for 10 ALD cycles. Figure

7.4.5 b, indicates the start of each ALD cycle. Calculating the slope or so called first derivative of the thickness curve at each data point gives the film growth rate. Whenever the growth rate exceeds a specified threshold value (in this case 2 nm/min), the beginning of a new ALD cycle is indicated correspondingly and the ALD cycle index is incremented by 1 as shown in figure 7.4.5 c.

Each irtSE experiment in this report was performed with 100 ALD cycles. But actually only last 10 ALD cycles of the homogeneuos growth region were superimposed to ob- tain one fine averaged optical layer thickness curve for progression of one ALD charac- teristic curve. Averaging reduces some random errors that will be described in section

7.4.5. This ALD averaged ALD characteristic curve was used in all other data evaluation processes.

7.4 Spectroscopic Ellipsometry

Figure 7.4.5 Schematic representation of 10 ALD cycles

a optical layer thickness of Al2O3 for continuous 10

ALD cycles with ALD process time

illustration not visible in this excerpt

ALD process time (min)

illustration not visible in this excerpt

ALD process time (min)

c ALD cycle index verus ALD process time

illustration not visible in this excerpt

ALD process time (min)

Kinetic ALD characteristic curve attributes were studied by analysing the time-resolved results obtained from irtSE. As described before, an averaged ALD characteristic curve was obtained from 10 continuous cycles in order to minimize the noise occured while acquisition of data. ALD characteristic curve atrributes; Mads was evaluated by taking a maximum value in an ALD characteristic curve, Lrem was evaluated by subtracting the minimum value from blue region in figure 3.5.1 from Mads and GPC was the global min- ima of fourth step in a complete ALD cycle (i.e. second Ar purge region in figure 3.5.1). In order to obtain the kinetic parameter of O3 reaction step ’KO3 ’ more accurately, a fit- ting function shown below was chosen

y =y0 + A.exp(−KO3t) (7.1)

where ’t’ is ALD process time, y0 and A are fit constants. This is an inbuilt exponenetial decay function available in mathematical software called OriginPro (Version 9.0). The fitting was only done for the blue region in figure 3.5.1 which indicates the O3 reaction step.

7.4.5 Sources of errors in SE

In general, there are always three broad categories of errors present in any analytical or measurement technique. These are hardware related errors, errors due to software or optical model and human errors. Generally in ellipsometry, errors caused by some failures in hardware, misalignment of optical elements, sample misalignments, sub- strate backside reflections etc. can propagate into the final parameter evaluations and must be taken into account. Software errors can be caused due to wrong algorithms in software, errors in model[ J.A. Woollam Co. 2011, p. 9-325 - 9-326 ] etc. Assuming, the software uses an accurate theory behind ellipsometry, this kind of errors can not be reduced without the help of manufacturer’s support. Human errors arise due to lack in knowledge or unintentional mistakes caused during evaluations and can be minimised by increasing the conceptual knowledge and experience.

In SE, hardware errors are either random or systematic. Random hardware errors are generally small, because SE measures intensity ratios. Hence, errors caused due to fluctuations in the absolute light lamp intensity or incomplete collection of light from the sample do not result in increase in the error[ Tompkins u. Irene 2005, p.474 ]. Con- sequently, SE measurements are very precise. This precision can be increased by aver- aging of several experiments done under similar process conditions, and then calculat- ing the standard deviation (SD) and mean[ Tompkins u. Irene 2005, p.266 ] (i.e. increas- ing the integration time). Thus making in-situ SE measurements more precise.

7.4 Spectroscopic Ellipsometry

Systematic hardware errors are probably the most important, but hard to identify. They are bigger in magnitude and are more significant as compared to random hardware er- rors. Systematic errors occur due to misalignment of optical elements, calibration drifts, angle of incidence and depolarization effects and are SE equipment dependent. One way to minimize these errors is to incorporate some parameters as fitting parameters like angle of incidence offsets, window effects and depolarization effects[ Tompkins u. Irene 2005, p.475 ].

In detail, misalignment of optical elements can cause errors in measuring (Ψ, Δ) values. Errors in Ψ can be due to misalignment in static or rotating elements like input polar- izer, rotating compensator and analyzer (i.e. polarizer on detector side), respectively. In RCE type of configuration, error in Ψ can be minimized by performing measurements in high accuracy mode or by adding Ψ offset. While performing measurements in high accuracy mode, the analyzer spins to two different positions and eliminates the error in

Ψ. While performing fast acquisitions, systematic errors can be reduced by adding Ψ offset in model options during post-data evaluation. On the other hand, error in Δ can occur due to undesired material deposition on window glasses, or due to slight thermal drifts compared to system calibration conditions. Windows at the FHR-ALD-300 tool are purged with Ar continuously, in order to keep them free from any depositions and thus, minimizing the measurement error in Δ. Temperature drifts can be optically modelled by adding some offset in Δ [ Johs 2000 ].

illustration not visible in this excerpt

ALD process time(min)

Figure 7.4.6 signal intensity of 250nm wavelength (red) and optical layer

thickness (black) versus ALD process time

Other source of systematic error can occur due to improper selection in range of wave- lengths. As one of our precursors was ozone. And ozone molecules absorb the light in ultraviolet range (wavelength below 300nm). It is shown in figure 7.4.6, where ozone and TMA exposure has been highlighted by yellow and green colours, respectively. The red curve shows the drop in light intensity at 250nm wavelength, only at those moments when ozone was introduced in chamber. And black curve shows the change in optical layer thickness versus time. This curve is superimposed on the red curve to match the exact point of drop in the intensity curve. Thus one can be really assure that ozone was the real cause for this periodic intensity drop. Consequently, all experiments in this research project used a wavelength range from 300nm to 1690nm. This was important to reduce the MSE by factor 15% to 20%.

8 RESULTS AND DISCUSSION

8.1 INTRODUCTION

In general, the overall ALD process can be divided into two different growth regimes: the initial growth mechanism on heterogeneous starting substrate, followed by homo- geneous film-by-film growth on a material itself[ Puurunen 2005 ][ Soto 1991, p.2695 ]. Al2O3 has been grown on HF-etched silicon substrates, without creating a silicon-oxide interface, but with a slight inhibition[ Nalwa 2002, p.131 ][ Delabie u. a. 2012, p.3 ]. In this report, the initial substrates used were c−Si with native oxide. As previous research has shown almost no incubation for Alumina growing on such hydroxylated silicon substrates[ Nalwa 2002, p.131 ]. An ideal linear growth initialization was assumed in this report, and the results presented in the following sections focus on homogeneous linear film-on-film growth region.

In section 8.2, the kinetic ALD charecteristic curves are studied by time-discrete iSE (as described in section 7.4.2) along with comparison of two gas shower head designs. This includes variation of the TMA exposure time, Ar purging time after TMA exposure, ozone exposure time and Ar purging time after ozone exposure, respectively. In section

8.3, the impact on charateristic ALD growth attributes and film properties by varying process parameters like total process pressure, Ozone flow and deposition temperature are investigated in more detail by time-continuos irtSE.

8.2 KINETIC ALD CHARACTERISTIC CURVES

illustration not visible in this excerpt

Figure 8.2.1 kinetic ALD characteristic curves for shower head design type I and

II at deposition temperature of 150°C and total process pressure of 150 Pa

ALD characteristic curve is the curve that provides the information (for example in terms of optical thickness, frequency drop as revealed by irtSE and QCM, respectively) regard- ing each step (four steps of ALD cycle as described in section 3.2) of an ALD cycle. Figure 8.2.1 is a time-continuous iSE measurement and compares the ALD character- sitic curve for both type I and type II shower head designs. Where green and blue high- lighted portions show the TMA and ozone exposure times. Figure shows a small differ- ence in O3 exposure and Ar purge after behaviour O3 exposure step for both shower head designs. However, each step in ALD characterstic curve is studied in detail with the help of exposure and purging time variations for both reactants, respectively.

8.2.1 TMA exposure

illustration not visible in this excerpt

Figure 8.2.2 variation in TMA exposure time

First and foremost step of every Al2O3 ALD process from TMA and O3 is the exposure of trimethylaluminium (as a first reactant). TMA exposure time was varied from 0.1 s to

3 s for a type I shower head design. A time-discrete irtSE measurement method was employed to study growth per cycle for each step variation and a TMA exposure time curve has been plotted in figure 8.2.2.

In figure 8.2.2, it can be seen that GPC tends to saturate right after 1.5 s of TMA exposure time. Thus a minimum amount of TMA dose needed is 1.5 s in order to have self-limited TMA adsorbed substrate surface.

The above experiments with same process conditions for shower head design of type II were not performed, and saturation of TMA after 1.5 s was assumed.

8.2.2 Argon purging after TMA exposure

illustration not visible in this excerpt

Figure 8.2.3 variation in Ar purging time after TMA exposure

Ar purging after TMA exposure is a necessary step to remove excessive TMA molecules and by-products from the process chamber. Thus leaving the substrate surface (with TMA adsorbates) ready for a second half-reaction with ozone. In figure 8.2.3, Ar purging after TMA exposure time curves have been plotted for shower head designs of type I and II.

A set of red data points shows the Argon purge variations were done with a type I shower head design installed in the chamber. It is found that the saturation point for type I shower head design was achieved after 30 s of Ar purging. For a type II shower head design this point is approximately 20 s. After these points of inflection, no significant change in growth per cycle has been seen and the curve tends to saturate beyond. It can be also seen that the Ar purging time less than these points of inflection tends to provide with GPC value more than 1Å/cycle. Because the amount of Ar gas needed to purge all excessive surface molecules is too less. If the Ar purging times are too less, then it may cause overlapping of two reactants. Which may introduce a CVD like effect, that manifests as increased growth rate and thickness non-uniformities[ Nalwa 2002, p.140 ].

In figure 8.2.3, the Al2O3 growth per cycle levels are significantly reduced in case of type II shower head design. Also the GPC levels after saturation point for type II shower head design are close to 1Å/cycle. While in case of type I shower head design GPC val- ues tends to settle around 1.2Å/cycle. This can be due to more effective purging of gas channels inside the shower head, where residual precursor could have been concealed.

8.2.3 Ozone exposure

illustration not visible in this excerpt

Figure 8.2.4 variation in O3 exposure for both type of shower head designs - a

time-continuous irtSE measurement

Ozone exposure completes the second half reaction and the Al2O3 layer is obtained. In figure 8.2.4, saturation curves of ozone for both shower head designs have been plot- ted from time-continuous irtSE. These curves are obtained by inverting the ozone expo- sure (growth decay portion in figure 8.2.1) from ALD characteristic curves for both type of shower head designs. The red and blue curves in the figure above is ozone exposure curve for shower head design I and II, respectively. For both shower head designs, the experiments were done at substrate temperature of 150◦C and 150 Pa of total process pressure. These curves indicates the minimum amount of ozone exposure needed to have a completed reaction with all the TMA adsorbates on the surface. It can be seen that, the trend for saturation is almost comparable for both shower head designs. How- ever, the blue curve shows a slight faster decay (also in figure 8.2.1) but this decay is small and can be neglected too. This means, the reaction of O3 was not limited by the transportation of fresh O3 molecules to the surface, but by reaction kinetics of the O3 molecules with TMA adsorbate itself.

8.2.4 Argon purging after ozone exposure

illustration not visible in this excerpt

Figure 8.2.5 variation in Ar purging time after ozone exposure

This Ar purging step is necessary to remove the reaction by-products and the excessive ozone molecules from the chamber. Reaction by products being ethyne and other formate complexes [ Goldstein u. a. 2008; Kim u. a. 2006 ].

For both type of shower head designs, all the experiments and chamber conditions

were unchanged thus ensuring the comparable results. Experimental conditions being the flow of gases, temperature of 150◦C and chamber pressure (200 Pa). Figure 8.2.5 shows that GPC level is around 1Å/cycle for both shower head designs. Also there is al- most no change in growth per cycle with varying Ar purging time after ozone exposure for both type of shower head designs, indicating a very fast removal of O3 out of the reaction zone.

8.3 IMPACT OF PROCESS PARAMETERS ON CHARACTERISTIC ALD GROWTH ATTRIBUTES AND FILM PROPERTIES

8.3.1 Total process pressure

Figure 8.3.1 variation in total process pressure

a ALD characteristic curves for Al2O3 depositions at different pressures

illustration not visible in this excerpt

Figure 8.3.1a shows the change in optical layer thickness versus the ALD cycle time at different pressures. Each colored curve in a plot denotes one Averaged ALD character- istic curve obtained from 10 cycles that was described before in section 3.5. From this figure it can be interpreted that, the chamber pressure has not much effect on various attributes of the process. All the curves are approximately superimposed on each other.

Figure 8.3.1b, elucidates the dependency of pressure on the various attributes of an ALD characteristic curve like TMA adsorption, GPC and growth kinetics ’KO3 ’. It can seen from this figure that, the effect of total process pressure on GPC is nearly con- stant. However, the TMA adsorption seems to increase by very little amount. With in- crease in pressure there can be more adsorption on the substrate surface because in- crease in pressure can can increase the kinetic energy of molecules that may facilitate TMA adsorbtion on the surface. Anyhow this change is so little and can be neglected too. Figure also shows that variation in total process pressure has no significant effect on growth kinetics ’KO3 ’. However, a little increase in growth kinetics can be favoured with the help of same reason given above. At low total process pressures (100 Pa, 150 Pa), all attributes have the minimum value among all values at various other pressures. This can be supported by an argument that, at low process pressure TMA and ozone molecules have less kinetic energy. Therefore showing minimum amount of TMA adsorption and slow reaction with ozone molecules, respectively.

8.3.2 Ozone flow

Figure 8.3.2 variation in O3 flow

a ALD characteristic curves for Al2O3 depositions at different O3 flows

illustration not visible in this excerpt

The ozone flow was varied manually by setting the needle valve position attached next to the ozone MFC (see table A.2 in appendix). The figure 8.3.2a, depicts the depen- dency of optical layer thickness for various ALD characteristic curves at different O3 flows.

The ozone flow was varied from 50 sccm upto 500 sccm as shown in table B.1. The fig- ure depicts the small trend while moving from 50 sccm to 500 sccm. This dependency was studied thoroughly with the help of figure 8.3.2b. It can be seen, that growth per cycle is not much changed with increase in ozone flow. And it was same with the case of TMA adsorption. However, the only thing that has been affected was the growth ki- netics ’KO3 ’. A small trend was found in ’KO3 ’ and it seems that growth kinetics tends to depend on the ozone flow. With increase in ozone flow from 50 sccm to 500 sccm the growth kinetics was increased from 0.14 sec−¹ upto 0.32 sec−¹.

This means that with increase in ozone flow the reaction with TMA adsorbate is faster and it keeps on increasing but the GPC was not changed much. More precisely, in this figure the decay part (blue part in figure 3.5.1) of the whole ALD cycle tends to decrement faster and thus attaining the GPC value much more quicker in higher flow rates than in the case of low flow rates. This can also lead to stabilization of particular GPC level much quicker for higher O3 flow rates.

Also it might be possible that, setting higher ozone flow rate can affect the ozone sat- uration curve. But we did not perform that experiment and that might be executed in future. The discussion done in this section is in favor with[ Nalwa 2002, p.140 ] which states that growth rate per cycle may be increased by increasing exposure time but not by increasing the precursor flux [ Ozeki 1992 ][ Jeong u. a. 1989 ][ Ritala u. a. 1993 ].

8.3.3 Figure Deposition temperature

8.3.3 variation in deposition temperature

illustration not visible in this excerpt

The most important parameter among all ALD process parameters is the temperature that relies with results in [ Yun 1997, p.2997 ]. Figure 8.3.3 shows that the deposition temperature was varied from 100°C to 500°C (see table B.1). From the figure 8.3.3a, it can be seen that an individual ALD characteristic curve tends to deviate from each other very significantly. As it was not the case for total process pressure and ozone flow variations, where all ALD characteristic curves superimpose on each other.

All attributes of ALD characteristic curve seems to be affected by deposition temperature. This can be realized better with the help of figure 8.3.3b where the TMA adsorption has been decreased slightly with increase in temperature but this dependency is small and can be neglected. It can be asserted that at higher temperatures, the TMA molecules have more kinetic energy and start to leave the surface. Hence in this report, we can neglect the effect of temperature on the TMA adsorption.

As compare to other process parameters GPC has been affected most by tempera- ture and this change is not small to neglect. In figure 8.3.3b, the GPC value has been reduced from 1.3Å per cycle to 0.8Å per cycle. This result was in accordance with [ Puurunen 2005 ], in which the decrease in GPC with increase in temperature was shown. This might be the case where at high temperatures, the newly formed aluminium oxide atoms do the self alignment and arrange into closely packed layers. It can also be the case where at higher temperatures, ozone molecules dissociates and forms the more reactive radicals and might had started etching the newly formed Al2O3 layer or can be some other reason or combination of many postulates.

Also, the effect of temperature on growth kinetics ’KO3 ’ has been shown in figure 8.3.3b, where ’KO3 ’ tends to show some increase with increase in temperature upto 300°C. Above this point the O3 reaction kinetics was stabilized or slightly decreased for 500°C. Thus from this figure it can be said that optimum temperature would be around 300°C where the GPC is also close to 1Å per cycle and the growth kinetics ’KO3 ’ is also the highest attaining the value of 1 s−¹.

XPS analysis

To understand better the effect of deposition temperature on ALD charateristic curve attributes, a quantitative analysis by X-ray photoelectron spectroscopy was done. It pro- vides the information on the composition of elements that are present on the surface of deposited material. In order to perform qualitative analysis, the sample substrate was transferred to the in-vacuo XPS analysis directly, thus keeping it unaltered. In this way, it was very promising to do the chemical composition analysis of the actual material de- posited.

Figure 8.3.4a depicts the effect of deposition temperature on carbon (C1s peak) contaminations. For 100°C, it can be seen that the peak is approximately at 292 e.V and is of maximum area as compare to all peaks at other temperatures. For higher temperatures, the area under the peak was reduced by a significant value.

Figure 8.3.4 XPS analysis at different deposition temperature

illustration not visible in this excerpt

This can be better seen in figure 8.3.4b, where right horizontal axis shows the change of carbon contamination level (at percentage) with varied deposition temperatures. At 100°C a carbon contamination level is maximum (3.2%) . This value has suddenly dropped to 1% for 200°C. This drop in itself is large enough to show the improvement in Al2O3 film quality over temperature. Carbon contaminations can still be seen de- creasing with rise in temperature, contamination level of almost 0.6% at 300°C to 0.3% approximately for 400°C and finally below the detection limit at 500°C. The resulting decrease in carbon contamination level with increase in deposition temperature is in favour of [ Yun 1997 ].

In addition, the figure 8.3.4b also shows the effect of temperature variation on aluminium to oxygen ratio. With increase in temperature, the aluminium to oxygen ratio was al- most constant around 40%. Being minimum of approximately 38% only for 100°C. This aluminium to oxygen ratio is close to the ideal case, where expected Al:O ratio is 40:60. Above all experiments were done at 200 Pa pressure with 100 cycles of ALD process and 5 seconds of TMA exposure, 60 seconds of Ar purging time after TMA exposure, 30 seconds of ozone exposure and 60 seconds of last Ar purge. With 75, 2000, 250 and 2000 sccm of respective gas flows.

AFM analysis

illustration not visible in this excerpt

Figure 8.3.5 effect of temperature on surface roughness

The final surface analysis technique used was an AFM for all temperature varied samples. The samples were analyzed in an area of 500nm by 500nm. Two kinds of roughness (min-to-max and rms) were evaluated. The min-to-max roughness is maximum z value plus the minimum z (tip position in z-direction) value.

Figure 8.3.5, shows an effect of substrate temperature on the surface (rms and min- to-max) roughness. This figure shows that temperature has almost no or little effect on rms roughness because it is pretty low approximately 0.2nm and stays almost con- stant. Nevertheless, peak roughness cannot be left unnoticed, the peak roughness seems to show some trend with temperature and the value has been changed from 2.8nm for 100°C to 1.7nm for 400°C. Then min-to-max roughness was increased again for 500°C to value 2nm approximately.

Figure 8.3.6 a) shows a 3-D plot of a aluminium oxide surface that was grown at 100°C

deposition temperature. The surface can be seen as various color spots, dark being the pits and light color means peaks. Other sub figures 8.3.6 b), c), d) and e) are for 200°C, 300°C, 400°C and 500°C deposition temperatures, respectively. 3-D plots, describe the surface impressions and depressions that are actually present on the sample surface. Each 3-D plot provides a topography of grown Al2O3 surface.

Figure 8.3.6 AFM topography after 100 Al2O3 ALD cycles at different

illustration not visible in this excerpt

8.4 REPRODUCIBILITY

In order to check the repeatability of the developed ALD process, figure 8.4.1, shows reproduced ALD characteristic curves for two process conditions. Both curves, red and blue (being reapeated) in each sub-figure overlap each other almost perfectly, denoting the reliability and reproducibility of the process, when performed again in same process conditions.

Figure 8.4.1 reproduced ALD cycle curves

illustration not visible in this excerpt

9 CONCLUSIONS AND OUTLOOK

In summary this research project, the capability of an in-situ SE to measure various thin film parameters was demonstrated and refined. The in-situ measurements from iSE are very effective while describing the various film attributes like temperature, thickness, etc. Various ALD characteristic attributes like TMA adsorption, ligand removal, growth kinetics ’KO3 ’, growth per cycle have been studied in variation of exposure and purge times, deposition temperature, total process pressure and ozone flow. Also the perfor- mance of two shower head designs was compared. ALD characteristic curves for all four steps have been plotted and studied individually. Saturation of TMA adsorption was found after 1.5 seconds. Saturation point for Argon purge time after TMA exposure was found to be 20 seconds and 30 seconds for type II and type I shower head designs, respectively. Also it was noticed that in case of type II shower head design the GPC levels have reduced approximately by 25% as compare to GPC levels for type I shower head design. In case of type II shower head design the GPC level after saturation was close to 1Å per cycle. For ozone exposure time, saturation takes place around 55 sec- onds for both shower head designs. After installing a new shower head design, there was no significant improvement found in the performance of Argon purge time followed by ozone exposure.

It has been found that the effect of substrate temperature on the various ALD cycle pa- rameters is most substantial than any other ALD process parameter variations. Tem- perature has affected almost all ALD characteristic curve attributes (TMA adsorption, growth kinetics KO3 , GPC, chemical composition and surface roughness). With increase in temperature, the carbon contaminations percentage levels were reduced to great extent. At the same time the aluminum to oxygen ratio was fairly constant for tempera- tures above 200°C. Min-to-max and rms roughness have been measured with the help of non-contact AFM method. And it was found that rms roughness stays at one level with unnoticeable change. While the min-to-max roughness minimized for temperature around 400°C to 500°C.

In a nutshell, it can be said that type II shower head design performs better than the type I shower head design. And the temperature was the major factor influencing the various attributes of ALD characteristic curve. Based on this research, I would suggest the optimized parameters for Al2O3 ALD process from TMA and Ozone (in ALD-300 tool from FHR anlagenbau Gmbh with shower head type II installed in it) with 200Pa total process pressure, 300°C set point temperature and 250 sccm of ozone flow. A pulsing sequence of 1.5 s, 20s, 30 s and 10 s for TMA exposure, purge, O3 exposure and purge, respectively, to produce carbon-free and smooth Al2O3 films.

ALD is very promising technique to deposit ultra thin and high quality films. With minia- turization according to need of future technologies, the quality has become of prime importance over speedy manufacturing. Methods for deposition and research of many more materials have been pipelined into researchers queue. Therefore in future, the use of ALD technology is very promising. The results of this project can be incorporated and same methodology can be applied to study other materials like metals, nitrides and dielectrics. Moreover, this research can be broadened by studying the initial phase of deposition and also by studying the electrical properties like leakage current and breakdown voltage of deposited Aluminum oxide.

REFERENCES

Andrade 1985

ANDRADE, Joseph D.: Surface and Interfacial Aspects of Biomedical Polymers: Volume

1 Surface Chemistry and Physics. Boston and MA : Springer US, 1985. - ISBN 978-1- 4684-8610-0

Aspnes u. Hauge 1976

ASPNES, D. E. ; HAUGE, P. S.: Rotating-compensator/analyzer fixed-analyzer ellipsometer: Analysis and comparison to other automatic ellipsometers. In: Journal of the Optical Society of America 66 (1976), Nr. 9, S. 949

B. Johs u. a. 1999

B. JOHS ; J. A. WOOLLAM ; C. HERZINGER ; J. HILFIKER ; R. SYNOWICKI ; C. BUNGAY:

Overview of Variable Angle Spectroscopic Ellipsometry (VASE), Part II: Advanced Applications. (1999), S. 29-58

Bernardi u. a. 2004

BERNARDI, M. I. B. ; CRISPIM, S. C. L. ; MACIEL, A. P. ; SOUZA, A. G. ; CONCEICAO,

M. M. ; LEITE, E. R. ; LONGO, E.: Synthesis and characterization of Al2O3/Cr2O3 -

based ceramic pigments. In: Journal of Thermal Analysis and Calorimetry 75 (2004),

Nr. 2, S. 475-480. http://dx.doi.org/10.1023/B:JTAN.0000027135.04033.e8. - DOI

10.1023/B:JTAN.0000027135.04033.e8. - ISSN 1388-6150

Binnig u. Quate 1986

BINNIG, G. ; QUATE, C. F.: Atomic Force Microscope. In: Physical Review Letters 56 (1986), Nr. 9, S. 930-933

cambridge nanotech July

CAMBRIDGE NANOTECH: Cambridge NanoTech ALD tutorial. http://www.

cambridgenanotechald.com/atomic-layer-deposition-overview.html. Version: July 2011

Dale E. Morton u. a. 2002

DALE E. MORTON ; BLAINE JOHS ; JEFF HALE: Optical Monitoring of Thin-films Using Spectroscopic Ellipsometry. In: Society of Vacuum Coaters (2002)

Delabie u. a. 2012

DELABIE, Annelies ; SIONCKE, Sonja ; RIP, Jens ; ELSHOCHT, Sven van ; POURTOIS, Geoffrey ; MUELLER, Matthias ; BECKHOFF, Burkhard ; PIERLOOT, Kristine: Reac- tion mechanisms for atomic layer deposition of aluminum oxide on semiconductor substrates. In: Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 30 (2012), Nr. 1, S. 01A127. http://dx.doi.org/10.1116/1.3664090. - DOI

10.1116/1.3664090. - ISSN 07342101

Elliott u. a. 2006

68 References

ELLIOTT, S. D. ; SCAREL, G. ; WIEMER, C. ; FANCIULLI, M. ; PAVIA, G.: Ozone-Based Atomic Layer Deposition of Alumina from TMA: Growth, Morphology, and Reaction Mechanism. In: Chemistry of Materials 18 (2006), Nr. 16, S. 3764-3773. http://dx. doi.org/10.1021/cm0608903. - DOI 10.1021/cm0608903. - ISSN 0897-4756

Ferlauto u. a. 2002

FERLAUTO, A. S. ; FERREIRA, G. M. ; PEARCE, J. M. ; WRONSKI, C. R. ; COLLINS, R. W. ; DENG, Xunming ; GANGULY, Gautam: Analytical model for the optical functions of amorphous semiconductors from the near-infrared to ultraviolet: Applications in thin film photovoltaics. In: Journal of Applied Physics 92 (2002), Nr. 5, S. 2424

Fujiwara 2007

FUJIWARA, Hiroyuki: Spectroscopic ellipsometry: Principles and applications.

Chichester and England and Hoboken and NJ : John Wiley & Sons, 2007. - ISBN 9780470060186

Goldstein u. a. 2008

GOLDSTEIN, David N. ; MCCORMICK, Jarod A. ; GEORGE, Steven M.: Al 2 O 3 Atomic Layer Deposition with Trimethylaluminum and Ozone Studied by in Situ Transmission FTIR Spectroscopy and Quadrupole Mass Spectrometry. In: The Journal of Physi- cal Chemistry C 112 (2008), Nr. 49, S. 19530-19539. http://dx.doi.org/10.1021/ jp804296a. - DOI 10.1021/jp804296a. - ISSN 1932-7447

Gross u. a. 2009

GROSS, L. ; MOHN, F. ; MOLL, N. ; LILJEROTH, P. ; MEYER, G.: The Chemical Structure of a Molecule Resolved by Atomic Force Microscopy. In: Science 325 (2009), Nr. 5944,

S. 1110-1114. http://dx.doi.org/10.1126/science.1176210. - DOI 10.1126/sci- ence.1176210. - ISSN 0036-8075

H. Suzuki u. a. 1989

H. SUZUKI ; N. FUJIMAKI ; H. TAMURA ; T. IMAMURA ; S. HASUO: A 4K Josephson memory - Magnetics, IEEE Transactions on. (1989)

Hans sundstrom 2005

HANS SUNDSTROM, M. I.: Ozone as the Oxidizing Precursor in Atomic Layer Deposition. (2005)

Hasuo 1992

HASUO, S.: High quality niobium Josephson junction for ultrafast computers. In: Thin Solid Films 216 (1992), Nr. 1, S. 21-27. http://dx.doi.org/10.1016/0040-6090(92) 90862-6. - DOI 10.1016/0040-6090(92)90862-6. - ISSN 00406090

Herzinger u. a. 1998

HERZINGER, C. M. ; JOHS, B. ; MCGAHAN, W. A. ; WOOLLAM, J. A. ; PAULSON, W.: El- lipsometric determination of optical constants for silicon and thermally grown silicon

References

dioxide via a multi-sample, multi-wavelength, multi-angle investigation. In: Journal of Applied Physics 83 (1998), Nr. 6, S. 3323

Himcinschi u. a. 2001

HIMCINSCHI, C. ; FRIEDRICH, M. ; MURRAY, C. ; STREITER, I. ; SCHULZ, S. E. ; GESSNER,

T. ; ZAHN, D. R. T.: Characterization of silica xerogel films by variable-angle spectroscopic ellipsometry and infrared spectroscopy. In: Semiconductor Science and Technology 16 (2001), Nr. 9, S. 806-811

J. Spiegelman u. J. Sundqvis 2011

J. SPIEGELMAN ; J. SUNDQVIS: Comparison of Water Vapor to Ozone for growth of ALD Films Comparison of Water Vapor to Ozone for growth of ALD Films. In: Gases&Instrumentation (2011), S. 20-25

J.A. Woollam Co. 2011

J.A. WOOLLAM CO., Inc.: Complete Ease: Data analysis Manual version (4.63). 645 M STREET, SUITE 102 LINCOLN, NE 68508-2243 USA : J.A. Woollam Co.,Inc., 2011

Jakschik u. a. 2003

JAKSCHIK, Stefan ; SCHROEDER, Uwe ; HECHT, Thomas ; KRUEGER, Dietmar ;

DOLLINGER, Guenther ; BERGMAIER, Andreas ; LUHMANN, Claudia ; BARTHA, Jo-

hann W.: Physical characterization of thin ALD-Al2O3 films. In: Applied Surface Sci-

ence 211 (2003), Nr. 1-4, S. 352-359. http://dx.doi.org/10.1016/S0169-4332(03) 00264-2. - DOI 10.1016/S0169-4332(03)00264-2. - ISSN 01694332

Jellison 1998

JELLISON, JrG.E: Spectroscopic ellipsometry data analysis: measured versus calculated quantities. In: Thin Solid Films 313-314 (1998), Nr. 1, S. 33-39

Jeong u. a. 1989

JEONG, Weon G. ; MENU, E. P. ; DAPKUS, P. D.: Steric hindrance effects in atomic layer epitaxy of InAs. In: Applied Physics Letters 55 (1989), Nr. 3, S. 244

Johs 2000

JOHS, Craig M. Blaine D.; Herzinger H. Blaine D.; Herzinger: Methods for uncorrelated evaluation of parameters in parameterized mathematical model equations for window retardence, in ellipsometer and polarimeter systems, United States Patent 6034777. 2000

Junige u. a.

JUNIGE, Marcel ; GEIDEL, Marion ; KNAUT, Martin ; ALBERT, Matthias ; BARTHA, Johann W.: Monitoring atomic layer deposition processes in situ and in real-time by spectroscopic ellipsometry.

Kim u. a. 2002

70 References

KIM, J. B. ; KWON, D. R. ; CHAKRABARTI, K. ; LEE, Chongmu ; OH, K. Y. ; LEE, J. H.: Im- provement in Al2O3 dielectric behavior by using ozone as an oxidant for the atomic layer deposition technique. In: Journal of Applied Physics 92 (2002), Nr. 11, S. 6739. http://dx.doi.org/10.1063/1.1515951. - DOI 10.1063/1.1515951. - ISSN 00218979

Kim u. a. 2003

KIM, Jaebum ; CHAKRABARTI, Kuntal ; LEE, Jinho ; OH, Ki-Young ; LEE, Chongmu: Ef- fects of ozone as an oxygen source on the properties of the Al2O3 thin films pre- pared by atomic layer deposition. In: Materials Chemistry and Physics 78 (2003),

Nr. 3, S. 733-738. http://dx.doi.org/10.1016/S0254-0584(02)00375-9. - DOI

10.1016/S0254-0584(02)00375-9. - ISSN 02540584

Kim u. a. 2006

KIM, Seong K. ; LEE, Suk W. ; HWANG, Cheol S. ; MIN, Yo-Sep ; WON, Jeong Y. ;

JEONG, Jaehack: Low Temperature (<100°C) Deposition of Aluminum Oxide Thin Films by ALD with O[sub 3] as Oxidant. In: Journal of The Electrochemical Soci- ety 153 (2006), Nr. 5, S. F69. http://dx.doi.org/10.1149/1.2177047. - DOI

10.1149/1.2177047. - ISSN 00134651

Manchanda u. a. 2001

MANCHANDA, L. ; MORRIS, M.D ; GREEN, M.L ; DOVER, R.B van ; KLEMENS, F. ;

SORSCH, T.W ; SILVERMAN, P.J ; WILK, G. ; BUSCH, B. ; ARAVAMUDHAN, S.: Multi- component high-K gate dielectrics for the silicon industry. In: Microelectronic Engi- neering 59 (2001), Nr. 1-4, S. 351-359. http://dx.doi.org/10.1016/S0167-9317(01) 00668-2. - DOI 10.1016/S0167-9317(01)00668-2. - ISSN 01679317

Mizuguchi u. a. 2005

MIZUGUCHI, M. ; SUZUKI, Y. ; NAGAHAMA, T. ; YUASA, S.: Atomically flat aluminum-

oxide barrier layers constituting magnetic tunnel junctions observed by in situ scanning tunneling microscopy. In: Applied Physics Letters 87 (2005), Nr. 17, S. 171909. http: //dx.doi.org/10.1063/1.2108121. - DOI 10.1063/1.2108121. - ISSN 00036951

Nalwa 2002

NALWA, Hari S.: Handbook of thin film materials. San Diego : Academic Press, 2002. - ISBN 0125129092

Ozeki 1992

OZEKI, M.: Atomic layer epitaxy of III-V compounds using metalorganic and hydride sources. In: Materials Science Reports 8 (1992), Nr. 3, S. 97-146

Park u. a. 2003

PARK, Hong B. ; CHO, Moonju ; PARK, Jaehoo ; LEE, Suk W. ; HWANG, Cheol S. ; KIM,

Jong-Pyo ; LEE, Jong-Ho ; LEE, Nae-In ; KANG, Ho-Kyu ; LEE, Jong-Cheol ; OH, Se-Jung: Comparison of HfO[sub 2] films grown by atomic layer deposition using HfCl[sub 4]

References

and H[sub 2]O or O[sub 3] as the oxidant. In: Journal of Applied Physics 94 (2003), Nr. 5, S. 3641

Patnaik 2003

PATNAIK, Pradyot: Handbook of inorganic chemicals. New York : McGraw-Hill, 2003 (McGraw-Hill handbooks). - ISBN 0070494398

Perrella 2004

PERRELLA, Andrew C.: Ballistic Electron Transport in Aluminum Oxide, Cornell University, Diss., 2004

Puurunen 2005

PUURUNEN, Riikka L.: Surface chemistry of atomic layer deposition: A case study for the trimethylaluminum/water process. In: Journal of Applied Physics 97 (2005), Nr. 12,

S. 121301. http://dx.doi.org/10.1063/1.1940727. - DOI 10.1063/1.1940727. - ISSN 00218979

Ritala u. a. 1993

RITALA, Mikko ; LESKELA, Markku ; NIINISTO, Lauri ; HAUSSALO, Pekka: Titanium isopropoxide as a precursor in atomic layer epitaxy of titanium dioxide thin films. In: Chemistry of Materials 5 (1993), Nr. 8, S. 1174-1181

Schmidt u. a. 2008

SCHMIDT, D. ; STREHLE, S. ; ALBERT, M. ; HENTSCH, W. ; BARTHA, J.W: Top injection reactor tool with in situ spectroscopic ellipsometry for growth and characterization of ALD thin films. In: Microelectronic Engineering 85 (2008), Nr. 3, S. 527-533

Siah u. a. 2013

SIAH, S.C ; HOEX, B. ; ABERLE, A.G: Accurate characterization of thin films on rough surfaces by spectroscopic ellipsometry. In: Thin Solid Films 545 (2013), S. 451-457

Skoog 1985

SKOOG, Douglas A.: Principles of instrumental analysis. 3. Philadelphia : Saunders College Pub., 1985 (Saunders golden sunburst series). - ISBN 9780030012297

Soto 1991

SOTO, C.: The reaction pathway for the growth of alumina on high surface area alu-

mina and in ultrahigh vacuum by a reaction between trimethyl aluminum and water. In: Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 9 (1991), Nr. 5, S. 2686

Tompkins u. Irene 2005

TOMPKINS, Harland G. ; IRENE, Eugene A.: Handbook of ellipsometry. Norwich and NY and Heidelberg and Germany : William Andrew Pub. and Springer, 2005. - ISBN 0815514999 Wagner u. Muilenberg 1979

WAGNER, C. D. ; MUILENBERG, G. E.: Handbook of x-ray photoelectron spectroscopy: a reference book of standard data for use in x-ray photoelectron spectroscopy. Physical Electronics Division, Perkin-Elmer Corp, 1979 http://books.google.de/books?id=

oY5TAAAAYAAJ

Yeong Kwan Kim u. a.

YEONG KWAN KIM ; SEUNG HWAN LEE ; SUNG JE CHOI ; HONG BAE PARK ; YOUNG DONG SEO ; KWANG HYUN CHIN ; DONGCHAN KIM ; JAE SOON LIM ; WAN DON KIM ; KAB JIN NAM ; MAN-HO CHO ; KI HYUN HWANG ; YOUNG SUN KIM ; SEOK SIK KIM ; YOUNG WOOK PARK ; JOO TAE MOON ; SANG IN LEE ; MOON YONG LEE: Novel ca- pacitor technology for high density stand-alone and embedded DRAMs. http: //dx.doi.org/10.1109/IEDM.2000.904332. - DOI 10.1109/IEDM.2000.904332

Yun 1997

YUN, Sun J.: Dependence of atomic layer-deposited Al2O3 films characteristics on growth temperature and Al precursors of Al(CH[sub 3])[sub 3] and AlCl3. In: Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 15 (1997), Nr. 6,

S. 2993. http://dx.doi.org/10.1116/1.580895. - DOI 10.1116/1.580895. - ISSN 07342101

Zhong u. a. 1993

ZHONG, Q. ; INNISS, D. ; KJOLLER, K. ; ELINGS, V.B: Fractured polymer/silica fiber sur- face studied by tapping mode atomic force microscopy. http://www.sciencedirect. com/science/article/pii/016725849390906Y. Version: 1993

Zhu u. Park 2006

ZHU, J. ; PARK, C.: Magnetic tunnel junctions. In: Materials Today 9 (2006), Nr. 11, S.

36-45. http://dx.doi.org/10.1016/S1369-7021(06)71693-5. - DOI 10.1016/S1369- 7021(06)71693-5. - ISSN 13697021

ACKNOWLEDGMENTS

The research studies reported in this project were carried out during the year 2012-

2013 in the Institute of Semiconductors And Microsystems at university of Technology Dresden, Germany.

First, I owe my deepest gratitude to my supervisors, Dipl.-Ing. Marcel Junige and Dipl.Ing. Martin Knaut for their invaluable contribution and help to this study. Marcel, without your guidance this work would not have been done. Both of you had always time for me and your rapid feedback concerning my ongoing project was just unbelievable. I had a great time during the complete year, working with you.

My special thanks go to my Professor Dr.rer.nat. Johann W. Bartha, for his valuable support and advice during this year in and outside of the laboratory. I would also like to appreciate Dr.-Ing. Matthias Albert for his advice and timely evaluation of my presentations. I owe warm thanks to Zulfija Ritter for her kind advice and help in AFM and ex-situ SE measurements.

III APPENDIX

A REFERENCE TEMPERATURES AND OZONE FLOW

The table A.1 shows the actual substrate temperature obtained after fitting the SE model parameters (temperature in Si_TEMP_JAW(Temp_Library) ) and measured data in corresponding to set point temperatures. Set point temperature is the temperature of substrate heating plate. But the actual temperature is the temperature of substrate sur- face. There is a significant temperature difference between the substrate heating plate and the surface. This difference is due to loss in transfer of heat from heating plate to small coupons of substrate, where actual depositions had taken place.

illustration not visible in this excerpt

Table A.2 O2 and O3 gas flows with their corresponding needle valve positions

80 Appendix A Reference temperatures and ozone flow

B PROCESS PARAMETERS

Tables B.1, B.2, B.3, B.4 and B.5 show the variation in various process parameters.

Various process parameters being total process pressure, deposition temperature, O3 flow, and variation in TMA, Ar, O3 exposure and purging times. All these variations have been shown in coloured text, while the rest of the process parameters being unaltered. Where tTMA, tAr1 , tO3 , tAr are the TMA exposure, Ar purge time after TMA exposure,

O3 exposure time and Ar purge time after O3 exposure. All variations in table B.1, have been performed with type II shower head design installed. While the rest of the parameters were varied for both type I and type II shower heads installed in the chamber. But the tTMA variations were only carried with type I shower head design.

illustration not visible in this excerpt

Table B.1 process parameter variations

illustration not visible in this excerpt

Table B.2 variations in TMA exposure time

82 Appendix B Process parameters

illustration not visible in this excerpt

Table B.3 variations in Ar purge time after TMA exposure

illustration not visible in this excerpt

Table B.4 variations in O3 exposure time

illustration not visible in this excerpt

Table B.5 variations in Ar purge time after O3 exposure