process concept and realisationprocess concept and realisation proefschrift ter verkrijging van de...

I

Plasma deposition of nanocomposite thin films

Process concept and realisation

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven,

op gezag van de Rector Magnificus, prof.dr. R.A. van Santen, voor een commissie aangewezen door het College voor Promoties in het openbaar te verdedigen op

maandag 10 mei 2004 om 16.00 uur

door

Gregory Robert Alcott

geboren te Basingstoke, Engeland

II

Dit proefschrift is goedgekeurd door de promotoren: prof.dr.ir. M.C.M. van de Sanden en prof.dr.ir. D.C. Schram The work described in this thesis was supported by The Netherlands Ministry of Economic Affairs within the framework of the Innovation Directed Research Program (IOP Oppervlaktetechnologie, IOT 99005). CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN Alcott, Gregory Robert Plasma deposition of nanocomposite thin films: Process concept and realisation / by Gregory Robert Alcott. – Eindhoven : Technische Universiteit Eindhoven, 2004. - Proefschrift. ISBN 90-386-1905-7 NUR 926 Trefwoorden: plasma / plasma processing / plasma depositie / plasma chemie / stoffig plasma / dunne film karakterisatie/ nanogestructureerde dunne films / hybride deklagen Key words: plasma / plasma processing / plasma deposition / plasma chemistry / dusty plasma / thin film characterisation / nanocomposite materials / hybrid coatings Printed and bound by Universiteitsdrukkerij Technische Universiteit Eindhoven Cover realisation by Paul Verspaget.

III

Table of contents Chapter 1: General introduction 1 1.1 Introduction 1 1.1.1 Wet chemical routes 2 1.1.2 Current “state of the art” and applications 3 1.2 PVD processes 7 1.3 CVD processes 7 1.3.1 Plasma activated (enhanced) CVD 8 1.4 Hybrid films via gas phase processes: a literature review 10 1.4.1 Silicon-based polymers 11 1.4.2 Gas phase polymerisation of hydrocarbons 14 1.4.3 Gas phase particle synthesis 15 1.5 Conclusions 18 1.6 Scope of this thesis 19 Chapter 2: Anisotropic plasma etching of SiOxCyHz layers 25 2.1 Introduction 25 2.1.1 Reactive ion etching 26 2.2 Experimental 28 2.2.1 Ion saturation current measurements 30 2.3 Results 32 2.3.1 DC bias measurements 33 2.3.2 Etch results 37 2.4 Conclusions 39 Chapter 3: Synthesis and characterisation of SiOxCyHz layers 41 3.1 Introduction 41 3.1.1 Problem description 41 3.2 Process design and reactor concept 44 3.2.1 Diagnostics and substrate preparation 47 3.3 Polymer synthesis in the capacitive plasma (Zone II) 47 3.4 Inorganic synthesis in the inductive plasma (Zone I) 48 3.5 Film adhesion 49 3.6 Mechanical/optical properties of hybrid layers 52 3.7 Discussion 53 3.8 Conclusions 53 Appendix A: Vapour pressures of precursors for the CVD of

silicon-based films 55 Appendix B: Cross cut tape test 61 Chapter 4: Synthesis, monitoring and characterisation of

nanoparticles for incorporation into nanocomposite layers 63 4.1 Introduction 63 4.1.1 Particle formation in r.f. discharges 64

IV

4.2 Forces on particles 65 4.3 Experimental 75 4.4 Infrared analysis 77 4.4.1 Rayleigh scattering 80 4.5 Physical characterisation of particles 81 4.6 Discussion and conclusions 83 Appendix C: Infrared analysis of particles 86 Chapter 5: Reaction kinetics and mechanisms in the plasma deposition of SiO2 from a TEOS/O2 gas mixture 91 5.1 Introduction 91 5.2 Molecular processes in TEOS/O2 discharges 92 5.3 Experimental 94 5.4 Mechanisms in the TEOS/oxygen plasma 96 5.5 Experimental results 100 5.6 Conclusions 102 Appendix D: Table of infrared absorption frequencies 105 Chapter 6: Deposition of nanocomposite layers for ultra low dielectric applications 107 6.1 Introduction 107 6.2 Reactor design 111 6.2.1 Experimental 113 6.3. Experimental results 114 6.3.1 Dielectric constant 115 6.4 Conclusions 116 Appendix E: Optical emission spectroscopy 118 Appendix F: Dielectric constant determination 122 Chapter 7: General Conclusions 123 7.1 Scratch resistant layers 125 7.2 Particle monitoring 125 7.3 Plasma chemistry 126 7.4 Potential application as low dielectric layers 126 Symbols and Abbreviations 127 Summary 129 Samenvatting 131 Acknowledgement 133 Curriculum Vitae 135 Publications related to this work 137

Chapter 1 General introduction

1

Chapter 1: General introduction

Abstract Relatively new developments in materials technology, fuelled by the growing interest surrounding nanotechnology have given rise to a new breed of materials known as nanocomposites. Nanocomposite are a subgroup of hybrid materials and represent a two phase material in which one phase appears in the form of nanometre sized clusters or particles. The concept of hybrid materials and their origins from polymer and wet chemical synthetic techniques is presented. The suitability of these materials in some key applications is reviewed and the current “state of the art” regarding material properties discussed. Advantages and disadvantages of the wet chemical versus gas phase depositions techniques are given, followed by a brief introduction to gas phase deposition techniques with particular attention paid to plasma processes.

1.1 Introduction In general, the worlds of ceramic and polymer science have developed independently from one another with little overlap between the two areas. Traditionally, the field of ceramics concerns itself with the preparation of inorganic matrix glass materials through sintering or firing processes while the area of polymers relates to the synthesis and processing of polymerizing network forming high molecular weight organic molecules. It is quite evident that ceramic and organic polymer materials strongly differ in chemical and mechanical properties, and it is for precisely this reason why combining these materials is of great interest. Typical advantages of organic polymers include flexibility, low density, toughness, and formability whereas ceramics have excellent mechanical and optical properties such as surface hardness, modulus, strength, transparency and high refractive index. If these materials can be effectively combined, a new class of high performance and highly functional organic-inorganic hybrid materials may be achieved. The field of hybrids and nanocomposites covers a wide range of materials and scientific disciplines ranging from polymer chemistry to plasma physics. An introduction into the subject must therefore start with some definitions. The terms and classifications given to hybrid materials differ depending on which discipline they are studied. Classification schemes vary depending on chemical interaction between the polymeric and ceramic phases, intrinsic dimensions and the preparation technique used. For clarification, the hybrid materials described in this thesis are grouped into two main classifications as pictorially presented in Fig. 1.1.


2

R

R

R

Organically modified ceramic: Derived from adding organic functionality to a ceramic material.

Nanocomposites: Particles mixed into a polymeric or ceramic material. Might or might not be chemically bonded.

= Ceramic matrix = Polymer chain = Particles R = Organic functionality

R

R

R

R

R

R

Organically modified ceramic: Derived from adding organic functionality to a ceramic material.

Nanocomposites: Particles mixed into a polymeric or ceramic material. Might or might not be chemically bonded.

= Ceramic matrix = Polymer chain = Particles R = Organic functionality

Figure 1.1: Classification of material forms as described in this thesis.

Organically modified ceramics are principally ceramic materials that have been modified with organic compounds so that some intrinsically organic attributes, such as hydrophobicity, are bestowed upon them. As the material is predominantly ceramic it exhibits similar mechanical and physical properties of ceramic materials such as high hardness and mechanical strength but poor flexibility. As the organic content increases the material properties evolve from those of ceramic to those attributed to plastics, such as toughness and flexibility. In this study the term organically modified ceramic is also extended to include plasma polymers deposited from siloxane precursors, as is described in chapter 3, despite the fact that the structure of the plasma siloxane films is sometimes closer to that of silicone polymers. Ceramic functionality has also been imparted into polymeric materials by the addition of nanometre sized ceramic particles. These materials, herewith referred to as nanocomposites, exhibit improved toughness and reduced gas permeability than the original plastics. Terms such as CERAMER and NANOMER have not been adopted in this study due to the inconsistency of use and wide variety of materials that they have been used to describe in the literature. In addition to this, phrases such as ‘nano-structured’ and ‘nano-crystalline’ are often used in the context of thin film morphology [1, 2] where the term ‘nano’ refers to the presence of nanometre sized crystallites within one material composition (cf. µc-Si [3, 4]). Such materials are not included within the scope of this work and the term nanocomposite refers to materials composed of nanometre sized ceramic particles embedded in polymeric layers. 1.1.1 Wet chemical routes The traditional synthetic route to hybrid materials is the sol-gel process. Sol-gel is a method to produce ceramic materials by preparation of a colloidal suspension of precursors (sol), gelation of the sol and removal of the solvent. The precursors are typically metal alkoxide compounds and derivatives thereof. One of the most studied sol-gel chemistries is that of the alkoxide tetraethoxysilane (TEOS) [5, 6, 7, 8]. Under aqueous conditions, the ethanoic groups on TEOS are hydrolysed to form


3

silanol groups. Application of heat then drives the reaction to completion, liberating H2O and forming SiO2. This reaction is depicted in Fig. 1.2. ROHOHSiOHORSi 2 +−≡→+−≡

OHSiOSiSiHOOHSi 2+≡−−≡→≡−+−≡ ………or

ROHSiOSiSiROOHSi +≡−−≡→≡−+−≡

Figure 1.2: Hydrolysis of TEOS, R represents an arbitrary hydrocarbon chain, in the case of TEOS, R = C2H5.

During the gelation phase the sol-gel solution can easily be cast into a complex shape or used to coat suitable substrates. Coating is achieved via spinning or dipping techniques where the final coating thickness is determined from spinning/dipping speeds and the viscosity of the gel solution. After casting or coating, the sol-gel solution is cured to complete the chemical reactions and drive out the remaining water or alcoholic by products and densify the material. Curing is typically achieved by baking the sample in an oven at temperatures up to and exceeding 400°C, but exposure to plasma or UV radiation from a lamp have also shown to be effective curing methods [9]. For an extensive review of sol-gel technology see J. Brinker et al. [10]. 1.1.2 Current state of the art & Applications The sol-gel process can be modified via one of two routes to produce hybrid materials. Functionalised alkoxide precursors such as (3-glycidoxypropyl) trimethoxysilane (Glymol) [11, 12] or isocyanatopropyl triethoxy silane [13] are used in place of, or in conjunction with conventional alkoxides. These react to form materials containing organic bridging units. Considering Glymol as an example, hydrolysis reactions occur on only three of the available positions surrounding the silicon atom and polymerization occurs at the epoxide group located at the end of the carbon chain. The resulting silica structure, containing organic bridging units, is depicted in Fig. 1.3.

SiO

O

SiO

O

R OOSi

SiSi

O O

OR

Si

RO

O

R

Si R

ORSi

OO

O SiSi

O

O

SiO

O

R OOSi

SiSi

O O

OR

Si

RO

O

R

Si R

ORSi

OO

O Si

Figure 1.3: Theoretical structure of organically modified ceramic synthesized from (3-glycidoxypropyl) trimethoxysilane.


4

In addition to improving the flexibility and toughness of the final material, the organic bridging units also lower the curing temperature necessary to achieve material densification. This is believed to be because the added pliancy of the material enables it to ‘relax’ into a denser structure with lower energy loads [14].

An alternative to using a modified precursor is to prepare a precursor solution where the organic and inorganic components are simultaneously synthesized from two different precursors. Materials synthesized from two component solutions (a precursor such as TEOS for the inorganic phase and glymol for the organic matrix) have been developed by the Fraunhofer Institute and are know by the trade name ORMOCER® [15]. Other authors have referred to these materials as nanocomposites, perhaps partly due to the copyright, but also as slight phase separation can occur during curing, resulting in the formation of SiO2-like clusters embedded in the predominantly organic matrix.

Inherently weak polymers are also reinforced by blending particulate fillers into the polymer resin prior to it being cured into a network structure [16, 17, 18, 19, 20, 21]. The disadvantages of this approach are the invariable coalescence of the filler particles into large aggregates and poor chemical interaction between the filler and polymer. This results in poor dispersion and consequently limited improvements in mechanical properties. Since the first nanocomposite materials were prepared shortly after the major interest in synthetic plastics started in the 1950’s, a variety of materials have been investigated as fillers including, clays [22], carbon nanotubes and cellulose whiskers.

Composite materials also have important applications in areas such as catalysis and electronics. In a recent paper Mohallem et al. [23] synthesized magnetic NiFe2O4/SiO2 nanocomposite films using sol-gel techniques. The films consisted of 6 to 10 nm sized particles of crystalline NiFe2O4 dispersed in an amorphous SiO2. Yano et al. [24, 25] compared various mechanical properties, including stress-strain curves, sorption isotherms and dynamic viscoelastic properties of hydroxypropyl cellulose (HPC) and poly (vinyl acetate) combined with silica prepared via sol-gel synthesis (with TEOS) to blended samples with untreated silica particles (diameter 20-40µm). They discovered that even with minimal amounts of silica, the sol-gel prepared samples exhibited drastic improvements in mechanical properties, whereas a comparative test with blended samples showed only minimal improvements. A decrease in overall tensile strength of the blended samples as the silica content increased, was attributed to aggregation of the silica particles with in the HPC matrix to form weak points, a phenomena which has also been observed by others [26, 27, 28]. Other chemistries such as (n-butyl methacrylate)/titania [29] and poly(vinyl alcohol)/alumina [30] have also been evaluated with respect to mechanical properties with similar results. Improvements in material properties and particle dispersibility have been demonstrated by subjecting the particles to a chemical pre-treatment, before mixing with the polymer matrix [31].

More recently, nanometre scale structuring of organic/inorganic hybrid materials has been synthesized using Polyhedral Oligomeric Silsesquioxane (POSS) based precursors for the inorganic component [32]. The prefabricated silica cage structures of the POSS compounds (Fig. 1.4) provide more control over the


5

structure and content of the SiO2 phase. The organic side groups (R) can be varied so as to maximise bonding to the organic network (Fig. 1.4). An overview of optimum mechanical properties and applications for wet chemically synthesised hybrid thin films is given in Table 1.1.

Figure 1.4: Structural representation of Polyhedral Oligomeric Silsesquioxane (POSS). R can easily be tailored to maximise bonding to organic network.

Table 1.1: Optimal scratch resistant properties achieved for different wet chemically synthesized hybrid materials.

Reference Material After 500 taber cycles Application

Li [33] HPO/TMOS >98 % Glazing Hard Coats

Urreaga [34] TEOS/MPS/MMA Barrel abrasion Ophthalmic

Al(iPA)3/glymol >99 % Ti(iPA)4/glymol 95-96 % Haas [35] Zr(C3H7O)4/glymol <97 % (200 cycles)

Lacquers

These materials have been evaluated as low dielectric constant (low-k) layers in semiconductor devices [36, 37]. Porous SiO2 layers synthesized from various POSS precursors [38, 39, 40] and prepared using sol-gel deposition techniques have produced thin films with dielectric constants typically between 2-3. Sol-gel processes with siloxane precursor have also been extensively studied [41, 42, 43, 44], with dielectric constants as low as 1.7 [45] being reported using TEOS as a precursor. These studies show that the film porosity and dielectric constant decrease as the curing temperature and time increase. Gas phase technologies are an attractive alternative to wet chemical processes as they utilize the same vacuum reactors and toolsets as are already used in the semiconductor industries. Synthesis of porous SiO2 layers from siloxane precursors using expanding thermal plasma [46], PECVD [47, 48] and reactive evaporation of SiO [49] have all been investigated.


6

Sol-gel processes are a popular choice for fabricating low-k films because they inherently produce porous materials. Dielectric layers with void volume fraction as high as 95 % have been reported [44]. Although such materials have dielectric constants around 1, their susceptibility to copper diffusion causes problems when attempting to implement them into semiconductor processes. On the other hand, gas phase deposition techniques inherently deposit dense materials and can easily be implemented into existing semiconductor processes. The complimentary nature of these two techniques has been recognised by a number of researchers who used PECVD [50] or PVD [51] technology to encapsulate wet chemically synthesized xerogels. Several authors have investigated the mechanical properties and other issues concerning the implementation of these porous materials in semiconductor processes [52, 53, 54, 55].

Wet chemical synthesis of hybrid films typically involves several process steps. A coating solution needs to be prepared, the substrate coated via a dip or spinning technique followed by post-deposition curing procedure. Solvents are usually required to prepare the coating solution, and high temperatures (or other energy load) to complete the curing. Gas phase deposition processes are typically one-step solvent free processes and are hence more attractive for the large-scale coating of hybrid thin films. Advantages and disadvantages of wet chemical and gas phase deposition processes are summarized in Table 1.2.

Table 1.2: Advantages and disadvantages of gas phase and wet chemical processing techniques for hybrid materials.

Process Advantage Disadvantage

Wet chemical

Good control of process chemistry

Relatively low

investment costs

Use of potentially harmful solvents

Long curing time and high

temperatures

Additional processing of nano-powders necessary to

achieve dispersion and bonding to plastic matrix

Gas Phase

Single step deposition process

No solvents

No post deposition

curing

Poor control over process chemistry

Relatively high investment

cost


7

Gas phase deposition processes are generally classified as either Chemical Vapour Deposition (CVD) or Physical Vapour Deposition (PVD) processes. 1.2 PVD processes Classical PVD processes differ from CVD processes in that no chemical reactions occur, instead a sample ingot of the desired composition is atomised (either by e-beam, plasma sputtering, thermal evaporation or some other method), transported/processed and deposited onto a suitable substrate (Fig. 1.5). Low pressures are used (10-5 to 10-8 Torr) to ensure mean free path lengths are in the order, or greater than the reactor dimensions and thus sputtered or evaporated particles reach the substrate without being scattered to the reactor walls by the background gas. Film properties and homogeneity are affected by the application of an electric field (causing ionisation of the gas) and varying the process pressure by the introduction of inert or sometimes reactive gas. In the plasma version of the technique, magnetic fields are used to enhance plasma density and thus deposition rate [56]. As no gas collisions occur, PVD technologies are line of sight processes and therefore difficult to implement for three dimensional substrates, requiring the use of mechanised substrate holders. CVD techniques differ from PVD techniques in that chemical reactions occur so that the final film is of a different composition to that of the precursors.

Moltenpool

Pumps

e-beam

Substrate holder

Separationplate

Pumpingport

Vacuumchamber

e-beamsource

Ingotrod

Moltenpool

Pumps

e-beam

Substrate holder

Separationplate

Pumpingport

Vacuumchamber

e-beamsource

Ingotrod

Figure 1.5: Schematic of e-beam evaporation apparatus.

1.3 CVD processes. Chemical vapour deposition is a process whereby solid films are synthesised from chemically reactive gas mixture. The energy required to drive the chemical reactions can be supplied in a number of different forms the most common of


8

which are heat (thermal CVD), electrical discharge (plasma enhanced or activated CVD), photons (photo CVD) and even sound (acoustic CVD). A typical CVD system consists of three components, a gas handling system where suitable precursors are vaporized and dosed, a reactor vessel where precursors react and an exhaust system where harmful reaction waste products are scrubbed and the harmless gases vented. The physical and chemical properties of the precursor are one of the most important considerations when designing a CVD system. A suitable CVD precursor would have the following characteristics: • Volatile and stable during transport to the CVD reactor. • Readily decompose under reaction conditions to form the desired coating. • Low toxicity. • Cost effective. • Not damaging to deposition apparatus.

The choice of activation method depends on the substrate material used. For thermally stable substrates, thermal activation is preferred as it usually produces superior coating properties for all but polymer-type films. 1.3.1 Plasma activated (enhanced) CVD Plasma is (at least for physicists) ionised gas consisting of a mixture of free electrons, ions and neutral atoms (depending on the degree of ionisation) and is formed when sufficient energy is put in to a gas. More than 99 % of matter in the universe exists in the plasma state though nature rarely produces plasma on the earth's surface. In nature plasma is formed when gas is exposed to high temperatures (inside of stars) or high energy radiation (UV radiation in the ionosphere). Under laboratory conditions however, where such high energies are not available, ionised gas is generated using high electric fields and is known as cold plasma or electric discharge. Electrical discharge is the most practical means of creating and sustaining low temperature plasma in the laboratory and many methods of coupling electrical energy into gases to generate plasma have been developed using both d.c. as well as a.c. power sources. The two most common methods for coupling electrical energy into a gas discharge are via capacitive coupling (CCP) (Fig. 1.6a), or via induction as is done with inductively coupled plasma (ICP) (Fig. 1.6b). Inherently different plasma conditions are created with CCP and ICP discharges and the choice of ICP or CCP discharge depends on the application.


9

~

Gas inlet

Pumps

r.f.

Substrate table

r.f.~

Pumps

Gas inlet PressureGauges

a. b.

~

Gas inlet

Pumps

r.f.

Substrate table

r.f.~

Pumps

Gas inlet PressureGauges

a. b.

Figure 1.6: Typical apparatus for CVD experiments, a. direct capacitively coupled parallel plate reactor, b. remote inductively coupled plasma. Capacitively coupled plasmas are characterised by relatively low plasma

density but high energy ion bombardment of the substrate surface. Most of the chemistry occurs on the surfaces within the reactor due to the high energy ion bombardment. High energy ions causes relatively unselective fragmentation of the surface adsorbed species and deposition or etch processes result depending on the power and molecules involved. These plasmas are typically used for thin film deposition/modification and, due to the high energy uni-directional ions, are particularly useful for high aspect ratio anisotropic etching. Remote plasma processes (such as ICP [57], electron cyclotron resonance (ECR) [58], helicon [59, 60], corona [61] and expanding thermal plasmas [62] sources) differ in that they produce higher density plasmas and the precursors are injected downstream of the plasma source. High density and absence of high energy ions at substrate surface (unless additional biasing is applied) results in a higher proportion of the chemistry occurs in the gas phase. These sources are useful in applications where high densities and temperatures are needed without high levels of ion bombardment to the surface.

Gas discharges are intrinsically none equilibrium systems and consequently need to be characterised by several parameters. Plasma is usually described in terms of electron and ion temperatures, Te, Ti, and the electron and ion number densities (number of species per cubic metre of gas) ne and ni. As electrons are point charge carriers with no vibrational or rotational energy levels, their temperature directly equates to their translational energy and is measured in electron volts (eV)*. Low temperature plasma can only be sustained by the continuous power input to compensate for losses. In the case of an electrical * Where 1 eV = e/kB = 11604 K, in which e is the electron charge and kB the Boltzman constant.


10

discharge, this energy is used to sustaining electric fields that accelerate and hence heat the free electrons and ions. If the gas pressure is low enough and hence the mean free path (λmfp) long enough, the electrons reach sufficient energy that they either undergo ionising collision with background neutral gas or go straight to the vessel walls where they recombine with positive ions to reform the neutral gas. If no chemical reactants are injected into the plasma then a majority of the electrical energy put into the discharge leaves the system in the form of heat, sputter processes at the vessel wall and light emission from the plasma glow. For a more thorough description of plasma processing see Chapman [63] or Lieberman et al. [64]. 1.4 Hybrid films via gas phase processes: a literature review A review of literature pertaining to the deposition of hybrid thin films via gas phase technology is sought. Due to the broad project description and the variety of scientific disciplines it encompasses, a strategy in which to approach the literature is imperative. For this reason literature reviewing thermal and plasma activated gas phase deposition techniques, synthesis of ceramic and polymeric materials and the current state-of-the-art regarding nano-composites is put into a literature database. From this database any presence of an overlap in process techniques and condition for the synthesis of ceramic and polymeric phases can be found. If however no overlap in process conditions exists, then it will be necessary to develop a dual source process whereby the ceramic and polymer phases can be independently activated before being combined to form the hybrid film. This approach to the literature is schematically represented in Fig. 1.7.

Deposition oforganic/polymer

coatings

Deposition ofinorganiccoatingsPrecursors

Deposition systemPlasma ( Type )Thermal

Deposition conditionsCoating characteristics

StructureProperties

Overlaps in conditions?Multipleprocess

conditions

One set ofprocess

conditions

Figure 1.7: Flowchart representing approach to the literature review

Literature specific to gas phase processing of 3 main material groups is therefore reviewed, namely,


11

• Silicon based polymers. Gas phase polymerisation of alkoxy silicon

precursors as possible synthetic routes to silica/silicone hybrid materials. • Gas phase polymerisation of hydrocarbons. Gas phase synthesis of

compounds from purely carbon based precursors. • Gas phase particle synthesis. Synthesis and control of particles in gas

phase processes. 1.4.1 Silicon based polymers The plasma polymerisation of siloxane precursors under various conditions is a subject that has received a great deal of attention by researchers and has been extensively reported in the literature. Many authors however, especially those seeking to deposit SiO2, do not classify the plasma polymerised siloxane thin films as organically modified ceramics despite the presence of carbon and hydrogen. Plasma deposition processes can be separated into two classes, direct and remote plasma (see Fig. 1.6). When a siloxane monomer is introduced into plasma, it is activated and fragmented by high-energy electrons and ions. At low pressure and hence long mean free path, activated molecular fragments and ions are more likely to collide with the vessel walls then with other gas phase species (unless they are negatively charged) and hence absorb and react on surfaces within the reactor to form thin films of similar composition and structure to the injected monomer. This process has been used to deposit hybrid films from a variety of organometallic precursors with applications ranging from biomedical films for implants [65], barrier coatings for food packaging [66, 67] to dielectric layers in integrated circuits [68, 69]. The most popular precursors for the deposition of organically modified ceramics are hexamethyldisiloxane (HMDSO) and tetramethoxysilane (TEOS) often in combination with oxygen and/or a variety of dopants. The physical and chemical properties of the resulting films depend heavily on the process parameters, such as, the way in which the power is coupled into the gas (inductive/capacitive), precursor injection point, the power per unit monomer mass flow*, and whether or not an oxidant is added. Coatings produced in this way from siloxane-based precursors are known as Plasma Polymer (PP) coatings and have the general composition SiOxCyHz where x ≤ 2 and y and z can be anything up to the stoichiometry of the injected precursor. If the deposition is performed at relatively low discharge power and without the addition of oxidant, then the film generally has a similar elemental composition to the precursor. A number of studies have been performed which compare film composition to process conditions and precursors [70, 71, 72]. A comparison between the IR spectra of the pure 1,2- bis(trimethyl)siloxyethane (TMSE) precursor and subsequent plasma polymer thin film can be seen in Fig. 1.8. At higher discharge powers, Fracassi et al. [73] found that while the elemental composition of the film was similar to that of the precursor, the concentration of specific molecular groups (C-H, C-O, O-C-O) and hence functionality differed. * Calculated by W/FM where W = power input (Watts), F = monomer mass flow (g/s) and M = relative molecular mass of monomer.


12

0.20.30.40.50.60.70.80.9

1

3800 3500 3200 2900 2600 2300 2000 1700 1400 1100 800 501

Wavenumber [cm-1]

Tran

smis

sion

(0-1

)

CO2

Si-CSi-CHx

C-H

Liquid

Flim

Figure 1.8: Retention of chemical functionality between thin film and precursor for plasma polymerised created from TMSE (from this work, see chapter 3).

The chemistry of the siloxane plasma process is complicated by the wide variety of compounds formed by both carbon and silicon, however some general observations can be made. Inagaki [74] concluded from a study of purely organic monomers, unsaturated compounds exhibit higher deposition rates than their saturated counter parts. The higher deposition rate is attributed to the unsaturated structure of the monomers, and the deposited films appear to contain some of the monomers conjugation. The readily polymerising nature of unsaturated hydrocarbon systems over the saturated counterparts is also true for siloxane precursors containing unsaturated carbon ligands [65, 70].

Addition of oxygen to the siloxane plasma vastly complicates the process chemistry and has been observed in films even when no oxygen is deliberately injected. A majority of authors attribute this to irremovable water molecules adsorbed on the surface of the substrate and deposition system. However, in some cases the hydroscopic nature of the precursor itself can be a cause of oxygen incorporation. A number of authors have reported adding O2 during the plasma polymerisation of a number of siloxane precursors [65, 71, 75, 76, 77, 78, 79] in attempts to deposit films with a SiO2 composition. Oxygen is readily fragmented by the plasma state and forms very reactive atomic oxygen and ozone which react with the siloxane monomers. At low oxygen concentrations, the action of atomic oxygen and ozone dramatically increases film growth rate but can cause powder formation at higher concentrations. Oxygen also readily etches the growing film surface to remove carbon and retards deposition rate. As the oxygen fraction in the process gas increases, the resulting coating composition goes from being similar to the monomer precursor composition, to that of SiO2. However, high purity SiO2 is difficult to attain because of silanol (SiOH) incorporation. Silanol incorporation is not a result of incomplete precursor oxidation, but of reactions between growth precursors and OH radicals that are inevitably present when hydrocarbon and oxygen chemistries are mixed. As an example of such a complex system, a more


13

extensive investigation into the chemistry of the siloxane and oxygen plasma process is presented in chapter 5.

To achieve dense stoichiometric (silanol free) silica from siloxane precursors requires both high density plasma and high substrate temperatures, or alternatively high energy ion bombardment is required. Menichella et al. [80] deposited a “gradient layer” coating by starting with a pure TEOS gas mixture and gradually increasing the oxygen concentration during the deposition process. By gradually increasing the oxygen concentration the deposited the film gradually became more SiO2 like at the top surface. The intermediate layer between the SiO2-like surface and bulk plastic substrate provided a smooth transition in thermal expansion coefficients, while the top SiO2 provided the scratch resistance and hardness required. Similar processes have been reported where alternate plasma polymer-SiO2 layers are deposited to form a multilayer stack. An overview of some silicon based silica/silicone hybrid coating chemistries and applications are summarised in Table 1.3 below.

Table 1.3: Precursors used in the deposition of organically modified ceramic films.

Precursors Dopant/oxidant Properties/ application

HMDSO [65, 75, 76, 77, 72, 69, 78, 79, 71,81, 82, 83, 84, 85, 86, 87,

88] O2, C2F4

Barrier coatings, Adhesion promoters

Low-k TMDSO [65, 76, 84] O2 Barrier coatings

VpMDSO [84] O2 Barrier coatings TMTSO [84] O2 Barrier coatings

TVTMTSO [84] O2 Barrier coatings

TMMOS [89] CF3(CF2)nC2H4Si(OCH3)3, n=0, 5, 7

Hydrophobic coatings

Silane [90] O2 Ophthalmic

coatings Methyl silane [91] -

HMDS [68] - Low-k [92] Ophthalmic

DVS [68] - Low-k TMS [93] O2 Low-k MTS [68] - Low-k


14

Metroline et al. [94] have a number of patents for apparatus and processes for depositing SiO2/Si-polymer-like coatings for protecting metal [95, 96, 97, 98]. 1.4.2 Gas phase polymerisation of hydrocarbons As the plasma will fragment most organic compounds, plasma polymers can be deposited from almost any organic monomer. Polymeric films have been deposited from a number of unsaturated and saturated hydrocarbon compounds such as the styrene monomer [99, 100], ethylene [101], naphthalene [102], ethylene oxide [103, 104], methane [105] and a mixture of fluorocarbons [105, 103, 104, 106]. Typically an organic precursor will undergo numerous fragmentations in the plasma, and the resulting film will form from most of them. Unlike wet chemical synthetic methods, plasma polymers differ in chemical composition from the monomers, even if the same monomers are used in the two processes. Plasma polymers are in most cases highly branched, highly cross-linked, insoluble and adhere to solid surfaces. The chemical and physical properties of the plasma polymers depend on the precursor gas and the type of discharge in a similar way to the siloxane chemistry. An overview of some hydrocarbon based coating chemistries and applications are summarised in Table 1.4 below.

Table 1.4: Some plasma polymers and their applications.

Precursor Additives Process Application

Styrene [100, 107, 99] O2, H2 a.c./d.c., r.f. Electrical components

Ethane [101, 108] - r.f. Barrier/protective coating

1-Fluoro naphthalene[109] - r.f. Not specified

Hexane [103, 104] CH2CO/H2 r.f. Biocompatible coatings

Fluoro ethene, Fluoro ethane, Fluoro butane

[106] H2

40 kHz parallel

plate

Hydrophobic protective coatings

Methane/tetrafluoromethane [105] SiH4/H2 r.f. ICP Anti reflection

coatings

In general, plasma polymers find applications as protective coatings on metals, hydrophobic coatings and insulating layers in electrical applications. The use of highly unsaturated hydrocarbon sources such as C2H2, in combination with high energy plasma, results in the deposition of extremely dense and highly cross linked films. These materials are known as diamond like carbon films (DLC), and are extremely hard. The Monsanto Company hold patents on DLC films [110], for


15

scratch resistant coating applications on glass substrates. Diamonex Inc. also holds a number of patents regarding the deposition of DLC coatings [111]. A number of excellent reviews regarding plasma polymers and plasma polymerisation techniques can be found elsewhere [112, 113, 114].

The only commercial, thermally driven process for applying purely carbon based polymers is the Parylene® processes. Here, one of the parylene precursors (in the form of a xylylene dimer) is vaporized, pyrolized in an oven at 600°C and transported to a cooled substrate where the xylylene radicals condense and react to form a layer. The advantages of the parylene process are that deposition can be done at atmospheric pressure and temperature. The resulting polymers are however, rather soft and porous compared to plasma polymers, but have nevertheless been considered for application such as encapsulation layers for poly LED devices [115] and a substitute for SiO2 as low dielectric in chip manufacturing [116, 117]. In a unique study Senkevich et al. [118], synthesised and incorporated silica nano powders into growing parylene thin films. SiO2 particles were generated by thermal decomposition of diacetoxy-di-tert-butoxysilane (DADBS) in oxygen and transported, via predominately gas flow, to the deposition chamber and incorporated into Parylene-C films. Parylene-C was synthesised from dichloro-di-paraxylylene (DPXC) and both DPXC and DADBS were simultaneously decomposed and activated at 615°C and 600°C respectively prior to entering the deposition zone, which was at room temperature.

1.4.3 Gas phase particle synthesis In the CVD process, a feed source gas is first decomposed by heat or plasma, followed by diffusion of the decomposition products onto the substrate to form a thin film. The decomposition products can also be used to form particles via either homogeneous gas phase nucleation or heterogeneous nucleation at surfaces within the reactor. These particles can then go on to be incorporated into the growing film. In general particle or powder formation in CVD processes is undesirable, and in the semiconductor industry great lengths are taken to prevent particles from entering the reactors [119]. In some processes, however, homogeneous gas phase nucleation is encouraged as the incorporation of particles into the growing film can offer some beneficial properties. Particle precipitation during CVD results in porous layers that can be used as supports for catalysts, as membranes or porous electrodes. If excess homogenous nucleation occurs, and a majority of the precursor is consumed before reaching the substrate, then loose powdery deposits are formed instead of a coating. A summary of some synthetic processes is listed in Table 1.5.


16

Table 1.5: Precursors and processes used for the gas phase synthesis of particles.

A majority of gas phase powder forming processes are in fact designed for synthesising particles, which are later sintered into bulk ceramic materials. In most of these methods gaseous precursors are introduced into a reactor system whereby energy is supplied causing homogenous nucleation reactions to occur which result in the formation of particles. Energy used to drive the chemical reactions can be supplied in a number of different forms including high power lasers [128], microwave discharge [129] and r.f. magnetron sources [130].

E. Whitby et al. [121] found that particle nucleation is more favourable where high temperature gradients exist. This is a result of high monomer activation in the high temperature region, followed by high concentrations in the low temperature region. This principle was used by Okuyama et al. [131] where they used a six-zone furnace to synthesise SiO2 particles. A similar effect occurs in the CO2 laser excitation processes [128, 123, 124] as depicted in Fig. 1.9. As the precursor gas flow passes through the CO2 laser beam, rapid thermal excitation of the gas occurs.

Particle Precursor Process

SiO2

TEOS [120], TRIES [120], TMOS [120], OMCTS

[120], SiH4 [121] + O2/O3 Thermal excitation

Mo Mo(CO)6 [129] Microwave Plasma

CeO2/CexY1-xO2-y CeCl3, YCl3 + O2 [122] Thermal excitation (size 9-80 nm)

Si SiH4 [128, 123] CO2 laser excitation Si3N4 SiH4 + NH3 [124] CO2 laser excitation

SiC SiH4 + CH4 [124] CO2 laser excitation

Pt-TiO2 Pt, Ti + O2 [130] Joule heating / Magnetron

sputtering (size 30-150 nm)

Pt-Al2O3 Pt, Al + O2 [130] Joule heating / Magnetron sputtering.

TiN TiCl4 NH3 [125] Thermal excitation

ZrO2 Zr(CO)4 [126],

Zr(n-C3H7O)4 [127] + O2

Inert gas condensation / flame pyrolysis (size 0.3-1

µm).


17

The excited gas then leaves the laser beam and rapidly cools, nucleates and forms particles.

Water cooled copper beam stop

Laser

CO2 laser

Inert gas inlet

Precursor inlet

Reaction flame

Filter/Particle collection

Figure 1.9: Laser activated particle-forming processes, as used by Cannon et al. [124].

Once formed the behaviour of very small particles (>20 nm) is primarily

governed by diffusion. For larger particles, thermophoretic or electrophoretic forces begin to govern the particles behaviour. In order to understand how particles behave in CVD reactors, a number of authors have attempted to model the processes [132]. Whitby et al. [121] describes a number of plausible explanations for the formation and movement of particles in a thermal CVD reactor. Amongst other things their numerical studies indicated that particle nucleation shows a steep dependency on pressure. Whitby et al. and Fujimoto et al. [120] proposed mechanisms for the formation/incorporation of particles and the transport of particles during the deposition of SiO2 films. Bai et al. [122] thermally decomposed CeCl3 and YCl3 in the presence of O2 to form CeO2 and Yttrium doped CeO2 powders. They characterised the collected powders to find they had a size distributions of either 9-30 or 30-80 nm depending on whether they collected the material in a cooled container outside of the reactor zone (9-30 nm), or an actively cooled susceptor situated inside the reactor (30-80 nm).

In the semiconductor and micro electronic deposition systems, particle production is an undesirable by-product of some plasma processing steps. Device failure as a result of dust contamination is a familiar issue and has prompted research into particle formation and transport mechanisms in the hope that it might be avoided [133]. A prominent area of literature regarding particle contamination deals with silane chemistries [134, 135, 136, 137, 138, 139] due to its wide spread application in industrial processes. Although particles have been observed in TEOS and HMDSO [140] plasma discharges, very little has been done in elucidating the mechanism by which they are formed. Fujimoto et al. [141] investigated particulate


18

formation from four siloxane precursors (triethoxysilane, tetramethylorthosilicate, octamethylcyclotetrasiloxane and TEOS) in an atmospheric pressure, thermally activated CVD process and determined TEOS, with the lowest overall reaction activation energy, would most readily form particles. A more detailed study on the formation and transport of nanometre sized particles formed in a TEOS/oxygen plasma is presented in chapter 4.

Thin film and particle properties depend on the synthesis technique. Wet chemical processes produce very porous soft materials that require some form of curing to achieve densification. PECVD and PVD materials are inherently denser, but rarely produce fully dense crystalline materials without additional substrate heating. Atmospheric pressure, high temperature CVD processes generally produce the hardest and most crystalline materials. It is therefore reasonable to assume that particle generated in low pressure discharges will also exhibit some level of porosity.

Porosity in PECVD synthesized silica and siloxane layers can be controlled by applying a d.c. bias potential (as demonstrated in chapter 3). The particles however, are formed from negative ions trapped in the plasma glow and the sheath surrounding them, and hence the energy of ion impinging on them, is much lower than the sheath above the electrode (as described in chapter 4). As the formation of dense fully oxidized SiO2 networks requires high energy input, (as seen with SiO2 deposition in chapter 3) the energy is not available to form dense particles. 1.5 Conclusions Very little has been reported about the gas phase deposition of nanocomposite materials, especially where inorganic and organic phases are combined. A great deal of literature is available regarding the plasma polymerisation of various siloxane precursors for a wide variety of applications, including scratch resistant layers and low-k materials.

• The synthesis of dense siloxane films is done using capacitively coupled plasma configurations where self biasing helps densify the layer.

• Films composition and structure can be characterised using FTIR

spectroscopy, XPS and Auger spectroscopy.

• The bonding of these films has to some extent been revealed using techniques such as FTIR and X-ray photon spectroscopy. Unfortunately, elucidation of the exact connection between the inorganic and organic components of these materials has not been fully characterised.

Although the authors rarely describe their films as hybrid materials, they are

never the less composed of both inorganic and organic components. From a hybrid-material point of view, a majority of these coatings are perhaps best described as organically modified ceramics, primarily inorganic with organic functionality. In cases where no oxygen is added, and relatively low energy plasma are used, the materials resemble highly cross-linked silicon polymers.


19

In order to maintain a good control of film composition, the inorganic and organic phases need to undergo activation separately and the active species combined in the deposition zone housing the substrate. Using this method, the inorganic and organic phases can be separately optimised before they are combined to generate the single coating. The reactor must also be flexible, so that it can be used to deposit the organically modified ceramic layers as well as nanocomposite layers as described in Fig. 1.1. 1.6 Scope of this thesis The primary aim of this work is to develop a process for the synthesis of hybrid materials using solely gas phase synthetic methods. To achieve this, an understanding of gas phase deposition technologies is necessary, brief introduction to which has been presented. From this understanding, process conditions required to synthesize nanocomposite materials can be identified. The strategy to achieve this is based on the continuous adaptation of a dual plasma reactor which is used throughout this study in various configurations. The dual plasma system provides independent control over conditions for the synthesis of both phases of the final material.

The thesis is divided into seven chapters. An introduction to hybrid materials has already been presented detailing the various classifications of hybrid materials, their origins from wet chemical processes and the most promising techniques from which they might be deposited using purely gas phase deposition technology. This is followed in chapter 2, by an introduction into plasma physics in a study of the helicon etch machine. This work is performed at the Australian National University in Canberra and investigates the feasibility of etching silicone based hybrid materials using SF6, CF4 and O2 as active etch gases.

Knowledge obtained from reviewing the literature is applied in chapter 3 where the design of a dual plasma reactor is discussed along with preliminary deposition results of the polymer and ceramic layers. The synthesis and characterisation of nanometre sized silica particles for incorporation into polymer layers is discussed in chapter 4. Particles formed and trapping in an electric field are monitored using in-situ FTIR. Ex-situ characterisation is performed using XPS and ESEM and the formation of macroscopic structures on the electrode is studied with the aid of fluid dynamic models. An in depth look at the intrinsic kinetics of the TEOS/O2 plasma system is presented in chapter 5 with a view to determining mechanisms for the formation of SiO2 particles/film and gaseous by-products of the process.

Particles synthesized in the low pressure discharge turned out to be porous and offered little to no mechanical enhancement of the hybrid layers. However, reports of nano-porous materials for as dielectric materials prompted a study into synthesizing porous layers for low-k applications. A twin capacitive plasma system was developed for this and is the subject of chapter 6. The thesis is concluded in chapter 7.


20

References [1] Veprek S., Thin Solid Films, 297, 145-153, (1997). [2] Karvankova P., Veprek-Heijman M.G.J., Zindulka O., Bergmaier A., Veprek S., Surface

and Coatings Technol., 163-164, 149-156, (2003). [3] Hadjadi A., Beorchia A., Roca I Cabarrocas P., Boufendi L., Thin Solid Films, 403-404,

139-143, (2002). [4] Huet S., Viera G., Boufendi L., Thin Solid Films, 403-404, 193-196, (2002). [5] Wu G., Wang J., Shen J., Yang T., Zhang Q., Zhou B., Deng Z., Bin F., Zhou D., Zhang.,

J. Non-Cryst. Solids, 275, 169-174, (2000). [6] Haereid S., Dahle M., Lima S., Einarsrud M.-A., J. Non-Cryst. Solids, 186, 96-103,

(1995). [7] Chaudhury N.K., Bhardwaj R., Murari B.M., Current Appl. Phys., 3, 177-184, (2003). [8] Yu M., Lin J., Wang S.B., J. Alloys Comp., 344, 212-216. (2002). [9] Medda S.K., Kundu D., De G., J. Non-Cryst. Solids, 318, 149-156, (2003). [10] Sol-Gel Science: The physics and chemistry of sol-gel processing, Brinker J., Scherer

G.W., Academic Press, (1990). [11] Innocenzi P., Brusatin G., Guglielmi M., Bertani R., Chem. Mat., 11, 1672-1679, (1999). [12] Matejka L., Dukh O., Brus J., Simonsick-Jr. W.J., Meissner B., J. Non-Cryst. Solids,

270, 34-47, (2000). [13] Noell J.L.W., Wilkes G.L. Mohanty D.K., J. Appl. Polym. Sci., 40, 1177-1194, (1990). [14] Laczka M. Cholewa-Kowalska K., Kogut M., J. Non-Cryst. Solids, 287, 10-14, (2001). [15] Frohlich L., Houbertz R., Jacob S., Popall M., Mueller-Fiedler R., Graf J., Munk M., von

Zychlinski H., Mat. Res. Soc. Symp. Proc., 726, 349-354, (2002). [16] Polmanteer K.E., Lentz C.W., Rubber Chem. Technol., 48, 795, (1975). [17] Dannenberg E.M., Rubber Chem. Technol., 48, 410, (1975). [18] Wagner M.P., Rubber Chem. Technol., 49, 703, (1976). [19] Kraus G., Rubber Chem. Technol., 51, 297, (1978). [20] Boonstra B.B., Polymer, 20, 691, (1979). [21] The Chemistry of Silica, Iler R.K., Wiley, New York, (1979). [22] Alexandre M., Dubois P, Mater. Sci. Eng., 28, 1-63, (2000). [23] Mohallen N.D.S., Seara L.M., Appl. Surf. Sci., 214, 143-150, (2003). [24] Yano S., Nakamura K., Kodomari M., Yamauchi N., J. Appl. Polym. Sci., 54, 163-176,

(1994). [25] Yano S., Iwata K., Kurita K., Mater. Sci. Eng., C, 6, 75-90, (1998). [26] Ning Y.-P., Tang M.-Y., Jiang C.-Y., Mark J.E., J. Appl. Polym. Sci., 29, 3209-3212,

(1984). [27] Sohoni G.B., Mark J.E., J. Appl. Polym. Sci., 45, 1763-1775, (1992). [28] Landry C.J.T., Coltrain B.K., Landry M.R. Fitzgerald J.J., Long V.K., Macromolecules,

26, 3702-3712, (1993). [29] Mauritz K.A., J. Appl. Polym. Sci., 40, 1401-1420, (1990). [30] Suzuki F., Onozato K., Kurokawa Y., J. Appl. Polym. Sci., 39, 371-381, (1990). [31] Bauer F., Sauerland V., Ernst H., Glasel H.-J., Naumov S., Mehnert R., Macromol.

Chem. Phys., 204, 375-383, (2002). [32] Tsai M-H., Whang W-T., Polymer, 42, 4197-4207, (2001). [33] Li C., Jordens K., Wilkes G.L., Wear, 242, 152-159, (2000). [34] Urreaga J.M., Matias M.C., Lorenzo V., de la Orden M.U., Mat. Lett., 45, 293-297,

(2000).


21

[35] Haas K.-H., Amberg-Schwab S., Rose K., Schottner G., Surf. Coat. Technol., 111, 72-79,

(1999). [36] Jeng S.P., Havemann R.H., Chang M.C., Mater. Res. Soc. Symp. Proc., 35, 337, (1994). [37] Ryan J.G., Geffken R.M., Poulin N.R., Paraszczak J.R., IBM J. Res., Dev., 39, 371,

(1995). [38] Liu P.T., Chang T.C., Sze S.M., Pan F.M., Mei Y.J., Wu W.F., Tsai M.S., Dai B.T.,

Chang C.Y., Shih F.Y., Huang H.D., Thin Solid Films, 332, 345-350, (1998). [39] Baklanov M.R., Mogilnikov K.P., Microelectron. Eng., 64, 335-349, (2002). [40] Xu J., Moxom J., Yang S., Suzuki R., Ohdaira T., Appl. Surf. Sci., 194, 189-194, (2002). [41] Chang T.C., Tsai T.M., Liu P.T., Mor Y.S., Chen C.W., Mei Y.J., Sheu J.T., Tseng T.Y.,

Thin Solid Films, 420-421, 403-407, (2002). [42] Uchida Y., Katoh T., Oikawa M., Mater. Sci. Semicond. Process., 5, 259-264, (2003). [43] Jain A., Rogojevic S., Ponoth S., Agarwal N., Matthew I., Gill W.N., Persans P.,

Tomozawa M., Plawsky J.L., Simonyi E., Thin Solid Films, 398-399, 513-522, (2001). [44] Kim R.-H., Park H.-H., Hyun S.-H., Yeom G.-Y., Thin Solid Films, 332, 444-448,

(1998). [45] Liu P.-T., Chang T.C., Hsu K.C., Tseng T.Y., Chen L.M., Wang C.J., Sze S.M., Thin

Solid Films, 414, 1-6, (2002). [46] Creatore M., Barrell Y., Kessels W.M.M., van de Sanden M.C.M., Mat. Res. Soc. Symp.

Proc., 766, (2003). [47] Shaniryan D., Weidner K., Gray W.D., Baklanov M.R., Vanhaelemeersch S., Maex K.,

Microelectron. Eng., 64, 361-366, (2002). [48] Borvon G., Goullet A., Mellhaoui X., Charrouf N., Granier A., Mater. Sci. Semicond.

Process., 5, 279-284, (2003). [49] Si J.-J., Show Y., Banerjee S., Ono H., Uchida K., Nozaki S., Morisaki H.,

Microelectron. Eng., 60, 313-321, (2002). [50] Schulz S.E., Koerner H., Murray C., Streiter I., Gessner T., Microelectron. Eng., 55, 45-

52, (2001). [51] Iacopi F., Tokei Zs., Stucchi M., Brongersma S.H., Vanhaeren D., Maex K.,

Microelectron. Eng., 65, 123-131, (2003). [52] Volinsky A.A., Vella J.B., Gerberich W.W., Thin Solid Films, 429, 201-210, (2003). [53] Travaly Y., Eyckens B., Carbonel L., Rothschild A., Le Q.T., Brongersma S.H., Ciofi I.,

Struyf H., Furukawa Y., Stucchi M., Schaekers M., Bender H., Rosseel E., Vanhaelemeersch S., Maex K., Gaillard F., Van Autryve L., Rabinzohn P., Microelectron. Eng., 64, 367-374, (2002).

[54] Lanckmans F., Maex K., Microelectron. Eng., 60, 125-132, (2002). [55] Lee H.J., Oh K.S., Choi C.K., Surf. Coat. Technol., 171, 296-301, (2003). [56] Jehn H.A., Surf. Coat. Technol., 112, 210-216, (1999). [57] Zhang J.L., Yu S.J., Ma T.C., Vacuum, 65, 327-333, (2002). [58] Bulkin P., Bertrand N., Drevillon B., Thin Solid Films, 296, 66-68, (1997). [59] Boswell R.W., Chen F.F., IEEE T. Plasma Sci., 25, (1997). [60] Hayden D.B., Juliano D.R., Neumann M.N., Allain M.C., Ruzic D.N., Surf. Coat.

Technol., 120-121, 401-404, (1999). [61] Thyen R., Weber A., Klages C.-P., Surf. Coat. Technol., 97, 426-434, (1997). [62] Kessels W.M.M., Severens R.J., Smets A.H.M., Korevaar B.A., Adriaenssens G.J.,

Schram D.C., van de Sanden M.C.M., J. Appl. Phys., 89, (2001). [63] Glow Discharge Processes, Chapman B.N., Wiley & Son, New York, (1980). [64] Principles of plasma discharges and materials processing, Lieberman M.A., Lichenberg

A.J., Wiley, New York, (1999).


22

[65] Garbassi F., Occhiello E., Macromol. Symp., 139, 107-114, (1999). [66] Mutikainen R., Thin Solid Films, 238, 248-257, (1994). [67] Schuster F., Peit G., US patent Nr. US5437725, (1995). [68] Nguyen V.S., Underhill J., Fridmann S., Pan P., J. Electrochem. Soc.: Solid-State Sci.

Technol., 132, 1925-1932, (1985). [69] Shirafuji T., Miyazaki Y., Nakagami Y., Hayashi Y., Nishino S., Jap. J. Appl. Phys., 38,

4520-4526, (1999). [70] Gengenbach T.R., Griesser H.J., Polymer, 40, 5079-5094, (1999). [71] Alexander M.R., Short R.D., Jones F.R., Michaeli W., Blomfield C.J., Appl. Surf. Sci.,

137, 179-183, (1999). [72] Grundmeier G., Stratmann M., Bunsenges B., Phys. Chem., 11, 1387-1392, (1995). [73] Fracassi F., d’Agostino R., Favia P., J. Electrochem. Soc., 139, 2636-2643, (1992). [74] Plasma surface modification and plasma polymerisation, Inagaki N., Technomic

Publishing Company Inc., (1996). [75] Korzec D., Traub K., Werner F., Engemann J., Thin Solid Films, 281-282, 143-145,

(1996). [76] Morra M., Occhiello E., Garbassi F., J. Appl. Polym. Sci., 48, 1331-1340, (1993). [77] Agres L., Segui Y., Delsol R., Raynaud P., J. Appl. Polym. Sci., 61, 2015-2022, (1996). [78] Taylor C.E., Boerio F.J., Ward S.M., Ondrus D.J., Dickie R.A., Brutto M.M., J. Adhes.,

69, 237-261, (1999). [79] Taylor C.E., Boerio F.J., J. Adhes., 69, 217-236, (1999). [80] Menichella S., Misiano C., Simonetti E., De Carlo L., Carrabino M., Society of Vacuum

Coaters, 37th Annual Technical Conference Proceedings, 37-40, (1994). [81] Poll H.-U., Meichsner J., Arzt M., Friedrich M., Rochotzki R., KreyBig E., Surf. Coat.

Technol., 59, 365-370, (1993). [82] Kusabirake M., Aozasa M., Thin Solid Films, 281-282, 533-535, (1996). [83] Fujii T., Hiramatsu M., Nawata M., Thin Solid Films, 343-344, 457-460, (1999). [84] Yu Q.S., Yasuda H.K., J. Polym. Sci.: Part A: Polym. Chem., 37, 1577-1587, (1999). [85] Bonnar M.P., Wilson J.I.B., Burnside B.M., Reuben R.L., Gengenbach T.R., Griesser

H.L., Beamson G., J. Mater. Sci., 33, 4843-4855, (1998). [86] Turner R.H., Segall I., Boerio F.J., Davis G.D., J. Adhes., 62, 1-21, (1997). [87] Creatore M., Barrell Y., Kessels W.M.M., van de Sanden M.C.M., Mat. Res. Soc. Symp.

Proc., 766, (2003). [88] High Rate Plasma Deposition of Silicon Oxide Like Films, van Hest M., PhD Thesis,

Technische Universitiet Eindhoven, (2002). [89] Takai O., Hozumi A., Sugimoto N., J. Non-Cryst. Solids, 218, 280-285, (1997). [90] Klug W., Gegenwart R., Herrmann R., Schneider R., SPIE Hard Mater. in Optics, 1275,

135-141, (1990). [91] Nault M., Weidman T., Sugiarto D., Mui D., Lee C., Yang J., J. Vac. Sci. Technol. B, 16,

3730-3733, (1998). [92] Samson F., Surf. Coat. Technol., 81, 79-86, (1996). [93] Shaniryan D., Weidner K., Gray W.D., Baklanov M.R., Vanhaelemeersch S., Maex K.,

Microelectron. Eng., 64, 361-366, (2002). [94] Metroline website http://www.metrolineindustries.com. [95] US Patent 5,298,587, issued March 29, (1994). [96] US Patent 5,320,875, issued June 14, (1994). [97] US Patent 5,433,786, issued July 18, (1995). [98] US Patent 5,494,712, issued February 27, (1996). [99] Sawa G., Yamanaka S., Nakamura S., Yamaguchi S., Jap. J. Appl. Phys., 20, 201-204,

(1981).


23

[100] Bui A., Carchano H., Sanchez D., Thin Solid Films, 13, 207-211, (1972). [101] Ooij W.J. van, Sabata A., Zeik D.B., Taylor C.E., Boerio F.J., Clarson S.J., Symposium,

6-42, (1993). [102] Kamo H., Yudasaka M., Kurita S., Matsui T., Kikuchi R., Ohki Y., Yoshimura S.,

Synth. Met., 68, 61-63, (1994). [103] Schue F., Clarotti G., A. Ait Ben Aoumar, Sledz J., Mas A., Geckeler K.E., Gopel W.,

Orsetti A., Marcromol Reports, A31, 1161-1175, (1994). [104] Clarotti G., A. Ait Ben Aoumar, Schue F., Sledz J., Macromol. Chem., 192, 2581-2590,

(1991). [105] Vanhulsel A., Dekempeneer E., Wegener W., Meneve J., Smeets J., Soc. Vac. Coat.

42nd Annual Technical Conference Proceedings, ISSN 0737-5921, (1999). [106] Lin Y., Yasuda H., Miyama M., Yasuda T., J. Poly. Sci.: Part A: Polym. Chem., 34,

1843-1851, (1996). [107] Stuart M., Proc. IEE., 112, 1614-1616, (1965). [108] Yasuda H., Hirotsu T., J. Polym. Sci.: Polym. Chem. Ed., 16, 313-326, (1978). [109] Munro H.S., Till C., J. Polym. Sci.: Polym. Chem. Ed., 24, 279-286, (1986). [110] US Patent, 5,635,245, issued June 3 (1997). [111] World Patent, WO 95/24275, issued September 14, (1995). [112] Yasuda H., Hirotsu T., J. Polym. Sci.: Polym. Chem. Ed., 16, 743-759, (1978). [113] Plasma surface modification and plasma polymerization, Inagaki N., Wiley, New York,

(1996). [114] Yasuda H., J. Polym. Sci.: Macromol. Rev., 16, 199-293, (1981). [115] Staring E.G.J., Braun D., Rikken G.L.J.A., Demandt R.J.C.E., Kessener Y.A.R.R.,

Bouwmans M., Broer D., Synth. Met., 67, 71-75, (1994). [116] Erjavec J., Sikita J., Beaudoin S.P., Raupp G.B., Mater. Lett., 39, 339-342, (1999). [117] Hanyaloglu B., Aydinli A., Oye M., Aydi E.S., Appl. Phys. Lett., 74, 606-608, (1999). [118] Senkevich J.J., Desu S.B., Chem. Mater., 11, 1814-1821, (1999). [119] Hu S.C., Chuah Y.K., Yen M.C., Building and Environment, 37, 201-208, (2002). [120] Fujimoto T., Okuyama K., Yamada S., Adachi M., J. Appl. Phys., 85, 4196-4206,

(1999). [121] Whitby E., Hoshino M., J. Electrochem. Soc., 143, 3397-3404, (1996). [122] Bai W., Choy K.L., Stelzer N.H.J., Schoonman J. Solid State Ionics, 116, 225-228,

(1999). [123] Botti S., Celeste A., Coppola R., Appl. Organometallic Chem., 12, 361-365, (1998). [124] Cannon W.R., Danforth S.C., Flint J.H., Haggerty J.S., Marra R.A., J. Am. Ceram. Soc.,

65, 324-340, (1982). [125] CVD Techniques for the synthesis or modification of porous ceramics, Dekker J.P., PhD

Thesis, Technische Universitiet Eindhoven, (1994). [126] Skandan G., Nanostruct. Mater., 5, 111-126, (1995). [127] Yuan F.L., Chen C.H., Kelder E.M., Schoonman J., Solid State Ionics, 109, 119-123,

(1998). [128] Botti S., Celeste A., Appl. Phys. A., 67, 421-424, (1998). [129] Liu B., Gu H., Chen Q., Mat. Chem. Phys., 59, 204-209, (1999). [130] Ying J.Y., J. Aerosol Sci., 24, 315-338, (1993). [131] Okuyama K., Fujimoto T., Hayashi T., AIChE J.: Ceramics Process., 43, 2688-2697,

(1997). [132] Tandon P., Rosner D.E., AIChE J., 42, 1673-1684, (1996). [133] Dusty Plasmas, Bouchoule A., Wiley, New York, (1999). [134] Stoffels E., Stoffels W.W., Kroesen G.M.W., Hoog F.J.de., J.Vac. Sci. Technol., A 14,

556-561, (1996).


24

[135] Boufendi L., Hermann J., Bouchoule A., Dubreuil B., Stoffels E., Stoffels W.W., Giorgi

M.L. de., J. Appl. Phys., 76, 148-153, (1994). [136] Cabarrocas P.R.I.,Gay P., Hadjadj A., J. Vac. Sci. Technol. A., 14, 655-659, (1996). [137] Dorier J.-L., Hollenstein Ch. Howling A.A., J. Vac. Sci. Technol. A., 13, 918-926,

(1995). [138] Watanabe Y., Shiratani M., Plasma Sources Sci. Technol., 3, 286-291, (1994). [139] Hollenstein Ch., Howling A.A., Courteille C., Magni D., Scholz S.M., Kroesen

G.M.W., Simons N., Zeeuw W de., Schwarzenback W., J. Phys. D: Appl. Phys., 31, 74-84, (1998).

[140] Courteille C., Magni D., Deschenaux Ch., Fayet P., Howling A.A., Hollenstein Ch., Soc. Vac. Coat. 41st Annual Technical Conference proceedings, (1998).

[141] Fujimoto T., Okuyama K., Yamada S., Adachi M., J. Appl. Phys., 85, 4196-4206, (1999).

Chapter 2 Anisotropic plasma etching of SiCxOyHz thin films

25

Chapter 2: Anisotropic plasma etching of SiCxOyHz thin films

Abstract This chapter looks at the characterization of a remote helicon plasma discharge using a Langmuir probe as part of a feasibility study for etching silicone like thin films. Etching is done using SF6, CF4, Ar and O2 or mixtures thereof. The plasma reactor is characterised using a Langmuir probe with a planar nickel tip and used to measure ion currents and floating potentials for the various etch gases used. Etch rates and etch profiles are determined using SEM images and step profile measurements.*

2.1 Introduction The potential ability to tailor the chemical and mechanical properties of hybrid materials makes them very attractive for a large number of applications, many of which were discussed in chapter 1. Applications such as micro electro mechanical systems (MEMS) [1], low-k dielectrics [2] and optical waveguides [3], however, require post deposition processing of the hybrid thin films in order to integrate them into these applications. One important post deposition process is anisotropic reactive ion etching (RIE) [4, 5] and is currently the only technique that offers chemically selective, anisotropic etching at sub-micron dimensions. RIE involves the ionization of selected etch gases to form reactive ions and radicals such as fluorine or oxygen. The reactive species react with the substrate surface to form volatile by-products that leave the surface and are pumped away. For Si and SiO2 this process is well understood and aspect ratios of 10:1 are readily achieved in industrial processes [6]. The etching gas mixture however, is specific for the material being processed and etch rate and etch profile are critically dependant on the plasma conditions and gases used.

In this study, a remote, inductively coupled r.f. helicon discharge is evaluated as a means of etching high aspect ratio nanometre size holes in organically modified ceramic thin films of SiCxOyHz composition. The plasma is characterised using a Langmuir probe and various process and plasma parameters are correlated to the etch rates. Etching is performed using either one, or combinations of SF6, CF4 and O2 as reactive gases. Etch rates and profiles are controlled by means of varying the discharge power in the helicon source and r.f. bias applied to the substrate in addition to the overall gas composition. Preliminary results of etch

* This work was performed at the Australian National University in Canberra as part of a three month traineeship.


26

profiles and substrate degradation are investigated using scanning electron microscopy (SEM). 2.1.1 Reactive ion etching Silicon, like other materials commonly used in the manufacture of micro electronic devices, forms volatile compounds when reacted with the halide elements. RIE processes take advantage of this and etching of Si and SiO2 thin films is achieved by forming the respective silicon fluoride, or chloride. Fluorine is used almost exclusively over chlorine in RIE processes as it has a higher reactivity and as chlorine has the tendency to form hydrochloric acid that corrodes metal surfaces in the reactor system and pumps. Fluorine is introduced into the reactor in the form of short chain fluorocarbons (CFC) such as CF4 [7, 8], CF3H [9] and/or C2F6 [10] but gases such as SF6 [11, 12] and NF3 [13] are also widely used. Hydrogen is added to the etch chemistry (in the form of H2 or as a partially hydrogenated CFC such as CF3H) for etching SiO2 to facilitate the formation of H2O and enhance the etch rate. In contrast, the addition of small amounts of oxygen to the Si etch chemistry is observed to enhance the etch rate as it oxidized the residual carbon layer resulting from fluorocarbon decomposition. Excess oxygen however, can cause oxidation of the photo resist mask and react with Si to form SiO2, retarding the etch rate. Figure 2.1 depicts some of the processes involved in the RIE of SiO2 using an CF4/H2 etch gas mixture. The standard etch model begins with dissociation of the gas via electron impact and ion bombardment of the substrate surface. The substrate surface is activated by the ion bombardment and ‘dangling’ bonds are formed. These ‘dangling bonds’ react with fluorine and oxygen species in the gas to form volatile compounds. Volatile compounds formed at the surface finally desorb and are pumped away.

CF4 + e → CF3 + F + e

SiF2 O

FSiF4

F

SiF4

SiO2 Substrate

Mask

H

H2O

OO

& H2 + e → H + H + ePlasma activation

CF4 + e → CF3 + F + e

SiF2 O

FSiF4

F

SiF4

SiO2 Substrate

Mask

H

H2O

OO

& H2 + e → H + H + ePlasma activation

Figure 2.1: Anisotropic reactive ion etching (RIE) of a silica substrate. Fluorine reacts to form volatile SF4 gas, while H2 reacts to remove oxygen.


27

The presence of both oxygen and carbon in silicone-based hybrid thin films complicates the plasma etch chemistry. Hydrocarbon materials, such as photo resist masks, and oily contaminations can be removed using oxygen plasma [14, 15]. Several authors have investigated the effect of oxygen plasmas on SiOxCyHz coating prepared using the sol-gel process [16, 17]. The magnitude of the overall effect of oxygen plasma treatment on these materials varies depending on the particular material density and structure. Impinging oxygen species react with the film to remove carbon and hydrogen moieties causing a change in the chemical composition. Ion bombardment also puts energy into the film and prolonged exposure causes film densification. Brief exposure however, has also been observed to give rise to the formation of silanol groups in the film [18].

In RIE processes the anisotropy of the etch profile is predominantly controlled by the high directionality of the impinging reactive ions. This condition is readily formed in parallel plate capacitively coupled reactor configurations, where the substrate is located on the driven electrode. In this configuration a net negative d.c. bias potential forms between the plasma and the electrode. Positive ions traversing this sheath are accelerated towards the substrate table and ion bombardment of the substrate occurs. In addition to increasing anisotropy of the etch profile and etch rate, r.f. biasing also subjects the substrate to higher heat load. By coupling more power into the discharge, higher reactive ion concentrations can be generated and the etch rate increased. Unfortunately, even at relatively low power levels (and therefore low ion density) high potentials are reached which can cause damage to sensitive substrates. In order to couple more power into the plasma and increase etch rate, it is necessary to actively cool the substrate. This removes some of the heat generated by the ion bombardment but is difficult to implement and does not compensate implantation and sputtering effects that occur at very high ion energies.

Capacitively coupled plasmas inherently exhibit much lower plasma densities than a corresponding inductive system. This is due to the high sheath potentials accelerating ions to higher energy before they collide with the electrode. As energy is used to accelerate ions through the sheath, less energy is available for ionisation and therefore the plasma density is lower than the corresponding inductive discharge. An ideal etch process is therefore governed by two criteria, 1. High plasma density to generate high concentration of reactive species. 2. Control of sheath potential above substrate to control ion energy.

Still higher plasma densities can be achieved by applying an axial magnetic field to an inductive plasma discharge. Axial magnetic fields increase the plasma density in two ways. As free electrons in the discharge cannot cross the magnetic field lines, electron losses to the reactor walls are reduced. The electron confinement results in a slight increase in the plasma density. However, more significantly the magnetic confinement enables waves to be launched in the plasma that provide additional electron heating mechanisms and couple energy into the discharge.

One successful magnetised plasma system is the helicon plasma [19, 20]. The helicon plasma source is comprised of a quart tube surrounded by a two loop


28

antenna diametrically placed on the outside of the source tube surrounded by confining magnetic coils. R.f. power supplied to the antenna produces a transverse r.f. magnetic field perpendicular to both the tube axis and a constant axial magnetic field. This r.f. field excites the m = 1 azimuthal mode of a helicon wave in the source tube. Energy is transferred from this wave to the plasma electrons via a mechanism known as Landau damping [21, 22, 23]. The additional electron heating, provided by the Landau damping mechanism, enables the helicon plasma sources to achieve plasma densities considerably higher than conventional inductive discharges. When used in etch applications [24, 25], the helicon source is typically placed above a diffusion chamber housing a substrate table. In this configuration the reactive ion concentration is determined by the helicon power, without generating high electric fields above the substrate. The addition of a separate r.f. bias to the substrate table is used to increase both the etch rate and achieve an anisotropic etch profile. 2.2 Experimental The helicon etch reactor consists of two sections, a source and a diffusion chamber (Fig. 2.2). The helicon source consists of a glass tube 15 cm in diameter and 30 cm long surrounded by the helicon antenna and two solenoids. A 13.56 MHz electric field is applied to the helicon antenna launching a plasma in the source that expands into a 35 cm diameter, 30 cm long aluminium diffusion chamber that is also surrounded by two solenoids. The substrate holder is situated at the bottom and in the centre of the diffusion chamber directly beneath the helicon plasma source. The water cooled substrate table is designed to accept 4 inch silicon wafers and is connected to a 13.56 MHz power supply to enable r.f. biasing of the substrate. Heating of the substrate, caused by ion bombardment, is transferred to the water cooled substrate holder via a helium gas pocket between the silicon substrate and cooled substrate holder. The reactor vessel is maintained at a base pressure of 10-6 Torr and substrates are entered and removed via a load lock system.

All gas flows are controlled using mass flow controllers (MKS) and are fed into the reactor via an inlet in the centre of the diffusion chamber. Experiments are performed using the gases SF6, CF4, O2 and Ar, or mixtures thereof, which are exhausted, along with reaction by-products, via the turbomolecular pump situated above the source. Further descriptions and details of this reactor configuration can be found elsewhere [26, 27, 28, 29, 30].


29

Matchingnetwork

Chamber solenoid

Water coolingto table

He

RF bias

RF supplyENI

OEM 25

RF supplyENI

ACG 5

Ni Probe

Matching networkand source cooling

antenna (outside vacuum)

wafer cartridge

Load lock and cartridgetransfer mechanism

sourcesolenoid

400l/sturbomolecular

pump

150l/sturboPhase locked to

Source supply

Gas inlet

Figure 2.2: Schematic of helicon etch machine.

Suitable masks for plasma etch processes are usually created using lithographic techniques, however, due primarily to time constraints a more robust and quicker masking method was needed. Mask requirements were fulfilled by drilling 315 µm holes into a 125 µm thick nickel plate. This provided a mask that was robust and virtually unaffected by the reactive plasma gases and could therefore be transferred and used for all samples. The substrate and mask set-up have been sketched in Fig. 2.3. A silicon wafer coated with a hybrid film is cut into six sections and each section is sandwiched between the nickel mask and a steel substrate table using two steel springs.


30

IPG coated Si wafer

Ni mask

Steel springs

Steel substrate holder

Figure 2.3: Schematic of substrate holder housing the substrate and nickel mask.

The organically modified ceramics used in this study are photo-initiated silicone polymer materials primarily designed for optical waveguide applications by Redfern Polymer Optics®. A schematic of IPG is drawn in Fig. 2.4.

Si Si SiO OO

R

R

R

R

R

R

SiSi SiO OO

R

R

RR

Where R is a polymerisable side chain e.g. or

Figure 2.4: Schematic of inorganic polymer glass. 2.2.1 Ion saturation current measurements When an insulated electrostatic probe is inserted into a gas discharge it will be bombarded with ions and electrons until it picks up a net potential equal to the plasma floating potential. If the tip of the probe is biased (VB) in relation to the plasma potential, it will either attract or repel positive ions or electrons depending if the probe is biased positively or negatively. A typical Langmuir probe measurement is depicted in Fig. 2.5. Here the current drawn from the plasma by the probe has been plotted as a function of the probe potential.

Langmuir probe characteristics are most easily described by assuming the probe is measuring a low pressure argon discharge. At large and negative bias, the probe will repel all electrons and a positive ion current is drawn. As the bias approaches a zero potential the high energy electrons are able to overcome the negative bias and strike the probe tip. This leads to a transition period where the probe is collecting both ions and electrons. During this transition the electron and ion fluxes generate opposite currents and results in net current of zero, corresponding to the floating potential of the plasma.


31

VB

I

ΦpΦf

Positive ion collection

Electron collection

-70 100

Electron & ion collection

VB

I

ΦpΦf

Positive ion collection

Electron collection

-70 100

Electron & ion collection

Figure 2.5: Typical I-VB characteristic for a Langmuir probe. Not to scale.

In this study the Langmuir probe was used in two configurations. In one configuration the probe tip was biased to a potential of –75 V (with respect to the reactor ground) and used as an ion probe. At this large negative potential the probe is assumed to repel all electrons and draw current equal to the ion saturation current. From standard texts [31], assuming a Maxwellian electron energy distribution and an ion velocity equal to the Bohm velocity (νB) [m.s-1], the ion current density (j+) at the probe and substrate is,

21

6.0

=+

i

eBe m

Tknj (2.1)

where Te is the electron temperature [K], kB the Boltzman constant [JK-1], and mi the ion mass [kg]. Equation (2.1) is only applicable to discharges where all of the negative charge is carried by the electrons.

The presence of negative ions significantly changes the plasma structure to the point where the standard treatment of probe data is no longer valid. When the plasma contains negative ions, a decrease in current in the electron collection region is observed at high positive probe bias. The electron and ion current regions become distorted and the Bohm criterion is modified as the negative ion density (n) increases. Variation in the Bohm criterion as a function of the n/n+ ratio [32], alter the relationship between the positive ion current and n+. In addition to these complications, a recent publication by Franklin et al. [33] reports further inconsistency in the standard treatment of probe data for negative ion plasmas and even suggests such plasmas are structured consisting of a positive/negative ion core surrounded by an electron/positive ion outer region. A detailed review of probe measurements in plasmas containing negative ions is not given, as the probe is


32

setup to measure the floating potential and ion current at high negative bias potential only. In this configuration the probe provides insufficient information upon which to start characterising the plasma. Instead, results of the probe measurements are used in a qualitative capacity where ion flux and potentials experienced by the probe are assumed to be similar to those experienced by the substrate during plasma etching.

The Langmuir probe is inserted into the plasma and a potential, VB, of -75 V applied to the tip (to collect positive ions only). The probe collection area consisted of a flat nickel disc with a total surface area A = 2.95·10-5 m2. Ion currents are measured by the potential drop across a 10 kΩ resistor placed between the probe tip and ground (cf. Fig. 2.6).

20 V

-75V

10kΩ Ni probe tip

Figure 2.6: Langmuir probe set up at -75 V bias supplied by eight 9 V batteries connected in series.

2.3 Results Power is coupled into an inductive discharge by the r.f. field heating the electrons and ions [34]. The hot electrons impart energy to the neutrals via ionising collisions, which also release photons. Ions are accelerated towards the walls due to the potential across the sheath and collide with the walls where upon they become neutralised and impart their kinetic energy to the walls in the form of heat. In a global scheme, energy is put into the plasma via the electric field heating the electrons, and is lost via light emission (from ionising collisions) and heat (via ion bombardment of walls). As neutral gas molecules collide with the walls they pick up some of the heat deposited by the ions.

Gases injected into the reactor are removed via the turbomolecular pump located on top of the helicon source. The increased velocity in which the gas molecules collide with the system walls, in conjunction with a decrease in the pumping efficiency with increased gas temperature result in an increased system pressure, as can be seen in Fig. 2.7. The slight pressure increase seen for argon is attributed solely to gas heating, whereas CF4 and SF6 both show larger pressure increases at higher helicon power inputs. The higher pressures seen for the molecular gases are assumed to be caused by an increase in the effective gas flow


33

caused by fragmentation of the gas. This argument is supported by the fact that the pressure increases are larger as the molecularity of the gas increases.

0 200 400 600 800 1000 1200

0

2

4

6

8

10

SF6

CF4

O2

Ar

Pres

sure

(mTo

rr)

Discharge power (W)0 200 400 600 800 1000 1200

0

2

4

6

8

10

SF6

CF4

O2

Ar

Pres

sure

(mTo

rr)

Discharge power (W) Figure 2.7: Pressure variations in the helicon discharge as a function of discharge power.

2.3.1 DC bias measurements The application of an r.f. voltage to the substrate table produces a d.c. field above the substrate provided the gas in this region is partly ionised. The relationship between the applied r.f. field and resulting d.c. bias was measured for the helicon etch machine and is presented in Fig. 2.8.

Different self-bias potentials are observed at the substrate depending on whether the bias is applied with or without the helicon plasma. At the same bias power input, the ion current to the biased electrode is increased due to the additional density generated by the helicon, and consequently the bias potential is reduced. If the ion current to the biased substrate remains constant, then the potential will scale linearly with the applied power. However, if the additional bias creates additional density above the biased substrate, the ion current will increase and the potential will scales with the P1/2 (Fig 2.8). The bias potential is seen to scale linearly with the applied biasing power suggesting that the plasma density does not increase with increasing bias power. The high biasing potentials observed are believed to be a consequence of the small surface area of the substrate table in relation to the grounded area of the steel expansion chamber.


34

10 1001000

100

10

Vbias

~ P

Bias power (W)

With helicon Without helicon

Bia

s pot

entia

l (-V

)

Vbias ~ P1/2

Figure 2.8: Negative bias as function of power applied to substrate table.

A Langmuir probe was used to measure ion currents and floating potentials in the centre of the diffusion chamber 18 cm above the substrate for each of the pure process gases at different pressures. In this configuration the probe measures the ion current (Fig 2.9) carried by the positive ions and as the plasma is assumed to be quasi neutral, n+ = ne. This criterion however, does not hold if appreciable concentrations of negative ion are present, in which case n+ = ne + n. Negative ions will shield the probe tip from some of the positive ions resulting in lower values of n+. In this situation the positive ion saturation current may well be more representative of ne then of n+. As argon is the most electro-positive of the gases, it produced the highest plasma density of positive ion with an estimated value of n+ = 1.2·1017 m-3 (assuming n = 0) The more electro negative gases all result in low currents to the probe tip suggesting that there are either fewer positive ions generated, and that a portion of the positive ions are being shielded from the probe by negative ions. The most electro-negative element present is fluorine, and as SF6 contained the most fluorine atoms it also produces the lowest positive ion current. Oxygen and CF4 generated comparable positive ion currents. For the argon discharge, plasma density scaled linearly with respect to discharge power, which is typical behaviour for a monatomic gas in an inductive discharge. The molecular gases O2, CF4 and SF6 all exhibited a jump in plasma density at between 100-200 W. This is attributed to a rapid increase in fragmentation as more power is coupled into the discharge and an associated transition from capacitive to inductive operation. This transition is more prominent at low pressures and was also observed in an argon discharge at > 1 mTorr.


35

0 200 400 600 800 1000 12000.0

0.5

1.0

1.5

2.0

2.5

Prob

e cu

rren

t (m

A)

Helicon discharge power (W)

Ar

CF4

O2

SF6

0 200 400 600 800 1000 12000.0

0.5

1.0

1.5

2.0

2.5

Prob

e cu

rren

t (m

A)

Helicon discharge power (W)

Ar

CF4

O2

SF6

Figure 2.9: Positive ion current as a function of helicon discharge power.

The floating potential of argon and oxygen discharges as a function of helicon power is plotted in Fig. 2.10. A negative and constant floating potential was seen to develop at a value of approximately -4 V in the argon discharge. The floating potential for the oxygen discharge was seen to be negative, and become more negative as the helicon power was increased.

The relationship between floating potential and helicon power is more complicated for the CF4 and SF6 discharges. In the CF4 discharge, the floating potential initially becomes increasingly negative as the helicon power increases, but at approximately 350 W begins to become less negative again (Fig. 2.11). According to OML theory, the floating potential depends on the ratios of Ti/Te and mi/me (see Eq. (4.4) and Fig. 4.2 in chapter 4). For argon, Te and Ti remain constant with increasing power and as no fragmentation can occur, mi/me ratio cannot change. The floating potential therefore remains constant as the discharge power is increased. The situation is somewhat more complicated for the molecular gases O2, CF4 and SF6. Fragmentation of oxygen and the other molecular gases at higher powers results in a slight reduction of mi. However, little in known about the plasma potential, and in addition to this, the presence of negative ions may influence Te, and as the total ion density increases ion-ion and ion-neutral processes may become significant. All of these factors in combination with other transport may effect Vf at the probe tip located downstream of the helicon source.


36

-18

-16

-14

-12

-10

-8

-6

-4

-2

00 200 400 600 800 1000

Discharge power (W)

Floa

ting

pote

ntia

l (V

)

Ar O2

-18

-16

-14

-12

-10

-8

-6

-4

-2

00 200 400 600 800 1000

Discharge power (W)

Floa

ting

pote

ntia

l (V

)

Ar O2

Figure 2.10: Floating potential as function of helicon power for Ar and O2 gases.

Fragmentation and negative ion formation are likely to be even more prominent processes in the CF4 and SF6 plasma discharges. Explanations for the behaviour of Vf as a function of helicon power (as shown in Fig. 2.11 and 2.12) therefore lies in an understanding of the molecular processes governing the production, transport and loss of these species in the plasma discharge.

-12

-10

-8

-6

-4

-2

0

0 200 400 600 800 1000 1200

Discharge Power (W)

CF4

Floa

ting

pote

ntia

l (V

)

-12

-10

-8

-6

-4

-2

0

0 200 400 600 800 1000 1200

Discharge Power (W)

CF4

Floa

ting

pote

ntia

l (V

)

Figure 2.11: Floating potential as function of helicon power for CF4 gas.


37

0 200 400 600 800 1000 1200-10

-5

0

5

10

15

Floa

ting

pote

ntia

l (V

)

Discharge power (W)

SF6

Figure 2.12: Floating potential as function of helicon power for SF6 gas.

2.3.2 Etch results Of the gases and gas mixtures used, SF6 produced the highest etch rates. Surprisingly the SF6/O2 gas mixture resulted in a lower etch rate than the pure SF6 despite the presence of carbon in the SiOxCyHz layers. Under the conditions used, CF4 and CF4/O2 plasmas both resulted in deposition on the substrate. A summary of the etch rates achieved for the various gases is presented in Table 2.1.

Table 2.1: Summary of etch results for different gases. Helicon power and substrate bias were set to 1000 W and 100 W respectively for all experiments.

Gas Ratio (%) Etch/Deposition (nm/min)

SF6 Pure SF6 524 etch SF6/O2 62/38 372 etch

CF4 Pure CF4 303 deposition CF4/O2 10/90 120 deposition

O2 Pure O2 30 etch

The dependence of etch rate on bias potential and helicon power were also studied (Fig. 2.13). The etch rate was seen to exhibit a linear dependence to


38

substrate bias potential but a relatively weak dependence on power supplied to the helicon antenna within the range studied.

-750 -600 -450 -300 -150 0050100150200250300350400450

Etc

h ra

te (n

m/m

in)

Bias voltage (V)0 300 600 900 1200 1500

050

100150200250300350400450

100W Bias

Helicon power (W)-750 -600 -450 -300 -150 0

050100150200250300350400450

Etc

h ra

te (n

m/m

in)

Bias voltage (V)0 300 600 900 1200 1500

050

100150200250300350400450

100W Bias

Helicon power (W)

Figure 2.13: Etch rate as a function of d.c. bias on substrate (left) and helicon power (right) for samples etched in a 1.8 mTorr SF6 plasma for 10 minutes.

The bias potential contributes to the ion energy at the substrate surface, and plasma density directly above. The combination of these effects resulted in a strong dependence of etch rate with substrate bias for the experimental conditions used in this study. The helicon plasma is responsible for dissociation of the reactant gas and generation of ion density in the discharge. The relationship between the etch rate vs. helicon power therefore is similar to that of helicon power vs. ion saturation current (cf. Fig. 2.9), although neutral fluorine atoms can also participate in the etch process.

a. b.a. b.

Figure 2.14: SEM images taken of the hole etched in the SiOxCyHz film, a.) shows a plan view b.) profile view.

315 µm

226 µm


39

Cross sectional and plan view SEM images (Fig. 2.14) taken after the coated substrate is exposed to a 10 minute SF6 etch plasma indicate that the holes made are considerably smaller than the holes in the nickel mask and exhibit poor anisotropy. This is caused by the crude mask arrangement used consisting of a nickel plate held in place with steel springs. The relatively thick nickel mask (125 µm) and poor contact with the substrate caused shadowing effects near the mask edge.

2.4 Conclusions Highest etch rates were achieved using pure SF6 gas with typical etch rates in the order of 500 nm/minute using a helicon power of 1000 W with 100 W substrate bias. SEM images taken of the substrate after etching indicated a relatively poor aspect ratio (Fig. 2.14). Poor etch profiles are attributed to the temporary mask design used, necessitated by time constraints. The nickel shim used to make the mask was 125 µm thick, which is more than 100 times thicker than an ideal mask made by lithographic techniques. The relatively high sidewalls this mask produced are likely to have “shadowed” the substrate from the impinging ion flux and hence reduced the diameter of the resulting holes etched. The samples also showed a “pebble dashed” effect around the edges of the etched hole. This was also attributed to the poor masking method employed, but also indicated the SiOxCyHz film to be relatively soft. Etch rate exhibited a greater dependence on d.c. potential formed from r.f. biasing the substrate then of power applied to the helicon source. References [1] Chen J., Liu L., Li Z., Tan Z., Xu Y., Ma J., Sens. Actuators A., 103, 42-47, (2003). [2] Homma T., Mater. Sci. Eng., R23, 243-285, (1998). [3] Tien P., Smolinsky G., Martin R., J. Appl. Opt., 11, 637, (1972). [4] Van der Drift E., Cheung R., Zijlstra T., Microelectron. Eng., 32, 241-253, (1996). [5] Lang W., Mater. Sci. Eng., R17, 1-55, (1996). [6] Stern M.B., Microelectron. Eng., 34, 299-319, (1997). [7] Cho B.-O., Hwang S.-W., Lee G.-R., Moon S.H., J. Vac. Sci. Technol. A, 18, 2791-2798,

(2000). [8] Balachova O.V., Alves M.A.R., Swart J.W., Braga E.S., Cescato L., Microelectron. J., 31,

213-215, (2000). [9] Tserepi A., Gogolides E., Cardinaud C., Rolland L., Turban G., Microelectron. Eng., 41-

42, 411-414, (1998). [10] Midha A., Murad S.K., Weaver J.M.R., Microelectron. Eng., 35, 99-102, (1997). [11] Camara N., Zekentes K., Solid-State Electronics, 46, 1959-1963, (2002). [12] Lishan D.G., Johnson D.J., Lee Y.S., Reelfs B.H., Westerman R.J., III-Vs Review, 15,

48-51, (2002). [13] Wang J.J., Lambers E.S., Pearton S.J., Ostling M., Zetterling C.-M., Grow J.M., Ren F.,

Solid-State Electronics, 42, 743-747, (1998). [14] Mozetic M., Zalar A., Vacuum, 71 (1-2), 233-236, (2003). [15] Choi K., Ghosh S., Lim J., Lee C.M., Appl. Surf. Sci., 206, 355-364, (2003). [16] Kim H.-R., Park H.-H., Hyun S.-H., Yeom G.-Y., Thin Solid Films, 332, 444-448,

(1998).


40

[17] Vallee C., Granier A., Aumaille K., Cardinaud C., Goullet A., Coulon N., Turban G.,

Appl. Surf. Sci., 138-139, 57-61, (1999). [18] Liu P.T., Chang T.C., Sze S.M., Pan F.M., Mei Y.J., Wu W.F., Tsai M.S., Dai B.T.,

Chang C.Y., Shih F.Y., Huang H.D., Thin Solid Films, 332, 345-350, (1998). [19] Boswell R.W., Chen F.F., IEEE T. Plasma Sci., 25, 1229-1244 (1997). [20] Chen F.F., Boswell R.W., IEEE T. Plasma Sci., 25, 1245-1257, (1997). [21] Landau L.D., J. Phys., 10, 25, (1946). [22] Boswell R.W., Ad. in Space Res., 1, 331-345, (1981). [23] Yun S.-M., Chang H.-Y., Phys. Lett. A, 248, 400-404, (1998). [24] Boswell R.W., Henry D., Appl. Phys. Lett., 47, 1095-1097, (1985). [25] Perry A.J., Boswell R.W., Appl. Phys. Lett., 55, 148-150, (1989). [26] Boswell R.W., Vender D., Plasma Sources Sci. Technol., 4, 534-540, (1995). [27] Charles C., Boswell R.W., J. Vac. Sci. Technol., A, 13, 2067-2073, (1995). [28] Perry A.J., Vender D., Boswell R.W., J. Vac. Sci. Technol. B, 9, 310-317, (1991). [29] Charles C., Boswell R.W., J. Appl. Phys., 78, 766-773, (1995). [30] Charles C., Boswell R.W., J. Appl. Phys., 84, 350-354, (1998). [31] Glow discharge Processes, Chapman B., Wiley, New York, (1980). [32] Amemiya H., J. Phys. D: Appl. Phys., 23, 999-1014, (1990). [33] Franklin R.N., Plasma Sources Sci. Technol., 10, 162-167, (2001). [34] Principles of plasma discharges and materials processing, Lieberman M.A., Lichenberg


Chapter 3 Synthesis and characterisation of SiCxOyHz thin films

41

Chapter 3: Synthesis and characterisation of SiOxCyHz thin films

Abstract The poor adhesion between plastic substrates and hard ceramic-like coatings is one of the major obstacles in the development of scratch resistant coatings for plastics. A new dual plasma CVD deposition system for depositing a novel polymer/inorganic hybrid films has been developed and constructed and is presented here. Mechanical properties of the ORMOCER coatings are characterised and compare favourably with wet chemically synthesized coatings of similar composition*.

3.1 Introduction The synthesis of organic/inorganic hybrid films on plastics substrates has traditionally been performed using wet chemical techniques in conjunction with dip coating or spin coating methods [1]. These methods require several processing steps, long curing times and need for large volumes of solvents in the synthesis of these coatings. The use of gas phase deposition techniques to synthesize hybrid films can potentially offer a solution to a number of these problems. Typically, gas phase deposition processes are solvent free, one step processes that do not require subsequent curing. Dense and pinhole free films can be deposited as no reaction by-products or solvents need to be removed after the deposition. The aim of this chapter is to develop a new process technique to deposit hybrid coatings using gas phase deposition technology, in particular plasma based synthesis techniques. 3.1.1 Problem description Many of the problems encountered when attempting to deposit scratch resistant films onto a plastic substrates are specific to the deposition technique involved, e.g. porous films are often encountered when using sol-gel deposition techniques [2, 3]. However, a number of issues arise as being inherent to the plastic-SiO2 combination. One issue of particular concern is that of adhesion between a polycarbonate substrate and SiO2-like layer in the context of plasma deposition systems. Poor adhesion is the result of one or a combination of reasons including; * The work contained in this chapter is the subject of two publications. Alcott G.R, Linden J.L., van de Sanden M.C.M., Mat. Res. Soc. Symp. Proc., 726, 297-302, (2002). Alcott G.R., Eijkman D.J., Schrauwen C.P.G., Linden J.L., van de Sanden M.C.M., 10th Neues Dresdner Vakuumtechnisches Kolloquium Proceedings, 55-60, (2002).


42

1) High or uncontrolled water content of the plastic [4]. 2) Weak boundary layer (WBL) [5], based on the assumption that the plastic

surface contains low molecular weight compounds and diffuse impurities. 3) A poor match of thermal expansion coefficients between the plastic substrate

and SiO2-like film [6]. 4) Low surface energy (e.g. polyethylene and polypropylene).

Notable improvements in film adhesion have been reported by using a variety

of pre-treatments [7]. Most pre-treatments involve some sort of cleaning procedure followed by either a brief exposure to plasma (Ar or O2) or to high temperature for anything from a few seconds up to several hours. Both physical and chemical changes in the plastic substrate can occur during these pre-treatments. Thermal pre-treatment serves to drive out absorbed water, whereas plasma (usually performed at reduced pressure) both drives out water and activates the plastic surface through ion bombardment. While most authors generally acknowledge ion bombardment improves adhesion because of cross linking of low molecular weight polymer chains at the surface, Vallon et al. [8], demonstrated that plasma treatment not only caused a greater extent of cross linking, but also caused photo-fries type rearrangements (Fig. 3.1). The photo-Fries rearrangements resulted in an increased number of active bonding sites for the impinging SiO2 precursor molecules to bond. Klemberg-Sapieha et al. [13] investigated the adhesion of SiNx to polycarbonate substrates when deposited using a plasma technique and sputtering. They attributed the improved adhesion of the plasma deposited SiNx to a 60 nm thick interface layer generated by ablation and re-deposition of the polycarbonate under ion bombardment in the initial stages of film deposition.

O C

O

O O• + C

O

O•

R1 R2

hv

PC

OH

H

O

OH

OH

O

OH

OH

recombinationchain breaking

photo-Frieshv

OH

CO

O

hv OH

COHO

photo-Fries

O C

O

O O• + C

O

O•

R1 R2

hv

PC

OH

H

O

OH

OH

O

OH

OH

recombinationchain breaking

photo-Frieshv

OH

CO

O

hv OH

COHO

photo-Fries

Figure 3.1: Reaction mechanisms following carbonate bond breaking in PC during Ar plasma treatment: photo-Fries rearrangements, chain scission and recombination.


43

Plasma deposition inevitably subjects the substrate to a heat load, the intensity of which depends on the type of plasma (high or low density), its proximity to the substrate (direct or remote) and the power dissipated by the discharge. For temperature sensitive substrates some of this heat load may be relieved by cooling the substrate table, however, the substrate surface in contact with the plasma will still experience elevated temperatures due to ion bombardment. The substrate cools when the plasma is switched off, and differences in thermal expansion coefficients between the inorganic coating and plastic substrate cause a stress build up in the coating. The coating relieves the stress by cracking and accelerated water and oxygen diffusion through the cracks leads to de-lamination. Stress induced adhesion failure occurs on a relatively short time scale from immediately after deposition up to a few days after depending on the stress.

The difference in expansion coefficients between plastics and SiO2 is a more difficult problem to resolve. Menichella et al. [9] attribute the poor adhesion of SiO2 (although they speak generally of any intrinsically inorganic coating) to CR-39* lenses to a difference in the thermal expansion. Their solution was to deposit what they termed a “gradient layer” (see chapter 1). By starting with a pure TEOS gas mixture they were able to deposit a silicone plasma polymer like film onto the CR-39 and by gradually increasing the oxygen concentration the top part of the film resembled SiO2. The interface layer provides a smooth transition in thermal expansion coefficients between the CR-39 and SiO2, while the top SiO2 provides the scratch resistance and hardness required. Silva Sobrinho et al. [10] attempted to show that through ablation and re-deposition processes, plasma treatment of polymers leads to a similar interface layer. They believe that the generation of this mixed interface region (“interphase”) is why plasma processes give superior adhesion over other deposition techniques. The nature of this interface region and why it improves adhesion has been the subject of a number of investigations, using techniques such as x-ray photoelectron spectroscopy (XPS) [8, 11, 12, 10, 13], attenuated total reflection infrared (ATR-IR) [8, 10], elastic recoil detection (ERD) [8], electron microprobe analysis (EMA) [10] and ellipsometry [11, 12, 14]. Results from compositional profile measurements across such interface regions should however be interpreted with some caution. Techniques such as EMA and ERD have beam apertures in the order of 30 nm and would therefore indicate a 60 nm thick interface region even if used to scan an infinitely narrow interface region (where polymer and SiO2 phases remain physically separate). In addition, ion induced mixing can occur for soft materials (plastics) further undermining the reliability of the results. However, these techniques are useful in a comparative capacity, and in this manner Silva Sobrinho et al. [10] demonstrated that PECVD films exhibit interface regions twice as thick as that of comparable samples prepared via PVD techniques.

Delamination can also be caused by diffusion of water and oxygen through the polymer from the uncoated side and then react with the polymer-film interface [15]. Adhesion failure caused by these effects usually occurs on much longer time scales

* Columbia Resin #39, allyl diglycol carbonate.


44

of up to several months, depending on the permeability and thickness of the polymer substrate and rate of reaction at the interface region.

Hydrocarbon rich polymer surfaces are often hydrophobic and exhibit low surface Gibbs energy. These surface conditions favour an island growth mechanism whereby growth precursors physisorb and diffuse on the substrate surface until they find an active site to chemically bond. This results in a film, which although displays homogenous surface coverage, might only be bonded to the substrate at a limited number of sites.

The deposition of soft polymeric films invariably gives better film adhesion to plastic substrates then hard inorganic layers and the problem therefore lies in combining the superior adhesion of polymeric layers with the hardness and barrier properties of inorganic films. An alternative to the deposition of gradient layers (as described above) is to deposit nanocomposite thin films consisting of ceramic particles embedded in a polymer layer. In this concept, the polymer layer provides good adhesion to the plastic substrate with added toughness and durability, while the ceramic particles (in this instance SiO2) provide hardness and an improved diffusion resistance for gases such as oxygen and water. The dense SiO2 particles improve the barrier properties by effectively increasing the diffusion length of the diffusing gas, as depicted pictorially in Fig. 3.2.

O2, H2OO2, H2O

PlasticSubstrate

PolymerLayer

CeramicParticles

Figure 3.2: Improved diffusion resistance of nanocomposite thin films. 3.2 Process design and reactor concept A reactor consisting of two separate reaction zones for generating the different components of the final hybrid material was designed and built (Fig. 3.3). Precursors for the polymer layer enter the lower half of the reactor (inlet 3) where they are activated by a capacitively coupled plasma formed between the r.f. powered substrate table and the grounded reactor walls. A low density (~300 W) plasma is used for the polymer layer formation to prevent substrate damage and maintain the functionality of the monomer used. Precursors used for the synthesis of the particles/inorganic fragments are directed into the upper reaction zone (inlet


45

1), where they are activated by the inductively coupled plasma (ICP). The intension here is for the high density ICP to fully oxidize the precursor and generate both active inorganic fragments and particulates which will then go on to be incorporated into the growing polymer layer. In addition to the siloxane precursors, oxygen can also be dosed into either Zone I or II by a pneumatically controlled gas dosing system.

Particulates are transported from Zone I to Zone II by a combination of physical transport effects but mainly gas flow and gravity. To avoid particles charging and becoming suspended in the electric field a pulsed plasma supply is used. Both plasmas were generated using Huttinger PFG 1000 RF generators, all gases were delivered using Bronkhorst HiTec mass flow controllers. Pressure regulation of the main reactor vessel is achieved using a MKS throttle valve and controller located in the exhaust line between the reactor and vacuum pump.

OO

OO

OO

~

~Pulsedr.f. 2

Substratetable

Ellipsometerwindows

~

~

Precursorinlet 2 (O2)

FTIR FTIR

Precursorinlet 1 (TEOS)

Pulsedr.f. 1

Precursorinlet 3

Zone II

Zone I

To pumps

OO

OO

OO

~

~Pulsedr.f. 2

Substratetable

Ellipsometerwindows

~

~

Precursorinlet 2 (O2)

FTIR FTIR

Precursorinlet 1 (TEOS)

Pulsedr.f. 1

Precursorinlet 3

Zone II

Zone I

To pumps

Figure 3.3: Schematic of dual plasma reactor showing inductive top plasma (Zone I) and capacitive bottom plasma (Zone II) with in situ FTIR facilities.


46

O2

Ar

Gas System Exhaust

Reactor

Pumps

Cold Trap

Bubbler 2Bubbler 1

CEM

Mas

s Flo

w C

ontro

llers

Inlet 1Inlet 2

Inlet 3

ReactorExhaust

To exhaust

O2

Ar

Gas System Exhaust

Reactor

Pumps

Cold Trap

Bubbler 2Bubbler 1

CEM

Mas

s Flo

w C

ontro

llers

Inlet 1Inlet 2

Inlet 3

ReactorExhaust

To exhaust

Figure 3.4: Gas dosing system showing bubbler and CEM systems for dosing of liquid precursors. All gas lines are heated.

The precursors were vaporised and delivered to the reactor using the gas delivery system depicted in Fig. 3.4. Liquid precursors were dosed using bubbler and Bronkhorst High-Tech CEM [16] (Controlled Evaporation & Mixing) systems with argon as a carrier gas. CEM systems utilises a liquid flow metre to determine precursor injection rate. In the bubbler system, the volumetric precursor flow is dependent on the precursor vapour pressure and flow rate of the carrier gas passed through the liquid. The precursor flow carried from the bubbler is calculated by,

carrierprecursor Fx

xF−

=1

(3.1)

where Fprecursor is the precursor flow in sccm, Fcarrier the carrier gas flow in sccm and x the fraction of the precursor vapour pressure divided by the system pressure. To calculate the precursor flow from the bubblers and to ensure sufficient carrier gas is dosed into the CEM system, the precursor vapour pressure needs to be know and have been measured for the precursors used (cf. Appendix A). An overview of the experiments performed in this chapter is presented in Table 3.1.


47

Table 3.1: Overview of experiments performed for this chapter.

Experiment Precursor Oxygen Plasma Pressure

Polymer synthesis

TPS, TMSE, TEOS, Glymol 0-60 % CCP 100-400 W 0.5-2.0

Torr

SiO2 synthesis TEOS 30-70 % ICP 300-1000W CCP-bias 0-170V

0.5-2.0 Torr

Adhesion tests TMSE - CCP 100-500W 1 Torr

3.2.1 Diagnostics and substrate preparation Deposition is performed on a number of different substrate materials, including crystalline silicon <100>, glass, polycarbonate (PC), steel and paper. Post deposition mechanical characterisation of the films is done using Taber abrasion (for scratch resistance) and Methyl Ethyl Ketone (MEK) solvent (for detection of pinholes). Visible transmittance is determined using a Perkin Elmer Lambda 900 photo spectrometer and infrared spectra are recorded using a Midac M2500-C FTIR spectrometer for films deposited on Si wafers. In situ diagnostics are made using the same FTIR spectrometer fitted with a mercury cadmium telluride (MCT) detector and interferometer mounted either side of the reactor. Post deposition particle size distribution was determined using a Beckman Coulter LS230 particle size analyser. Film thicknesses are measured using a Tencor P-10 step profiler.

The plasma pre-treatments are performed using the capacitively coupled r.f. discharge (Zone II), with Ar and O2 as active gases, while deposition was achieved using 1,2-bis(trimethyl)siloxyethane and Ar mixtures and sometimes O2 to produce SiO2 layers. Polycarbonate substrates (obtained from General Electric) are first rinsed with ethanol and dried in air. Adhesion performance was graded 0-5 using the cross cut tape test [17], where 5 represents no delamination of the coating and 0 indicated more than 65 % removal of the coating in accordance with the standard as represented in Appendix B. 3.3 Polymer synthesis in the capacitive plasma (Zone II) To characterise the deposition characteristics of the system, plasma polymerisation of TPS, TMSE and TEOS were optimised with respect to discharge power and precursor concentration using argon as carrier gas. Film deposition rate is dependant on the precursor concentration and the power dissipated by the discharge. The peak growth rate at 300 W was seen for TEOS precursor, with a partial pressure of 0.1 Torr corresponding to a peak deposition rate of 2 nm/s. At higher partial pressures particles were seen in the reactor and the film quality suffered, this was especially prominent for the TPS precursor. The addition of oxygen to the siloxane deposition accelerated the growth rate and lowered the threshold at which powder formation occurred. The increased growth rate is likely to be caused by atomic oxygen radicals in the discharge. Film hardness also


48

improved with addition of oxygen but adhesion to the polycarbonate substrates deteriorated as is usually observed with more SiO2-like properties.

IR measurements of the plasma polymerised TMSE layers (in the absence of O2) showed them to retain much of the functionality of the original precursor (Fig. 3.5) This is indicated by the large C-H stretching region seen between 2850-3000 cm-1 and the Si-CHx peaks found between 750-1400 cm-1. The CO2 band seen at approximately 2350 cm-1 is due to atmospheric CO2 in the ambient of the spectrometer. At high deposition powers denser more SiO2 type films were deposited as indicated by a reduction and broadening of the C-H stretching vibrations (2850-3000 cm-1 and at 1100 cm-1).

0.20.30.40.50.60.70.80.9

1

3800 3500 3200 2900 2600 2300 2000 1700 1400 1100 800 501

Wavenumber [cm-1]

Tran

smis

sion

(0-1

)

CO2

Si-CSi-CHx

C-H

Liquid

Flim

Tra

nsm

issio

n (0

-1)

Wavenumber (cm-1)

Film

Liquid

0.20.30.40.50.60.70.80.9

1

3800 3500 3200 2900 2600 2300 2000 1700 1400 1100 800 501

Wavenumber [cm-1]

Tran

smis

sion

(0-1

)

CO2

Si-CSi-CHx

C-H

Liquid

Flim

Tra

nsm

issio

n (0

-1)

Wavenumber (cm-1)

0.20.30.40.50.60.70.80.9

1

3800 3500 3200 2900 2600 2300 2000 1700 1400 1100 800 501

Wavenumber [cm-1]

Tran

smis

sion

(0-1

)

CO2

Si-CSi-CHx

C-H

Liquid

Flim

Tra

nsm

issio

n (0

-1)

Wavenumber (cm-1)

Film

Liquid

Figure 3.5: Plasma polymerised TMSE (film) retains much of the functionality of the liquid TMSE precursor (liquid).

3.4 Inorganic layer synthesis in the inductive plasma (Zone I) TEOS and O2 were activated using the ICP and the active precursor fragments flowed into Zone II where they formed a SiO2 layer. Increasing the O2 concentration resulted in more SiO2-like layers being deposited. Although stoichiometric SiO2 can be produced with the ICP, ion bombardment achieved by applying an r.f. bias to the substrate table, was necessary in order to achieve dense SiO2 layers with relatively low levels of SiOH. The broad absorption feature appearing between 2900-3800 cm-1 in the infrared spectrum shown in Fig. 3.6 indicates incorporation of SiOH moieties in the plasma deposited TEOS layers.

As the d.c. bias potential is decreased from 170 to 0 V the peak associated to SiOH groups increases indicating a reduction in density of the layer. The peaks centred at 1070 cm-1 represent Si-O stretching, while the rather broad absorption between 2700 and 3600 cm-1 indicates Si-OH character. Silanol bending peaks also appear at 925 cm-1. The higher spectral definition of these regions supports this assumption, as these peaks would be much broader if the H2O and CO2 were incorporated into the films themselves.


49

5001000150020002500300035000

20

40

60

80

100

120

140

Si-OSi-OSi-OH(s)

CO2

Si-OH(b)

dc

b IR

tran

smis

sion

(%)

Wavenumber (cm-1)

a

Figure 3.6: FTIR spectra of SiO2 deposited with 1kW ICP and various d.c. bias voltages, a. = 170 V, b =100, V c. = 80 V and d = 0 V.

Total pressure, total gas flow and precursor concentrations were all varied in an attempt to synthesize nanometre size particulates using the inductive plasma. However a number of issues prevented the synthesis of particles using the ICP. Firstly, the lack of any confining horizontal electric fields in the ICP required operating the discharge under high pressure/low flow conditions to maximise the residence time and encourage homogeneous nucleation. Due to the long mean free path, activated precursor fragments undergo numerous collisions with the ICP tube walls before reacting reactor Zone II. This resulted in rapid contamination of the ICP tube walls with black silicon/carbon deposits. These deposits delaminated as the tube cooled between plasma operations and contaminated the layers growing on the substrate beneath. In addition to this, the ICP was unstable at high precursor concentrations and a very small bright plasma in the centre of the tube (causing the quartz tube to melt in a matter of seconds) suggested a switched to a capacitive power coupling as the pressure was increased. 3.5 Film adhesion The water content of polycarbonate varies depending on the storage conditions. The typical water content for polycarbonate stored at room temperature and humidity is approximately 0.1-0.2 wt%. Some of the adsorbed water can be removed from the polycarbonate by heating in an oven for two hours at 120°C. The structural changes caused by the oven pre-treatment can be seen in the infrared spectra shown in Fig. 3.7.


50

5000 4000 3000 2000 1000

-0.06

-0.04

-0.02

0.00

0.02

0.04

O-H (H2O)Aromatic C-H stretching

IR a

bsor

ptio

n

Wavenumber (cm-1)

C-H stretching

Figure 3.7: FTIR spectrum recorded during a thermal pre-treatment of PC, negative absorption bands at 3600 cm-1 and 3100 cm-1 indicate a loss of H2O and saturated C-H respectively. Positive stretches at 2900 cm-1 indicate an increase in unsaturated C-H content.

The negative absorption bands between 3500 and 3750 cm-1 indicate loss of water from polycarbonate, as does the re-distribution of modes in the C-H stretching regions between 2800 and 3100 cm-1. Dehydration of the polycarbonate from oven treatment can also be seen by changes in the surface energy. Before oven pre-treatment polycarbonate has a contact angle with water droplets of 75°, samples exposed to oven temperature of 120°C for anything from two to twenty four hours exhibited more hydrophobic surfaces with contact angles ranging from 84 to 90°.

Although many authors report improved adhesion with mild plasma pre-treatment, in this study plasma pre-treatment over two minutes appeared to lead to polymer degradation and poor adhesion. In a series of experiments in which a silicone layers were deposited from 1,2-bis(trimethyl)siloxyethane onto PC substrates after a series of plasma pre-treatments, film adhesion was seen to decrease with increased pre-treatment. Fig. 3.8 shows how different exposure to an argon plasma pre-treatment affects the adhesion of a pp-TMSE layer (0 hours) and the longer term stability (336 hours).


51

0 2 4 6 8 100

1

2

3

4

5

0 Hours after deposition 336 Hours after deposition

Adh

esio

n (r

ated

0-5

)

Treatment time (minutes) Figure 3.8: Adhesion of plasma deposited silicone layer after various plasma pre-treatment times. Pre-treatment conditions were 300 W r.f. capacitive Ar discharge at 1 Torr.

0 5 10 15 20 25 300

1

2

3

4

5

Adh

esio

n (r

ated

0-5

)

Oxygen (vol%)

0 Hours after deposition 96 Hours after deposition

Figure 3.9: Decrease in adhesion as film composition moves from silicone like to silica-like with increasing O2 concentration in feed gases. Deposition performed using a 300 W capacitive r.f. plasma with 1,2-bis(trimethyl)siloxyethane, Ar and O2 as precursors.

Exposure times below two minutes do not appear to improve or degrade the immediate adhesion properties. Above two minutes exposure to the plasma may be


52

causing damage to the polycarbonate surface either in the form of high-energy ion bombardment or UV emission from the plasma. As it is common for commercial polycarbonate to contain UV stabilizers, degradation due to UV radiation from the plasma is unlikely to be responsible for the failing adhesion. Ion bombardment does, however, subjects the PC substrate to a heat load and longer pre-treatment times will therefore result in higher substrate temperatures. As the TMSE film is deposited immediately after the pre-treatment, longer pre-treatment times will result in a high substrate temperature during film growth, and consequently higher intrinsic stress in the final coating. As the deposited layer moves from a silicone structure towards that of SiO2, compressive stress in the layer causes delamination (Fig. 3.9). 3.6 Mechanical/optical properties of hybrid layers The optimum mechanical properties were measured for plasma polymer layers synthesized in Zone II using TEOS and oxygen as precursors. The films were pinhole free (good MEK test results) and showed good abrasion and optical properties. However, despite pre-treating the polycarbonate substrate at 120°C in an oven for 2 hours the films exhibited poor long term adhesion. After 2 weeks the films showed signs of cracking and delamination and were unsuitable for barrier measurements. Results of tests administered before film degradation can be seen in Table 3.2 together with a comparison with wet chemically in-house deposited hybrid film*. Table 3.2: Comparison of physical properties of hybrid coatings deposited via gas phase and wet chemical techniques.

* The in-house hybrid film was prepared from a sol-gel solution containing 50 wt% Glymol, 30 wt% Methytrimethylsilane and 20 wt% aluminium isobutoxide reacted with 1 equivalent water with isopropyl alcohol as a solvent. The coating was oven cured at 130°C for 3 hours and had a thickness of approximately 1 µm.

Characterisation Gas phase synthesis Wet chemical synthesis

Taber abrasion (% loss of transmission after 100 cycles) 1.4 5.0

Solvent resistance (MEK) >100 double rubs 100 double rubs Adhesion (cross cut tape test) Good Good

Impact Good Good Surface free energy 43.22 mJ/m2 17 – 39.1 mJ/m2

Contact angle water 69° 76 - 92°


53

3.7 Discussion Relating film properties to deposition parameters in PECVD processes is more difficult to achieve than in thermally activated CVD, and any relationships determined are usually only valid for a specific deposition apparatus. The interdependency of many plasma parameters (Te, ni, ne along with various electric fields) in conjunction with a low pressure system, make PECVD processes difficult to fully characterise to the extent where no one system parameter can be independently varied to determine its influence on the overall process. For example, here we attempted to determine the influence of discharge power on the film adhesion. Pre-treatment times exceeding two minutes adversely effected film adhesion and this was primarily attributed to increased heating of the substrate due to extended exposure to the plasma. However, it is possible that film adhesion suffered because of damage to the polycarbonate substrate caused by ion bombardment or exposure to UV radiation. Similar problems were encountered when correlating growth rate to plasma parameters. Growth rate is observed to increase as a function of the power applied to the discharge, but are films deposited at low power comparable to films deposited at higher power? At low power levels, and hence low density, the precursor will be only partly fragmented and much of the precursor structure will be present in the final layer. As the power is increased, the precursor may undergo several dissociations and the resulting film may differ from the film produced at lower power. However, in addition to ne, Te may also vary with power and increased dissociation will result in a reduction in the average ion mass, both of which affect the ion sheath and hence ion energy at the substrate (as can be seen from Eq (4.4) in chapter 4). Variations in the ion energy will also affect the film structure and composition. The substrate temperature may also change as a consequence of increased ion bombardment from the plasma, which may also influence growth rate and film properties. Does the introduction of oxygen, for example, increase the growth rate by affecting the discharge parameters or by accelerating the discharge chemistry? Some of these issues are the subjects of later work. The chemistry of the TEOS/O2 plasma system is investigated in chapter 5. 3.8 Conclusions This chapter set out to characterise the deposition system designed for the gas phase synthesis of hybrid thin films. Although silicone-like polymer films and SiO2 films have been deposited and characterised, a number of questions have arisen regarding the interaction of plasma on the substrate surface and chemical reactions. Some general conclusions however can be drawn.

The synthesis of inorganic SiO2 layers was successfully demonstrated using TEOS/O2 injected into an ICP. Additional r.f. biasing of the substrate was necessary to achieve reasonable deposition rates and to remove SiOH and densify the films. Infrared transmission spectra showed that by using high discharge power and relatively high oxygen concentrations, C and H free SiO2 could be generated in the inductive discharge, as demonstrated in Fig. 3.6. Although these layers exhibited excellent mechanical properties they also exhibited poor adhesion to polycarbonate substrates. All attempts at synthesizing particulates using the ICP


54

discharge were unsuccessful, probably due to the absence of confining electric fields and thus short residence times for negative ions (cf. chapter 4).

The synthesis of silicone polymer layers from various siloxane precursors was successfully demonstrated. Adhesion was observed to deteriorate as the siloxane layers became more inorganic (increase O2 content in plasma) and was therefore attributed to intrinsic stress caused by temperature increases during the plasma deposition process. Of the precursors evaluated, TMSE exhibited the best long term stability and adhesion to the polycarbonate substrate. This was attributed to the higher carbon content of this precursor that produced a softer SiOxCyHz material. The addition of oxygen to the siloxane discharge increased the film deposition rate and hardness, but reduced layer adhesion to the polycarbonate substrates. Plasma pre-treatments in excess of two minutes adversely affected layer adhesion to polycarbonate substrates. Acknowledgement I would like to thank M. Creatore for assistance with the deposition and characterisation of the inorganic SiO2 layers. References [1] Sol-Gel Science: The physics and chemistry of sol-gel processing, Brinker J., Scherer

G.W., Academic Press, (1990). [2] Conde A., Duran A., Damborenea J.J. de, Prog. Org. Coat., 46, 288-296, (2003). [3] Que W., Zhang Q.Y., Chan Y.C., Kam C.H., Comp. Sci. Technol., 63, 347-351, (2003). [4] Schafer M.M., Seidel C., Fuchs H., Voetz M., Appl. Surf. Sci., 173, 1-7, (2001). [5] Vallon S., Hofrichter A., Drevillon B., Klemberg-Sapieha J.E., Martinu L., Poncin-

Epaillard F., Thin Solid Films, 290-291, 68-73, (1996). [6] Fu Y., Du H., Sun C.Q., Thin Solid Films, 424, 107-114, (2003). [7] Baumgartner K.-M., Schneider J., Schulz A., Feichtinger J., Walker M., Surf. Coat.

Technol., 142-144, 501-506, (2001). [8] Vallon S., Hofrichter A., Guyot L., Drevillon B., Klemberg-Sapieha J.E., Martinu L.,

Poncin-Epaillard F., J. Adhes. Sci. Technol., 10, 1287-1311, (1996). [9] Menichella S., Misiano C., Simonetti E., De Carlo L., Carrabino M., Soc. Vac. Coat., 37th

Annual Technical Conference Proceedings, 37-40, (1994). [10] Da Silva Sobrinho A.S., Schuhler N., Klemberg-Sapieha J.E., Wertheimer M.R., J. Vac.

Sci. Technol., .A 16, 2021-2029, (1998). [11] Vallon S., Brenot R., Hofrichter A., Drevillon B., Gheorghiu A., Senemaud C., Sapieha

J.E.K., Martinu L., Poncin-Epaillard F., J. Adhes. Sci. Technol., 10, 1313-1332, (1996). [12] Bergeron A., Klemberg-Sapieha J.E., Martinu L., J. Vac. Sci. Technol., A 16, 3227-

3234, (1998). [13] Klemberg-Sapieha J.E., Poitras D., Martinu L., Yamasaki N.L.S., Lantman C.W., J. Vac.

Sci. Technol., A 15, 985-991, (1997). [14] Bergeron A., Klemberg-Sapieha J.E., Martinu L., J. Vac. Sci. Technol. A 16, 3227-3234,

(1998). [15] Roualdes S., Sanchez J., Durand J., J. Membr. Sci., 198, 299-310, (2002). [16] Boer H.J., J. de Physicque IV, 5, C5-961, (1995). [17] Annual Book of ASTM standards, 06.01, D-1 on Paint and Related Coatings, Materials

and Applications E-12 on Appearance, PCN: 01-060196-14, (1996).

Appendix A Vapour pressures of precursors for the CVD of silicon based films

55

Appendix A: Vapour pressures of precursors for the CVD of silicon-based films

Abstract Vapour pressure measurements of the precursors tetraethoxysilane (TEOS), tripropylsilane (TPS), (3-glycidoxypropyl) trimethoxysilane (Glymol), perfluoroctyltriethoxysilane (PTES) and 1,2-bis(trimethyl)siloxyethane (TMSE) in the temperature range 20 to 200°C are presented. Measured data was fit using the integrated Clausius-Clapeyron equation, pvap=p0 e-∆H/RT, where p0 is the integration constant and ∆H is the enthalpy of vaporization*.

Introduction Precursors suitable for CVD experiments need to be easily vaporized to form vapours stable during transport but readily reactor when activated. Assuming these criteria are satisfied a precursor is then selected based on the specific properties of the layer it produces [1], costs (including implementation), health and safety precautions, and environment impact. In practice, metal halides (usually F or Cl due to the poor volatility and stability of the heavier halide compounds), alkoxides, acetyl acetonates and organometallics (M-(CxHy)z, where M = central metal) and compounds with various combinations of these ligands generally satisfy the criteria. Since there has been considerable interest in the deposition of silica and silicone type materials (see chapter 1) a large number of suitable precursors have been found and tested. Silicon organometallic compounds (Si(CH3)4 and higher) and SiH4 although used are either extremely flammable or pyrophoric and considerable safety precautions need to be taken when using them. The silicon halides although stable, are very reactive and readily form corrosive by products such as HF or HCl on contact with water. In addition to the corrosive nature of the halides, the tendency to form environmentally harmful halogen-hydrocarbons (e.g. Freon) necessitates the implementation of expensive waste gas processing. Consequently, the most attractive compounds for silica/silicone deposition are the silicon alkoxides and derivates thereof. Tetraethoxysilane, 1,2-bis(trimethyl)siloxyethane were chosen as low cost silicon alkoxides. (3-glycidoxypropyl) trimethoxysilane is a compound used in the wet chemical synthesis of silica/silicone based hybrid thin films and was chosen to make comparisons between the gas phase and wet chemical processing techniques. Tripropylsilane was selected to deposit hydrophobic SiOxCyHz films as the presence of Si-O moieties is known to encourage H-bonding and absorption of water. * Alcott G. R., van de Sanden M.C.M., Kondig S., Linden J.L., Chem. Vap. Dep., 10, 20-22, (2004).


56

The controlled vaporization and subsequent transport of liquid and solid precursors is an essential part of gas phase process technology. To accurately control precursor feed rate when using conventional bubbler systems, it is imperative to know the precursor vapour pressure, thermal stability and in many cases the purity of the sample [2, 3, 4]. Here we report further measurements performed using a procedure and system developed in-house [1] designed to quantify vapour pressure, stability and the presence of volatile contaminants. The precursors investigated in this study are used for the deposition of silicon based inorganic/organic hybrid thin films. Experimental A detailed description of the vapour pressure apparatus can be found elsewhere [1], however, for convenience a brief description is given here. A sample is loaded into the sample vial and attached to the vacuum system situated inside an oven at valve V3 (Fig. A.1). Valves V1 and V2 are opened and the chambers C1 and C2, situated between the differential pressure gauge and V1 and V2 respectively, are pumped down to a base pressure of 3.0·10-2 Torr. After the system has been pumped down, V1 is closed and V3 opened for a sufficient time to remove all air from the sample. Solid samples are pumped for longer to allow desorption of water and other physisorbed contaminants from the sample surface. After removing contaminants from the sample, V3 is closed and remains closed for a predetermined time interval (t), while V1 and V2 are opened for a short time interval (10 seconds) and then closed. The measurement commences at the start of t when V3 is closed. At the end of the time interval t, V3 is opened and the vapour pressure of the sample expands into C2, and allowed to equilibrate for 10 minutes. The measurement is concluded by closing V3 and opening V2.

V1

V2 V3

Π

P

P

T Π

P

T

= Oil Diffusion pump

= Pneumatic valve

= Rotary vane pump

= Quartz thermocouple

= Pressure gauge

= Differential pressure Oven

Cold trap

C1

C2

= Sample vial

V1

V2 V3

Π

P

P

T Π

P

T

= Oil Diffusion pump

= Pneumatic valve

= Rotary vane pump

= Quartz thermocouple

= Pressure gauge

= Differential pressure Oven

Cold trap

C1

C2

= Sample vial

Figure A.1: Schematic of the vapour pressure measuring system.


57

Measurements at different temperatures are performed by setting the oven temperature (maximum 200°C) and repeating the aforementioned procedure after allowing the system and sample to equilibrate at a new oven temperature. A schematic of the apparatus is depicted in Fig. A.1. The number of measurement points was 29, 26, 19, 29 and 28 for TEOS, TPS, TMSE, Glymol and PTES respectively. Systematic errors are avoided by performing the measurements in a random order with respect to temperature. Some degassing of the samples was observed during initial sample measurements, but these results were discarded. Experimental error is based on the tolerance of the differential pressure gauge and taken to be approximately 2 %. Statistical error in the recorded data is calculated using least squares regression with a 95 % confidence interval. Precursor stability measurements The system is designed and operated in such a way as to differentiate effect such as precursor decomposition, degassing of lines and leakages from the vapour pressure data itself. This is achieved by varying the time interval t before each measurement and recording the pressure changes as a function of time both before and after opening the sample valve. A hypothetical measurement curve can be seen in Fig. A.2. The gradient of the curve measured during t, i.e. when sample is isolated from the system (curve A), provides a measure of the system leak rate plus any degassing of the walls. By subtracting curve A from curve B the pressure increase due to sample decomposition can be estimated (assuming no degassing of the sample). If the derivative at A and B are equal then no sample decomposition occurs. The pressure exerted by the sample is then measured by extrapolating curve B to the point where the sample is opened to the system (C). The true sample vapour pressure is then determined by plotting ∆p (C) values as a function of t and extrapolating back to t = 0.

0 10 20 30 40 50 60 70-202468

101214

t

C

B

∆p (T

orr)

Measurment time (minutes)

A

Figure A.2: Hypothetical measurement result for determining leak rate and sample decomposition.


58

(3-glycidoxypropyl) trimethoxysilane [CAS 2530-83-8] (98 %) and tripropylsilane [CAS 998-29-8] (99 %) were obtained from Aldrich®, 1H, 1H, 2H, 2H- perfluorooctyltriethoxysilane [CAS 51851-37-7] was obtained from ABCR (Brunschwig), 1,2-bis(trimethyl)siloxyethane [CAS 7381-30-8] (99 %) and tetraethoxysilane [CAS 78-10-4] were obtained from (Merck). Samples were used as delivered (no special sample preparation methods) and none of the chemicals were purified before use. Results & Discussion The vapour pressure curves for each of the precursors can be seen in Fig. A.1. The vapour pressures for TEOS, TMSE and TPS are virtually the same within the temperature range measured in this study. The enthalpy of vaporization can be calculated from the gradient of the vapour pressure curve using the Clausius-Clapeyron equation (cf. Eq. (A.1)) and are presented in Table A.1 along with the calculated values for the enthalpy of vaporization.

−∆−= *

* 11expTTR

Hppvap (A.1)

where p* is the vapour pressure at some temperature T*. ∆H is the enthalpy change of vaporization in kJ/mol, R is the gas constant and p is the vapour pressure in Torr. Calculated values ∆H are summarised in Table A.1.

2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6

1

10

100

1000

255075100125150175200

TEOS [22] TEOS [23] TEOS

Vap

our p

ress

ure

(Tor

r)

1000 / T (K-1)

(a)

T (oC)


59

2.55 2.70 2.85 3.00 3.15 3.30

10

100

60708090100110120(b)

TPSV

apou

r pre

ssur

e (T

orr)

1000 / T (K-1)

T (oC)

2.2 2.4 2.6 2.8 3.0 3.2

1

10

1005075100125150175200

(c)

Glymol. PTES TMSE

T (oC)

1000 / T (K-1)

Vap

our p

ress

ure

(Tor

r)

Figure A.3: Vapour pressure plots (a) comparison between Handbook [5], I & E Chemistry [6] and measured values for the vapour pressure curve of tetraethoxysilane, (b) tripropylsilane, (c) (3-glycidoxypropyl) trimethoxysilane, 1H, 1H, 2H, 2H-perfluoroctyltriethoxysilane and 1,2-bis(trimethyl)siloxyethane.

The TEOS vapour pressure curve measured in this study shows good agreement with data previous published in the literature [5, 6] (see Fig. A.3a).


60

Table A.1: Summary of evaporation enthalpies for various silicon based precursors Glymol, PTES, TPS, TMSE and TEOS. TEOS [6] and TEOS [5] refer to data reported in the literature.

None of the precursors showed any significant signs of decomposition within the temperature range they were measured, although signs of de-gassing and the presence of volatile impurities could be seen for all of the samples during initial measurements.

References [1] van Mol A.M.B, Driessen J.P.A.M., Linden J.L., de Croon M.H.J.M., Spee C.I.M.A.,

Schouten J.C., Chem. Vap. Dep., 7, 3, 101-104, (2001). [2] Spee C.I.M.A., Verbeek F., Kraaijkamp J.G., Linden J.L., Rutten T., Delhaye H., van de

Zouwen E.A., Meinema H.A., Mat. Sci. Eng., B17, 108, (1993). [3] Spee C.I.M.A., Linden J.L., Mackor A., Timmer K., Meinema H.A., Mat. Res. Soc. Symp.

Pro., 93, 425, (1996). [4] Waffenschmidt E., Musolf J., Heuken M., Heime K., J. Supercond., 119, 5, (1992). [5] CRC Handbook of Chemistry and Physics 1st student edition, R.C. Weast, (1987). [6] Industrial & Engineering Chemistry (Vapor Pressure of pure substances), D.R. Stull,

(1947).

* RMM refers to Relative Molecular Mass

Precursor ∆H (kJ/mol) RMM* (g/mol) Measuring range

(°C) Glymol 69.4 ± 1.7 236.33 80-180 PTES 66.1 ± .7 510.24 100-160 TPS 48.2 ± 0.5 158.35 50-110

TMSE 48.5 ± 0.7 206.42 50-100 TEOS 50.8 ± 0.6 208.32 20-100

TEOS [6] 45.8 ± 1.1 - 16-168.5 TEOS [5] 42.8 ± 0.0 - 25-110

Appendix B Cross cut tape test

61

Appendix B: Cross cut tape test The ASTM cross cut tape test [17] is designed for evaluating the adhesion of thin films with a thickness of less than 125 µm. The test is administered by making two sets of six parallel cuts perpendicular to one another on a selected area of the substrate free of blemishes and minor surface imperfections. The cuts are made using a sharp razor blade, scalpel, knife or other cutting device in conjunction with a straight edge and ensuring that the coating film has been penetrated. Tape is applied across the cuts and pressed into place by rubbing with the eraser end of a pencil. Within 90 ± 30 seconds of application the tape is removed and the coating rated according to a visual inspection of the coating. In this study, films subjected to the cross cut tape test were rated according to Table B.1.

5 – 0% area removed

3 – 5-15% area removed

2B – 15-35% area removed

1B – 35-65% area removed

0 – > 65% area removed

4 – <5% area removed

Figure B.1: Key to cross cut tape test results.

Chapter 4 Synthesis, monitoring and characterisation of nanoparticles...

63

Chapter 4: Synthesis, monitoring and characterisation of nanoparticles for incorporation into nanocomposite layers

Abstract This chapter deals with the synthesis and characterisation of nanometre sized SiO2 particles in a PECVD system for incorporation into polymer layers. Attempts to synthesize particles using the ICP discharge described in chapter 3 were unsuccessful due to the rather low residence times and absence of confining fields. Particle formation in capacitive plasma systems, notably in semiconductor processing, has however been extensively reported in the literature. Literature concerning the formation and behaviour of particles in low pressure gas discharges is reviewed with the goal of determining typical process conditions and formation mechanisms. Relevant theories and equations are taken and used in rudimentary calculations for the process conditions and reactor geometry used in this study. FTIR absorption spectroscopy coupled with XPS and ESEM measurements are used to determine the composition, size and number density of particles trapped in the discharge.

4.1 Introduction Recent trends, aided by the growing hype surrounding nanotechnology, have made the incorporation of nano-particles into growing layers desirable. Numerous examples where the controlled incorporation of nanometre-sized moieties into growing films produces materials with unique properties that can be tailored to suit particular applications, exist in the literature [1, 2]. The incorporation of nanometre sized silica particles into polymeric films has been shown to densify the material and consequently improve mechanical and physical properties. Improvements in such properties as tensile strength [3], thermal stability [4], scratch and wear resistance [5], diffusion barrier and low-k have been demonstrated. In addition to the synthesis of new and novel materials, the incorporation of particles into growing layers has also been shown to improve growth rates and precursor conversion yields in precipitation CVD processes [6, 7].

In deposition systems used in the semiconductor and micro electronic industries, particle production is an undesirable by-product of some plasma processing steps. Device failure as a result of dust contamination is a familiar issue and has prompted research into particle formation and transport mechanisms in the hope that it might be avoided [8]. A prominent area of literature regarding particle


64

contamination deals with silane chemistries [9, 10, 11, 12, 13, 14] due to its widespread application in industrial processes. Although particles have been observed in TEOS and HMDSO [15] plasma discharges, very little has been done in elucidating the mechanism by which they are formed. This is perhaps partly due to the added complexity of these gas mixtures brought on with the addition of the carbon chemistry.

Fujimoto et al. [16] investigated particulate formation from four siloxane precursors (triethoxysilane, tetramethylorthosilicate, octamethyl-cyclotetrasiloxane and TEOS) in an atmospheric pressure, thermally activated CVD process and determined Arrhenius type activation energies (Ea) for the overall conversion of precursor into particle. As no consideration for thermal or fluid dynamical effects were taken the precise values for the activation energies reported should be taken with care. Their work did however suggest that TEOS, with the lowest overall reaction activation energy, would most readily form particles. Before particles can be synthesized and incorporated into growing silicone layers, an understanding of particle formation and transport processes in CVD reactors is required. 4.1.1 Particle formation in r.f. discharges Two prominent mechanisms have been proposed to explain the manifestation of particles in low pressure discharges. At pressures typical for plasma processing the cross sections for chemical reactions between neutral species are too low to explain the plethora of particles found. Reaction cross sections are several orders of magnitude higher for ions, but positive ions are easily lost to the vessel walls where they either react to form a coating or pick up an electron and neutralise. Research into SiH4 plasmas [17, 18, 19, 20 ,21], has revealed that negatively charged ions can be formed in some gases and, being repelled by the opposing potential in the sheath at the reactor walls, become trapped in the discharge. Negative ions are predominantly the product dissociative attachment processes where electron impact dissociates a neutral species to form a neutral species and a negative ion. Trapped negative ions act as nucleation centres for homogeneous particle growth. The high electro-negativity of fluorine and oxygen, causing high negative ion concentrations, are believed to be responsible for particle production in etch processes [22, 23, 24, 25]. The presence of negative ions also affects the plasma properties and causes them to differ from those of electropositive plasmas as formed from noble gases such as argon. The loss of electron density through electron attachment processes drives the electron temperature up in order to sustain the ionization rate. At high negative ion concentration (sometimes a factor of ten higher then ne [26]) most of the negative charge is carried by the negative ions and the sheath potential drops, and as the current is still carried by the electrons, the plasma becomes more resistive.

Particles have also been observed in plasma systems where no reactive or electronegative gases have been introduced. In these systems, particles are thought to be ejected from the reactor wall as a result of ion bombardment. Once ejected particles rapidly attain negative charge from the plasma glow and become trapped. Suh et al. [27] proposed a more extensive mechanism and devised a model for cluster formation up to (SiO2)10 from SiH4 and O2 gas mixtures using quantum


65

chemical theory. Assuming clusters containing more than ten silicon atoms (SiO2)10 were irreversibly formed and that the nucleation and initial growth of these species was very rapid, they modelled the nucleation and coagulation of particles. The model predicted a fall in particle concentration immediately after the initial nucleation period due to coagulation processes. Immediately after nucleation two mechanisms govern the particle growth rate, coalescence, where particle size increase at the expense of particle number density, and the arrival of growth precursors at the particle surface. Coalescence dominates particle growth immediately after nucleation as growth precursors are depleted. However, as particle numbers decrease and surface charge increases with particle size causing particle to repel on another, the arrival of growth precursors begins to dominate.

At a certain point growing clusters of molecules will reach a size where they begin to experience additional forces to those they experienced as a gas. This point has been proposed to be once a cluster consisted of ten Si atoms or more (for SiO2 particles) [27] or when a cluster reaches a radius of about 10 nm [8].

In order to achieve controlled synthesis and incorporation of these macroscopic particles into growing layers, it is therefore necessary to understand the forces acting on particles in a plasma discharge and how these forces vary depending on the particle dimensions and plasma parameters. 4.2 Forces on particles In the electrode configuration used in this study, the electrostatic force and thermophoretic force act to push the particles away from the r.f. electrode (cf. Fig. 4.1). These forces are counteracted by the ion and neutral drag (Fid, Fnd) and gravitational force (Fg) (see Eqs. (4.8) to (4.12)) that act to push the particle downwards towards the electrode (Fig. 4.1). The relative magnitude of these forces determines the position of the particles since all forces are a function of the position in the reactor.

Fg + FNd + Fid+

+

++

+

+

+

e++

+

+

ee

e

Fes + Fth

Fg + FNd + Fid+

+

++

+

+

+

e++

+

+

ee

e

Fes + Fth

Figure 4.1: Pictorial representation of forces acting on trapped particles. Fg is the gravitational force, Fnd and Fid the neutral and ion drag forces, Fes the electrostatic force and Fth the thermophoretic force.


66

Child-Langmuir sheath theory states that the steady state positive ion and

electron flux to the surface of an insolated body suspended in a plasma has to be equal. To maintain equal electron and ion fluxes, the body gains a negative charge and becomes surrounded by a sheath of positive space charge (see Fig. 4.2).

e

ee

e

ee

e

e

eb.

ThickSheath

e

e

e

e

ThinSheath

a.

Isolated wall

- ----

⊕

⊕

⊕⊕

°°°

°⊕

⊕

⊕⊕

⊕⊕

° ⊕ ⊕ ⊕

⊕

⊕⊕

⊕

⊕

⊕ ⊕

⊕

⊕⊕

⊕⊕

⊕

⊕

°

°

⊕

°°

°° °°

VpVfVp

Vf

e

ee

e

ee

e

e

eb.

ThickSheath

e

e

e

e

ThinSheath

a.

Isolated wall

- ----

⊕

⊕

⊕⊕

°°°

°⊕

⊕

⊕⊕

⊕⊕

° ⊕ ⊕ ⊕

⊕

⊕⊕

⊕

⊕

⊕ ⊕

⊕

⊕⊕

⊕⊕

⊕

⊕

°

°

⊕

°°

°° °°

VpVfVp

Vf

Figure 4.2: a. Thin sheath model where object is larger than sheath dimensions and therefore assumed to be planar and b. Thick sheath model for small objects in the orbit motion limited (OML) regime, note that some of the electrons are not collected.

For a particle with radius (a) greater than the Debye length (λD) the thin or planar sheath model is used (Fig. 4.2a). In the model, ions entering the sheath are assumed to have been accelerated to the Bohm velocity in a pre-sheath region and, assuming a Maxwellian energy distribution function, the potential at the body surface (Vf) with respect to the bulk plasma (Vp) is,

+

−=− 1

2ln

2 e

ieBpf m

meTk

VVπ

(4.1)

where mi and me are the ion and electron mass and ions are first accelerated by ½kBTe/e in the pre-sheath region to the Bohm velocity. For argon this gives Vf - Vp = -5.2 (kBTe/e), and hence for Te = 1 to 3 eV gives Vf - Vp = -5.2 to -15.6 V. A full derivation of the planar sheath model is treated in standard texts [28, 29]. For very small particles however (λD >> a) not all incident ions are collected by the particle and Orbit Motion Limited Theory (OML) is needed. This is pictorially represented in Fig. 4.2b.

Using OML theory the ion current Ii (assuming Maxwellian ion distribution) and electron current Ie (assuming Boltzmann distribution) to the particle surface are,


67

( )

−−

=

iB

pf

i

iBii Tk

VaVem

TkenaI

)(1

8 21

2

ππ (4.2)

( )

−

−=

eB

pf

e

eBie Tk

VaVem

TkenaI

)(exp

8 21

2

ππ (4.3)

where ni is the ion density in the plasma bulk, Ti, the ion temperature, kB the Boltzman constant and Vf(a) the potential of the particle. At steady state Ie = Ii hence,

( ) ( )

−−

=

−

iB

pf

ie

ei

eB

pf

TkVaVe

mTmT

TkVaVe )(

1)(

exp21

(4.4)

The normalized floating potential [e(Vf(a)-Vp) /kBTe] depends only on the

electron to ion mass and temperature ratios. The numerical solution for e(Vf(a)-Vp) /kBTe (Equation 4.4) as a function of Ti/Te for Ar ions is plotted in Fig. 4.3.

0.01 0.1 11

2

3

4

-eV f(a

)-V p /

k BT e

(Ti / Te)2

Figure 4.3: Particle potential as a function of the ion/electron temperature ratio calculated for an ion mass of 40 (Ar).

From Fig. 4.3 we observe that the potential Vf-Vp at the particle surface (for a specific ion mass) depends only on the relative temperature of the ions and electrons and the particle radius.

Using the fit parameters form Fig. 4.3, the potential of the dust particle can be calculated (Eq. (4.4)), assuming the particle is isolated and has spherical geometry. The number of charge carriers is then calculated from the capacitance,


68

aVVeZ

VVQC

pf

D

pf04πε=

−=

−= (4.5)

where ε0 is the permittivity of free space, Q the value of the particle charge and ZD the number of charge carriers (electrons) on the surface of the particle. Applying OML theory to a typical capacitively coupled plasma with an electron temperature of 3 eV, a density of 1015 m-3 and neutral gas temperature of 398 K, gives a particle potential of approximately -7.4 V, which is lower than that derived from the planar sheath model under the same conditions. The electrostatic force (Fes) acting on a particle is then,

eEZQEF Des == (4.6) with E is the electric field vector that points from the bulk plasma towards the powered electrode. The magnitude of E can be very large in a parallel plate reactor configuration as the electrode typically reaches a self bias potential of several tens of volts and is separated from the bulk plasma by a sheath which is typically a few millimetres thick. The resulting electric field is then several tens of thousands of volts per meter as was shown in a recent study of the electric field in an r.f. discharge using fluorescence dip spectroscopy where fields as high as 105 V/m were measured [30].

In addition to the electrostatic force, positively charged ions travelling from the plasma bulk towards the electrode can also collide with the dust particles (see Fig. 4.1). This induces an ion drag (Fid) force that acts to push the particles towards the electrode. The ions pick up energy as they are accelerated through the sheath and some of this energy is converted into heat when the ions collide with the electrode. Heating of the electrode via ion bombardment produces a temperature gradient and a consequent thermophoretic force (Fth) that acts to push particles away from the hot electrode. This assumes that gas heating occurs via collisions with the hot electrode and that heating of neutral species through interaction with charged particles is negligible. The thermophoretic and ion drag forces are calculated using Eqs. (4.7) and (4.8) respectively.

( ) nTnth

th TvaF ∇

−+−= καπ 1

3251

1532

,

2

(4.7)

( ) ( )

−++

+

−≈+= ∞

eB

pfeBeBi

eB

pfCid

Oidid Tk

VaeVTknaTkn

TkVaeV

aFFF 132.611ln 22

22

β (4.8) where κT is the translational part of the thermal conductivity of the gas, ni is the density of the unperturbed plasma, vth,n the thermal velocity and Tn the temperature


69

of the neutral gas. The parameter β is calculated assuming the ion velocity υi is equal to the Bohm velocity using Eq. (4.9),

Di

D

m

eZ

λυπεβ

)21(4 2

0

2

= (4.9)

In addition to the electrostatic, ion drag and thermophoretic forces, the gas

flow also produces a neutral drag force (Fnd), and as the particles have mass they will experience a gravitational force Fg (cf. Fig 4.1). These forces are expressed as,

gaF pg ρπ 334= (4.10)

( )( )nDnthnnnd sHnmaF υυυπ −−= ,2

2 (4.11)

where H(s) is a function of s only and is defined by,

( ) ( ) ( )

−++−

+= s

sss

sssH s erf

411exp

21

2221 π (4.12)

and s is the ratio,

πυυυ 2

,nth

nDs−

= (4.13

ρp is the particle density, g the gravitational constant, mn the mass, nn the number density of the neutrals and erf(s) the error function of s. υD and υn represent the mean velocity of the dust particle and gas flow respectively. Fnd is calculated assuming the ‘kinetic’ or long mean free path’ regime. This regime is established by a high value of the Knudsen number Kn which is defined as the ratio of the mean free path of the neutral gas molecules to the dust particle radius (Kn = λmfp/a). For low values of Kn (high pressure hydrodynamic regime) the drag force can be obtained from the Stokes law and is proportional to the dust particle radius and velocity. For low pressure processes λmfp is in the order of a few hundred micrometers and a is a few µm, Kn is much larger than unity and the kinetic regime is chosen where the principles for the derivation of the neutral drag force are the same as those for the ion drag (see Eq. (4.8)). The gas velocity used for determining the magnitude of Fnd was taken from a fluid dynamic model of the reactor (see Fig. 4.8.) As the particle radius decreases the electrostatic force dominates pushing smaller particles to regions of low electric field (high plasma potential), usually situated in the centre of the plasma glow.


70

All of the forces depend on either the particle size/mass (as the density is assumed to be constant). The electrostatic and thermophoretic forces depend on Te and the temperature gradients inside the reactor respectively. The downwards forces Fg and Fid depend on the particle mass and plasma density (ne) while the Fnd depends on the gas pressure and velocity and acts in the direction of the flow.

0 5 10 15 20 25 3010-17

10-16

10-15

10-14

10-13

10-12

10-11

10-10

10-9

Forc

e (N

)

Particle size (µm)

Fg

Fes

Fth

Fid

Fnd

Figure 4.4: Calculated values for forces acting on dust particles trapped in an r.f. discharge. Fg is the gravitational force, Fnd and Fid the neutral and ion drag forces, Fes the electrostatic force and Fth the thermophoretic force.

The relative magnitude of these forces is plotted as a function of particle radius (a) in Fig. 4.4. The electrostatic force has a linear dependence on the electric field, the magnitude of which varies greatly in the plasma [30]. Bulk plasma is considered quasi-neutral and therefore the electric field in this region is relatively small. In the sheath regions however, large electric fields exist to confine the highly mobile electrons [30]. In this study, self biasing of the driven electrode plate was observed to reach 150 V during particle synthesis experiments. Assuming a neutral gas temperature of 400 K, Te = 3 eV and ne = 1015 m-3 the sheath above the driven electrode is calculated to be 8.3 mm, which is within the experimentally observed estimation of between 5 to 10 mm (from the electrode surface to the beginning of the glow). This gives an electric field above the electrode of approximately 18000 V/m* and consequently the electrostatic force, Fes is calculated to be 7.4·10-13 N for 100 nm particles. Under the same conditions, and assuming a bulk density of SiO2 * This is comparable to measurements performed by Takizawa [30] in an r.f. discharge.


71

to be 2000 kg/m3, the total force contribution from Fg and Fid, is equal to 6.51×10-16 N. The enormous electric field in the plasma sheath therefore represent an insurmountable barrier for the particles and so they become trapped between the plasma glow and the sheath edge.

The electrostatic force is the most dominant for small particles. As the particle radius increases the gravitational force begins to dominate and it is not until the particle diameter reaches 27.2 µm that the particle mass becomes large enough for the particle falls through the electrode sheath (Fig. 4.4). An overall balance of the ‘up’ and ‘down’ forces can be seen in Fig. 4.5.

0 5 10 15 20 25 30 35 4010-14

10-13

10-12

10-11

10-10

10-9

Forc

e (N

)

Particle Diameter (µm)

Total up force

Total down force

27.2µm

Figure 4.5: Balance of forces show that particle with diameter greater than 27.2 µm will be expelled from the plasma.

However, particles considerably smaller than 27 µm are observed buried in heaps of particles observed to form on the electrode during plasma operation. During preliminary particle forming experiments several unusual structures were observed on the driven substrate table under certain flow conditions. As the total gas flow rate entering the reactor decreased, the distribution of particles inside the reactor varied. At relatively high flow rates (> 200 sccm), particles were seen to heap in the centre of the electrode and form around the edges of the cage and conical inlet. At low flow rates (< 56 sccm) particles formed ‘mushroom-like’ structures in the centre of the electrode (Fig. 4.6).


72

Figure 4.6: Mushroom structures deposited onto a 10 cm silicon wafer during a particle synthesis experiment. Experimental conditions were, TEOS = 11.2 sccm, O2 = 28 sccm, Ar = 16.8 sccm, Power = 300 W, Pressure = 1 Torr.

A possible explanation for the observed unusual structures might be as follows. In order for the particles smaller than 27.2 µm to traverse this sheath the repulsive electrostatic force must be reduced. For Fes to be reduced either the field in the sheath must be reduced for a time long enough for the particles to cross, or the particle charge must be reduced. The influence of the various plasma parameters alone is also not sufficient to explain this phenomenon. The plasma density only affects Fid, and even if the calculation is performed taking Te = 1 eV, the forces still balance at a particle diameter of 17.6 µm. Figure 4.1 already holds some suggestions as to how the particles trapped on the sheath edge might overcome the sheath potential. In the centre of the plasma discharge the particle is completely surrounded by a positive space charge as described by Eqs. (4.2) to (4.5). As the particle approaches the electrode sheath, the particle sheath and electrode sheath can combine and part of the particle surface is shielded from electrons in the plasma glow. This causes the net electron flux to the particle surface to drop and consequently the net particle charge to drop. As the particle charge drops, so does Fes and the particle descends still further towards the electrode loosing more charge until eventually the gravitational and neutral drag forces cause it to fall onto the substrate table. In addition to shielding from the electrode, there are also shielding effects from other particles. As particles accumulate in the electrode sheath, the OML sheaths around each particle begin to overlap providing additional shielding from the plasma glow. This process is represented schematically in Fig. 4.7.


73

e

e

e

e

a): Particle grows. Equal e & ion flux to particle in plasma.

e

e

e

e

e

b): Reduced e flux as particle nears electrode sheath.

e

e

e

e

e

ee

c): Reduced Fes allows particle to penetrates sheath.

e

ee

Electrode Electrode Electrode

⊕

⊕ ⊕

⊕

⊕

⊕

⊕

⊕⊕

⊕ ⊕ ⊕ ⊕

⊕⊕

⊕⊕

⊕

⊕

⊕⊕

⊕

⊕

⊕ ⊕⊕ ⊕

⊕⊕

⊕

⊕

⊕⊕

⊕⊕

⊕

⊕

⊕ ⊕

⊕

⊕⊕

⊕

e

e

e

e

a): Particle grows. Equal e & ion flux to particle in plasma.

e

e

e

e

e

b): Reduced e flux as particle nears electrode sheath.

e

e

e

e

e

ee

c): Reduced Fes allows particle to penetrates sheath.

e

ee

Electrode Electrode Electrode

⊕

⊕ ⊕

⊕

⊕

⊕

⊕

⊕⊕

⊕ ⊕ ⊕ ⊕

⊕⊕

⊕⊕

⊕

⊕

⊕⊕

⊕

⊕

⊕ ⊕⊕ ⊕

⊕⊕

⊕

⊕

⊕⊕

⊕⊕

⊕

⊕

⊕ ⊕

⊕

⊕⊕

⊕

Figure 4.7: Particle loss mechanism.

The loss of particle charge as they enter the electrode sheath, and hence reduction of Fes, implies that the large electric field in this region is not responsible for confining the particles in the plasma. Assuming the electric field acting on the charged particles is just above the electrode sheath (where the particles are still immersed in the plasma), and that particles stop growing once the particle and electrode sheath start to combine. Balancing the up and down forces gives,

ththesndidgupdown FQEFFFFFFF +=+=++== (4.14) therefore,

QEFFFFF esthndidg ==−++ (4.15) re-arranging to give

QFFFF

E thndidg )( −++= (4.16)

Using Eqs (4.2) to (4.5), and a particle diameter of between 100 and 200 nm

(taken from ESEM images of particles collected after experiments) the confining electric field, E, becomes between 3.72 and 1.65 V/m. This field strength is associated more with the plasma bulk than the sheath region, suggesting that the particles are trapped just above the electrode sheath.

The particles or particle clusters are experimentally observed to breach the sheath in the centre of the electrode to form “heaps” of particles on the substrate. An understanding of the forces on the particles, combined with a fluid model of the reactor geometry can explain the different particle deposits at different flow conditions. From Fig. 4.8 it is evident that the position at which the particle heaps


74

and mushroom structures form coincides with a stagnation point in the flow profile where Fnd is perpendicular to the electrode surface. The very low flow, and hence long residence time, allows the particles to gradually loose their charge and traverse the sheath as described in Fig. 4.7. At higher flows the stagnation region is large and particles fall through a larger area of the sheath to form particle heaps. The stagnation point is smaller at low flows and a narrow column forms and grows vertically through the sheath as particles continue to breach the sheath-plasma boundary. Once the tip of the column reaches the sheath edge where the particles are suspended, particles begin to collide with the sides of the column tip and the structure begins to grow horizontally in all directions forming a mushroom shape. The importance of the stagnation point is evident from the fluid models shown in Figs. 4.8 and 4.9. In Fig. 4.9 the precursor and carrier gas mixture is injected into the side of the plasma and the stagnation point in the centre of the electrode no longer forms. When this configuration was tried experimentally, no particle heaps or mushroom structures were observed.

Figure 4.8: Fluid model of reactor showing contours in the velocity magnitude (m/s). Note, stagnation point in centre of the electrode.

0.0•10-5

2.0•10-4

4.0•10-4

6.0•10-4

8.0•10-4

1.0•10-3

1.7•10-7

3.8•10-6

7.5•10-6

1.1•10-5

1.45•10-5

1.8•10-5

Inlet flow

Inlet flow from above

Figure 4.8: Fluid model of reactor showing contours in the velocity magnitude (m/s). Note, stagnation point in centre of the electrode.

0.0•10-5

2.0•10-4

4.0•10-4

6.0•10-4

8.0•10-4

1.0•10-3

0.0•10-5

2.0•10-4

4.0•10-4

6.0•10-4

8.0•10-4

1.0•10-3

1.7•10-7

3.8•10-6

7.5•10-6

1.1•10-5

1.45•10-5

1.8•10-5

1.7•10-7

3.8•10-6

7.5•10-6

1.1•10-5

1.45•10-5

1.8•10-5

Inlet flow

Inlet flow from above


75

Figure 4.9: Fluid model of reactor showing contours in the velocity magnitude (m/s). Note, stagnation point no longer exists in centre of the electrode.

Inlet flow

5.5•10-7

5.4•10-5

1.1•10-4

1.6•10-4

2.1•10-4

2.6•10-4

0.0•10-5

2.0•10-4

4.0•10-4

6.0•10-4

8.0•10-4

1.0•10-3

Figure 4.9: Fluid model of reactor showing contours in the velocity magnitude (m/s). Note, stagnation point no longer exists in centre of the electrode.

Inlet flow

5.5•10-7

5.4•10-5

1.1•10-4

1.6•10-4

2.1•10-4

2.6•10-4

5.5•10-7

5.4•10-5

1.1•10-4

1.6•10-4

2.1•10-4

2.6•10-4

0.0•10-5

2.0•10-4

4.0•10-4

6.0•10-4

8.0•10-4

1.0•10-3

0.0•10-5

2.0•10-4

4.0•10-4

6.0•10-4

8.0•10-4

1.0•10-3

To summarize, we have looked at the influence and relative magnitude of the various forces acting on particles trapped in a low pressure r.f. discharge. The calculations suggest that the large electric field generated by the electrode sheath represents an almost insurmountable barrier for particles trapped in the plasma glow. Particles however are observed to penetrate this sheath at sized much smaller than predicted by the calculations. This observation is accounted for by shielding effects that occur when the particles approach the electrode sheath, and at high particle number densities. The formation of mushroom structures observed on the electrode under low gas flows is therefore a consequent of the stagnation point in the gas flow (vertical and positional contribution) combined with the confining electric field above the electrode (responsible for horizontal growth at the mushroom top). 4.3 Experiment Particles were synthesized, trapped and monitored using a capacitively coupled r.f. discharge as described in chapter 3 and Fig. 4.10. The steel substrate table acts as the driven electrode and is surrounded by a grounded steel cage to contain the plasma. No cooling systems were implemented and the substrate typically reached a temperature of 100°C during film growth or particle synthesis which typically lasted about 10 minutes.

Infrared spectroscopic measurements were done using a Midac M2500-C Fourier transform infrared spectrometer (FTIR) that consisted of two units. The


76

detector unit housed the Mercury Cadmium Telluride (MCT) detector, while the source unit housed the Michelson interferometer and glow bar source. The two instrument units were positioned either side of the reactor and the infrared beam (∅ = 2 cm) passed over the driven electrode clearing the substrate table by 3 mm. Spectra were recorded using a spectral resolution of 1 cm-1 and averaging 10 consecutive scans to improve the signal-to-noise ratio. Infrared spectra were de-convoluted using the Peak Fit (Systat Software) and heights of absorbance features were used to monitor variation in species concentration.

~

FTIR Detector

Trapped particles

Powered electrodeGrounded cage

Precursor inlet

FTIR Source

Figure 4.10: Schematic of particle synthesis and trapping reactor with in-situ FTIR spectrometer.

The tetraethoxysilane precursor [CAS 78-10-4] 98 % purity was evaporated and dosed using a Bronkhorst HiTec CEM system while argon and oxygen flows were controlled using Bronkhorst HiTec mass flow controllers. All gas lines to the reactor were heated to 80°C to prevent condensation of the TEOS precursor. A MKS throttle valve located in the reactor exhaust system regulated the reactor pressure, which was constant during each experiment.

The key experiment to deposit hybrid films in which SiO2-like particles are embedded in an organic matrix is as follows. Conditions from particle synthesis and trapping experiments were combined with conditions optimized for growing a-C:H layer. This was done by first synthesizing the particles, purging the precursor gases while holding the SiO2 particles in the plasma sheath above the electrode and then injecting methane. The methane would deposit on the substrate and particle surfaces to form an amorphous carbon layer, in which the particles would hopefully become embedded. A summary of the experiments and process conditions are presented in Table 4.1.


77

Table 4.1: Experimental condition for particle synthesis and trapping experiments.

Experiment

Power [Watt]

Fraction TEOS

Fraction O2

Fraction CH4

Flow [sccm]

Pressure [Torr]

Particle synthesis 100-500 0.05-0.3 0-0.7 0 100-300 1

Mushroom formation 200-300 0.1-0.2 0.5 0 56 0.5-1

Hybrid simultaneous 300 0.2 0.5 0.2 200 1

Hybrid sequential 300 0.1 0.3 0.4 100 1

4.4 Infrared analysis Particle formation in the capacitive plasma could be observed in-situ with FTIR spectroscopy [31]. The decomposition products of TEOS in the predominantly oxygen plasma were CO2 (2300 cm-1), CO (2200 cm-1), HCOOH (1776 cm-1), CH4 (2900 cm-1) and H2O or OH (3500 - 4000 cm-1 and 1500 - 2000 cm-1), and residual TEOS (1000 - 1250 cm-1). The transversal optical (T.O.) vibrational mode corresponding to solid SiO2 could also be identified at (1066 cm-1). By switching the precursor gas flow off and keeping the plasma discharge on, particles could be trapped a region close to the sheath above the powered electrode. It was found that particles could be trapped for an indefinite time period in the argon discharge. The solid SiO2 IR peak could be verified by comparison with an IR absorption spectrum taken of particles collected after the deposition experiment (Fig. 4.11).

The calculated extinction of the infrared beam resulting from either bulk or surface absorptions was simulated for SiO2 and SiO, assuming spherical particles and optical constants obtained from the literature [32], using Eqs. C.1 to C.6 in Appendix C, and plotted as a function of wavelength in Fig. 4.12.

It can be seen that the measured particle peak shows most similarity to the bulk and surface SiO2 synthetic peaks. XPS measurements taken of particles collected from after the experiment showed the particle composition to be almost that of bulk SiO2 at 32.5 % Si, 67.5 % O and 0.4 % C, which gives O/Si of 2.1 and 0.4 % carbon impurity.


78

1000 1500 2000 2500 3000 3500 4000 4500

0.0

0.2

0.4

0.6

0.8

1.0

Abs

orpt

ion

(arb

. uni

ts)

Wavenumber (cm-1)

Ex situ: FTIR of particles collected. In situ: FTIR of particles trapped in plasma.

SiO2

residual gas phase products in plasma

Figure 4.11: Comparison between FTIR spectra of particles trapped in a TEOS free plasma discharge and particles collected on a c-Si wafer afterwards.

600 700 800 900 1000 1100 1200 1300 14000.0

0.2

0.4

0.6

0.8

1.0

Extin

ctio

n (A

rb. u

nits

)

Wavenumber (cm-1)

As deposited Surface SiO2

Bulk SiO2

Surface SiO Bulk SiO

Figure 4.12: Comparison between calculated bulk and surface vibrations for SiO and SiO2 with measured IR spectra of particles synthesized.


79

Particles trapped due to the high electric field in the sheath region above the powered electrode were monitored with the FTIR spectrometer. Figure 4.13 shows the growth, trapping and loss of particles from the plasma as a function of time as monitored by the solid-state SiO2 peak as shown in Fig. 4.11. The plot is divided into three sections representing particle nucleation, trapping and expulsion from plasma stages of the experiment. During particle nucleation the SiO2 peak increases sharply and reaches a maximum two minutes after the precursor flows are directed into the reactor. The TEOS signal increase less rapidly during particle nucleation as the TEOS is being consumed in particle formation and growth reactions. After 2 minutes the first particles grow large enough to leave the plasma glow and an equilibrium is reached between particle formation/growth and particles dropping out of the infrared beam and onto the substrate. At four minutes the precursors flows are switched off and only argon is dosed into the discharge. A drop in the SiO2 absorption peak at this point indicates loss of some particles from the plasma sheath. On switching the precursors flows off, the plasma conditions change from that of an electro-negative discharge caused by the O2 with low ne caused by the TEOS to an electro positive with higher ne. This particle loss is attributed to a change in the plasma conditions as argon gradually replaces the precursor gases (possibly a result of higher ion drag).

-1

0

1

2

3

4

5

0 2 4 6 8 10

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

area

A [a

.u.]

[TE

OS

] uc [

mol

e%]

time [min]

particle formation Ar/TEOS/O2 plasma

particle trapping Ar plasma

particle expulsion no plasma

loss of particles

trapped particles

film deposition

Time (min.)

TEO

S (m

ole

%)

particle formationAr/TEOS/O2 plasma

Are

a (a

rb. U

nits)

particle trappingAr plasma

loss of particles

particle expulsionno plasma

trappedparticles

filmdeposition

-1

0

1

2

3

4

5

0 2 4 6 8 10

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

area

A [a

.u.]

[TE

OS

] uc [

mol

e%]

time [min]

particle formation Ar/TEOS/O2 plasma

particle trapping Ar plasma

particle expulsion no plasma

loss of particles

trapped particles

film deposition

Time (min.)

TEO

S (m

ole

%)

particle formationAr/TEOS/O2 plasma

Are

a (a

rb. U

nits)

particle trappingAr plasma

loss of particles

particle expulsionno plasma

trappedparticles

filmdeposition

Figure 4.13: Time dependent particle measurements of particle trapping using FTIR spectroscopy in a TEOS/O2 plasma.

A bulk of the particles remains trapped above the powered electrode until the r.f. field is switched off after nine minutes. At this point all remaining particles drop out of the infrared beam and onto the electrode. The residual SiO2 signal


80

observed between nine and eleven minutes is due to deposition of material on the KBr windows inside the reactor. 4.4.1 Rayleigh scatting The particle number density for a typical plasma experiment, as shown in Fig 4.13, can be estimated from the absorption data using Eqs. (C.1) to (C.12) presented in Appendix C. ESEM images of particles collected after an experiment (Fig 4.14) give a particle radius of approximately 100 nm. Combining this estimation of the particle size with the SiO2 absorption peak centred at 1066 cm-1 provides an estimation of the particle density np (Eq. (C.2) to (C.6) Appendix C), np of approximately 1014 m-3. This also explains why no scattering signal is observed for the particles synthesized in the pure TEOS/O2 plasma. Extinction of the infrared radiation when passing through a particle cloud of np = 1014 m-3 particles of a = 100 nm gives A(1066 cm-1) = 0.003. This value is smaller than the signal-to-noise ratio of the spectra recorded using a sample rate of 1 scan/s and averaging 10 consecutive scans, and hence no scattering signal could be detected. In addition to this, drift of the baseline signal occurred within the 12 minute time scale of the experiment, preventing the possibility to increase the signal acquisition time and improve the sensitivity.

Rayleigh scattering was observed however, in an experiment originally designed to deposit SiO2 particles in an amorphous carbon layer (as described in the experimental section). Particles synthesized in a TEOS/O2 plasma and suspended in a methane plasma grew to a size where significant light scattering could be observed. Fig. 4.15 shows the FTIR spectra recorded of the particles trapped in the methane plasma and two simulated lines calculated using the method described in Appendix C. The broad absorption feature centred at 2950 cm-1 is attributed to an amorphous carbon layer deposited on the particles.

Figure 4.14: Particle size distribution measured by ESEM image taken of particle clusters.


81

There is some uncertainty as in the literature as to exactly how the Rayleigh scattering model should be applied to infrared data exhibiting such baseline tilts. Courteille et al. [15] used a wide spectral range from 1000 to 4000 cm-1 to fit the baseline and derive values for the particle size and number density despite apparent deviation from the observed spectrum at 3000 cm-1. Stoffels et al. [21] however, applied the Rayleigh equations to a narrower spectral region, between 1000 and 3000 cm-1 to achieve a better correlation with the data. Deviations at higher wavenumbers were attributed to a breakdown of the model assumptions (namely 2πa/λ << 1 [33]) and at higher wavenumbers a Mie scattering model should be applied. Both of these approaches have been were investigated in this study and a comparison of the results is presented in Table C.1 in Appendix C.

The calculations presented in Appendix C assume that the particles are all of equal size and spherical. The optical constants for amorphous carbon are not exactly known in the literature, due to the strong dependence of the optical properties of these layers on the deposition conditions and the large variety of deposition conditions reported for this material. The particle size calculations presented here use values for a-C:H reported by Savvides et al. [34] and Pawlak et al. [35] of n = 6.1 + 3.4i. ESEM image taken of purely SiO2 particles collected after a light scattering experiment indicate the SiO2 particle radius (a1) to be between 50-150 nm (Fig. 4.14).

1000 2000 3000 4000 5000 60000.000

0.004

0.008

0.012

0.016

0.020

Recorded spectra Simulated spectra

IR A

bsor

ptio

n (a

rb. u

nits

)

Wavelength (cm-1)

Figure 4.15: Infrared spectrum recorded during the deposition of a nanocomposite material, along with simulated curves for Rayleigh scattering.

4.5 Physical characterisation of particles Thermal gravimetric analysis (TGA) is a useful technique that provides information regarding the thermal stability and to some extent structure of a material. The measurement is performed by placing a sample of known mass in an oven and


82

heating under a controlled atmosphere. The thermal stability is determined by monitoring variations in the sample mass under either inert (Ar or N2) or oxidizing (O2 or O2/N2) atmospheres. In addition to monitoring variation in sample mass, monitoring the energy absorbed by the sample at different temperatures can also indicate variation in the material structure. Results from TGA can help predict how a material will react during post deposition plasma or thermal processing as is typically encountered when integrating a new material into semiconductor process. TGA was performed on particles collected after film deposition in both pure N2 and 20 % O2 in N2 atmospheres, using a temperature ramp rate set to 5°C/minute. Both measurements indicate the particles to be thermally stable up until temperatures in excess of 800°C. A 14 % weight loss was observed for both measurements indicating no further oxidation of the particles took place in the oxidising atmosphere. A positive gradient in the temperature differential is indicative of exothermic reactions within the sample, and a negative gradient indicates endothermic chemical changes. The initial positive gradient is rather confusing as it suggests an exothermic reaction is occurring during the initial heat up. A possible explanation is that during this period any Si-OH groups in the sample might react to liberate water and form SiO2. The evaporation of water from the sample would be an endothermic process, but the overall formation of SiO2 and H2O would be exothermic. The sample showed no signs of blackening or decomposition after TGA measurements, again indicating that the plasma produced particles are very stable.

0 200 400 600 800 1000

20

40

60

80

100

-4

-3

-2

-1

0

1

2

3 % Weight loss O2

% Weight loss N2

Wei

ght c

hang

e (%

)

Temperature oC

Temp. Diff. O2

Temp. Diff. N2

Tem

p. D

iff. (

o C/m

g)

Figure 4.16: TGA analysis taken in pure N2 and Air of powder collected from Zone I after nanocomposite layer synthesis. Note the initial positive slope which might indicate hydroxyl loss.


83

4.6 Discussion and conclusions The synthesis, trapping of particles from a TEOS/O2 gas mixture was successfully demonstrated using a capacitively coupled discharge. The high thermal stability suggested by TGA, coupled with the dominance of Si-O groups in the infrared suggests the particles are silica-like. However, it is likely that the particles are highly porous. Porosity is known to exist in plasma deposited films deposited from siloxane precursors such as TEOS [36] and HMDSO [37] and high bias potential are used to increase film density (as demonstrated in Fig. 3.6). The particles however, are formed from negative ions trapped in the plasma glow and the sheath surrounding them, and hence the energy of ions impinging on them is much lower than in the sheath region above the electrode (cf. section 4.2). As the formation of dense fully oxidized SiO2 networks requires high energy input, (as seen with SiO2 deposition in chapter 3) the energy is not available to form dense particles. In addition to this, particles collected post-deposition coagulate into extremely light structures that are easily affected by electrostatic forces. Small angle x-ray diffraction (SAXS) measurements also show no long range order within the particle structure. All these observations suggest the particles consist of an extremely porous SiO2 network.

Particle number density and size could be estimated from in situ absorption FTIR measurements. The formation of particle heaps, and mushroom structures in the centre of the powered electrode was discussed in the context of OML theory combined with a fluid dynamic model of the reactor geometry used.

Although incorporating porous particles into silicone layers is unlikely to offer any improvement in mechanical or barrier properties, the controlled incorporation of porosity is interesting for low dielectric constant materials. Two limitations in the current reactor design have been identified from this work. Firstly the stagnation point in the current reactor geometry prevents the homogeneous distribution of particles throughout the growing layer. Secondly, the use of a single plasma source prohibits the simultaneous synthesis of a SiO2 and silicone layer as the two chemistries cannot be separately controlled. Using a modified reactor design, chapter 6 investigates the deposition and characterisation of nano-porous layers for low dielectric constant applications.

Acknowledgement I would like to thank A. Klinkenberg for his CFD modelling work.


84

References [1] Terranova M.L., Piccirillo S., Sessa V., Rossi M., Botti S., J. Phys. IV France, 9, 365-

371, (1999). [2] Senkevich J.J., Seshu D.B., Chem Mater. 11, 1814-1821, (1999). [3] Suzuki F., Onozato K., J. Appl. Polym. Sci., 39, 371-381, (1990). [4] Sohoni G.B., Mark J.E., J. Appl. Polym. Sci., 45, 1763-1775, (1992). [5] Sawyer W.G., Freudenberg K.D., Bhimaraj P., Schadler L.S., Wear, 254, 573-580,

(2003). [6] Hurt R.H., Allendorf M.D., AIChE. J., 37, 1485-1495, (1991). [7] Tandon P., Rosner D.E., AIChE J., 42, 1673-1684, (1996). [8] Dusty Plasmas, Bouchoule A., Wiley, New York, (1999). [9] Stoffels E., Stoffels W.W., Kroesen G.M.W., Hoog F.J.de., J. Vac. Sci. Technol. A,

14, 556-561, (1996). [10] Boufendi L., Hermann J., Bouchoule A., Dubreuil B., Stoffels E., Stoffels W.W.,

Giorgi M.L. de., J. Appl. Phys., 76, 148-153, (1994). [11] Cabarrocas P.R.I.,Gay P., Hadjadj A., J. Vac. Sci. Technol. A, 14, 655-659, (1996). [12] Dorier J.-L., Hollenstein Ch. Howling A.A., J. Vac. Sci. Technol. A, 13, 918-926,

(1995). [13] Watanabe Y., Shiratani M., Plasma Sources Sci. Technol., 3, 286-291, (1994). [14] Hollenstein Ch., Howling A.A., Courteille C., Magni D., Scholz S.M., Kroesen

G.M.W., Simons N., Zeeuw W de., Schwarzenback W., J. Phys. D: Appl. Phys., 31, 74-84, (1998).

[15] Courteille C., Magni D., Deschenaux Ch., Fayet P., Howling A.A., Hollenstein Ch., Soc. Vac. Coat. 41st Annual Technical Conference Proc., (1998).

[16] Fujimoto T., Okuyama K., Yamada S., Adachi M., J. Appl. Phys., 85, 4196-4206, (1999).

[17] Optical Studies of micron-sized particles immersed in a plasma, Swinkels G.H.P.M, PhD. Thesis, Technische Universitiet Eindhoven, (1999).

[18] Boufendi L., Bouchoule A., Plasma Sources Sci. Technol., 3, 262, (1994). [19] Hollenstein Ch., Dorier J.L., Dutta J., Sansonnens L., Howling A.A., Plasma Sources

Sci. Technol., 3, 278, (1994). [20] Dorier J.L., Hollenstein Ch., Howling A.A., J. Vac. Sci. Technol. A, 13, 918, (1995). [21] Stoffels E., Stoffels W.W., Kroesen G.M.W., Hoog F.J. de, Electron Technology, 32,

255, (1998). [22] Buss R.J., Hareland W.A., Plasma Sources Sci. Technol., 3, 268-272, (1994). [23] Steinbruchel C., Yoo W.J., Plasma Sources Sci. Technol., 3, 273-277, (1994). [24] Stoffels W.W., Stoffels E., Kroesen G.M.W., Haverlag M., den Boer J.H.W.G., de

Hoog F.J., Plasma Sources Sci. Technol., 3, 320-324, (1994). [25] Kamata T., Kakuta S., Yamaguchi Y., Makabe T., Plasma Sources Sci. Technol., 3,

310-313, (1994). [26] Amemiya H., J. Phys. D: Appl. Phys., 23, 999-1014, (1990). [27] Suh S.-M., Zachariah M.R., Girshick S.L., J. Vac. Sci. Technol. A, 19, 940-951,

(2001). [28] Introduction to Plasma Physics and Controlled Fusion, Vol., I Chen F.F., Plenum

Press, New York, (1984). [29] Fundamentals of Plasma Physics, Bittencourt J.A., Pergamon Press, Oxford, (1986). [30] Takizawa K., Sasaki K., Kono A., Appl. Phys. Lett., 84, 185-187, (2004).


85

[31] Courteille C., Magni D., Howling A.A., Nosenko V., Hollenstein C., Soc. Vac. Coat.

40th Annual Technical Conference Proc., (1997). [32] Handbook of optical constants of solids, Palik E.D., Academic Press, London,

(1998). [33] Light scattering by small particles, van de Hulst H.C., Chapman & Hall, New York

& Dover, (1981). [34] Savvides N., J. Appl. Phys., 59, 4133, (1986). [35] Pawlak F., Balanzat E., Dufour Ch., Laurent A., Paumier E., Perriere J., Stoquert J.P.,

Toulemonde M., Nucl. Instr. Meth. In Phys. Res. B, 122, 579-582, (1997). [36] Jeong S.H., Nishii J., Park H.-R., Kim J.-K., Lee B.-T., Surf. Coat. Technol., 168, 51-

56, (2003). [37] Borvon G., Goullet A., Mellhaoui X., Charrouf N., Granier A., Mater. Sci. Semicond.

Process., 5, 297-284, (2002).

Appendix C Infrared analysis of particles

86

Appendix C Infrared analysis of particles A molecule can absorb radiation if the energy of the photon matches the transition between energy states in the molecule. For vibrational energy transitions, the energy difference corresponds to wavelengths in the infrared region of the electromagnetic spectrum. The selection rules for the transition state that there must be a change in the molecules dipole moment during the transition. Therefore, almost all except homonuclear diatomic molecules absorb infrared radiation, making IR spectroscopy ideal for identifying chemical systems. For a detailed description of infrared spectroscopic methods see Griffiths [1] or Pelikan [2]. The application of the FTIR absorption spectroscopy technique for monitoring particles trapped in a plasma sheath is presented here.

Infrared spectroscopy can be used to quantify constituent concentrations in a sample by application of the Beer-Lambert law (Eq C.1) in which the absorbance A of a species is related to the ratio of the intensity of the incident (I0) and transmitted (I) light, sample path length (l) and concentration (c),

( ) ( )( ) ( )lc

IIA να

ννν =−=

0log (C.1)

where α is the coefficient of absorption. If scattering can be neglected then, the absorption of infrared radiation by SiO2 particles can be used to determine particle number density. The shape and intensity of Si-O absorption features centred at 1066cm-1 differs depending on whether the sample consists of a single bulk of SiO2 or as fine powder dispersion. Using bulk values for the optical constants n and k, the extinction of the IR beam due to absorption can be calculated assuming spherical particles using Eq. (C.2),

( ) ( )( )

dkeIIA νπ

ννν 4

0log −=−= (C.2)

where ν is the wavenumber (cm-1), k the extinction coefficient of bulk SiO2 and d is a parameter relating to the effective combined thickness of material (particles) traversed by the light beam. In this particular case where we consider a cloud of particles suspended in the plasma, the “material” is represented by the thin cloud of particles, with number density nP. The parameter d can be related to the particle radius a and number density by the following derivation. The volume of the cloud of particles within the infrared beam is given by,

drV bcloud2π= (C.3)

where rb is the and diameter of the infrared beam in the plasma. The SiO2 particles are assumed to be spherical and hence Vparticle is given by,


87

3

34

TParticles aV π= (C.4)

where aT is the particle radius (the reason for the subscript T will become evident later in Fig C.1). The number density nP can now be calculated based on the volumes calculated above:

343

Tbparticlebeam

cloudP al

dVV

Vnπ

=⋅

= (C.5)

where,

bbbeam lrV 2π= (C.6) and lb is the length of the infrared beam. Consequently, nP·aT

3 can be given as a function of the parameter d, which is a constant for one simulation:

13

43 C

ldan

bTP ≡=⋅

π (C.7)

where C1 is a constant. d is determined from fitted simulated spectra using SiO2 bulk values for n and k (Eq. C.2). In addition to absorption, if the particles trapped in the infrared beam are large enough (2πaT >> λ) they will also lead to Rayleigh and Mie scattering. This can be seen as a tilting of the baseline in the spectrum that becomes stronger at higher wavenumbers as both Rayleigh and Mie scattering are wavelength dependant. The extinction of light due to scattering can be calculated by defining the scattering cross section Csca. From Eq. C.2,

( )10ln

bPsca lnCA =ν (C.8)

where,

2~1~

2

2

4

65

3128

+−=

nnaC T

sca λπ

(C.9)

where aT is the particle radius, λ the wavelength of incident light and n~ the complex refractive index of the particle. This theory assumes particles are spherical and of sufficiently low number density that scattering is independent, i.e. distance between particles is larger then three times the particle radius. Equation (C.9) reflects the well known λ4 dependence of Rayleigh scattering.


88

At this point we have two expressions (C.5 and C.8) and two unknowns (aT and nP). It is therefore possible to determine both the particle diameter and density provided we have both scattering and absorption data. Unfortunately no scattering was observed during the synthesis of pure SiO2 particles. In order to solve for nP, aT was estimated from ESEM images taken of particles collected after the experiment (cf. Fig 4.14). Taking aT to be between 100-150 nm, nP is calculated to be approximately 1014 m-3.

For cases where the particle is composed of more than one material (e.g. a coated particle), the expression (Eq. C.8) is modified so that aT becomes the total particle radius (Fig. C.1).

a1

a2aT

SiO2 corea-C:H coating

a1

a2aT

SiO2 corea-C:H coating

Figure C.1: Structure of SiO2 particle coated with a-C:H layer.

n~ is replaced with the effective refractive index approximation effn~ as determined by Eq. (C.10), the so called Bruggeman effective medium approximation [3].

22

221

12 ~~~ nVVn

VVn

tottoteff += (C.10)

where V1 and V2 are the volumes of materials with refractive indices 1

~n and 2~n .

The total extinction due to Rayleigh scattering is therefore,

( ) 42

462

25

3128

2~1~

10lnννπν Can

nnl

A TPeff

effb =+−

= , (C.11)

where C2 is a constant and aT = a1 + a2. Because C2 can be determined by simulating the baseline tilt, 6

TP an ⋅ can be determined from this:

3

2

25

3128

262

2~1~

10ln

C

nnl

Can

eff

effbP =

+−

=⋅π

, (C.12)


89

where C3 is a constant. The two equations derived (Eqs. C.9 and C.12) contain three unknown

parameters a1, a2 and n. However, with these equations, a2 can be determined as a function of a1:

21

61

11

32 a

CCa

= . (C.13)

and a2 is determined as a function of a1.

SiO2 particle coated with a-C:H did cause significant scattering and absorption of the infrared radiation, but it is still not possible to solve due to the addition of a new variable a2. Assuming a1 and nP from the pure SiO2 case we can estimate a2 to be between 100 - 120 nm, giving a total particle diameter, aT, of to be between 200-270 nm (see Fig. 4.14). References [1] Chemical Infrared Fourier Transform Spectroscopy, Griffiths P.R., Wiley & Sons, New

York,(1975). [2] Applications of Numerical Methods in Molecular Spectroscopy, Pelican P., Ceppan M.,

Liska M., CRC Press, London, (1993). [3] Petrik P., Lehnert W., Schneider C., Lohner T., Fried M., Gyulai J., Ryssel H., Thin Solid

Films, 383, 235-240. (2001).

Chapter 5 Reaction kinetics and mechanisms during PECVD from a TEOS/O2 gas mixture

91

Chapter 5: Reaction kinetics and mechanisms during PECVD from a TEOS/O2 gas mixture

Abstract The electron impact dissociation of TEOS and reaction with atomic oxygen radicals are investigated and compared with a view to elucidate the TEOS decomposition mechanism in an r.f. capacitively coupled discharge. Rate constants for a wide variety of electron impact and neutral chemistry processes are calculated using data taken from the literature. Dominant reactions for the decomposition of TEOS and formation of hydrocarbons and CHOOH are identified and compared with FTIR absorption and optical emission spectroscopy measurements made during a TEOS/O2 PECVD process.

5.1 Introduction The thermal and plasma chemistry of the siloxane/O2 deposition process is complicated by the wide variety of compounds both silicon and carbon readily react to form. However, these reactions are of considerable interest in the field of organometallic chemistry and are vital to the understanding of the chemical vapour deposition (CVD) and plasma enhanced chemical vapour deposition (PECVD) of silicon, silica, silicon nitride and not least, silicone based hybrid thin films [1]. Determining the intrinsic kinetics and mechanisms of plasma driven CVD processes using organosilicon precursors is not only complicated by the chemical nature of the elements present in the precursor, but also due to the nature of the plasma process itself. Plasmas are non-equilibrium systems that are typically characterised by three temperatures, the electron temperature (Te) that governs the electron impact kinetics, the ion temperature and the neutral gas temperature upon which the neutral chemistry depends. For plasmas with a relatively low degree of ionization, like the capacitively coupled plasma used in this study, the ion and background gas temperature are usually very close and are often assumed to be equal. The non-equilibrium nature of plasma processes results in a complex chemistry where the global conversion of precursors into products and by-products is the result of a complex interplay between these processes.

Intrinsic reaction kinetics and mechanisms are also effected by the fluid dynamics and diffusion processes specific to any given reactor configuration. A preliminary insight into these processes can be derived by correlating experimental observation with accepted chemical or physical ideology. An example of this would be to deriving a global activation energy by correlating growth rate with


92

substrate or reactor temperature. Rudimentary reaction kinetics describing such processes, such as precursor conversion to film with one rate constant (and hence one step mechanism), are derived from such measurements. However, the oversimplification of the chemistry in these models often limits their applicability to the reactor system in which the measurements were made. A number of authors have used this approach to investigate the tetraethoxysilane (TEOS) and oxygen plasma system with sometimes contradicting findings [2, 3]. In an XPS and mass spectroscopic study Fracassi et al. [4] concluded that the TEOS decomposition was primarily the result of electron impact dissociation and that oxygen only slightly influenced the deposition rate even when the O2/TEOS ratio was over 2.5. Other authors’ report a more dramatic increase in deposition rate as the oxygen concentration is increased, suggesting oxygen plays a more dominant role in the TEOS decomposition [5, 6, 7]. In reality, the relative influence of oxygen on the film deposition rate and process chemistry is a consequence of the electron energy and energy distribution function, which in itself depends on the specific plasma system.

More sophisticated models based on Monte Carlo simulations offer a greater insight into the intrinsic processes and offer greater applicability at the cost of computational expense. This approach was employed by Stout et al. [8] in a Monte Carlo model of the TEOS/O2 plasma. The results of this model corroborated the surface atomic oxygen mechanism proposed by Raupp et al. [5]. In addition to the computation requirements, these models also rely heavily on thermochemical data derived from either experimental or theoretical studies and is often difficult to find. In this work, rate constants for a wide variety of reactions are calculated from data taken from literature sources. This data is evaluated with the aim of deriving mechanisms for the decomposition of TEOS and formation of several of the observed by-products. Fourier transform infrared (FTIR) absorption spectroscopy and optical emission spectroscopy (OES) are used to monitor precursor and by-product concentrations and substantiate the mechanisms proposed. Data used for the rate calculations presented here was obtained from work published by Allendorf et al. [9, 10], Meeks [11], Ho [12] for the oxygen and argon chemistry, Holtgrave et al. [13] Coltrin et al. [14], Morgan et al. [31] and Ho [15] for the electron impact TEOS rates and various authors for the neutral hydrocarbon chemistry [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30]. 5.2 Molecular processes in TEOS/O2 discharges The relative importance of the following two possible initiation processes for subsequent decomposition of TEOS is considered, namely, the electron impact dissociation and the reaction with atomic oxygen as outlined in Eqs. (5.1) and (5.2). e + Si(OC2H5)4 → Products (5.1) Si(OC2H5)4 + O → Products (5.2)

Very little has been published regarding the collision cross section for electron impact with TEOS. In an experimental study Holtgrave et al. [13] reported electron


93

impact cross section of TEOS ionization reactions derived from Fourier-transform ion mass spectrometry measurements. Their measurements suggested that TEOS fragments preferentially to form ions of mass 149, 193, and 179 corresponding to Si(OC2H5)2OCH3

+, Si(OC2H5)3OCH2+ and Si(OC2H5)3O+ respectively. These

fragments are formed from the loss of ethyl, ethoxy and methyl radicals. They did not observe the formation of negative ions and as neutral products cannot be trapped, the Fourier-transform ion mass spectrometry technique cannot measure them.

In a more recent study Morgan et al. [31] calculated the cross section for electron impact excitation and dissociation for TEOS to form neutral products. Results from both these studies suggest that the threshold electron energy for dissociation or ionization is in excess of 10 eV and that electrons with lower energy would only cause excitation. Electron impact dissociations mechanisms suggested by Morgan, Holtgrave and Coltrin [14] are listed in Table 5.1.

Table 5.1: Electron impact dissociation mechanisms for TEOS.

Reaction Energy (eV) Ref.

Si(OC2H5)4 + e ↔ Si(OC2H5)3OCH2˙ + CH3˙ + e 3.6 [31] Si(OC2H5)4 + e ↔ Si(OC2H5)3O˙ + C2H5˙ + e 4.0 [31] Si(OC2H5)4 + e ↔ Si(OC2H5)3˙ + OC2H5˙ + e 5.2 [31] Si(OC2H5)4 + e ↔ Si(OC2H5)2OCH3

+ + OC2H5˙+ CH2˙ + 2e - [13] Si(OC2H5)4 + e ↔ Si(OC2H5)3OCH2

+ + CH3˙ + 2e - [13] Si(OC2H5)4 + e ↔ Si(OC2H5)3O+ + C2H5˙ + 2e - [13] Si(OC2H5)4 + e ↔ Si(OC2H5)3(OH) + C2H4 + e - [14]

Morgan et al. reported only one rate constant for the electron impact dissociation of TEOS and did not specify which of the possible decomposition mechanisms listed in Table 5.1 this was attributed to. The lower activation energy associated with the first mechanism (3.6 eV) suggests that would be the preferred mechanism, although here all reactions are assumed to have an equal probability.

The second reaction for the initiation of TEOS (Eq. (5.2)) requires atomic oxygen that is formed from the electron impact dissociation of O2 as described in Eq. (5.3). O2 + e → 2 O + e (5.3)

The rate constant for this process has been reported by Meeks et al. [11] in a comprehensive study of the O2/Ar plasma system. The reaction scheme considers over 35 separate excitation and dissociation processes including electron impact


94

reactions with Ar and O2, metastable and ion reactions and neutral radical reactions with O2. A comparison between the rate constant for the electron impact dissociation of oxygen and TEOS at different electron energy distributions is plotted in Fig. 5.1.

0 2 4 6 8 10 12 1410-11

10-10

10-9

10-8

10-7

10-6

10-5

Morgan [31]: TEOS + e ions + e Holtgrave [13]: TEOS + e ions + e Morgan [31]: TEOS + e neutrals + e Meeks [11]: O2 + e O + O* + e

Meeks [11]: O2 + e O + O-

k (c

m3 /s

)

Te (eV) Figure 5.1: Electron impact dissociation rates for oxygen and TEOS as a function of electron temperature. Rate coefficients for the electron impact reactions are calculated by multiplying the electron impact cross sections with the electron density for each electron energy assuming a Maxwellian electron energy distribution function.

The total rate constant for electron impact ionisation calculated from Morgan et al. [31] is slightly lower, but agrees well with the total ionisation rate calculated from the experimental data published by Holtgrave et al. [13]. The datasets suggest that the CH3, CH2, OC2H5 and C2H5 species are the most dominant by-products of electron impact dissociation and electron impact ionisation reactions, with a slight preferential decomposition into ions as opposed to neutral species. 5.3 Experimental Experiments were performed in a capacitively coupled plasma driven by a 1 kW r.f. power supply operated at 13.56 MHz (Fig. 5.2). The steel substrate table acts as the driven electrode and is surrounded by a grounded steel cage to contain the plasma. No cooling systems were implemented and the substrate typically reached a temperature of 100°C during film growth or particle synthesis experiments that typically lasted 10 minutes (cf. chapters 4 and 6). Precursors are evaporated and dosed using a Bronkhorst HiTec CEM system while argon and oxygen flows are


95

controlled using Bronkhorst HiTec mass flow controllers. All gas lines to the reactor were heated to 80°C to prevent condensation of the TEOS precursor. A MKS throttle valve located in the reactor exhaust system regulated the reactor pressure, which is maintained at 1 Torr during all experiments.

~

FTIR detectorFTIR source

Plasma glow


OES

Precursor inlet

~

FTIR detectorFTIR source

Plasma glow


OES

Precursor inlet

Figure 5.2: Reactor configuration indicating in situ FTIR absorption and optical emission spectroscopy. The substrate table is also the driven electrode and is surrounded by a grounded cage to confine the plasma glow.

Tetraethoxysilane [CAS 78-10-4] 98 % purity was supplied by Merck and used as delivered. Experimental flows and conditions are listed in Table 5.2.

Table 5.2: Experimental condition for particle synthesis of nanocomposite films.

Objective Plasma Power [W]

TEOS Flow [sccm]

O2 Flow [sccm]

Total Flow [sccm]

Pressure [Torr]

Actinometry 0 - 400 - 0 – 90 100 0.12 – 1

Chemistry 0 – 400 0 – 30 0 – 50 100 1

Infrared absorption spectroscopic measurements were done using a Midac M2500-C Fourier transform infrared spectrometer (FTIR) consisting of two units. The detector unit housed the Mercury Cadmium Telluride (MCT) detector, while the source unit housed the Michelson interferometer and glow bar source. The two instrument units were positioned either side of the reactor and the infrared beam (∅


96

= 2 cm) passed over the driven electrode clearing the substrate table by approximately 3 mm. Spectra were recorded using a spectral resolution of 1 cm-1 and averaging ten consecutive scans to improve the signal-to-noise ratio. Infrared spectra were de-convoluted using Peak Fit (Systat Software) and heights of absorbance features were used to monitor variation in species concentration.

Optical emission spectroscopy was done using an Ocean Optics USB-2000 spectrometer with a spectral range of between 280 to 800 nm with a resolution of approximately 1 nm. Details of the OES measurements and treatments of the data are presented in Appendix E. 5.4 Mechanisms in the TEOS/oxygen plasma system FTIR absorption spectroscopy was used to identify the dominant decomposition products of the TEOS/O2/Ar chemistry. If the oxygen concentration is in excess, and the chemistry allowed to go to completion, the expected by products would be SiO2, CO2 and H2O as these are the most thermodynamically stable compounds formed from the available elements. As the chemistry is kinetically limited, a range of intermediates in the oxidation process are formed. A typical spectrum recorded during particle synthesis is shown in Fig. 5.3 and the various molecular vibrations listed in Table D.1 in Appendix F.

500 1000 1500 2000 2500 3000 3500 4000

0.00

0.01

0.02

0.03

0.04

0.05

0.06

H2O

CH4

TEOS

CO2CO

CHOOH

H2O

CH4

CHOOH

CO2

C2H4

C2H4SiOx

IR a

bsor

ptio

n (a

rb. u

nits

)

Wavenumber (cm-1)

TEOS

Figure 5.3: FTIR absorption spectrum recorded during SiO2 particle synthesis using 30 % TEOS, 50 % O2, 30 % Ar at 1 Torr pressure and 300 W discharge.

TEOS decomposes via electron impact dissociation (Eq. (5.4)) and reaction with atomic oxygen radicals (Eq. (5.5)). Evaluating which of these mechanisms dominates the TEOS decomposition is complicated by the uncertainty in


97

determining the concentration of the other reactants, namely the plasma density and atomic oxygen concentration when TEOS is injected.

eCH)HOC(SiOCH)HOC(Sie 33522

k

452

TEOSe

++→+ (5.4)

OH)HOC(OSiHCO)HOC(Si 35242

k

452

TEOSO

+→+ (5.5)

Reaction between TEOS and atomic oxygen exhibits a rather weak temperature dependence and within the temperature range 0 to 200°C, TEOS

Ok varies between 2.87·10-13 and 2.16·10-12 cm3/s. The corresponding electron impact induced decomposition of TEOS is a thousand times higher with an associated rate constant of 10-8 to 10-9 cm3/s. However, Langmuir probe measurements for pure argon and argon/oxygen gas mixtures give ne to be approximately 1015 m-3 (assuming Te = 3 eV) whereas actinometry (see Appendix E, chapter 6) give nO to be approximately 1020 m-3 in an O2/Ar plasma. The overall rate for these two processes ( TEOS

eR , TEOSOR ) is the product of the rate constant and concentration. The

initial TEOS decomposition rate resulting from these two processes is described by Eq. (5.6), for electron impact dissociation and Eq (5.7) for reaction with atomic oxygen.

TEOSeTEOSe

TEOSe nnkR = (5.6)

TEOSO

TEOSO

TEOSO nnkR = (5.7)

The decomposition rate for electron impact (Eq. (5.6)) and reaction with

atomic oxygen (Eq. (5.7)) under the typical conditions considered here both rates are estimated to be 102 mol s-1. These calculations suggest that both electron impact decomposition and reaction with atomic oxygen are equally important in the TEOS decomposition process. However, in these calculations ne and nO are measured from pure Ar and Ar/O2 plasmas and the addition of TEOS to the plasma will affect both ne and nO. However, in the actinometric study, nO is the balance of atomic oxygen production (via electron impact on O2) and loss. The addition of TEOS to this system will provide an additional loss mechanism and consequently shift the production/loss equilibrium towards loss and nO will be lower than that observed in the absence of TEOS. Based solely on the electron impact cross section data, the dominant initiation reaction for TEOS is likely to be via electron impact resulting in the formation of both radical or ion fragments of similar composition. Once activated, the TEOS fragments will go onto react with O2 (as it is the most abundant molecule) and other unreacted TEOS molecules. The following reaction mechanisms and rate constants presented in Table 5.3 are either taken directly or estimated by comparing work published by Holtgrave et al. [13], Morgan et al.


98

[31], and Ho et al. [15]. The mechanisms and rate constants for the hydrocarbon chemistry was taken from a number of literature sources [16-30].

Table 5.3: Reaction of TEOS in a TEOS/O2 PECVD process. (r/i) indicates the species can be either a radical or ion, x represents the number of ethoxy groups (0-4) and n the number of electrons for ionizing collisions.

Initiation via electron impact TEOS (k ≈ 10-8 10-9 cm3/s) Ref. Si(OC2H5)x + e → Si(OC2H5) x-2OCH3

(r/i) + (r/i)CH2 + (r/i)OC2H5 + (n)e [13] Si(OC2H5)x + e → Si(OC2H5) x-1OCH2

(r/i) + (r/i)CH3 + (n)e [31] Si(OC2H5) x + e → Si(OC2H5) x-1O(r/i) + (r/i)C2H5 + (n)e [31]

Initiation via oxygen attack (k ≈ 10-12 cm3/s) Si(OC2H5) x + ·O → Si(OC2H5)x-1OC2H4

· + ·OH [15] Si(OC2H5)x + O3 → Si(OC2H5) x-1OC2H4

+ OH + O2 [15] Si(OC2H5) x + ·OH → Si(OC2H5) x-1OC2H4

·+ H2O [15]

Oxidation reactions (k ≈ 5·10-11 cm3/s) Si(OC2H5)xOCH2

· + O2 → Si(OC2H5)xO(r/i) + CHOOH Est. Si(OC2H5)xOC2H4

· + ·O → Si(OC2H5)xO(r/i) + CH3HCO [15] Si(OC2H5)xOC2H4

(r/i) + O3 → Si(OC2H5)xO(r/i) + CH3HCO + O2 [15] Si(OC2H5)xOC2H4

(r/i) + ·OH → Si(OC2H5)xOH + CH3HCO [15]

Proton transfer (k ≈ 10-11 cm3/s) Si(OC2H5)xO(r/i) +Si(OC2H5)x → Si(OC2H5)xOH + Si(OC2H5)x [15] Si(OC2H5)xOCH2

(r/i) +Si(OC2H5)x → Si(OC2H5)xOCH3 + Si(OC2H5)x Est.

Starting with an O2/TEOS ratio of 50/20 the TEOS initiation is predominantly via electron impact to form a variety of TEOS ion and radical fragments. These fragments then go on to react with O2 and other oxygen species or again with unreacted TEOS (proton transfer reactions). Side reactions between C2H4OSi(OC2H5)3 or OSi(OC2H5)3 radicals with TEOS can also occur to form dimer and trimer species. These species are of similar chemical and structural composition to TEOS and are therefore likely to participate in the same electron impact and atomic oxygen reactions as TEOS. If the clusters become large enough they can pick up a net negative charge in the electronegative plasma and become trapped (see chapter 4). This mechanism is believed to be responsible for the formation of the particles observed in the plasma glow. TEOS can also react with Si(OC2H5)OH to form a dimer, however, the associated rate constant and lack of ethanol observed in the absorption FTIR suggests that this route does not significantly contribute to the production of clusters. TEOS oxidation reactions with O, O2 and OH produce CHOOH and CH3HCO (acetaldehyde). CH3HCO can


99

also go on to react further with O to form CHOOH and CH3 as depicted in Eq. (5.8). CH3HCO + ·OH → CH3

· + CHOOH (5.8)

Evidence for CH3HCO in the infrared is difficult to identify due to extensive overlap with peaks from CHOOH (around 1600 - 1900 cm-1), with CH4 and TEOS (2850 to 3150 cm-1) and with Si-O bulk and surface vibrations (1000 - 1300 cm-1). CH3HCO is identified by a strong CO stretch at 1743 cm-1 and the absence of this peak is believed to be due to the high reactivity of this compound with atomic oxygen (at 300 K k ≈ 8.5·10-12 cm3/s) and therefore short lifetime under deposition conditions.

The CH3 radical is also a product of the electron impact dissociation of TEOS and other TEOS fragments. It is central to the subsequent hydrocarbon chemistry along with CH2 formed from electron impact on TEOS. CH2 can go onto react with H to form CH. Unsaturated hydrocarbons such as C2H4 and C2H2 are also by-products of the TEOS chemistry, despite the relatively high oxygen concentration (50 vol%). The formation of C2H2 is a direct product of the CH2 and C2H3 radicals, as depicted by the mechanism presented in Fig. 5.4.

CH2 + CH2 → C2H2 + H + Hk = 6.6·10-11

CH + CH3 → C2H3 + H

CH3 + OH → CH2(s) + H2Ok = 1.6·10-11

CH3 + OH → CH2 + H2Ok = 2.4·10-11

C2H3 + OH → C2H2 + H2Ok = 3.3·10-11

CH3 from CH3CHO + OH

CH2 formation

C2H3 + H → C2H2 + H2

k = 1.5·10-10

C2H3 + CH → C2H2 + CH2

k = 8.3·10-11

C2H3 + CH3 → C2H2 + CH4

k = 3.3·10-11

C2H3 + O → C2H2 + CO

CH2 + H → CH + H2

k = 1.6·10-10

CH2(s) + H → CH + H2

CH formationk = 5.0·10-10

k = 5.0·10-11

k = 1.7·10-10

Figure 5.4: Mechanism for formation of acetylene (rate coefficients are in cm3/s).


100

5.5 Experimental results Mechanisms for the formation of compounds observed in the infrared have been presented. These mechanisms can also be corroborated with trends in the concentration of by-products as a function of the process parameters.

Figure 5.5 shows the concentrations of several decomposition products as a function of the TEOS inlet concentration. At low TEOS concentrations, TEOS fragments formed from initial electron impact processes and reactions with atomic oxygen produced from electron impact dissociation of O2 initiate decomposition of the TEOS. When O2 is in excess, oxygen is available to react with the hydrocarbons formed from the electron impact reactions on TEOS to form secondary by-products such as CO2 and H2O. As the TEOS concentration increases, more oxygen is consumed with TEOS initiation and less is available to form secondary reaction by-products. Consequently the CO2 concentration decreases as the TEOS concentration increases. Formic acid, however, is formed early on during the TEOS decomposition cycle (from reactions with primary TEOS fragments and CH3CHO), along with CH3. As more TEOS decomposes, the CHOOH concentration increases along with products from the hydrocarbon chemistry resulting from the additional CH3 production, namely CH4 and C2H2.

5 10 15 20 25 30

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0.035

CHOOH CH

4

C2H

2

CO2

IR p

eak

heig

ht (a

rb. u

nits)

TEOS concentration (%)

Figure 5.5: Infrared peak height for CHOOH (1105 cm-1), CH4 (3017 cm-1), C2H2 (730 cm-1) and CO2 (668 cm-1) as function of TEOS concentration. The other plasma conditions are discharge power = 300 W, O2 concentration = 50 %, total pressure = 1 Torr and total flow rate = 100 sccm. Similar trends are seen if the TEOS concentration is fixed and the oxygen

varied as shown in Fig. 5.6. When no oxygen is added the chemistry is completely governed by electron impact dissociation of TEOS to predominantly produce


101

mostly CH3 that goes on to form CH4 and C2H2 similar to that observed in Ar/CH4 plasmas [32]. As the oxygen concentration increases atomic oxygen starts to initiate the decomposition of TEOS forming CHOOH and CH3 react with the initial TEOS fragments. At 30 vol% oxygen, excess oxygen begins to participate with the hydrocarbon chemistry to form CO2 and the hydrocarbons are consumed. Finally, the oxygen concentration is high enough that the initial stable products (CHOOH and CH4) begin to be consumed.

0 10 20 30 40 50 60 70 80

0.000

0.005

0.010

0.015

0.020

0.025

0.030

IR p

eak

heig

ht (a

rb. u

nits

)

Oxygen concentration (%)

CHOOH CH4

C2H2

CO2

Figure 5.6: Infrared peak height for CHOOH (1105 cm-1), CH4 (3017 cm-1), C2H2 (730 cm-1) and CO2 (668 cm-1) as function of oxygen concentration. The other plasma conditions are discharge power = 300 W, TEOS concentration = 20 %, total pressure = 1 Torr and total flow rate = 100 sccm.

The influence of discharge power was also investigated but was not seen to have a large influence on the TEOS/O2 chemistry. Increasing the discharge power affects the process in two ways. Firstly, it increases the plasma density, which drives the electron impact reactions and favours the production of by-products related to electron impact processes. Secondly, increasing the discharge power increases the potential across the electrode sheath above the substrate. The effect of this is to increase the energy picked up by positive ions as they cross the electrode sheath and collide with the substrate table. The increased ion energy available at the substrate surface favours heterogeneous film forming processes (or at high energy, sputtering at the substrate). Figure 5.7 shows the variation in CH4, CO2, C2H2 and CHOOH concentration as a function of discharge power. CHOOH is readily polymerised and the decrease in CHOOH with increasing discharge power is likely to be the result of increased ion energy available at the substrate surface to


102

drive the polymerisation reactions. CH4, CO2 and C2H2 are less polymerisable and the slight increase in plasma density does not significantly influence the production or consumption of these compounds. The weak dependency of these product concentrations on the discharge power also suggests that their production is limited by reactions between neutral species and less dependant on the electron impact processes.

100 200 300 400 500

0.005

0.010

0.015

0.020

0.025

0.030

0.035

IR p

eak

heig

ht (a

rb. u

nits

)

Discharge power (W)

CHOOH CH4

C2H2

CO2

Figure 5.7: Infrared peak height for CHOOH (1105 cm-1), CH4 (3017 cm-1), C2H2 (730 cm-1) and CO2 (668 cm-1) as function of discharge power. The other plasma conditions are, O2 concentration = 70 %, TEOS concentration = 20 %, total pressure = 1 Torr and total flow rate = 100 sccm.

5.6 Conclusions In the TEOS/O2 plasma process the two dominant initiation steps for TEOS decomposition are electron impact dissociation and reaction with atomic oxygen. Calculated rate constants for electron impact with O2 and TEOS suggest that the process is initiated by electron impact of TEOS to form Si(OC2H5)2OCH3, Si(OC2H5)3OCH2, Si(OC2H5)3O ions and neutral species. These fragments then go on to react with O, O2, O3 or OH to form CHOOH and CH3CHO. TEOS decomposition can also be initiated by attack of oxygen radicals and the relative magnitude of these two processes depends mainly on the relative concentrations of TEOS and oxygen, but also on the plasma conditions (ne and Te). Despite the apparent complex chemistry the presence of C2H2, H2O and CHOOH appear to be the most kinetically favourable products found by following reactions with the highest rate constants presented in the literature.

The electron impact data on TEOS suggests an almost 1:1 mixture of ions and neutrals are formed in the initiation processes. However, a majority of the TEOS and hydrocarbon chemistry reviewed here deals with reactions between neutral


103

species. The ability to account for many of the species observed in the infrared, by reviewing data for primarily neutral reaction suggests that the ions undergo similar reactions to the radicals reactions reviewed.

Acknowledgement I would like to thank Pauline Ho and Ellen Meeks for providing rate data for the neutral chemistry and Lowell Morgan for providing electron impact cross section data for TEOS. Special thanks goes to Ton van Mol for his assistance in finding and help in evaluating the data. References [1] Alcott G. R., Linden J. L., van de Sanden M.C.M., Mat. Res. Symp. Proc., 726, 297-

302, (2002). [2] Buchta C., Wagner H. G., Wittchow W., Zeitschrift fur Physikalische Chemie, 190,

167-181, (1995). [3] Janca J., Talsky A., Zvonicek V., Plasma Chem. Plasma Proc., 16, 187-194, (1996). [4] Fracassi F., d’Agostino R., Favia P., J. Electrochem. Soc., 139, 2636-2643, (1992) [5] Raupp G. B., Cale T.S., Peter H., Hey W., J. Vac. Sci. Technol. B., 10, 37-45, (1992). [6] Dobkin D.M., Mokhtari S., Schmidt M., Pant A., Robinson L., Sherman A., J.

Electrochem. Soc., 142, 2332-2340, (1995). [7] Kim M. T., Thin Solid Films, 360, 60-68, (2000). [8] Stout P. J., Kushner M. J., J. Vac. Sci. Technol. A., 11, 2562-2571, (1993). [9] Allendorf M. D., Melius C. F., J. Phys. Chem., 96, 428-437, (1992). [10] Allendorf M. D., Melius C. F., Ho P., Zachariah M.R., J. Phys. Chem., 99, 15285-

15293, (1995). [11] Meeks E., Larson R. S., Ho P., Apblett C., Han S. M., Edelberg E., Aydil E. S., J.

Vac. Sci. Technol. A, 16, 544-563, (1998). [12] Ho P., Melius C. F., J. Phys. Chem., 99, 2166-2176, (1995). [13] Holtgrave J., Riehl K., Abner D., Haaland D., Chem. Phys. Lett., 215, 548-553,

(1993). [14] Coltrin M. E., Ho P., Moffat H. K., Buss R. J., Thin Solid Films, 365, 251-263,

(2000). [15] Chemical Reactions in TEOS/Ozone Chemical Vapour Deposition, Ho P., Sandia

National Laboratories Report, SAND2000-0217, (2000). [16] Wilson C., Balint-Kurti G.C., J. Phys. Chem. A, 102, 1624-1631, (1998). [17] Bauerle S., Klatt M., Wagner H.Gg., Ber. Bunsenges Phys. Chem., 99, 870-879,

(1995). [18] Baulch D.L., Cobos C.J., Cox R.A., Esser C., Frank P., Just Th., Kerr J.A., Pilling

M.J., Troe J., Walker R.W., Warnatz J., J. Phys. Chem. Ref. Data, 21, 411-429, (1992).

[19] Devriendt K., Van Poppel M., Boullart W., Peeters J., J. Phys. Chem, 99, 16953-16959, (1995).

[20] Tsang W., Hampson R.F., J. Phys. Chem. Ref. Data, 15, 1087, (1986). [21] Mebel A.m., Diau E.W.G., Lin M.C., Morokuma K., J. Am. Chem. Soc., 118, 9759-

9771, (1996). [22] Fahr A., Laufer A.H., Tardy D.C., J. Phys. Chem A, 103, 8433-8439, (1999).


104

[23] Atkinson R., Baulch D.L., Cox R.A., Hampson R.F. Jr., Kerr J.A., Rossi M.J., Troe

J., J. Phys. Chem. Ref. Data, 26, 521-1011, (1997). [24] Marinov N.M., Int. J. Chem. Kinetics, 31, 183-220, (1999). [25] Prada L., Miller J.A., Combust. Sci. Technol., 132, 225-250, (1998). [26] Opansky B.J., Seakins P.W., Pedersen J.O.P., Leone S.R., J. Phys. Chem., 97, 8583-

8589, (1993). [27] Pedersen J.O.P., Opansky B.J., Leone S.R., J. Phys. Chem., 97, 6822-6829, (1993). [28] Knyazev V.D., Slagle I.R., J. Phys. Chem., 1995, 2247-2249, (1995). [29] Tsang W., J. Phys. Chem. Ref. Data, 17, 887, (1988). [30] Tsang W., J. Phys. Chem. Ref. Data, 20, 221-273, (1991). [31] Morgan W.L., Winstead C., McKoy V., J. Appl. Phys., 92, 1663-1667, (2002). [32] Masi M., Cavallotti C., Carra S., Chem. Eng. Sci., 53, 3875-3886, (1998).

Appendix D Table of infrared absorption frequencies

105

Appendix D: Table of infrared absorption frequencies

Table D.1: Products of the TEOS/O2/Ar plasma chemistry.

Molecule Vibration Frequency Vibration Frequency H2O bending 1400-1800 cm-1 stretching 3500-4000 cm-1 CHOOH C-O stretch 1105 cm-1 C=O stretch 1776 cm-1 CO2 bending 668 cm-1 stretching 2340 cm-1 CO stretching 2025-2265 cm-1 C2H4 bending 949 cm-1 C2H2 bending 730 cm-1 CH4 bending 1305 cm-1 stretching 3017 cm-1 SiO2 Bulk 1000-1250 cm-1 Surface 1100-1250 cm-1 TEOS Si-O sym str. 792 cm-1 Si-O asym str. 1087 cm-1 CH2 rocking 800 cm-1 CH3 rocking 964 cm-1 C-O sym str. 1087 cm-1 C-O asym str. 1116 cm-1 CH3 scissor 1170 cm-1 CH2 twisting 1302 cm-1 CH2 wagging 1394 cm-1 C-H asym str. 2895 cm-1 (CH2) C-H sym str. 2937 cm-1 (CH3) C-H asym str. 2980 cm-1 (CH3)

Chapter 6 Deposition of nanocomposite layers for ultra low dielectric applications

107

Chapter 6: Deposition of nanocomposite layers for ultra low dielectric applications

Abstract The introduction of controlled porosity has been shown to dramatically reduce the dielectric constant of a material and porous thin films are currently of great interest in the semiconductor industry. In this chapter results from chapters 4 and 5 are used to design a new reactor configuration for the controlled synthesis and incorporation of nanoparticles into silicone polymer layers. Structure and composition of the nanocomposite layer produced are characterised using ESEM and Infrared spectroscopy. Thermal stability and electrical properties are determined to evaluate the thin films suitability in low-k dielectric applications. Dielectric constants as low as 1.82 ± 0.02 were achieved at 1 MHz.

6.1 Introduction As the dimensions of integrated circuit devices scale to smaller feature sizes, the resistance-capacitance (RC) delay of the metal interconnect will increasingly limit the performance of high speed logic chips [1, 2]. The integration of ultra low dielectric materials* (k < 2.5) can reduce this problem and there are a number of candidates with k in the range of 2-3 [3].

The dielectric constant or relative permittivity of a material is dependant on the material structure, polarization and polarizability [4]. There are several microscopic mechanisms of polarization in a dielectric material. Electronic dipole moments are induced by external electric fields or can be permanent due to the presence of polar bonding between the elements. Induced dipoles result from the electron cloud surrounding the atoms being ‘pulled’ out of place (electronic polarization), or by a distortion of the material structure (distortion polarization) by the external electric field. A permanent dipole moment can also exist if the elements of different electronegativity are bonded within the material and this is known as orientation polarization. In a d.c. electric field all of these phenomena contribute to the dielectric constant. When the applied field is alternating the frequency of the signal comes into play. Electronic polarization follows the electric field almost instantaneously as only the displacement of electrons is involved. The distortion polarization involves the motion of atoms and consequently cannot * Throughout this thesis the symbol k has been chosen to represent the dielectric constant εr, in line with nomenclature used in the semiconductor industry.


108

respond so rapidly. Both induced polarizations are subject to counter-active restoring forces, which give rise to a resonant frequency. In contrast, orientation polarization requires the motion of complete molecules and there is no counteractive restoring force and hence no characteristic resonance. The orientation polarization is, however, opposed by thermal disorder.

At low frequencies all three components contribute to the real part of the dielectric constant. The maximum frequency for orientation polarization is approximately 109 Hz. Above this frequency only the induced polarization affects the dielectric constant (distortion and electronic polarization). The threshold for distortion polarization is approximately 1013 Hz, beyond which only electronic polarization is defining the dielectric constant. The resonant frequency of the electronic polarization is typically beyond the frequency of visible light at approximately 1015 Hz. The overall affect of this is that the dielectric constant decreases as the frequency increases. The various contributions to the dielectric constant as a function of frequency are represented schematically in Fig. 6.1.

Electronicpolarization

Distortionpolarization

Orientationpolarization

109 1013 1015

Frequency (Hz)

Die

lect

ric c

onst

ant

Electronicpolarization

Distortionpolarization

Orientationpolarization

109 1013 1015

Frequency (Hz)

Die

lect

ric c

onst

ant

Figure 6.1: Dielectric constant as function of frequency.

The dielectric constant is therefore high if the molecules are polar and easily polarizable. Due to the frequency dependence, it has become standard to measure and report the dielectric constant in an electric field of 1 MHz. Although most semiconductor devices requiring ultra low-k materials run at much higher frequency (i.e. the Intel® Pentium® IV processor running at > 1 GHz), the negative dependence of k with respect to frequency ensures RC delays will not occur as the frequency is ramped up.

There are several ways in which the dielectric constant of a material can be modified to suit a certain application. All aspects of the polarizability can be affected by changing the elemental and structural composition of the film. A majority of low-k materials are based on SiO2 or a-C:H [5, 6, 7] or hybrid layer somewhere between the two. Carbon is less electronegative than oxygen, and hence the introduction of hydrocarbon moieties into SiO2 reduces the dielectric constant.


109

In a similar way, Si-OH groups create a large dipole, and although they indicate porosity, they increase orientational polarization and hence the dielectric constant increases. The introduction of voids into the material effectively reduces the material density and has a much stronger affect on the dielectric constant than trying to alter the polarizability. Porous SiO2 layers synthesized from various POSS precursors [8, 9, 10] and prepared using sol-gel deposition techniques have produced thin films with dielectric constants typically between 2-3. Sol-gel processes with siloxane precursor have also been extensively studied [11, 12, 13, 14], and reports of dielectric constants as low as 1.7 [15] have been reported using TEOS as a precursor. The ultra low-k properties of these materials are achieved by a combination of low film density and a lowering of the polarizability by incorporation of CH moieties. Post deposition curing of these films can reduce the dielectric [16] constant by driving polar OH groups out of the material, or increase it as film densification reduces porosity.

Gas phase technologies are an attractive alternative to wet chemical processes as they utilize existing vacuum reactors and toolsets in the semiconductor industries. Synthesis of porous SiO2 layers from siloxane precursors using expanding thermal plasma sources [17], PECVD [18, 19] and reactive evaporation of SiO [20] have all been investigated. More recently Grill et al. [21, 22, 23, 24] attempted to synthesize porous low-k films by incorporating volatile organic hydrocarbon fragments into siloxane layers. By removing the volatile hydrocarbon fragments in a 400°C post deposition annealing procedure, Grill et al. were able to produce porous layers exhibiting dielectric constants as low as 2.1. Unfortunately, gas phase technologies offer less control over the deposition chemistry and therefore less control over film structure and consequently polarizability.

While there are a number of potential ultra low-k dielectrics candidates [25], they all share some common properties. Firstly, all are porous; the dielectric constant of silica ksilica = 4, while air has kair ≈ 1 and there are currently no dense low-k dielectric material for semiconductor applications available with a dielectric constant below 2.5 [3]. The porosity of these materials already introduces technical challenges when implementing into copper based interconnect ICs. Porous ultra low-k materials are also mechanically weak compared to SiO2 and are susceptible to damage during subsequent processing. In addition to this, porous materials are sensitive to wet and dry cleaning chemicals and hold volatile compounds which become free to contaminate subsequent processing steps [26]. A number of author have investigated the mechanical properties and other issues concerning the implementation of these porous materials in semiconductor processes [27, 28, 29, 30]. An overview of the current “state of the art” for low-k and ultra low-k materials and processes is presented in Table 6.1.


110

Table 6.1: Overview of “current state of the art” for ultra low-k materials.

Reference Technique Precursors Porosity (%) Dielectric Liu [8] Sol-gel HSQ 2.7-3.3 Baklanov [9] Sol-gel HSQ/MSQ 2.5 ±0.1 Xu[10] Sol-gel MSQ 43-56 % 2.0-2.1 Chang [11] Sol-gel Not specified 2.65-3.36 Uchida [12] Sol-gel BTM/BTE 56 % 1.9-2.7 Jain [13] Sol-gel TEOS/Glymol 30-80 % Liu [15] Sol-gel TEOS 1.70-3.70 Lee[30] PECVD BTMSM 62 % 2.1 Murray [31] Sol-gel Not specified 36-55 % 1.7-2.3 Creatore [17] ETP HMDSO 2.9 Grill [16] PECVD TMS 3.0-4.4 Grill [21-24] PECVD TMCTS 0-29 % 2.05-2.8 Shaniryan [18] PECVD TMS 5.5-7.5 % Si [20] PVD SiO 1.89-2.11 Frohlich [32] Sol-gel ORMOCER® - 3.1

Sol-gel processes are a popular choice for fabricating low-k materials because they inherently produce porous materials and materials with void volume fraction as high as 95 % have been reported [14]. In addition to this, wet chemical processes offer greater control over the process chemistry and polarization effects can be minimised in the final film. Although highly porous materials have dielectric constants around 1, their susceptibility to copper diffusion makes them unsuitable for implementation into semiconductor processing. On the other hand, gas phase deposition techniques inherently deposit dense materials and can easily be implemented into existing semiconductor processes. The complimentary nature of these two techniques has been recognised by a number of researchers who have used PECVD [33] or PVD [34] technology to encapsulate wet chemically synthesized xerogels.

The necessity for separate reaction conditions for the synthesis and processing of dual phase hybrid materials has been the central theme to this work. Chapter 3 investigated the feasibility of depositing a dual phase polymer/SiO2 layer using an ICP/CCP plasma reactor. The failure to synthesize nanometre sized particles in chapter 3 lead to study of particle synthesis and trapping phenomena presented in chapter 4. This lead to the discovery that particles formed in low pressure capacitively coupled discharges are porous, similar to thin films deposited under such conditions. Consequently incorporating these particles into silicone layers to improve scratch resistance, as discussed in chapter 1, is somewhat pointless as the particles themselves are mechanically weak. TGA results however, showed these particles to be thermally stable (cf. Fig. 4.16) and thermally stable, nano-porous particles are interesting for low-k applications.


111

Building on the finding presented in chapters 4 and 5, a new dual capacitively coupled plasma reactor is designed and used to synthesize nano-porous particles and incorporate them into a simultaneously synthesized silicone layer. The nano-porosity of these particles is used to introduce controlled porosity into the silicone layers with the aim of producing ultra low-k layers. 6.2 Reactor design As in chapter 3, the concept of a dual plasma system for the deposition of a mixed phase material is deemed necessary, as the two material phases require separate process conditions. Here, oxygen is required for the synthesis of SiO2 particles, but would be detrimental to the polymer chemistry causing oxidation of the hydrocarbon content of the polymer. The first dual plasma concept presented in chapter 3 relied upon an ICP for particle synthesis. However, plasma stability issues and the absence of any confining electric fields needed to trap negative ions and nucleate particle growth, resulted in this concept being abandoned. Subsequently a modified dual plasma system has been conceived and built to overcome the limitations of the first.

The new reactor concept consists of two capacitively coupled plasma systems connected to two 13.56 MHz 1 kW power supplied through matching networks (Fig. 6.2). Both plasmas can be pulsed. TEOS and oxygen, for particle synthesis, are fed through the top inlet into Zone I where they are activated by the plasma formed between the r.f. electrode mesh and the top and bottom plates of the chamber. The self biasing of the r.f. electrode mesh acts to trap negative ions so that they may nucleate particle growth in this region. Both plasmas can be modulated so that particles trapped in Zone I can traverse the r.f. electrode mesh and flow into Zone II where they are combined with the polymer forming chemistry. A mesh with a hole diameter of 1 mm was chosen for the top electrode in order to insure the sheath thickness was larger than the hole diameter and that no hollow cathode discharges would ignite in the electrode and disrupt the trapping electric field. The precursors were evaporated and dosed using the gas delivery system described in chapter 3.

A computation fluid model (CFD) of the particle forming region was made to identify recirculation or stagnation regions caused by the geometry (Fig. 6.3). The mesh electrode and grounded plate separating the two plasma regions have a open area of 38 %. The effect of this is to produce a more laminar flow at the substrate (Electrode II) and consequently a more homogenous film deposition profile.


112

~

~

Zone I

Zone II OES

Adjustable FTIRwindow

TMSE

TEOS/O2

R.F.

R.F.

Electrode I(wire mesh)

Electrode II

Separatingmesh

~

~

Zone I

Zone II OES

Adjustable FTIRwindow

TMSE

TEOS/O2

R.F.

R.F.

Electrode I(wire mesh)

Electrode II

Separatingmesh

Figure 6.2: Dual CCP deposition system.

3.0·10-4

1.5·10-5

2.4·10-4

1.8·10-4

1.2·10-4

6.0·10-5

0.0

Electrode I

Electrode II

3.0·10-4

1.5·10-5

2.4·10-4

1.8·10-4

1.2·10-4

6.0·10-5

0.0

3.0·10-4

1.5·10-5

2.4·10-4

1.8·10-4

1.2·10-4

6.0·10-5

0.0

Electrode I

Electrode II

Figure 6.3: CFD model of reactor flows in m/s.


113

6.2.1 Experiments Low-k films were synthesized using both the upstream (Zone I) and downstream (Zone II) plasmas. TEOS was dosed using a Bronkhorst HiTec CEM system and combined with O2 and Ar flows in the heated gas lines (80°C) and fed into the upstream plasma. TMSE was dosed using a bubbler system (cf. section 3.2) using Ar as a carrier gas and fed via heated gas lines (80°C) in the to downstream plasma. The injection of oxygen during the synthesis of the plasma polymer layer causes the layer to become more glass like and consequently the adhesion to decrease (see chapter 3). It is therefore necessary to ensure that all of the oxygen is consumed in the particle forming plasma where TEOS and oxygen are injected. Optical emission spectroscopy was done using an Ocean Optics USB-2000 spectrometer with a spectral range of between 280 to 800 nm and resolution of approximately 1 nm. Light from the reactor was fed into the OES spectrometer via an optical fibre. The concentration of O atoms in an Ar/O2 plasma was monitored under various conditions using optical emission spectroscopy (OES) using an approach reported by Han et al. [35] and Lieberman [36]. Details of this study can be found in Appendix E. This study indicated that 0.05 TEOS mole fraction injected into the 0.5 O2 mole fraction discharge (cf. Fig E.3 Appendix E).

The flow and conditions for the experiments presented in this chapter are listed in Table 6.2.

Table 6.2: Experimental conditions for particle synthesis of nanocomposite films.

Objective Top Plasma [W]

Bottom Plasma [W]

TEOS [sccm]

O2 [sccm]

TMSE [sccm]

Flow [sccm]

Pressure [Torr]

Oxygen depletion study

0-200 0-400 0-30 0-50 - 100 1

Low-k materials 200-400 100-400 10-35 50 10-20 200 1

The plasmas are modulated with on and off times of 475 ms and 25 ms respectively to prevent particle trapping in the upstream plasma and above the substrate. When operating at 200 W, a self-bias potential of 150 V is observed to develop on the substrate table. Depositions are performed using a deposition time of 5 minutes resulting in film thicknesses ranging from 0.88 to 2.04 µm depending on the precursor flows and discharge power. Film thicknesses are determined using a Tencor P-10 surface profiler with an estimated accuracy of approximately ± 10 nm.

Boron doped silicon wafers with a resistivity of between 7 and 21 ohm cm were placed in the centre of the r.f. biased substrate table and used as substrates for


114

determining the dielectric constants. All substrates were used as delivered and no attempt was made to clean or remove the native oxide.

The dielectric constants are determined from measurements of the materials capacitance using two independent methods described in Appendix F. 6.3 Experimental results A variety of deposition conditions were used to deposit the low-k layers with deposition rates varying from 3.0 to 6.8 nm/s. Increasing the TEOS and oxygen flows to the top plasma resulted in hazy coatings exhibiting high surface roughness as shown in the ESEM image in Fig. 6.4. The haze and high surface roughness observed was attributed to the increased synthesis and incorporation of particles and particle clusters into the growing TMSE layers.

Figure 6.4: ESEM images of nanocomposite layer.

Some of the samples were cleaved and cross sectional ESEM images taken in an attempt to identify the individual particles inside the TMSE layer however, no particles could be seen. The absence of particles in the cross sectional ESEM images suggests that the particle are mechanically weak and the structure is not that of dense silicate. If the particles are dense and the intermolecular Si-O bonds within the particle are strong, then the particles should not break and would therefore be visible in the cross sectional SEM image. However, the absence of particles in the cross sectional ESEM image suggests the particles break apart internally instead of separating from the coating. This process is depicted schematically in Fig. 6.5.


115

a. SiOxCyHz Particles

Silion substrate

TMSEPolymer

b.

a. SiOxCyHz Particles

Silion substrate

TMSEPolymer

b.

Figure 6.5: Schematic of particles encapsulated in pp-TMSE layer, a. strong particles bonding to pp-TMSE severed curing cleavage, b. weak particles.

This is consistent with the particle structure discussed in chapter 4 where, similar to SiO2 films deposited in r.f. discharges, the low electric field surrounding particles in a plasma results in high porosity. 6.3.1 Dielectric constant The dielectric constant of a material is affected by the material composition and porosity as discussed in the introduction to this chapter. The influence of both these phenomena on the dielectric constant can be seen in Fig. 6.6.

100 150 200 250 300 350 4001.5

2.0

2.5

3.0

3.5

4.0

4.5

Nanocomposite: Method A Nanocomposite: Method B Pure TMSE layer: Method A

Die

lect

ric c

onst

ant

Bias power (W)

a.

0 5 10 15 20 25 30 35 401.5

2.0

2.5

3.0

3.5

4.0

4.5

Nanocomposite: Method A Nanocomposite: Method B

Die

lect

ric c

onst

ant

TEOS concentration (% total flow)

b.

100 150 200 250 300 350 4001.5

2.0

2.5

3.0

3.5

4.0

4.5

Nanocomposite: Method A Nanocomposite: Method B Pure TMSE layer: Method A

Die

lect

ric c

onst

ant

Bias power (W)

a.

0 5 10 15 20 25 30 35 401.5

2.0

2.5

3.0

3.5

4.0

4.5

Nanocomposite: Method A Nanocomposite: Method B

Die

lect

ric c

onst

ant

TEOS concentration (% total flow)

b.

Figure 6.6: Dielectric constant as a function of a. substrate bias and b. oxygen concentration added to particle synthesis plasma.


116

Increasing the r.f. bias potential (Fig. 6.6a) is known to enhance film density by increasing the energy of ions bombarding the film surface. The subsequent drop in porosity caused the dielectric constant to increase. Films deposited under identical conditions but without the TEOS/O2 particle plasma (Fig 6.6a) exhibit much higher k values indicating that the lower k values observed for the nanocomposite films is indeed due to the incorporation of porous particles.

The effect of film composition on the dielectric constant is illustrated in Fig. 6.6b. At low oxygen concentrations silicone-like particles are produced retaining some of the organic functionality of the original TEOS precursor. As the oxygen concentration in the particle forming plasma increases, so does the inorganic nature of the particles. Inorganic materials are characterised by polar bonding, this increases the concentration of permanent dipoles present in the film.

Identical trends in k as a function of bias power and TEOS concentration are observed independent of the technique used to measure k. A critical assessment of the two techniques used showed Method A to yield the most reliable results (cf. Appendix F). 6.4 Conclusions The successful synthesis of nanometre sized silica/silicone particles from a TEOS/O2 plasma, and subsequent incorporation into plasma polymerised TMSE layers was demonstrated. Thin films exhibiting ultra low dielectric constants (values as low as 1.82 ± 0.02 are obtained as measured using Method A) were deposited. The dielectric constant increased with increasing oxygen added to the particle plasma, and increasing r.f. bias applied to the substrate table. This is believed to be due to an increase in the SiOH content of the particles, and the subsequent increase in the dielectric constant is therefore attributed to increased polarity of the layer. Increasing the power applied to the substrate table increases the bias potential above the substrate and the resulting increase in ion energy causes film densification. The consequent reduction in film porosity leads to an increase in the dielectric constant at high bias powers. Acknowledgement I would like to thank Adriana Creatore, Herman de Jong and Bertus Husken for their help in the dielectric constant measurement. References [1] Jeng S.P., Havemann R.H., Chang M.C., Mater. Res. Soc. Symp. Proc., 35, 337, (1994). [2] Ryan J.G., Geffken R.M., Poulin N.R., Paraszczak J.R., IBM J. Res., Dev., 39, 371,

(1995). [3] http://public.itrs.net [4] Maex K., Baklanov M.R., Shamiryan D., Iacopi F., Brongersma S.H., Yanovitskaya Z.S.

J. Appl. Phys., 93, 8793-8841, (2003). [5] Grill A., Meyerson B.S., Patel V.V., IBM J. Res. Develp., 34, 849-857, (1990). [6] Grill A., Thin Solid Films., 398-339, 527-532, (2001). [7] Grill A., Diamond and Relat. Mater., 10, 234-239, (2001).


117

[8] Liu P.T., Chang T.C., Sze S.M., Pan F.M., Mei Y.J., Wu W.F., Tsai M.S., Dai B.T.,

Chang C.Y., Shih F.Y., Huang H.D., Thin Solid Films, 332, 345-350, (1998). [9] Baklanov M.R., Mogilnikov K.P., Microelectron. Eng., 64, 335-349, (2002). [10] Xu J., Moxom J., Yang S., Suzuki R., Ohdaira T., Appl. Surf. Sci., 194, 189-194, (2002). [11] Chang T.C., Tsai T.M., Liu P.T., Mor Y.S., Chen C.W., Mei Y.J., Sheu J.T., Tseng T.Y.,

Thin Solid Films, 420-421, 403-407, (2002). [12] Uchida Y., Katoh T., Oikawa M., Mater. Sci. Semicond. Process., 5, 259-264, (2003). [13] Jain A., Rogojevic S., Ponoth S., Agarwal N., Matthew I., Gill W.N., Persans P.,

Tomozawa M., Plawsky J.L., Simonyi E., Thin Solid Films, 398-399, 513-522, (2001). [14] Kim R.-H., Park H.-H., Hyun S.-H., Yeom G.-Y., Thin Solid Films, 332, 444-448,

(1998). [15] Liu P.-T., Chang T.C., Hsu K.C., Tseng T.Y., Chen L.M., Wang C.J., Sze S.M., Thin

Solid Films, 414, 1-6, (2002). [16] Grill A., Patel V., J. Appl. Phys., 85, 3314-3318, (1999). [17] Creatore M., Barrell Y., Kessels W.M.M., van de Sanden M.C.M., Mater. Res. Soc.

Symp. Proc., 766, E6.9.1, (2003). [18] Shaniryan D., Weidner K., Gray W.D., Baklanov M.R., Vanhaelemeersch S., Maex K.,

Microelectron. Eng., 64, 361-366, (2002). [19] Borvon G., Goullet A., Mellhaoui X., Charrouf N., Granier A., Mater. Sci. Semicond.

Process., 5, 279-284, (2003). [20] Si J.-J., Show Y., Banerjee S., Ono H., Uchida K., Nozaki S., Morisaki H.,

Microelectron. Eng., 60, 313-321, (2002). [21] Grill A., Patel V., J. Appl. Phys. 79, 803-805, (2001). [22] Grill A., J. Appl. Phys., 93, 1785-1790, (2003) [23] Grill A., Patel V., Rodbell K.P., Huang E., Baklanov M.R., Mogilnikov K.P., Toney M.,

Kim H.-C., J. Appl. Phys., 94, 3427-3435, (2003). [24] Grill A., Neumayer D.A., J. Appl. Phys., 94, 6697-6707, (2003). [25] Jin C., Luttmer J.D., Smith D.M., Ramo T.A., MRS Bull., 22, 39, (1997). [26] Mogsi K., Jacobs T., Brennan K., Rasco M., Wolf J., Augur R., Microelectron. Eng., 64,

11-24, (2002). [27] Volinsky A.A., Vella J.B., Gerberich W.W., Thin Solid Films, 429, 201-210, (2003). [28] Travaly Y., Eyckens B., Carbonel L., Rothschild A., Le Q.T., Brongersma S.H., Ciofi I.,

Struyf H., Furukawa Y., Stucchi M., Schaekers M., Bender H., Rosseel E., Vanhaelemeersch S., Maex K., Gaillard F., Van Autryve L., Rabinzohn P., Microelectron. Eng., 64, 367-374, (2002).

[29] Lanckmans F., Maex K., Microelectron. Eng., 60, 125-132, (2002). [30] Lee H.J., Oh K.S., Choi C.K., Surf. Coat. Technol., 171, 296-301, (2003). [31] Murray C., Flannery C. Streiter I., Schulz S.E., Baklanov M.R., Mogilnikov K.P.,

Himeinschi C., Friedrich M., Zahn D.R.T., Gessner T., Microelectron. Eng., 60, 133-141, (2002).

[32] Frohlich L., Houbertz R., Jacob S., Popall M., Mueller-Fiedler R., Graf J., Munk M., von Zychlinski H., Mat. Res. Soc. Symp. Proc., 726, 349-354, (2002).

[33] Schulz S.E., Koerner H., Murray C., Streiter I., Gessner T., Microelectron. Eng., 55, 45-52, (2001).

[34] Iacopi F., Tokei Zs., Stucchi M., Brongersma S.H., Vanhaeren D., Maex K., Microelectron. Eng. 65, 123-131, (2003).

[35] Han S.M., Aydil E.S., Thin Solid Films, 290-291, 427-434, (1996). [36] Principles of plasma discharges and materials processing, Lieberman M.A., Lichenberg


Appendix E Optical emission spectroscopy and actinometry

118

Appendix E: Optical emission spectroscopy and actinometry

Optical emission spectroscopy was done using an Ocean Optics USB-2000 spectrometer with a spectral range of between 280 to 800 nm and resolution of approximately 1 nm. Light from the reactor was fed into the OES spectrometer via an optical fibre. The concentration of O atoms in an Ar/O2 plasma was monitored under various conditions using optical emission spectroscopy (OES) using an approach reported by Han et al. [35] and Lieberman [36]. Argon is added as an inert tracer gas to the oxygen plasma and the transition λ = 750.4 nm (2p1 → 1p0) was monitored. Changes in the atomic oxygen concentration were followed using the λ = 777.4 nm transition arising from the decay, O(3p5P) → O(3s5S) + hv(777.4nm). The O(3p5P) state is populated via both electron impact excitation of O and as a result of electron impact dissociation of O2,

ePpPpeek

+→+ )3(O)2(O 534O

(E.1)

ePpedek

++→+ )3(OOO 52

O

(E.2) where O

ek and Odek represent the production rate due to electron impact excitation

and electron impact dissociation excitation respectively. The Ar 750.4 nm transition has a similar excitation cross-section and threshold dependency to electron temperature as the O emissions at λ=844.6 nm (3p3P → 3s3S transition) and λ = 777.4 nm (3p5P → 3s5S transition). The relative error associated with these measurements was estimated to be 55 % and 20 % for the 777.4 nm and 844.6 nm lines respectively when compared to VUV measurements [1]. Unfortunately due to the restricted wavelength range of the emission spectrometer (300-800 nm) the λ = 844.6 nm transition was not available. From work published by Pagnon et al. [1]

Oek / O

dek is estimated to be approximately 10-3, i.e. the dissociative excitation process contributes approximately 0.1 % to the total O(3p5P) population and is therefore ignored for simplicity. The atomic oxygen concentration can be related to the argon concentration by,

Ar

OAr

O

Ar

O

nn

kk

II

e

e= (E.3)

where O

ek and Arek are the excitation rate coefficients. The ratio O

ek / Arek ≈ 0.1 [1] is

insensitive to changes in plasma operating conditions, due to the similarity in the respective process dependencies on the electron energy distribution. The O(3p5P) excited state can also decay non radiatively by quenching collisions with other gas atoms and reactor walls. These processes have not been considered as their


119

magnitude depends on the reactor system and therefore would require a complimentary calibration technique such as laser induced fluorescence (LIF) [2]. As the quenching losses have not been considered, the subsequent results are an underestimate of the O density and should therefore only be interpreted as order of magnitude estimations. Results As might be expected the atomic oxygen concentration in the plasma increases as the discharge power is increased (Fig. E.1). The close agreement in atomic oxygen concentrations for measurements made at 1 Torr and 0.5 Torr indicates that there is no significant change in the production (i.e. no change in Te or ne) or quenching of the atomic oxygen excited state within this pressure region. The higher scatter in the data measured at 1 Torr is attributed to hollow cathode formation at higher pressures (caused by a slight drop in λD with increasing p).

0 100 200 300 400 5005

10

15

20

25

1.0 Torr 0.5 Torr

n O x

1020

(m-3

)

Discharge power (W)

0 100 200 300 400 5005

10

15

20

25

1.0 Torr 0.5 Torr

n O x

1020

(m-3

)

Discharge power (W)

Figure E.1: Variations in atomic oxygen density as a function of the discharge power, O2 flow = 10 sccm, Ar flow = 90 sccm.

Under certain operating conditions, hollow cathode discharges appeared in the grounded containment cage surrounding the plasma and were observed to ‘flicker’. In areas where hollow cathode discharges were seen, higher levels of O emissions were also observed. The increased O emission at hollow cathode sites is most likely a result of higher electron density in these regions causing an increased concentration of excited O atoms. The flickering of these hollow cathode discharges appeared to be most severe at 1 Torr and 0.1 mole fraction O2 as can be seen in the increased scatter of the data in Fig. E.2. Hollow cathodes appear when the dimensions of hollow cathode hole are of the order of a few Debye lengths (λD) and electron confinement occurs. For ne = 1015 m3 and Te = 3 eV the Debye length is approximately 0.4 mm whereas the grid holes are 2 mm in diameter. For high


120

oxygen concentrations the plasma is electronegative, where electron attachment and negative ion formation cause ne to drop and, in order to maintain the ionization rate, Te to increase. Decreasing the oxygen concentration causes the plasma to become more electropositive and ne increases and Te drops. The Debye length is proportional to ne

1/2 and inversely proportional to Te1/2 and thus λD also decreases

until a hollow cathode emission occurs.

0.0 0.2 0.4 0.6 0.8 1.0

2

4

6

8

10

12

14

16

0.12 Torr 0.50 Torr 1.00 Torrn O

x10

20 (m

3 )

Ar fraction (0-1)

Hollow cathode region

0.0 0.2 0.4 0.6 0.8 1.0

2

4

6

8

10

12

14

16

0.12 Torr 0.50 Torr 1.00 Torrn O

x10

20 (m

3 )

Ar fraction (0-1)

Hollow cathode region

Figure E.2: Oxygen atom density (nO) as function Argon volume fraction at different total pressures. Scatter in data at 0.9 fraction Ar is due to hollow cathode discharge flickering.

0.00 0.05 0.10 0.15 0.200.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

n O x

1020

(m-3

)

TEOS Fraction (0-1)0.00 0.05 0.10 0.15 0.20

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

n O x

1020

(m-3

)

TEOS Fraction (0-1)

Figure E.3: Atomic oxygen concentration in bottom plasma as a function of TEOS fraction in top plasma.


121

The dramatic drop in the atomic oxygen concentration as TEOS is injected into the top plasma is evidence that all of the oxygen is consumed in the TEOS/O2 plasma (Fig. E.3). The consumption of atomic oxygen is already complete at relatively low discharge powers (Fig. E.4).

0 50 100 150 200 250 300

0.0

0.5

1.0

1.5

2.0

300 Watts + TEOS

400 Watts

300 Watts

200 Watts

n O x

1020

(m-3)

Top Plasma Power (W)

100 Watts

0 50 100 150 200 250 300

0.0

0.5

1.0

1.5

2.0

300 Watts + TEOS

400 Watts

300 Watts

200 Watts

n O x

1020

(m-3)

Top Plasma Power (W)

100 Watts

Figure E.4: Atomic oxygen concentration in bottom plasma as a function of discharge intensity of top plasma.

References [1] Pagnon D., Amorim J., Nahorny J., Touzeau M., Vialle M., J. Phys. D: Appl. Phys., 28,

1856-1868, (1995). [2] George A., Engemann J., Brockhaus A., J. Phys. D: Appl. Phys., 35, 875-881, (2002).

Appendix F Dielectric constant determination

122

Appendix F: Dielectric constant determination

Method A: 1 cm2 samples are prepared from the Si substrate and sputtered with a 60 nm gold front contact layer. Silver paint is applied to the backside of the sample to ensure a good electrical contact to the silicon wafer. The sample is then mounted between two contacts and connected to the Hewlett Packard 4284A Precision LCR meter (20 Hz to 1 MHz) (Fig. F.1) and the capacitance measured at 1 kHz and 1 MHz oscillating potentials of 50 and 100 mV.

HP 4284A LCR meter

Au top contact

SiO2 nanocomposite

Si wafer

Ag paint

HP 4284A LCR meter

Au top contact

SiO2 nanocomposite

Si wafer

Ag paint

Figure F.1: Circuit setup for determining the dielectric constant using Method A.

Method B: A contact pattern of aluminium is sputtered onto the top of the sample as shown in Fig. F.2a and F.2b. The sample is then placed into an oscillating circuit consisting of two comparators and a flip-flop reset switch (Fig. F.2c). The negative and positive inputs of comparators one and two (Cp1

- and Cp2+) are set at 1/3 and

2/3 the line voltage respectively, by R2. The other two comparator inputs (Cp1+ and

Cp2-) are charge via resistors R4 and R5 until the comparator outputs switch and the

flip-flop and transistor grounds the comparator inputs (Cp1+ and Cp2

-) and resets the circuit.

Low-k layerAl sputteredelectrode

Si Wafer

a.

C Cp1

Cp2+

-

+

-

R1

R2

R3

R4

R5

n

Flip

-Flo

p R

eset

V

C

b.

c.

Low-k layerAl sputteredelectrode

Si Wafer

a.

C Cp1

Cp2+

-

+

-

R1

R2

R3

R4

R5

n

Flip

-Flo

p R

eset

V

C

b.

c.

Figure F.2: Method B for determining the dielectric constants. a. plan view of sample b. cross sectional sample view showing Al layer, dielectric material and Si wafer and c. resonance circuit.


123

For the idea case, the overall capacitance (CT) relates to the circuit frequency (f),

( ) fRRCT

54

44.1+

= (F.1)

The sample preparation technique used in Method B actually generates two

series capacitances within the sample, C1 and C2 [F]. The relative contribution to the overall capacitance (CT) scales with the respective areas A1 and A2 [m2] as described in Eq. (F.2).

1

21

0

1

20

10

1

21

111111−

−

−

+=

+=

+=

AAddA

dACC

C r

rr

Tεε

εεεε (F.2)

where d is the thickness of the dielectric layer [m] with permittivity εr, and ε0 is the permittivity of free space [F.m-1].

The oscillating frequency is determined by R4, R5 and C. The presence of internal and parasitic capacitance are accounted for by calibrating the circuit resonant frequencies with a series of fixed capacitors (1 % tolerance).

100 1000 10000

10

100

1000

f [kH

z]

C [F]

Standard fit Calibration

Figure F.3: Calibration of circuit used in Method B for determining the capacitance of nanocomposite samples.

For both measurement techniques, the dielectric constant k is determined from

the capacitance by,


124

ACdk

0ε= (F.3)

where C is the sample capacitance in [F].

Errors in the dielectric constant measurements are predominantly caused by the uncertainty in the film thickness and uniformity and are estimated to be approximately 2-6 %. Method B is likely to be less reliable than A for several reasons namely,

1. The circuitry is not mounted in any form of enclosure and hence not protected from surrounding environmental influences (electromagnetic interference, moisture and mechanical vibrations).

2. No subsidiary circuitry is included to compensate temperature variations or fluctuations in the power supply.

3. Each sample is effectively measured at a different voltage frequency as the circuit resonance depends on the capacitance of the sample.

Chapter 7 General Conclusions

125

Chapter 7: General Conclusions The primary aim of this work is to develop a process for the synthesis of hybrid materials using solely gas phase synthetic methods. Such a broad project objective required putting into context and the initially objectives became to focus on the mechanical properties of the layers for barrier and scratch resistant applications. A dual plasma reactor system utilizing both inductive and capacitive plasma configurations to generate a high plasma density for synthesizing inorganic particles and low density for polymeric layers was designed. Limitations in the initial reactor concept prompted investigations into the plasma chemistry and physics behind dusty plasmas. A deeper understanding of these processes lead to a change in the focus of the work and a new reactor configuration was designed for the deposition of low-k layers. The main conclusions resulting from the various investigation contained in this work are summarized below. 7.1 Scratch resistant layers PECVD techniques offer several advantages over wet chemical synthetic methods for producing single phase hybrid films of the general form SiOxCyHz. However, adhesion to polymeric substrates degrades with the increasing inorganic nature of the layer and almost complete de-lamination occurs as the material structure approaches that of stoichiometric thermal silica. Wet chemical studies show that mechanical properties can be achieved while maintaining good adhesion particles if inorganic particles are embedded into the soft organic layers. The conclusion at this stage is therefore to focus on the synthesis of nanocomposite materials starting with a study of dusty plasmas. 7.2 Particle monitoring Particle synthesis was attempted in both inductively coupled and capacitively coupled plasma discharges. The failure to synthesize particles in the inductive plasma was attributed to the low residence time in the ICP tube caused by the absence of confining electric fields, and plasma instability attributed to negative ions. The successful synthesis of particle in the capacitively coupled plasma was attributed to the confining effects of the electrode sheath. It was found that the huge potential present in the ion sheath above a driven electrode represents a theoretically insurmountable barrier for trapped particles. OML theory predicts that the electrostatic force dominates for small particles, pushing them to the lowest potential (centre of the plasma discharge). As the particle mass scales with a3, the gravitational force dominates as the particle radius increases. Despite this, particles several orders of magnitude smaller than those predicted by OML theory are capable of breaching the sheath to arrive at the electrode surface. Particles are able to overcome this electric field when the electron flux to the particle surface is reduced by shadowing from the other particles and the electrode itself.

Chapter 7 General Conclusions

126

7.3 Plasma chemistry Despite the apparent complex chemistry the decomposition of TEOS and formation of the unexpected by-products acetylene, formic acid and water can be explained using relatively simple reaction mechanisms (typically involving less then 4 steps) with relatively high rate constants (> 10-16 cm3/s) involving purely neutral species. Although electron impact dissociation data suggests that the process is initiated by electron impact on TEOS, reactions between neutral species are sufficient to elucidate much of the deposition process chemistry. 7.4 Potential application as low dielectric layers The deposition of ultra low-k layers using the new dual capacitively coupled plasma reactor was successfully demonstrated. Thin films exhibiting ultra low dielectric constants (values as low as 1.82 ± 0.02) were deposited. The dielectric constant was found to be dependent on the oxygen concentration in the particle synthesis plasma and bias power applied to the substrate table. This is effects were attributed to an increase in polarity of bonds within the film (caused by the incorporation of SiOH groups) and film densification due to ion bombardment respectively.

Symbols and Abbreviations

127

Symbols and Abbreviations Chemical abbreviations ORMOCER organically modified ceramic PP plasma polymer R arbitrary hydrocarbon side group Al(iPA)3 aluminium tri isopropoxide Glymol (3-glycidoxypropyl) trimethoxysilane HPC hydroxypropyl cellulose MPS 3-methacryloxypropyltrimethoxysilane MMA methyl methacrylate TEOS tetraethoxysilane Ti(iPA)4 titanium tri isopropoxide TMOS tetramethyloxysilane TMTSO 1,3,5,7-tetramethylcyclotetrasiloxane TVTMTSO 2,4,6,8-tetravinyl 2,4,6,8-tetramethylcyclotetrasiloxane TMMOS tetramethylmetoxysilane HMDS hexamethyldisilazane TMSE 1,2- bis(trimethyl)siloxyethane TMDSO tetramethyldisiloxane VpMDSO vinylpentamethyldisiloxane DVS 1-3 divinyl tetramethyldisiloxane MTS methyl trimethoxysilane DLC diamond like carbon DADBS diacetoxy-di-tert-butoxysilane DPXC dichloro-di-paraxylylene TRIES triethoxysilane TMOS tetramethyloxysilane OMCTS octamethyltrisiloxane HMDSO hexamethyldisiloxane HSQ hydrogen-silsesquioxane MSQ methyl-silsesquioxane POSS polyhedral oligomeric silsesquioxane Abbreviations UV ultraviolet radiation CVD chemical vapour deposition PVD physical vapour deposition CCP capacitively coupled plasma ICP inductively coupled plasma ECR electron cyclotron resonance WBL weak boundary layer PC polycarbonate XPS x-ray photon spectroscopy ERD elastic recoil detection

Symbols and Abbreviations

128

ATR-IR attenuated total reflection infrared EMA electron microprobe analysis PECVD plasma enhanced chemical vapour deposition FTIR fourier transform infrared XPS x-ray photon spectroscopy ESEM environmental scanning electron microscopy OML orbital motion limited RC resistance-Capacitance TGA thermal gravimetric analysis a.c. alternating current d.c. direct current Symbols Ti ion temperature ne electron density ni ion density λmfp mean free path Ea activation energy λD Debye length a particle radius Vf potential of a body with respect to plasma mi ion mass me electron mass kB Boltzman constant ε0 permittivity of free space e unsigned charge on an electron (1.602 x10-19 C) Te electron temperature in units of volts (V) ns ion density at the sheath/plasma interface M ion mass (Kg) I current (A) Ii ion current (A) Poff pressure when plasma is switched off (mTorr) Pon pressure when plasma is switched on (mTorr) Vf floating potential (V) Vp plasma potential (V) υB Bohm velocity, speed of sound in plasma (m.s-1) Fes electrostatic force Fg gravitational force Fid ion drag force Fth themophoretic force Fnd neutral drag force κT translational component of the thermal conductivity of a gas

Summary

129

Summary Recent developments in materials technology, fuelled by the growing hype surrounding nanotechnology, have given rise to a new breed of materials known as nanocomposites. Nanocomposite materials (a subgroup of hybrid materials) are formed from standard polymers impregnated with nanometre sized ceramic or inorganic particles. Bulk polymeric properties, such as toughness and low weight are retained, while the incorporation of ceramic/inorganic nano-particles adds additional functionality such as improved scratch resistance and reduced permeability and makes hybrid materials interesting for a wide variety of applications. In this thesis, the possibility to incorporate controlled porosity into nanocomposite thin films as low-k material for dielectric interlayers in semiconductor devices is successfully investigated.

Wet chemical process technologies are currently the only methods used to synthesize nanocomposite materials on an industrial scale. However, high curing temperatures, multiple process steps and use of environmentally harmful solvents make wet chemical techniques unattractive for production on the scale necessary to fulfil industrial requirements. In contrast, gas phase process technologies are typically one-step solvent free processes and therefore offer an attractive alternative to wet chemical processes. In addition to these benefits, the extensive use of vacuum and gas phase processing equipment in the semiconductor processing industry make gas phase synthesis of nanocomposite films for semiconductor applications particularly attractive.

The primary aim of this work has been the conception and realisation of a process for synthesizing hybrid materials using solely gas phase synthetic methods. The emphasis is on plasma-based synthesis of nanocomposites since for these processes the substrate temperature can be kept low. The main theme through the thesis has been the continued design, testing and adaptation of a reactor system conceived from reviewing literature regarding a range of materials and deposition processes. The development of the reactor concept progresses with increased understanding of the plasma physics and chemistry as obtained from applying in situ plasma diagnostics to monitor the plasma and particle formation, and from techniques related to film formation. Alongside the reactor development, the research has also investigated reactive ion etching and adhesion of hybrid coatings, chemical kinetics and the physics of dusty plasmas.

The thesis begins with an extensive review of relevant literature regarding a variety of hybrid film deposition techniques and material descriptions. Results of this review prompted the design of the initial reactor concept consisting of a dual inductively coupled/capacitively coupled (ICP/CCP) plasma reactor, i.e. the initial approach chosen was to separate the production of nanoparticles spatially from the film formation process on the substrate. Due to its ability to form both hard silica and flexible silicone based materials, silicon based process chemistry was chosen as the main chemistry with which to deposit the hybrid films. Several silicon

Summary

130

precursors were evaluated for producing silica and silicone materials using the ICP and CCP plasmas. Preliminary findings showed the precursor combinations O2/Tetraethoxysilane (TEOS) and 1,2-bis(trimethyl)siloxyethane (TMSE) were optimal for synthesizing the inorganic and organic film fractions respectively. Both inorganic and polymeric thin films were successfully synthesized and compared well to wet chemically synthesized films.

With the early success in synthesizing hybrid layers, the more challenging task of synthesizing nanocomposite materials became the objective of the research. Unfortunately all initial attempts at synthesizing nanoparticles with the ICP were unsuccessful. An investigation into dusty plasma physics and the plasma chemistry of the O2/TEOS precursor system identified negative ion formation and trapping as being the key steps in the formation of particles in low pressure discharges. The low residence time in the ICP tube caused by the absence of confining electric fields was the primary reason why particles were not formed in this plasma. On the other hand, in the CCP, due to the presence of confining electric fields, successful generation of dust particles was demonstrated. Characterisation of these particles, both in situ using Fourier transform infrared absorption spectroscopy (FTIR) and of particles collected after the experiments using a range of physical and chemical characterization techniques, indicated the particles to be porous and silica-like. It was demonstrated that these particles could be grown, manipulated and kept for a long time above the electrode of a CCP.

Combining the finding of the dusty plasma and silicon chemistry studies resulted in the design of a second dual plasma reactor consisting of two capacitive plasmas. In addition, the nanoparticle production and the film deposition were temporally separated by the use of pulsed plasma operation. This reactor was then used in combination with the O2/TEOS and TMSE precursor systems to deposit nanocomposite films with controlled porosity for low dielectric constant films in semiconductor devices. This approach proved to be successful and low-k films exhibiting dielectric constants as low as 1.82 ± 0.02 were demonstrated. On this finding a patent application has been based.

Samenvatting

131

Samenvatting Recente ontwikkelingen in de materiaaltechnologie hebben, onder invloed van de groeiende aandacht voor de nano-technologie, tot een nieuwe type materialen geleid die, vanwege hun structuur, bekend staan als nanocomposieten. Nanocomposieten (een subgroep van de hybride materialen) worden gevormd uit standaard polymeren die geïmpregneerd zijn met anorganische nanodeeltjes. In het geval van de nanocomposiet materialen blijven de polymeren eigenschappen zoals de taaiheid en het lage gewicht behouden, terwijl door het inbouwen van de anorganische of keramische nanodeeltjes additionele functionaliteit wordt toegevoegd zodat bijvoorbeeld de krasvastheid wordt verbeterd en de gasdoorlaatbaarheid wordt verlaagd. De mogelijkheid om unieke eigenschappen te combineren maakt hybride materialen erg interessant voor een groot aantal verschillende toepassingen. In dit proefschrift wordt ondermeer succesvol aangetoond dat films aangebracht kunnen worden met een controleerbare (nano-)porositeit, waardoor dit type hybride materiaal geschikt is als een isolerende tussenlaag met lage diëlectrische constante (zgn. low-k materialen) in halfgeleidertoepassingen.

Natchemische procestechnologieën zijn momenteel de enige methoden die gebruikt worden voor de synthese van nanocomposieten op een industriële schaal. Door de hoge droogtemperaturen, de vele processtappen en het gebruik van schadelijke oplosmiddelen is deze produktiemethode echter niet erg aantrekkelijk voor de hoge eisen voor de produktie op een industriele schaal.Een aantrekkelijk alternatief voor natchemische productie zou gasfase procestechnologie zijn. Dit is een enkelstapsproces, waar geen oplosmiddelen aan te pas komen. Bovendien is het gebruik van vacuüm- en gasfase depositie-apparatuur standaard in bijvoorbeeld de halfgeleiderindustrie.

De primaire doelstelling van het promotie-onderzoek betreft het ontwerp en de realisatie van een proces dat hybride materialen kan synthetiseren op basis van gasfase processen. De nadruk ligt met name op plasma geassisteerde processen omdat deze processen het gebruik van een lage substraattemperatuur mogelijk maakt. Het belangrijkste thema in dit proefschrift is het ontwerpen, testen en aanpassen van een reactorsysteem, op basis van een diepgaand onderzoek van het plasma en de deeltjes gevormd in het plasma middels geavanceerde in situ plasmadiagnostieken. Het ontwerp van het gasfase depositieproces en de reactor wordt telkens verbeterd ten gevolge van een beter begrip van de plasmafysica en chemie en de processen die tot een hybride filmvorming leiden. Parallel aan de reactorontwikkeling, is er onderzoek verricht naar het plasma etsen van hybride materialen, en de hechting tussen hybride coatings en een polymeren substraat.

Op basis van uitgebreid literatuuronderzoek op het gebied van hybride materialen is een eerste reactor concept ontworpen bestaande uit zowel een inductief gekoppeld plasma (ICP) als een capacitief gekoppeld (CCP) plasma. Deze eerste benadering was gebaseerd op de gedachte de produktie van nanodeeltjes en

Samenvatting

132

het film depositieproces ruimtelijk te scheiden. Er is gekozen voor een op silicium gebaseerde chemie, omdat dit de mogelijkheid opent zowel hard silica als flexibel siliconenachtige films te deponeren. Verschillende precursoren zijn geëvalueerd voor de synthese van silica en siliconen met behulp van de ICP/CCP plasma’s. Voorbereidend onderzoek liet zien dat een combinatie van O2/tetraethoxysilaan (TEOS) en 1,2-bis(trimethyl)siloxyethaan (TMSE) het meest geschikt waren voor het synthetiseren van respectievelijk de anorganische en de organische fractie. Zowel de anorganische als de organische films zijn succesvol gesynthetiseerd en zijn qua eigenschappen vergelijkbaar met lagen gemaakt met behulp van natchemische technieken.

De volgende stap in het promotie-onderzoek was de synthese van nanocomposiet materialen. Helaas bleek het niet mogelijk om nanodeeltjes te maken met behulp van het ICP plasma. Onderzoek naar de fysica en de chemie van het plasma met TEOS en O2 als precursor systeem, wees uit dat de vorming en het opsluiten van negatieve ionen in het plasma de belangrijkste stap is bij de vorming van deeltjes in een lage druk plasma. In een ICP is de verblijftijd van de precursoren te kort door de afwezigheid van opsluitende elektrische velden waardoor deeltjesvorming in het plasma nauwelijks plaatsvindt. In tegenstelling tot het ICP werd in het CCP, ten gevolge van de aanwezigheid van opsluitende electrische velden, met succes deeltjesvorming aangetoond. De deeltjes konden over een langere tijd gemanipuleerd en bewaard worden boven de electrode van het CCP. Zowel in situ als ex situ karakterisatie van de nanodeeltjes gemaakt in het CCP met behulp van o.a. Fourier Transform Infrarood (FTIR) absorptiespectroscopie, wees uit dat de deeltjes silica-achtig en poreus zijn.

De resultaten uit het onderzoek naar poeder-plasma’s en de siliciumchemie hebben tot het ontwerp van een tweede plasmareactor geleid bestaande uit twee capacitief gekoppelde plasma’s (CCP). De scheiding van deeltjesvorming en filmdepositie werd nu verwezenlijkt in de tijd door het gebruik van gepulste plasma’s. In combinatie met de precursorsystemen O2/ TEOS en TMSE zijn in deze reactor nanocomposiet films gemaakt met een controleerbare porositeit en met een laagst bereikte dielektrische constante van 1.82 ± 0.02. De potentie van deze films, als low-k films in de halfgeleiderindustrie is groot en een patent op deze vinding is ingediend.

Acknowledgement

133

Acknowledgement It goes without saying that the research presented in this thesis, performed over the last four years, is the result of teamwork and not only the efforts of a single person. I wish then to thank all those who have contributed, in whatever way, to the realization of this manuscript.

I would first like to thank my supervisor, Richard van de Sanden for his guidance, inexhaustible energy and enthusiasm. Thanks also for sticking up for me. Many thanks also to the members of the ETP group, Adriana for your valuable input, not only during the brief period we shared and office, but throughout the duration of the project. The two students Sascha Kondic and Dimitri Eijkman, thank you for your valuable contributions.

A majority of the experimental work presented in this thesis was conducted at the TNO TPD and I would like to express my thanks to all my colleagues who helped things run smoothly and provided morel support. In particular I would like to mention Gerwin and Frank for your help around the laboratory and especially Hans for your support and unique ability to simply know the right people to get things done! Special thanks must go to Ton for his invaluable support, friendship and always finding time no matter how trivial the question. Thanks for introducing me to the Zeelste kermis and making me feel right at home with all the great people who inhabit that small town. And not least, thanks for waiting, if it were not for you I’d still be standing at Eindhoven station!

For the three months of this project I spent in Australia I would like to thank all the members of the RSPhysSE group for making me feel welcome. Rod Boswell for allowing me to attend his insightful plasma course entitled “my friend the electron” that has formed the foundation to my understanding of plasma physics. Rob and Christine for all weekend surfing trip, Christmas dinners and generally making my time in Australia so unforgettable. Thanks.

This would all have been much harder without my close friends outside of my work environment who have offered moral support over the years. In particular Akif and John. The Bommel crowd Kornel, Richard, Vivianne, Huub (on occasion), Hans, Frank and Danny for putting up with me talking nonsense. And of course Ursula for putting up with so much more.

Finally I would like to thank my family who have always stood by me. My parents, George and Neil for helping me move and make my flat a home, Sally for your feng shui advice and artistic input and Simon for making so many trips over to visit and keeping me company.

David Sheel and Karel Spee, thanks for believing in me.

Acknowledgement

134

Curriculum vitae

135

Curriculum vitae

4 December 1974 Basingstoke, UK

September 1993 - July 1997 BSc Hons Chemistry, University of Bath, UK

Twelve months traineeship at Pilkington Technology Centre,

Ormskirk, UK

November 1997 – November 1999 Marie Curie fellowship, TNO TPD, Eindhoven, NL

November 1999 – February 2004 PhD, Department of Applied Physics, Eindhoven University

of Technology

Three months working visit to the group of Prof. R. Boswell at the Research School of Physical Sciences and Engineering,

Australian National University, Canberra, AU

February 2004 - Present Research Scientist, TNO TPD, Eindhoven, NL

Curriculum vitae

136

Publications related to this work

137

Publications related to this work Alcott G.R., van Mol A.M.B., Eijkman D., Linden J.L., van de Sanden M.C.M., “Reaction kinetics and mechanisms during PECVD from a TEOS/O2 gas mixture”, submitted to Plasma Sources Science and Technology (2004). Alcott G.R., van de Sanden M.C.M., Kondic S., Linden J.L., “Vapour pressures of precursors for the chemical vapour deposition of silicon based films”, Advanced Materials, Chem Vap. Deposition, 10 (1), 20-22, (2004). Alcott G.R., Linden J.L., van de Sanden M.C.M.,” Gas phase deposition of hybrid coatings”, Materials Research Society Symposium Proceedings, Vol 726, Q9.9.1-Q9.9.6, (2002). Alcott G.R., Eijkman D.J., Schrauwen C.P.G., Linden J.L., van de Sanden M.C.M., 10th Neues Dresdner Vakuumtechnisches Kolloquium Proceedings, 55-60, (2002). Patents Alcott G.R., van de Sanden M.C.M., Linden J.L., Hamers E.A.G., “CVD van Hybride materialen”, Application Nr PCT/NL03/0037, Priority date 18/01/2002, International filing date 17/1/2003. Alcott G.R., Creatore M., Linden J.L., van de Sanden M.C.M., “Method of manufacturing ultra low dielectric layers”. Application date 13/2/2004.

process concept and realisationprocess concept and realisation proefschrift ter verkrijging van de...

Documents