Tantalum- and Ruthenium-Based Diffusion Barriers/Adhesion

Home , Diffusion barrier

TANTALUM- AND RUTHENIUM-BASED DIFFUSION BARRIERS/ADHESION

PROMOTERS FOR COPPER/SILICON DIOXIDE AND

COPPER/LOW κ INTEGRATION

Xiaopeng Zhao, B.E., M.E.

Dissertation Prepared for the Degree of

DOCTOR OF PHILOSOPHY

UNIVERSITY OF NORTH TEXAS

December 2004

APPROVED:

Jeffry A. Kelber, Major Professor William E. Acree, Jr., Committee Member Mohammad Omary, Committee Member Angela Wilson, Committee Member Ruthanne D. Thomas, Chair of the Department of Chemistry Sandra L. Terrell, Dean of the Robert B. Toulouse School of Graduate Studies Zhao, Xiaopeng, Tantalum- and ruthenium-based diffusion barriers/adhesion promoters

for copper/silicon dioxide and copper/low κ integration. Doctor of Philosophy (Analytical

Chemistry), December 2004, 84 pp., 2 tables, 32 illustrations, 105 references.

The TaSiO6 films, ~8Å thick, were formed by sputter deposition of Ta onto ultrathin SiO2 substrates at 300 K, followed by annealing to 600 K in 2 torr O2. X-ray photoelectron spectroscopy (XPS) measurements of the films yielded a Si(2p) binding energy at 102.1 eV and

Ta(4f7/2) binding energy at 26.2 eV, indicative of Ta silicate formation. O(1s) spectra indicate

that the film is substantially hydroxylated. Annealing the film to > 900 K in UHV resulted in

silicate decomposition to SiO2 and Ta2O5. The Ta silicate film is stable in air at 300K. XPS data show that sputter-deposited Cu (300 K) displays conformal growth on Ta silicate surface

(TaSiO6) but 3-D growth on the annealed and decomposed silicate surface. Initial Cu/silicate

interaction involves Cu charge donation to Ta surface sites, with Cu(I) formation and Ta

reduction. The results are similar to those previously reported for air-exposed TaSiN, and

indicate that Si-modified Ta barriers should maintain Cu wettability under oxidizing conditions

for Cu interconnect applications.

XPS has been used to study the reaction of tert-butylimino tris(diethylamino) tantalum

(TBTDET) with atomic hydrogen on SiO2 and organosilicate glass (OSG) substrates. The results on both substrates indicate that at 300K, TBTDET partially dissociates, forming Ta-O bonds with some precursor still attached. Subsequent bombardment with atomic hydrogen at 500K results in stoichiometric TaN formation, with a Ta(4f7/2) feature at binding energy 23.2 eV and

N(1s) at 396.6 eV, leading to a TaN phase bonded to the substrate by Ta-O interactions.

Subsequent depositions of the precursor on the reacted layer on SiO2 and OSG, followed by atomic hydrogen bombardment, result in increased TaN formation. These results indicate that TBTDET and atomic hydrogen may form the basis for a low temperature atomic layer deposition

(ALD) process for the formation of ultraconformal TaNx or Ru/TaNx barriers.

The interactions of sputter-deposited ruthenium with OSG at 300 K have been studied by

XPS for Ru coverages from ~ 0.1 monolayer to several monolayers, using in-situ sample transfer between the deposition and analysis chambers. The results indicate Stranski-Krastanov (SK) type growth, with the completion of the first layer of Ru at an average thickness corresponding to

1 monolayer average coverage. Ru(0) is the only electronic state present. XPS core level spectra

indicate weak chemical interactions between Ru and the substrate. A less pronounced tendency

towards SK growth was observed for Ru deposition on parylene. Deposition of Ru on OSG

followed by electroless deposition of Cu resulted in the formation of a shiny copper film that

failed the Scotch® tape test. Results indicate failure mainly at the Ru/OSG interface.

ACKNOWLEDGMENTS

The author wishes to express his gratitude to Prof. Jeffry A. Kelber for his guidance and careful instruction. Special thanks are extended to Dr. William E. Acree,

Jr., Dr. Mohammad Omary, Dr. Angela Wilson and Dr. Robert M. Wallace, for their informative discussions. Financial supports provided by the Semiconductor Research

Corporation (SRC) through the Center for Advanced Interconnect Science and

Technology (CAIST), SRC through a Novellus/SRC customized research program, and from the Robert Welch Foundation (grant no. B-1356), are gratefully acknowledged.

Finally, the author would like to express sincere appreciation to his wife, Liqin Zhang for her constant encouragement and support.

ii TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ...... ii

LIST OF TABLES...... vi

LIST OF ILLUSTRATIONS...... vii

LIST OF ABBREVIATIONS...... x

CHAPTER

1. INTRODUCTION ...... 1

1.1. RC Delay...... 3

1.2. Copper Metallization for Interconnection Technology ...... 4

1.3. Diffusion Barriers ...... 6

1.4. Thin Film Deposition ...... 8

1.4.1. Physical Vapor Deposition (PVD)-DC Magnetron Sputtering.. 8

1.4.2. Chemical Vapor Deposition (CVD) ...... 9

1.5. Low-κ Dielectric...... 11

1.6. Nucleation Modes Characterized by XPS...... 14

1.7. Experimental Aspects...... 15

1.7.1. Apparatus ...... 15

1.7.2. X-Ray Photoelectron Spectroscopy (XPS) ...... 16

1.8. Chapter References ...... 22

iii 2. COPPER INTERACTION WITH A TANTALUM SILICATE SURFACE:

IMPLICATIONS FOR INTERCONNECT TECHNOLOGY ...... 26

2.1. Introduction...... 26

2.2. Experimental Details...... 28

2.3. Results...... 30

2.3.1. Ta Silicate Formation...... 30

2.3.2. Thermal Decomposition...... 34

2.3.3. Air-Exposure of Ta Silicate Film ...... 35

2.3.4. Cu/Film Interactions ...... 35

2.4. Discussion...... 42

2.5. Conclusions...... 43

2.6. Chapter References ...... 44

3. CHEMICAL VAPOR DEPOSITION OF TANTALUM NITRIDE WITH

TERTT-BUTYLIMINO TRIS(DIETHYLAMINO) TANTALUM AND

ATOMIC HYDROGEN ...... 47

3.1. Introduction...... 47

3.2. Experimental Details...... 49

3.3. Results...... 50

3.3.1. TaN Formation on SiO2 ...... 50

3.3.2. TaN Formation on OSG...... 56

3.4. Discussion...... 62

iv 3.5. Conclusions...... 64

3.6. Chapter References ...... 65

4. RETHENIUM SPUTTER DEPOSITION ON ORGANOSILICATE GLASS

AND ON PARYLENE: AN XPS STUDY OF INTERFACIAL

CHEMISTRY, NUCLEATION AND GROWTH ...... 68

4.1. Introduction...... 68

4.2. Experimental Detail ...... 69

4.3. Results...... 69

4.4. Discussion...... 74

4.5. Conclusions...... 75

4.6. Chapter References ...... 76

BIBLIOGRAPHY...... 78

v LIST OF TABLES

Table Page

1.1. Properties of possible interconnect metals...... 5

1.2. Requirements for low dielectric constant intralayer-dielectric materials ...... 13

vi LIST OF ILLUSTRATIONS

Figure Page

1.1. Comparison of intrinsic gate delay and interconnect (RC) delay as a function of

feature size...... 4

1.2. Schematic diagram of a basic Cu interconnect structure...... 6

1.3. The DC magnetron sputtering process ...... 9

1.4. Deposition profiles of (a) PVD and (b) CVD ...... 10

1.5. Principle of MOCVD process...... 11

1.6. OSG schematic bonding structure ...... 13

1.7. Intensity vs. time plots of (a) Frank−Van der Merwe (layer-by-layer);

(b) Stranski−Krastanov growth (monolayer then islanding) and (c) Volmer−Weber

(islanding) ...... 15

1.8. Combined XPS and DC magnetron PVD/CVD apparatus ...... 16

1.9. The XPS emission process for a model atom ...... 17

1.10. Schematic drawing of a X-ray photoelectron spectrometer ...... 19

1.11. Angle resolved XPS ...... 21

1.12. Schematic of the Auger process of a model atom ...... 21

2.1. Formation of SiO2 and Ta silicate. (a) Si(2p); (b) O(1s) and (c) Ta(4f) Spectra ...32

2.2. XPS intensity ratio as a function of annealing temperature...... 34

2.3. XPS spectra of Ta silicate at room temperature and after annealing at 900 K for 30

min in UHV. (a) Si(2p) Spectra and (b) Ta(4f) Spectra...... 36

vii 2.4. XPS spectra of Ta silicate before and after air exposure. (a) Si(2p) spectra; (b)

O(1s) spectra and (c) Ta(4f) spectra ...... 37

2.5. X-ray excited Cu(L3VV) Auger spectral evolution as a function of Cu deposition on

the unannealed silicate film...... 39

2.6. Cu(2p)/Si(99) XPS intensity ratio vs. deposition time (a) before and (b) after

annealing the silicate film at 1000K ...... 40

2.7. Ta(4f) XPS spectra at Cu deposition for 1min and 5.5 min...... 41

3.1. Tert-butylimino tris(diethylamino) tantalum (TBTDET) structure...... 48

3.2. XPS spectral changes upon reaction with TBTDET precursor. (Bottom traces) SiO2

sample before precursor exposure; (bottom middle traces) precursor adsorption at

120 K; (top middle traces) after annealing to 300K; (top traces) after H/H2

exposure. (a) Si(2p); (b) O(1s); (c) Ta(4f); (d) N(1s) and (e) C(1s)...... 52

3.3. XPS spectra change after annealing without and with subsequent H2/H flux

exposure on SiO2 sample. (a) Ta(4f) and (b) N(1s)...... 54

3.4. XPS spectra change with cycles of TBTDET dose and H2/H flux exposure on SiO2

sample. (a) Ta(4f) and (b) N(1s)...... 55

3.5. Ta(4f) XPS spectra after TBTDET dose and H2/H flux exposure on SiO2. (a)

Normal incidence and (b) Grazing incidence ...... 56

3.6. XPS spectral changes upon reaction with TBTDET precursor. (Bottom traces) OSG

sample before precursor exposure; (middle traces) precursor adsorption at 120 K

followed by annealing to 300K; (top traces) after H/H2 exposure. (a) Si(2p); (b)

O(1s); (c) Ta(4f); (d) N(1s) and (e) C(1s)...... 59

viii 3.7. XPS charge shift during various process stages of TaN formation, stage 1: OSG

sample as received; 2: cool down to 120K; 3: TBTDET dose at 120K; 4: anneal to

300K; 5: hydrogen flux exposure for 90 min; 6: hydrogen flux exposure for 270

min ...... 60

3.8. XPS spectra change with cycles of TBTDET dose and H2/H flux exposure on OSG

sample. (a) Ta(4f) and (b) N(1s)...... 61

3.9. Ta(4f) XPS spectra after TBTDET dose and H2/H flux exposure. (a) on SiO2 and

(b) on OSG...... 63

4.1. Core level XPS spectra for the OSG film before (bottom) and after (top) annealing

at 500 K for 30 min. in UHV ...... 70

4.2. Ru(3d)/C(1s) region, showing the partial overlap of the Ru(3d) and C(1s) core

level features. Spectrum corresponds to about 30 min Ru deposition

(approximately 3 monolayers). Ru(3d3/2) component indicated by dashed line was

determined from the intensity and binding energy of the Ru(3d5/2) component....71

4.3. The evolution of the Ru(3d5/2)/C(1s) intensity ratio as function of Ru deposition vs

time at 300 K. (Left) Deposition of Ru on OSG. The film thickness at 10 min

corresponds to 1 monolayer. (Right) Deposition of Ru on Parylene...... 72

4.4. XPS C(1s)/Ru(3d) spectra (left), and O(1s) spectra (right) ...... 73

LIST OF ABBREVIATIONS

ALD Atomic layer deposition

AP Auger parameter

ARXPS Angle resolved X-ray photoelectron spectroscopy

CAE Constant analyzer energy

CHA Concentric hemispherical analyzer

CVD Chemical vapor deposition

FAT Fixed analyzer transmission

FM Mode Frank - Van der Merve mode

FWHM Full width at half maximum

HSA Hemispherical sector analyzer

IC Integrated circuit

ILD Intralayer dielectric

IMFP Inelastic mean free path

MOCVD Metal-organic chemical vapor deposition

OSG Organosilicate glass

PVD Physical vapor deposition

RC Delay Resistance – capacitance delay

SEM Scanning electron microscopy

SK Mode Stranski - Krastanov mode

x TBTDET Tert-butylimino tris(diethylamino) tantalum

UHV Ultra-high vacuum

ULSI Ultra-large scale integrated

VW Mode Volmer - Weber mode

XPS X-ray photoelectron spectroscopy

xi CHAPTER 1

INTRODUCTION

With the shrinkage of integrated circuit geometry, the resistance-capacitance (RC) delay in interconnects starts to become one of the major problems as the width of interconnect lines becomes < 1 µm [1, 2]. Strategies to reduce the delay include incorporating metals with lower resistivity and higher electromigration resistance, such as copper, and providing isolation with low dielectric constant materials [3]. Copper (Cu) is an attractive substitute for aluminum (Al)

because it offers a lower resistivity (1.67 µΩ⋅cm versus 2.66 µΩ⋅cm for Al) and better

electromigration resistance. Cu integration is challenged, however, by its rapid diffusion into

silicon (Si) through silicon dioxide (SiO2) and poor adhesion to SiO2 [4]. The diffusion of Cu

into Si results in adverse effects, including the increase of junction leakage currents and a degradation of oxide quality if it is present in device regions [5, 6]. Consequently, diffusion

barriers and adhesion promoters that do not alloy with Cu, adhere well to both Cu and dielectrics,

and exhibit electrical stability in high temperature are required for Cu integration [4, 7].

Diffusion barriers such as tantalum (Ta), tungsten (W), ruthenium (Ru) and corresponding

nitrides or carbides have been proposed as barriers [8-14]. The ability of Cu to wet, or grow

conformally, on most surfaces during deposition is extremely sensitive to oxygen (O)

contamination at the Cu/substrate interface. A partial monolayer of oxygen at the Cu/barrier

interface will significantly degrade wettability and step coverage, and allow facile agglomeration

at temperatures at or above 300K [8-10]. This sensitivity to low-level oxygen contamination

poses a significant threat to processing reliability, since Cu deposition is typically not done under

UHV conditions [4]. An air-stable diffusion barrier is desirable for Cu deposition.

1 At the same time, the signal processing times and signal cross talk between adjacent lines

in ultra-large scale integrated (ULSI) circuits can be improved by replacing SiO2 with materials having low dielectric constants than that of SiO2 (κ ~ 4). The surface chemical state and structure of low-κ dielectric materials are different from SiO2. The potential for relatively poor adhesion between low k dielectric materials and barrier layers is one of the main issues involved in the integration of Cu and low k materials into a standard microfabrication process [15].

The studies presented in this dissertation focus on the interactions of diffusion barriers/adhesion promoters with Cu, SiO2 and low-κ dielectric materials. X-ray photoelectron spectroscopy (XPS) was used for surface analysis. Thin film deposition technologies used in these studies include DC magnetron sputtering and metal-organic chemical vapor deposition

(MOCVD). These studies include:

(1) Cu interaction with a Ta silicate surface: implications for interconnect technology.

(2) Chemical vapor deposition of TaN with tert-butylimino tris(diethylamino) tantalum

(TBTDET) and atomic hydrogen.

(3) Ruthenium sputter deposition on organosilicate glass and on parylene: an XPS study of interfacial chemistry, nucleation and growth.

This dissertation consists of four chapters. The first chapter provides background information on the fundamental interconnect integration, namely, role of RC delay, necessity of

Cu and low-κ dielectric materials, diffusion barriers, thin film deposition methods, nucleation

modes, as well as experimental methodologies. Chapter 2 presents the studies on Cu interaction

with a Ta silicate surface and its implications for interconnect technology. Chapter 3 includes

studies of chemical vapor deposition of tantalum nitride (TaN) with Tert-butylimino

2 tris(diethylamino) tantalum and atomic hydrogen. Chapter 4 contains the study of ruthenium

sputter deposition on organosilicate glass and on parylene.

1.1. RC Delay

With the shrinking of transistor size in integrated circuits (IC) to improve chip

performance, speed and device density have increased. The effective measure of device

performance or total circuit delay is a combination of two components – the switching time of an

individual transistor, known as intrinsic gate delay, and the signal propagation time between

transistors, known as interconnect (RC) delay [16-18]:

2 2 2 2 RC =2ρκε0(4L /P +L /T ), (1-1) where R is metal wire resistance, C is intralayer dielectric (ILD) capacitance, ρ is metal

resistivity, κ and ε0 are dielectric constant of the intralayer dielectric and vacuum respectively, L,

T are the length and thickness of the conductor respectively, P is the distance between two conducting lines.

For feature size below 1 µm, the intrinsic gate delay decreases continuously with decreasing size of the transistors and the RC delay becomes the dominant factor (Fig.1.1) [1, 2].

New materials are being utilized by industry to reduce the RC delay by either lowering the interconnect wire resistivity, or by reducing the dielectric constant of the ILD. A significant improvement has been achieved by replacing the Al interconnects with Cu, which has ~30% lower resistivity than that of Al [2]. Since no other metal can offer lower resistance without other disadvantages (e.g., low electromigration resistance), Cu will remain the choice for interconnect.

Therefore, further advances can be achieved by reducing the dielectric constant of the ILD, i.e., by replacing SiO2 with a low-κ material.

3 3 Intrinsic Gate Delay 2.5

)

c Interconnect Delay (RC)

s 2

–

1 1.5

(

m 1

y 0.5

D 0 00.511.522.533.5 Feature Size (µm)

Fig. 1.1 Comparison of intrinsic gate delay and interconnect (RC) delay as a function of feature size.

1.2. Copper Metallization for Interconnect Technology

Al has been widely used in the past due to its relatively low resistivity, good corrosion resistance in air and the formation of stable Al/SiO2 interfaces [4, 19]. Despite these favorable characteristics, poor resistance to electromigration and hillock formation of Al leads to lower lifetime and shorts between levels. In addition, the need to lower interconnect delay (Fig. 1.1) has lead to the examination of other metals as replacements to Al [2, 4, 20].

A comparison of Al to some other lower-resistivity metals that are potential replacements is listed in Table 1.1. Applications of silver (Ag) and gold (Au) to replace Al have been suggested in the last three decades. Although Au has very low corrosion in the air, it is far more costly and has a higher resistivity and lower electromigration resistance than Cu [4, 20, 21]. The

4 advantage of Ag is its low resistivity, about 5% lower than that of Cu. However, it is more costly

than Cu, has a poor electromigration resistance, and diffuses into SiO2 at a much faster rate than

Cu, especially under electrical bias [4, 20, 21]. Cu is the best choice to replace Al due to its low resistivity, high melting point and high electromigration resistance [4, 20, 21]. Because Cu is more conductive than Al and smaller interconnect lines can provide the same current-carrying capability, Cu makes it possible to reduce the number of levels of metal and reduce the capacitance. However, several processing and reliability issues need to be addressed before Cu technology can be fully integrated with device manufacteruring. Cu will react with Si very

strongly to form Cu silicide at temperatures as low as 200 C, and the reaction is detrimental to

the electrical performance of Si [22, 23]. Cu can also move easily through the oxide to the Si/O

interface, under conditions of combined elevated temperature and electric field (“bias/thermal

stress- BTS”) [5, 24]. For the low κ replacement of SiO2, the potential interaction between Cu and low κ materials is a serious reliability issue [25, 26]. Therefore, diffusion barriers that can

block the Cu transport and adhesion promoters that can improve Cu adhesion are required for Cu

integration into device manufacturing.

Table 1.1 Properties of possible interconnect metals.

Properties Cu Ag Au Al

Resistivity 1.67 1.59 2.35 2.66 (µΩ.cm) Electromigration Very Low High Low Resistance High Corrosion in Air High High Very Low Low

1.3. Diffusion Barriers

A diffusion barrier/adhesion promoter (Fig. 1.2) is very important for Cu–based metallization due to the high diffusivity of Cu into SiO2 and low k dielectric materials [4]. A successful diffusion barrier must form strong chemical bonds to the dielectric substrate and to

Cu. The chemical bonding at these interfaces should be the result of self-limiting reactions, otherwise reactions will continue to occur with time and the properties of substrates, barriers and metals may be compromised. Thus the ultimate objective is to find an adhesion promoter that will also perform as a diffusion barrier.

Diffusion Barrier /Adhesion Promoter

ILD ILD

Substrate

Fig. 1.2 Schematic diagram of a basic Cu interconnect structure. “ILD” is the intralayer dielectric: currently SiO2 or a lower k dielectric material.

The refractory metals, and their binary and ternary compounds, such as Ta, W, Ru, molybdenum (Mo), vanadium (V), TaN and tantalum silicon nitride (TaSiN), have been widely

6 investigated as diffusion barriers [8-12, 26]. These refractory metals are not miscible with Cu at

low temperature [4, 27, 28]. In addition, they have high melting points for thermal stability

during processing [4, 27, 28].

Titanium (Ti) and titanium nitride (TiN) are widely used in Al-based interconnect

systems [27, 29]. Unfortunately, they cannot provide enough Cu barrier performance since Cu

and Ti form a bulk alloy, and a Cu grain boundary diffusion proceeds readily in TiN [30, 31].

For Cu-based interconnect systems, Ta-based diffusion barriers possess an inherent advantage

over Ti-based barriers because Ta is thermodynamically stable with respect to Cu. Cu and Ta are

almost completely immiscible up to their melting point and do not react [32]. In addition, the

Ta/Si interface is stable up to ~920K [33]. TaN has been widely recognized as an attractive diffusion barrier due to its thermal stability and high conductivity [12-14]. Ta nitride is highly stable with respect to Cu due to the absence of Cu-Ta and Cu-N compounds [27]. Furthermore, the TaNx/Si interface is more stable than Ta/Si interface because higher activation energy is

needed to dissociate Ta-N bond before silicide formation [27]. It has been shown that

stoichiometric TaN fails through Cu diffusion along grain boundaries, but at a higher

temperature than Ta [34, 35]. TaSixNy has been investigated as a diffusion barrier for Cu since being amorphous over a broad temperature range lacks grain boundaries that can act as diffusion pathways [10, 27, 36]. Addition of a third element into a Ta nitride matrix disrupts the crystal lattice and leads to the formation of a stable amorphous ternary phase TaSixNy with significantly higher recrystallization temperature than TaN.

Ru-based materials are of growing interest as diffusion barriers for Cu due to ruthenium’s high melting point (2310°C) and immiscibility with Cu [11, 37]. It shows negligible solid solubility with Cu even at 900°C [32, 38]. Strong adhesion between Cu/Ru is confirmed by

7 scribe and tape-peel tests [11, 39]. Ruthenium is relatively noble, allowing for facile electrodeposition of Cu films [11, 37].

1.4. Thin Film Deposition Methods

In these studies, tantalum, copper and ruthenium depositions were performed by direct current (DC) magnetron sputtering deposition. Tantalum nitride was deposited by metal-organic chemical vapor deposition (MOCVD) using tert-butylimino tris(diethylamino) tantalum

(TBTDET) as precursor.

1.4.1. Physical Vapor Deposition (PVD) - DC Magnetron Sputtering

The process of DC magnetron sputtering process is shown in Fig. 1.3. The electrons that are ejected from the cathode are accelerated away from the cathode. A magnetic field is imposed in such a way that the electrons can be made to circulate on a closed path near the target surface.

This high flux of electrons creates a high density argon (Ar) plasma from which Ar ions can be extracted to sputter the target material. The cathode (negative) potential attracts Ar ions from near the edge of the plasma region and they are accelerated across the cathode fall region to impinge on the cathode target, resulting in atoms of target materials ejected from the surface.

Once sputtered, the target atoms travel until they reach the substrate. The principal advantage of this magnetron sputtering configuration is that a dense plasma can be formed near the cathode at low pressures so that ions can be accelerated from the plasma to the cathode without loss of energy due to physical and charge-exchange collisions. This allows a high sputtering rate with a low potential on the target, lower pressures and less gas incorporation into the sputtered film.

8 This DC sputtering cannot be used to sputter dielectric target materials, since charge buildup on the cathode surface will prevent bombardment of the surface.

Target Atoms

e Electrons

Ar Ar Atoms “E” Field

Ar+ Ar ions “B” Field

Target (Cathode) Negative High Voltage Ar+ e e e Ar+ Ar+ Ar Plasma region Ar e e Ar+ Ar e Ar e Ar Ar

Ar Growing film Ar

Substrate (Anode)

Fig. 1.3 The DC magnetron sputtering process.

1.4.2. Chemical Vapor Deposition (CVD)

PVD is not expected to be applied beyond 45 nm technology nodes due to the poor (non- conformal) step coverage caused by the shadowing effect in the small feature size, high aspect ratio via geometries (Fig. 1.4a) [40]. Accordingly, CVD is of interest since it can provide more uniform (conformal) step coverage over high aspect ratio structures (Fig. 1.4b) [27, 40]. In CVD, the constituents of a vapor phase react at a hot surface (typically higher than 300˚C) to deposit a solid film. CVD process is schematically illustrated in Fig. 1.5. Reactants are first transferred

9 from the reactor inlet to the deposition zone. Gas phase reactions can occur, resulting in the deposition of low molecular weight clusters and usually poor adhesion, low density and high- defect film. If significant gas-phase reactions are avoided, film precursors and reactants are absorbed on the growth surface and surface reactions (heterogeneous) occur on the heated surface, following by surface migration of film formers and nucleation on the growth sites. By- products of the surface reaction are then transported away from the deposition zone toward the reactor exit. The sample surface chemistry, temperature and thermodynamics determine the compound deposited. The most favorable end product on the substrate is a clean, stoichiometricly-correct film. MOCVD is a CVD method, in which the precursor is a metal- organic species. MOCVD can provide thickness control within one atomic monolayer [27, 40]. It has become a cost-effective manufacturing process for a variety of compound semiconductor devices [4, 27, 40].

(a) PVD, Non-conformal (b) CVD, Conformal

Fig. 1.4 Deposition profiles of (a) PVD and (b) CVD.

Carrier Gas Flow

Gas phase reactions + + = Transport to surface Redesorption of film precursor Desorption of by-product Surface diffusion Surface reaction Substrate

Fig. 1.5 Principle of CVD process.

1.5. Low-κ Dielectric

Since Cu has replaced Al and become the common metallization material, one of the

main challenges to further decrease RC delay today is to replace SiO2 (κ ~ 4) with lower κ dielectrics [15]. Power consumption is another major concern for interconnects besides signal delay. The increasing frequencies and densities lead to a dramatic increase in power consumption. The dynamic power dissipation can be decreased by a lower wire capacitance, resulting from the low κ dielectrics [15, 41]. The low k dielectrics can also be used to reduce cross-talk noise between metal wires [4, 15, 41].

The dielectric constant (κ), or relative permittivity (εr) is represented in equation:

ε = εr ε0 = κε0 = (1+χe)ε0, (1-2)

11 where ε0 is the permitivity of vacuum, and χe is the electric susceptibility, describing the

tendency of a material to permit an applied electric field to induce dipoles in the material. The

microelectronics community has adopted κ in contrast to the scientific community using εr.

Thermal stability is another important requirement for low k dielectrics. Low k materials must be able to withstand the temperature up to 670-720 K for several hours during Cu metallization and the following annealing step to ensure void free copper deposits [4]. New low k materials should also have a lot additional properties (Table 1-2) related to performance and reliability for use as an intralayer dielectric material [42, 43].

A low κ value relative to that of SiO2 (κ ~ 4) can be achieved by either compositional

changes (e.g. less polarizable chemical bonds), or reduction of density through change of

network structure (e.g. via addition of bulky terminal groups) or introduction of porosity [15].

Organosilicate glass (OSG, κ = 2.8-3.3) is one of the leading low-k materials for use as an intralayer dielectric [44]. The OSG schematic bonding structure is shown in Fig. 1.6. In OSG,

CH3 terminal groups that cannot network are introduced in the silicate glasses to decrease dielectric constant. Replacement of O with organic groups also lowers the ionic contributions to

3 κ. Typical densities of OSG are between 1.2 and 1.4 g/cm , much lower than that of SiO2 (2.1-

3 2.3 g/cm ) [15]. The κ value of OSG (κ = 2.8-3.3) depends on the number of CH3 groups

because they lower both polarity and density of the material by steric hindrance [44]. The surface

chemical state and structure of OSG are different from SiO2. Potential relatively poor adhesion between low k dielectric materials and barrier layers is one of the main issues when low k materials are integrated into a standard microfabrication process [15].

12 Table 1.2 Requirements for low dielectric constant intralayer-dielectric materials.

Low κ dielectrics properties Value

Dielectric Constant < 3

Thermal Stability >400°C

Adhesion (to metal, self-adhesion) Pass tape test after thermal cycles to 450°C

Breakdown Strength >2 MV.cm-1

Moisture Adsorption <1%

Etch Rate >3nm/s

Chemical Strong acid/base resistance, no solvent

adsoption

Gap Fill No voids at 0.35 µm, aspect ratio = 2

H CH3 Si O Si O O O

O Si Si CH3 CH O 3 O Si Si O H H C 3 O

Fig. 1.6 OSG schematic bonding structure.

13 1.6. Nucleation Modes Characterized by XPS

The thermodynamics of the growth mode of film on the substrate can be expressed by the

surface energy criteria [45], as a balance between the free energies of the film, substrate and

interface (γF, γS and γFS, respectively) :

∆γ = γF + γFS - γS. (1-3)

∆γ < 0 will favor the covering over the substrate by spreading of the condensate film (i.e. the

film “wets” the substrate), leading to the layer-by-layer growth (Frank - van der Merve, or FM)

Mode. If the lattice spacing of the film and substrate are different, there will be an increase of strain and the value of ∆γ with film thickness under epitaxial growth conditions. To release the strain, the growth mode transforms from layer-by-layer growth to islanding after the formation of the first (usually) 1-3 monolayers. This is known as Stranski - Krastanov, or SK mode. ∆γ ≥ 0 will favor the non-wetting or 3D islanding growth (Volmer - Weber, or VW) mode.

XPS can be used to characterize the growth mode of an overlayer on a substrate [46-48].

The change in metal film and substrate XPS intensities with deposition time, the so-called

“uptake” curve, is used to analyze the growth modes. Typical uptake curves for the three growth modes are shown in Fig. 1.7. The FM growth mode (Fig. 1.7a) typically has a plot with several breakpoints, with each breakpoint indicating the completion of a monolayer. The SK growth mode (Figure 1.7b) is characterized by one or two breaks, followed by level off at higher coverage. There are no further breaks when islanding occurs with the increase of metal deposition. For VW growth (Fig. 1.7c), the uptake curve increases linearly with no changes or

breaks in slope.

y y

t t

i i

s s n

n 1st 2nd 3rd 1st

e e

t t n

n layer layer layer layer

I I

S S

P P

X X

S S

/ /

M M

Deposition Time

Metal(M) Metal(M) Substrate(S) Substrate(S) Substrate(S)

(a) (b) (c)

Fig. 1.7 Intensity vs. time plots of (a) Frank−Van der Merwe (layer-by-layer);

(b) Stranski−Krastanov (monolayer then islanding) and (c) Volmer−Weber (islanding) growth.

1.7. Experimental Aspects

1.7.1. Apparatus

The combined XPS/PVD/CVD system used in this study is shown schematically in figure

1.8. The UHV main chamber (with a base pressure of 7 x 10-10 torr after bake out) is equipped

with a hemispherical analyzer (VG AX100), an unmonochromatized Mg/Al x-ray source (PHI)

for XPS and Ar+ sputter-cleaning capabilities (PHI). PVD/CVD chamber (with a base pressure of

1 x 10-8 torr after bake out) is isolated from the analysis chamber and sputtering target/sample distance is adjusted so that an appropriate deposition rates can be achieved.

15 XPS PVD Tower PVD(Magnetron PVD Sputter Tower #2 Deposition) e - Tower (backside) Be am

Turbo Sample Introduction Pump Introductionand Manipulation TPD r ke Capability ac Cr n CVD Precursor ge ro Introduction System yd CVD H RGA Precursor Introduction System

-10 -8 Main Chamber (7*10 ) torr PVD/CVD Chamber (1*10 ) torr Fig. 1.8 Combined XPS and DC magnetron PVD/CVD apparatus.

1.7.2. X-Ray Photoelectron Spectroscopy (XPS)

XPS is performed by analyzing the energy distribution of photoelectrons emitted from a solid surface as a result of X-ray excitation. The process of XPS emission is shown schematically in Fig. 1.9. Photons with sufficient energy strike the surface and are absorbed,

resulting in ionization and emission of core (inner shell) electrons. The kinetic energy (KE) of the electron is dependent on the photon energy of the X-ray employed and not an intrinsic material property. The binding energy of the electron (EB) is a parameter identifying the electron specifically in terms of its parent element and atomic energy level. The binding energy of the electron is given by [49]:

EB = hv - KE- ϕs, (1-4)

16 where hv is the energy of the incident X-rays and ϕs is the spectrometer work function (a constant for a given analyzer). As all these three parameters are known or measurable [49], the binding energy can be determined using Eq. (1-4).

Photoelectron EB = hv -KE-ϕs

Al Ka or Mg Ka

Fig. 1.9 The XPS emission process for a model atom.

X-ray photoelectron spectroscopy requires an X-ray source, electron energy analyzer, electron detector and a recorder. The choice of anode materials for the X-ray source must obey two rules [50]. First, the source must have high enough photon energy and intensity for sufficient core electrons to be photoejected from all elements. Second, the source must possess narrow X- ray line width that will not broaden the resultant spectrum excessively. Al Kα (1486.6eV), Mg Kα

(1253.6eV) or monochromatic Al Kα (1486.7eV) X-ray sources [49] are usually used due to their relatively narrow linewidths and high photon energies. Bombarding an Al or Mg anode material with high-energy electrons generates X-rays. The electrons are emitted from a thermal source,

17 usually in the form of an electrically heated tungsten filament. An electron energy analyzer measures the kinetic energy distribution of the emitted photoelectrons. A concentric hemispherical analyzer (CHA) (Fig. 1.10), otherwise known as hemispherical sector analyzer

(HSA) is currently the most popular analyzer. There is a gap between a pair of concentric hemispherical electrodes for the electrons to pass. The lens between the sample and analyzer retards the photoelectrons to produce sufficiently high resolution, and focuses photoelectrons emitted from the substrate onto the entrance slit to the analyzer. A potential difference is applied across the two hemispheres with the outer voltage being more negative than the inner one. The kinetic energy of the electrons that reach the detector is given by [50]:

E=ke∆V, (1-5) where e is the charge on the electron, ∆V is the potential difference between the hemispheres and

2 2 k is the spectrometer constant and equal to R1R2/(R2 -R1 ) (R1 and R2 are the radii of the inner and outer hemispheres, respectively). Fixed analyzer transmission (FAT), sometimes known as constant analyzer energy (CAE) is a popular operation mode for CHA [50]. In the FAT mode, absolute resolution (∆E) is constant and electrons are accelerated or retarded to a user-defined energy that the electrons possess as they pass through the analyzer (Pass Energy Ep). Only electrons with energy E= Ep +∆E can pass through the spectrometer. The energy resolution

(±∆E) is fixed across the entire spectrum by retarding the electrons to a constant kinetic energy.

Channel electron multipliers (channeltrons) are used to count the electrons arriving at the detector [50]. They consist of a glass tube with a collector at one end and an anode at the other.

The internal surface of the detector is coated with a material that will emit many secondary electrons when hit by an electron with an energy more than the threshold kinetic energy. When a large potential difference is applied across the channeltron, each electron arriving at the detector

18 will typically result in ~108 electrons reaching the anode. Channeltrons can detect up to 3*106 counts/s [50]. A computer based recorder plots the number of detected electrons versus the binding energy in the resulting spectrum.

Outer Inner hemisphere hemisphere

-V2

R2 R1 -V1

Multi-channel Detector Lens

e c r u o s y a r - X

Sample

Fig. 1.10 Schematic drawing of an X-ray photoelectron spectrometer.

For every element, there will be a characteristic binding energy associated with each core atomic orbital, resulting in a characteristic set of peaks in the photoelectron spectrum [49, 50].

The exact binding energy of an electron depends on the level from which photoemission is occurring, the formal oxidation state of the atom, and the local chemical and physical environment. Changes in any of these give rise to a small shift in the peak positions (chemical shift) in the spectrum. A higher binding energy of the core-level photoelectron is generally

19 observed for an atom of a higher positive oxidation state due to the extra coulombic interaction between the photo-emitted electron and the ion core [50]. Variations in the binding energies arising from differences in the chemical potential and polarizability of compounds can be used to identify the chemical state of the materials. XPS can also be used to identify and determine the concentration of the elements in the surface since each element has a unique set of binding energies [50].

XPS also provides a non-destructive analysis for compositional information as a function of depth. Angle resolved XPS (ARXPS) is shown in Fig. 1.11. The escape depth within the solid sample is proportional to λcosθ [50], where λ is the inelastic mean free path (IMFP) and θ is the angle between the sample surface normal and analyzer axis. As θ increases, the effective electron escape depth decreases and the analyzed region becomes more surface localized, leading to an increase of the surface sensitivity. There are two advantages for ARXPS. First, it can be applied to films that are too thin to be analyzed by conventional depth profiling techniques or polymers that are irretrievably damaged by such methods. Second, unlike sputtering, it is a non- destructive technique that can provide chemical state information. By comparing relative intensities at different θ, a species can be determined to be enriched or depleted in the surface region.

X-ray sources also excite Auger electrons. The Auger process for the KLL transition is shown schematically in Fig. 1.12. After the K level is ionized by the creation of a hole in the level K, the atom relaxes by filling the hole via a transition from higher level (with energy EL1).

As a result of the transition, the energy difference (EK-EL1) becomes available as excess kinetic energy. A fraction of this energy is given to another electron either in the same level or in a more shallow level (e.g. L23), whereupon the second electron is ejected. The kinetic energy (KE)

20 Analyzer A n Surface normal a ly ze hν e- hν e- r

Escape θ Mean free path Mean free path depth

Fig. 1.11 Angle resolved XPS.

of Auger electron is approximated by [49]:

KE = (EK – EL1) – EL3. (1-6)

The Auger electron KE is independent of the exciting radiation and characteristic of the element in the sample. Chemical bonding information can be obtained from Auger peak energy positions and lineshapes in some cases, although the greater complexity of the Auger peaks often complicates the extraction of chemical information.

Auger Emission

h v M, etc.

L*2,3

Fig. 1.12 Schematic of the Auger process of a model atom.

21 1.8. Chapter References

[1] Jeng, S.-P., Havemann, R. H. and Chang, M.-C., Mat. Res. Soc. Symp. Proc. 337 (1994)

25.

[2] http://nepp.nasa.gov/index_nasa.cfm/934, retrieved the information on July 25, 2004.

[3] Treichel, H. and Goonetilleke, C., Adv. Engin. Mater. 3 (2001) 461.

[4] Murarka, S. P., Mater. Sci. Eng. R 19 (1997) 87.

[5] Mcbrayer, J. D., Swanson, R. M. and Sigmon, T. W., J. Electrochem. Soc. 133 (1986)

1242.

[6] Wendt, H., Cerva, H., Lehmann, V. and Pamler, W., J. Appl. Phys. 65 (1989) 2402.

[7] Shacham-Diamond, Y. and Dubin, V. M., Microelctronic Eng. 33 (1997) 47.

[8] Chen, L., Magtoto, N. P., Ekstrom, B. and Kelber, J. A., Thin Solid Films 376 (2000)

115.

[9] Ekstrom, B. M., Lee, S., Magtoto, N. and Kelber, J. A., Appl. Surf. Sci. 171 (2001) 275.

[10] Shepherd, K. and Kelber, J., Appl. Surf. Sci. 151 (1999) 287.

[11] Chyan, O., Arunagirl, T. N. and Ponnuswamy, T., J. Electrochem. Soc. 150 (2003) C347.

[12] Kim, H., Kellock, A. J. and Rossnagel, S. M., J. Appl. Phys. 92 (2002) 7080.

[13] Wu, Y. Y., Kohn, A. and Eizenberg, M., J. Appl. Phys. 95 (2004) 6167.

[14] Lee, W.-H., Lin, J.-C. and Lee, C., Mater. Chem. Phys. 68 (2001) 266.

[15] Maex, K., Baklanov, M. R., Shamiryan, D., Lacopi, F., Brongersma, S. H. and

Yanovitskaya, Z. S., J. Appl. Phys. 93 (2003) 8793.

[16] Bohr, M. T., Solid State Technol. 9 (1996) 105.

[17] Grill, A., Diamond and Related Mater. 10 (2001) 234.

22 [18] Treichel, H., Ruhl, G., Ansmann, P., Wurl, R., Muller, C. and Dietlmeier, M.,

Microelectron. Eng. 40 (1998) 1.

[19] Strausser, Y. E., Scheibner, E. J. and Johannssen, J. S., Thin Solid Films 52 (1978) 203.

[20] Murarka, S. P. and Hymes, S. W., Crit. Rev. Solid State Mat. Sci. 20 (1995) 87.

[21] Murarka, S. P., Gutmann, R. J., Kaloyeros, A. E. and Lanford, W. A., Thin Solid Films

236 (1993) 257.

[22] Torres, J., Appl. Surf. Sci. 91 (1995) 112.

[23] Laurila, T., Zeng, K., Kivilahti, J. K., Molarius, J. and Suni, I., Superficies y Vacio 9

(1999) 199.

[24] Chiang, C., Tzeng, S., Raghavan, G., Villasol, R., Bai, G., Bohr, M., Fujimoto, H. and

Fraser, D. B., Proc. VMIC (1994) 414.

[25] Tompkins, H. G. and Tray, C., J. Electrochem. Soc. 136 (1989) 2331.

[26] Zeng, Y., Russell, S. W., McKerrow, A. J., Chen, L. and Alford, T. L., J. Vac. Sci.

Technol. B 18 (1999) 221.

[27] Kaloyeros, A. E. and Eisenbraun, E., Annu. Rev. Mater. Sci. 30 (2000) 363.

[28] Li, J., Shacham-Diamand, Y. and Mayer, J. W., Mater. Sci. Rep. 9 (1992) 1.

[29] Ting, C. Y. and Wittmer, M., Thin Solid Films 96 (1982) 327.

[30] Wang, S. Q., Mater. Res. Soc. Bull. 19 (1994) 30.

[31] Chang, C. and Uh, C., Appl. Phys. Lett. 57 (1990) 617.

[32] Massalski, T. B., Binary Phase Diagram. Westerville, OH: American Society For Metals

(1990).

[33] Holloway, K., Fryer, P. M., Cabral, C., Harper, J. M., Bailey, P. J. and Kelleher, K. H., J.

Appl. Phys. 71 (1992) 5433.

23 [34] Min, K., Chun, K. and Kim, K., J. Vac. Sci. Technol. B 14 (1996) 3263.

[35] Kuo, Y.-L., Huang, J.-J., Lin, S.-T., Lee, C. and Lee, W.-H., Mater. Chem. Phys. 80

(2003) 690.

[36] Nicolet, M.-A., Suni, I. and Finetti, M., Solid State Technol. 26 (1983) 129.

[37] Gottlieb, U., Laborde, O., Giauque, P. H., Nicolet, M.-A. and Madar, R., Microlectron.

Eng. 60 (2002) 107.

[38] Madhavaram, H., Idriss, H., Wendt, S., Kim, Y. D., Knapp, M., Over, H., Asman, J.,

Loffler, E. and Muhler, M., J. Catalysis 202 (2001) 296.

[39] Ganesan, P. G. and Eizenberg, M., Mat. Sci. Eng. 103 (2003) 213.

[40] Wang, Q., Ekerdt, J. G., Gay, D., Sun, Y.-M. and White, J. M., Appl. Phys. Lett. 84

(2004) 1380.

[41] Maier, G., Prog. Polym. Sci. 26 (2001) 3.

[42] Singer, P., Semicond. Int. 5 (1996) 88.

[43] Lee, W. W. and Ho, P. S., Mater. Res. Soc. Bull. 22 (1997) 19.

[44] Cheng, Y. L., Wang, Y. L., Wu, Y. L., Liu, C. P., Liu, C. W., Lan, J. K., O'Neil, M. L.,

Ay, C. and Feng, M. S., Thin Solid Films 447-448 (2004) 681.

[45] Zhang, L., Persaud, R. and Madey, T. E., Phys. Rev. B 56 (1997) 549.

[46] Stampanoi, M., Vaterlans, A., Aeschlimann, M. and Pescia, D., J. Appl. Phys. 64 (1988)

5321.

[47] Argile, S. and Rhead, G. E., Surf. Sci. Repts. 10 (1989) 277.

[48] Slaugther, J. M., Weber, W., Guntherodt, G. and Falco, C. M., MRS Bull. 17 (1992) 39.

[49] Moulder, J. F., Sticle, W. F., Sobol, P. E. and Bomben, K. D., Handbook of X-ray

Photoelectron Spectroscopy, Physical Electronics, Inc., Eden Prairie, Minnesota, 1995.

24 [50] Seah, M. P., In Auger and X-ray Photoelectron Spectroscopy, D. Briggs and M. P. Seah

(New York, 1990), Vol. 1, 245.

25 CHAPTER 2

COPPER INTERACTION WITH A TANTALUM SILICATE SURFACE: IMPLICATIONS

FOR INTERCONNECT TECHNOLOGY [1]*

2.1. Introduction

A critical issue in copper (Cu) interconnect technology is the wetting and adhesion of Cu to the barrier substrate. Cu will wet — grow conformally — on many barrier surfaces (e.g., tantalum (Ta) and tungsten (W)) under ultra-high vacuum (UHV) conditions [2-4]. Cu wettability, however, is very sensitive to oxygen contamination; even a partial monolayer of oxygen at the Cu/barrier interface will significantly degrade wettability and step coverage, and allow facile agglomeration at temperatures at or above 300 K [2-4]. The sensitivity of Cu wetting to low-level contamination is of practical importance for reliable processing since Cu deposition is typically not done under UHV conditions [5]. Tantalum silicon nitride (TaSiN) is an exception to such behavior. Recent X-ray photoelectron spectroscopy (XPS) studies [2] have demonstrated that sputter-deposition Cu will wet air-exposed TaSiN at 300K. This is in agreement with scanning electron microscopy (SEM) studies [6] reporting superior Cu thermal stability on TaSiN relative to Ta and tantalum nitride (TaN) for deposition under non-UHV conditions. XPS data of air-exposed TaSiN revealed a silicon (Si) (2p) binding energy of 102 eV

[2], consistent with metal silicate formation [7].

* Reproduced with permission from [Zhao, X., Magtoto, N. P., Leavy, M. and Kelber, J. A., Thin

Solid Films 415 (2002) 308]. Copyright [2004] Thin Solid Films.

26 Silicate thin films are of increasing interest as high dielectric materials for gate oxide [8,

9] and system-on-a-chip [10] applications. Although high dielectric materials in bulk form are suitable for interconnect applications due to capacitance-induced coupling between interconnects and resulting delays in signal propagation [11], the formation of an extremely thin silicate interface between Cu and the barrier substrate may not prove a hindrance. Evidence of this is the finding that TaSiN [12] is a potential diffusion barrier with good electrical properties.

In this study, we report the formation of Ta silicate and the Cu wetting behavior on this silicate film. The growth mode of a film on the substrate is governed by the surface free energies of the film (γF), substrate (γS) and film-substrate interface (γFS) according to Eq. (1-3) [13].

Wetting will occur for ∆γ<0. Since most surface free energies of metals (γF) are positive and larger than those of oxides (γS), γFS must be very large and negative for wetting to occur [13].

The data reported here demonstrate that Ta silicate films were formed by sputter deposition of Ta onto ultrathin silicon dioxide (SiO2) substrates at 300 K, followed by annealing in oxygen (O) (600 K, 2 torr). The XPS data of the films were characterized by an Si(2p) binding energy at 102.1 eV and Ta(4f7/2) binding energy at 26.2 eV, in agreement with findings for other silicate materials [2, 7]. Annealing Ta silicate films to > 900 K in UHV results in silicate decomposition to SiO2 and tantalum oxide (Ta2O5). Sputter deposition of Cu onto the unannealed Ta silicate substrate at 300 K results in the initial formation of Cu(I) and a linear increase in Cu signal vs. deposition time (uptake curve) until an average Cu thickness of ~1 Å is achieved. At this point, the formation of Cu(0) is observed, coincident with a change in the slope of the uptake curve. These findings are similar to those previously reported for air-exposed

TaSiN [2].

2.2. Experiment Details

Experiments were carried out in a UHV main chamber equipped with a hemispherical analyzer (VG AX100), an unmonochromatized Mg/Al x-ray source (PHI) for XPS, argon (Ar) sputter capabilities (PHI), and a base pressure of 7 x 10-10 torr. All XPS data reported here were acquired at 25 eV constant pass energy. XPS data analysis was carried out using commercially available software (ESCA Tools) that utilizes Gaussian-Lorentzian functions and Shirely background subtraction to synthesize peak components [14]. XPS spectra were acquired with the sample aligned normal to the analyzer lens axis (normal emission) and at 60° with respect to the normal emission (grazing emission). Atomic concentrations were calculated with atomic sensitivity factors specific for the hemispherical analyzer (VG100 AX) and were obtained directly from the manufacturer (VG Microtech). Relative atomic concentrations were derived from the XPS intensities according to [15]:

NA/NB = (IASB)/(IBSA), (2.1) where N, S and I are, respectively, the atomic concentrations, atomic sensitivity factors and XPS signal intensities.

The film thickness dA was determined by the XPS intensity ratio of the film to substrate

(IA/IB) [16] according to:

∞ ∞ -dA/(λAcosθ) -dA/(λBcosθ) IA/IB =( IA /IB )[(1-e )/e ], (2.2) where λA and λB are the photoelectron inelastic mean free path (IMFP) in the overlayer and substrate, respectively, dA is the thickness of the overlayer, and θ is the angle between

∞ ∞ photoelectron analyzer and sample surface normal. IA and IB are XPS intensities of infinitely

28 ∞ ∞ thick overlayer and substrate, respectively. IA and IB are usually unavailable, therefore

∞ sensitivity factors of the overlayer and substrate are normally used, which are proportional to IA

∞ and IB . Calculated [17, 18] IMFP value for Si(2p) electrons in SiO2 is 31.3 Å. For Cu overlayers, the Cu(2p3/2) intensity was monitored (λA= 19.8 Å). The calculated SiO2 and silicate thickness is thinner than the actual thickness due to the enhanced Si(2p) photoelectron emission of substrate along the Si(100) single crystal [16]. Previous study [19] has shown that the actual

SiO2 and silicate thickness is larger than those calculated by ~43%. Therefore, the actual SiO2 and silicate thickness obtained from multiplying the calculated thickness (Eq (2.2)) by 1.4, is reported in this study.

The UHV main chamber was attached to a dual magnetron (Maxteck) sputtering chamber

(base pressure 1 x 10-8 torr) capable of sputter depositing either Ta or Cu. Sputter deposition was carried out using a commercial water-cooled magnetron source (MiniMak) and Argon plasma with a partial pressure of 15 mtorr of Argon. Argon of 99.999% purity, tantalum target of 99.999% purity and cooper targets of 99.999% purity were used for sputter deposition. The deposition rate could be controlled by adjusting the plasma power, with a constant power of 50W for Ta and 25W for Cu. Ta deposition rates of 0.2 Å/sec and Cu deposition rates of 0.2 Å/min were achieved, allowing a study of the film/substrate interface. All Ta and Cu depositions reported in this paper were done at room temperature (~300K). Sample temperature in either chamber could be varied between 100 K and 1300 K by a combination of liquid nitrogen cooling and resistive heating of the sample holders. A chromel-alumel thermocouple between the sample and Ti holder was used to monitor the temperature. Sample transport between chambers was accomplished under UHV conditions.

29 A 1cm2 silicon sample cut from an n-doped Si (100) wafer was first cleaned successively in dilute HF solution, acetone and deionized water to remove surface oxide, and then was inserted to the UHV chamber. Adventitious carbon and oxygen were removed by heating the sample to 1100 K in UHV. Sample cleanliness was verified by XPS.

2.3. Results

2.3.1. Ta Silicate Formation

Ta silicate films were formed by a two-step process. First, a 6 Å (Eq. (2.2)) thick SiO2

9 film was grown by direct oxidation of a Si(100) substrate (1.7 x 10 L O2 at 2 torr, 600 K).

Second, 6 Å (Eq. (2.2)) of Ta were deposited on the SiO2 film at 300 K, and annealed at 600 K in

9 1.7 x 10 L O2 at 2 torr. The evolution of Si(2p) spectra during SiO2 formation and subsequent reaction with Ta is displayed in fig. 2.1a. Corresponding O(1s) spectra are displayed in Fig.

2.1b.

The formation of an SiO2 film is marked by a feature at 103.1 eV in the Si(2p) spectrum

(Fig. 2.1a), in good agreement with accepted binding energy values for SiO2 [15]. Analysis of the relative intensity of the Si103/Si99 relative intensity yields an average SiO2 film thickness of 6

Å (Eq. (2.2)). The corresponding O(1s) spectrum (Fig. 2.1b) has a full width at half-maximum

(FWHM) of 2.5 eV, significantly broader than the FWHM of 2.1 eV observed under these conditions for non-hydroxylated stoichiometric SiO2 films. Assigning one component (FWHM

= 2.1 eV) to the reported value of 532 eV for O in SiO2 [17], the O(1s) peak (Fig. 2.1b) is well fit by the addition of a second component (FWHM = 2.1 eV) at 533 eV, corresponding to Si-OH

[20], indicating that the ultrathin SiO2 film has observable hydroxyl content.

30 Deposition of Ta followed by annealing in oxidizing conditions results in changes to both Si(2p) and O(1s) spectra. A new feature at 102.1 eV is observed in the Si(2p) spectrum

(Fig. 2.1a). A corresponding spectrum, acquired at grazing emission (Fig. 2.1a) shows an enhanced intensity for this feature, indicating that the new feature is associated with the surface region of the film. The FWHM of the O(1s) spectrum (Fig. 2.1b) increases from 2.5 eV to 2.7 eV, and the peak maximum shifts from 532.2 eV to 531.1 eV. A Si(2p) feature at ~102.1 eV has been reported for aluminum silicate formation [7], and for stoichiometric zirconium silicate

(ZrSiO4)[21]. A comparison of O(1s) spectra for ZrSiO4 and SiO2 reveals a trend similar to that shown in Fig. 2.1b; a broader peak for the silicate than for SiO2 (2.3 eV vs 2.0 eV), and a lower silicate binding energy (531.1 eV vs 532.2 eV for SiO2) [21]. The similarity of Si(2p) spectra shown in Fig. 2.1a indicates that Ta reaction with the SiO2 film results in silicate formation. The large FWHM for the O(1s) spectrum (Fig. 2.1b) indicates two oxygen environments, and the peak is well fit by two components (FWHM = 2.1 eV) at 530.6 eV and 531.9 eV. Previous studies of silicate films [22] indicate that the component at 530.6 eV can be assigned to oxygens bridging between Ta and Si sites. The component at 531.9 eV is consistent with SiO2 [2, 20], and with metal hydroxide Ta-OH [2, 15]. An assignment to SiO2, however, is inconsistent with the fact that there is negligible Si(2p) intensity at 103 eV after reaction with Ta (Fig. 2.1a), and that the Si103/O531.9 intensity ratio corresponds to a Si/O atomic ratio of 1:10. The O(1s) component at 531.9 eV (Fig. 2.1b) is therefore assigned to hydroxylated Ta. A comparison of

O(1s) spectra acquired at normal and grazing emission indicates that the hydroxyl and bridging oxygen components is constant through the thin film (O531.9/O530.6 ~2/3). This may be due to the extremely thin nature of the film (6Å) or may indicate that the film is uniformly hydroxylated.

31 The Ta(4f) spectra acquired after Ta deposition and annealing are shown in Fig. 2.1c. The spectra include both the Ta(4f7/2) and the Ta(4f5/2) photoelectron lines, which are present in a 4:3

(a) SiOx(102.1 eV) Si(99.1 eV) silicate at grazing

SiO2 (103.1 eV) emission

silicate at normal emission

SiO2 at normal emission

108 104 100 96 92

(b) 530.6 eV silicate at grazing 532 eV emission 533 eV

silicate at normal emission XPS Intensity XPS SiO2 at normal emission

538 536 534 532 530 528 526

Silicate at grazing emission

Silicate at normal emission

34 32 30 28 26 24 22 20 18 16

Binding Energy (eV)

Fig. 2.1 Formation of SiO2 and Ta silicate. (a) Si(2p); (b) O(1s) and (c) Ta(4f) Spectra.

32 intensity ratio, with a 1.9 eV separation [15]. The binding energies reported in this paper refer to the Ta(4f7/2) photoelectron line. Ta spectra acquired at normal emission and grazing emission

(Fig. 2.1c) are well fit by a doublet with an individual component FWHM of 2.3 eV and a

Ta(4f7/2) binding energy of 26.2 eV. Similar results were reported [2] for air-exposed TaSiN and attributed to a homogeneous tantalum silicate (TaxSiyOz) mixture.

The Si(2p) at 102.1 eV, Ta(4f) at 26.2 eV and O(1s) at 530.6 eV are consistent with the formation of a Ta silicate film [2, 7, 21, 22]. The O531.9/O530.6 intensity ratios for normal emission and grazing emission (Fig. 2.1b) are both 2/3, indicating a film with uniform hydroxyl composition. A Ta/Si102 /O elemental ratio of 1:1:6 was obtained by Eq. (2.1) for the grown film, suggesting either a Ta-deficient silicate or a Ta oxide component, or possibly a hydroxylated silicate. The latter possibility is consistent with the O(1s) spectrum (Fig. 2.1b). The estimated thickness of the silicate film was ~8Å (Eq. (2.2)). This estimation is an upper bound to the thickness, since the electron inelastic mean free path (IMFP) in Ta silicate may be less than in

SiO2, due to heavy atom scattering [15, 16].

For Ta reaction with an ultrathin SiO2 substrate, a Si(2p) feature at ~ 102.1 eV might arguably be assigned to a Si-suboxide, due to Ta removal of the oxide overlayer, revealing the

Si/SiO2 interface region. In order to explore this possibility, similar experiments were carried out on ultrathin SiO2 films, but with annealing to 600 K in the absence of O2. Negligible growth in Si(2p) intensity at ~102.1 eV and Ta(4f) intensity at ~26.2 eV were observed. Subsequent anneals in 2 torr O2, however, immediately resulted in a significant increase of Si(2p) intensity at 102.1 eV and Ta(4f) intensity at 26.2 eV, indicating that these features are formed only under oxidizing conditions. The features of Si(2p) at 102.1 eV and Ta(4f) at 26.2 eV are therefore assigned to a Ta silicate film.

33 2.3.2. Thermal Decomposition

The silicate film was annealed in UHV in 100 K increments starting from 300 K.

Following each anneal period (30min), the sample was cooled to room temperature and XPS spectra were acquired. Annealing of the silicate film up to 900 K results in a significant increase of the Si103/Si99 XPS intensity ratio, and a decrease of the Si102.1/Si99 intensity ratio (Fig. 2.2).

Si(2p) spectra of the sample acquired at 300K and after annealing to 900K are shown in Fig.

2.3a. The Si(2p) spectrum shifts from 102.1 eV to 103 eV, indicating the formation of SiO2 [15].

Additionally, the Ta(4f7/2) spectra at 300K and annealing to 900K(Fig. 2.3b) were observed to shift from 26.2 eV to 26.7 eV. The feature of Si(2p) at 103 eV is assigned to Si in SiO2 [15] and

Ta(4f) at 26.7 eV is assigned to Ta in Ta2O5 [15]. These changes are consistent with silicate decomposition to form SiO2 and Ta2O5, in accord with the published phase diagram for this system [23, 24].

0.4

0.35 o

t a 0.3 R Si /Si

102.1 99

i 0.25

t 0.2

S 0.15

XP 0.1 Si103/Si99

0.05

0 0 200 400 600 800 1000 1200 Temperature (K)

Fig. 2.2 XPS intensity ratio as a function of annealing temperature.

34 2.3.3. Air-Exposure of Ta Silicate Film

A Ta silicate sample formed as described above was exposed to air for 2 hours, then transferred to the UHV main chamber for XPS analysis. Si(2p) spectra of the sample before and after air-exposure are shown in Fig. 2.4a. No shift for Si(2p) spectra is observed upon air exposure. Similarly, air-exposure induces no shift for O(1s)(Fig. 2.4b) and Ta(4f)(Fig. 2.4c) spectra. However, the intensities of Si(2p)(Fig. 2.4a), O(1s)(Fig. 2.4b) and Ta(4f) (Fig. 2.4c) decrease after air-exposure. C(1s) intensity (not shown) at 284.5 eV increases after air- exposition. The component at 284.5 eV is assigned to adventitious C [15]. This is in accord with the decrease of the Si(2p), O(1s) and Ta(4f) intensities of the Ta silicate film. This result suggests that the silicate surface is stable in the air and can act as a “robust” diffusion barrier since Cu deposition is typically not done under UHV conditions [4].

2.3.4. Cu/Film Interactions

Cu was sputter deposited onto an unannealed Ta silicate film in sequential depositions at

300K. After each deposition, the sample was transferred from the deposition chamber back to the main chamber for XPS analysis. The Cu(L3VV) Auger line shape (X-ray excited) as a function of deposition time, displayed in Fig. 2.5, provides a “fingerprint” of the oxidation state of the deposited Cu on the silicate film [2, 25]. The feature at 914.9 eV Auger kinetic energy is assigned to Cu(I), whereas the feature at 917.6 eV is assigned to Cu(0) [26, 27]. The oxidation state of the Cu evolves with deposition time from Cu(I) to Cu(0). Cu depositions at short (1– 4.5 minutes) deposition times (Fig. 2.5) yield a Cu(L3VV) Auger feature at 914.9 eV, indicative of formation of Cu(I). At longer deposition times (Fig. 2.5), a new component at 917.6 eV was

(a) Si(99.1 eV) Si (102.1eV) Annealed at Si 900K for (103eV) 30 min

Room temperature

n 108 104 100 96 92

S (b)

XP 26.7 eV 26.2 eV

Annealed at 900K for 30 min

Room temperature

34 32 30 28 26 24 22 20 18 Binding Energy (eV)

Fig. 2.3 XPS spectra of Ta silicate at room temperature and after annealing at 900 K for 30 min in UHV. (a) Si(2p) Spectra and (b) Ta(4f) Spectra.

(a) Si(2p) After air exposure

Before air exposure

108 104 100 96 92 (b) O(1s) After air exposure

Before air exposure

536 534 532 530 528 526

Before air exposure

34 32 30 28 26 24 22 20 18 Binding Energy (eV)

Fig. 2.4 XPS spectra of Ta silicate before and after air exposure. (a) Si(2p) spectra; (b) O(1s) spectra and (c) Ta(4f) spectra.

37 observed, indicative of formation of Cu(0). The Cu Auger parameter (AP) was calculated according to the following [28, 29]:

AP = KE(CuL3VV)+ EB (Cu2p), (2.3)

KE(CuL3VV) = hν– EB(CuL3VV), (2.4) where KE and EB are kinetic energy and binding energy, respectively. The calculated Auger parameter associated with the Cu(I) spectral feature is 1848.2 eV(Fig. 2.5). An Auger parameter value of 1848.2 eV is in the range of Auger parameters reported for other Cu(I) compounds [15].

The Auger parameter associated with Cu(0) (Fig. 2.5) is 1850.5 eV, lower than that reported for bulk Cu that is 1851.3 eV. Similar results were reported for Cu growth on oxidized TaSiN [2],

Cu0.6Al0.4 growth on SiO2 [30] and Cu growth on Al2O3 [31], and were attributed to the ineffective screening of the holes in the Auger final state for small Cu particles [2, 15, 30, 31].

The growth mode of the sputter deposited Cu on the substrate can be characterized by plotting the increase in the Cu(2p3/2) intensity (relative to Si(2p)) ratio as a function of deposition time [32]. Fig. 2.6a and 2.6b includes the growth mode of sputter deposited Cu on the unannealed and decomposed (Annealed to 1000K) Ta silicate film respectively. For unannealed

Ta silicate film, the Cu(2p3/2)/Si(2p) intensity ratio increases linearly with deposition time, but exhibits a change in slope at 4.5 min deposition time (Fig. 2.6a). This change in slope coincides with termination of Cu(I) formation and the initiation of Cu(0) growth (Fig. 2.5). Such behavior indicates conformal growth of a Cu(I) ad-layer, followed by formation of a Cu(0) layer [33]. For

Cu depositions on the annealed and decomposed Ta silicate film, the linearity without change in slope observed in Fig. 2.6b indicates that copper deposition occurs with the formation of 3-D islands consistent with the behavior of Cu on SiO2 [2]. The X-ray excited Auger data for

38 Cu on the annealed, decomposed film (not shown) show little Cu(I) formation, indicating negligible Cu/surface charge transfer.

Cu(I) Cu(0)

VV) XPS Intensity 11.5 3 9.5 Cu(L 8.5 5.5 4.5

1.0 Deposition Time (min)

905 910 915 920 925

Kinetic Energy (eV)

Fig. 2.5 X-ray excited Cu(L3VV) Auger spectral evolution as a function of Cu deposition on the unannealed silicate film.

2.5 (a) Before Silicate Decomposition Cu(0) 2.0 2nd layer

1.5 Cu(I)

1.0 1st layer )/Si(2p) XPS Intensity Ratio XPS Intensity )/Si(2p)

3/2 0.5 Cu(0)

Cu(2p (b) After Silicate Decomposition 0.0 0 2 4 6 8 10 12 14 Cu Deposition Time (min)

Fig. 2.6 Cu(2p)/Si(99) XPS intensity ratio vs. deposition time (a) before and (b) after annealing the silicate film at 1000K.

The data in Figures 2.5 and 2.6a demonstrate that initially deposited copper reacts with the Ta silicate film to form Cu(I) at 300K. At higher coverage, Cu(0) formation is observed.

Figs. 2.5 and 2.6a indicate that formation of Cu(0) coincides with the change in slope in the uptake curve. This indicates SK (conformal) growth. Cu “wets” the silicate surface. The deposited Cu grows conformally as Cu(I) to a maximum average thickness of ~1Å (Eq. (2.2)).

This corresponds to a surface coverage of ~0.5 monolayers, assuming a Cu+ ion diameter of

1.92Å [34]. Similar effects have been observed for Cu (0.5 ML) on polycrystalline aluminum oxide surface [35], Cu (~0.35ML) on hydroxylated α–Al2O3(0001) [30] and Cu (~0.4 ML) on air-exposed TaSiN samples [2]. In these systems, the deposited Cu exhibits conformal growth

[33], with the initial formation of a Cu(I) ad-layer, followed by growth of Cu(0) over the Cu(I)

40 interface. It is not known whether maximum Cu(I) coverage is limited by Cu(I)-Cu(I) repulsive interactions, or by the surface concentration of appropriate active sites.

The Ta(4f7/2) XPS spectra are shown in Fig. 2.7, acquired after Cu depositions for 1 min and 5.5 min respectively. The Ta(4f7/2) spectrum shifts from 27.2 eV to 26.7 eV with Cu depositions from 1 min to 5.5 min, indicating that Ta is reduced as the deposited Cu metal is oxidized to Cu(I). No changes were obtained for corresponding Si(2p) and O(1s) spectra (not shown).

27.2 eV 26.7 eV

Cu deposition for 5.5 min

ity

P Cu deposition X for 1 min

34 32 30 28 26 24 22 20 18 Binding Energy (eV)

Fig. 2.7 Ta(4f) XPS spectra at Cu deposition for 1min and 5.5 min.

41 2.4. Discussion

The XPS data (Fig. 2.1) indicate the formation of an 8Å thick film of uniform composition with the stoichiometry TaSiO6. The Si(2p) intensity at 102.1 eV indicates that the Si atoms are bonded to oxygen atoms bridging between the Si and metal, as expected for a silicate structure [7, 21]. The O(1s) spectrum, however, indicates at least two oxygen chemical environments within the film, at 531.9 eV and at 530.6 eV, with an intensity ratio of ~2:3. A binding energy of 531.9 eV could be assigned to SiO2, but this contradicts the Si(2p) data which indicates a main component at binding energy of 102.1 eV and a negligible component at binding energy at 103.1 eV. A possible assignment of O(1s) spectrum is to assign the 531.9 eV feature to OH groups located (presumably) on the Ta sites. Other evidence suggesting that this is a hydroxalated structure derives from the fact that the film is stable to air exposure (Figs. 2.4a,

2.4b and 2.4c), with the only change (other than adventitious carbon) resulting from a slight increase in the relative O(1s) intensity near 532 eV (Fig. 2.4b). On this basis, a structure such as displayed below is tentatively proposed:

O OH

Si Ta

The above structure is consistent with two oxygen environments, and with the formation of a formal Ta(V) species. Such a formal structure, however, would yield a stoichiometry of TaSiO5 and an O531.9/O530.6 intensity ratio of only 1:4 instead of the observed 2:3. In a film only 8 Å thick, however, considerable non-stoichiometry may be expected due to the high proportion of interfacial sites between the film and the Si substrate. In addition, the choice of Ta as a

42 hydroxylation site is reasonable but purely conjectural. It is not known, for example, whether OH bonded to Si would yield an O(1s) binding energy distinguishable from OH/Ta under existing experimental conditions. Obviously, a complete determination of film structure requires additional measurements (e.g., FTIR).

The x-ray excited Cu(L3VV) Auger spectra (Fig. 2.5) and uptake curve (Fig. 2.6a) demonstrate that copper reacts with Ta silicate film in the first layer and forms Cu(I) at 300K to a maximum coverage of 0.5 ML. At higher coverage, Cu(0) is formed. The behavior of Cu on the

Ta silicate film can be compared with Cu on the decomposed Ta silicate film, a mixture of SiO2 and Ta2O5 (Fig. 6b). Copper interacts only weakly with the SiO2 and Ta2O5 surface, forming

Cu(0) even at low coverage [2]. The copper uptake curve (Fig. 2.6b) for the decomposed film does not display the change in slope that is characteristic of layer-by-layer growth [2, 33]. In contrast to Cu on the decomposed film, Cu on the unannealed Ta silicate film displays initial conformal growth with formation of Cu(I). This demonstrates significant charge exchange between the initial Cu atoms and the Ta silicate surface. This could be the result of the more ionic Ta cation and more covalent ionic Si formation on the Ta silicate surface [21, 34]. There is a strong interaction between initially deposited Cu and Ta silicate surface. Cu “wets” the surface because of the significant charge exchange between Cu and Ta silicate film.

2.5. Conclusions

9 A 6Å SiO2 film is formed by exposing clean Si(100) to 1.7 x 10 L O2 at 1000K. An 8Å

Ta silicate film is formed after Ta deposition and oxidation in 2 torr oxygen at 600K. The Si(2p) component at binding energy of 102.1 eV ,Ta(4f7/2) component at binding energy of 26.2 eV and

43 O(1s) at binding energy of 530.6 eV demonstrate a Ta silicate environment. The Ta silicate film is stable up to 900K in UHV and stable at room temperature in the air. These data indicate Ta silicate film can act as a “robust” diffusion barrier since Cu deposition is typically done under non-UHV environment. Ta silicate film decomposed to form SiO2 and Ta2O5 ~900K.

Copper was sputter deposited onto silicate film before and after annealing the silicate sample ~1000K. Copper grows conformally on the unannealed Ta silicate surface and is characterized by a feature in the Cu(L3VV) Auger line shape at 914.9 eV indicative of Cu(I).

Further deposition of copper results in the Cu(L3VV) Auger line shape at 917.6 eV indicative of

Cu(0). These data indicate that the initially deposited copper reacts with the Ta silicate surface and forms Cu(I) for the first ad-layer. Subsequent copper deposition results in Cu(0) formation.

2.6. Chapter References

[1] Zhao, X., Magtoto, N. P., Leavy, M. and Kelber, J. A., Thin Solid Films 415 (2002) 308.

[2] Shepherd, K. and Kelber, J., Appl. Surf. Sci. 151 (1999) 287.

[3] Chen, L., Magtoto, N. P., Ekstrom, B. and Kelber, J. A., Thin Solid Films 376 (2000)

115.

[4] Ekstrom, B. M., Lee, S., Magtoto, N. and Kelber, J. A., Appl. Surf. Sci. 171 (2001) 275.

[5] Murarka, S. P., Mater. Sci. Eng. R 19 (1997) 87.

[6] Jiang, Q. T., Faust, R., Lam, H. and Mucha, J., Proceedings of the 1999 International

Interconnect Technology Conference (IEEE), New York, USA (2000) 125.

[7] Klein, T. M., Niu, D., Epling, W. S., Li, W., Maher, D. M., Hobbs, C. C., Hedge, R. I.,

Baumvol, I. J. R. and Parsons, G. N., Appl. Phys. Lett. 75 (1999) 4001.

[8] Wilk, G. D. and Wallace, R. M., Appl. Phys. Lett. 74 (1999) 2854.

44 [9] Wilk, G. D. and Wallace, R. M., Appl. Phys. Lett. 76 (2000) 112.

[10] Gill, P., Miller, M. and Nguyen, B. Y., Microelectron. Eng. 56 (2001) 169.

[11] Grill, A., Diamond Related Mater. 10 (2001) 234.

[12] Angyal, M. S., Shacham-Diamand, Y., Reid, J. S. and Nicollet, M.-A., Appl. Phys. Lett.

67 (1995) 2152.

[13] Zhang, L., Persaud, R. and Madey, T. E., Phys. Rev. B 56 (1997) 549.

[14] ESCA Tools, Surface/Interface, Mountain View, CA.

[15] Moulder, J. F., Stickle, W. F., Sobol, P. E., Bomben, K. D., Chastain, J. and King, R. C.,

Handbook of X-Ray Photoelectron Spectroscopy, Physical Electronics, Eden Prairie,

MN.

[16] Seah, M. P., Pratical Surface Analysis, in: D. Briggs, M.P. Seah (Eds), Auger and X-ray

Photoelectron Spectroscopy 1, Second Ed., Wiley, New York (1990) 201.

[17] Lu, Z. H., McCaffery, J. P., Brar, B., Wilk, G. D., Wallace, R. M., Feldman, L. C. and

Tay, S. P., Appl. Phys. Lett. 71 (1997) 2764.

[18] Powell, C. J., Jablonski, A., Tilinin, I. S., Tanuma, S. and Penn., D. R., J. Electron

Spectrosc. Relat. Phenom. 98 (1999) 1.

[19] Garza, M., Liu, J., Magtoto, N. P. and Kelber, J. A., Appl. Surf. Sci. 222 (2004) 253.

[10] Miller, M. L. and Linton, R. W., Anal. Chem. 57 (1985) 2314.

[21] Guittet, M. J., Crocombette, J. P. and Salim, M., Phys. Rev. B 63 (2001) 125117.

[22] Mekki, A., Holland, D., McConville, C. F. and Salim, M., J. Non-Cryst. Solids 208

(1996) 267.

[23] Beyers, R., J. Appl. Phys. 56 (1984) 147.

45 [24] Alers, G. B., Werder, D. J., Chabal, Y., Lu, H. C., Gusev, E. P., Gustafsson, T. and

Urdahl, R. S., Appl. Phys. Lett. 73 (1998) 1517.

[25] Nuesca, G. M. and Kelber, J. A., Thin Solid Films 262 (1995) 224.

[26] Klein, J. C., Proctor, A., Hercules, D. M. and Black, J. F., Anal. Chem. 55 (1983) 2205.

[27] Kowalczky, S. P., Pollak, R. A., Mcfeely, F. R., Ley, L. and Shirley, D. A., Phys. Rev. B

8 (1973) 2387.

[28] Wagner, C. D., Gale, L. H. and Raymond, R. H., Anal. Chem. 151 (1979) 466.

[29] Garenstroom, S. W. and Winograd, N., J. Chem. Phys. 67 (1977) 3500.

[30] Shepherd, K., Niu, C., Martini, D. and Kelber, J. A., Appl. Surf. Sci. 158 (2000) 1.

[31] Niu, C., Shepherd, K., Martini, D., Tong, J., Kelber, J. A., Jennison, D. R. and Bogicevic,

A., Surf. Sci. 465 (2000) 163.

[32] Stampanoi, M., Vaterlans, A., Aeschlimann, M. and Pescia, D., J. Appl. Phys. 64 (1988)

5321.

[33] Peden, C. H. F., Kidd, K. B. and Shinn, N. D., J. Vac. Sci. Technol. A9 (1991) 1518.

[34] Huheey, J. E., Keiter, E. A. and Keiter, R. L., Inorganic Chemistry, Principles of

Structure and Reactivity, Fourth Ed., Harper Collins, New York (1993) 114.

[35] Ohuchi, C. F. S., French, R. H. and Kasowski, R. V., J. Appl. Phys. 62 (1987) 2286.

46 CHAPTER 3

CHEMICAL VAPOR DEPOSITION OF TANTALUM NITRIDE WITH TERTT-

BUTYLIMINO TRIS(DIETHYLAMINO) TANTALUM AND ATOMIC HYDROGEN

3.1. Introduction

Due to the high melting point, chemical and thermal stability, and excellent conductivity, refractory metal nitrides are widely recognized as diffusion barriers in metal-semiconductor interconnects. Among these metal nitrides, Tantalum nitride (TaN) is one of the most extensively studied materials for silicon (Si) device fabrication [1, 2]. TaN films are usually deposited using physical vapor deposition (PVD), such as reactive sputtering [1, 3-7]. PVD, however, is not expected to be applied beyond 45 nm technology nodes due to the poor step coverage caused by the shadowing effect in the small feature size, high aspect ratio contact and via holes [8].

Accordingly, chemical vapor deposition (CVD) and atomic layer deposition (ALD) processes are being developed for deposition of transition metal nitride thin films with high-quality step coverage.

Only a few results have been reported on TaN CVD films since there are only a few tantalum (Ta) source gases that have a high vapor pressure convenient for CVD processing. Ta halides such as tantalum chloride (TaCl5) [2] and TaF5 [9, 10] have been used as Ta sources, but these can result in the incorporation of chlorine (Cl) and fluorine (F) impurities in the growing films [11]. In order to avoid problems associated with the presence of halides, attention has shifted to halide-free precursors, including tert-butylimino tris(dimethylamino) tantalum

(TBTDET) (Fig. 3.1) [12-14]. TBTDET is liquid at the room temperature and has a high vapor

47 pressure (0.1 torr at 363 K). Ta nitride films with good step coverage and electrical properties have been deposited on Si and SiO2 by TBTDET CVD at ~720K – 920K [12, 13, 15]. This temperature range is relatively high for most low-κ materials. Carbon-containing low κ materials are often thermally unstable above ~670K [16]. Ta nitride was formed on SiO2 by TBTDET and hydrogen radicals at low temperature of ~530K [14], but there are few reports about TaN CVD deposition on low κ materials at low temperature. Additionally, previous studies [17] of PVD Ta on silicon(Si):oxygen(O):carbon(C) low-k substrates indicate that sputter deposition results in a relatively diffuse (~ 50 Å) interfacial layer containing tantalum carbide (TaC). PVD Cu will not wet this TaC interfacial layer at 300 K, thus imposing a lower limit on the thickness of reliable

PVD Ta barriers for PVD Cu seed. tBu N

Ta NEt Et2N 2 NEt2

Fig. 3.1 Tert-butylimino tris(diethylamino) tantalum (TBTDET) structure.

In this study, XPS has been used to characterize the reaction of TBTDET with atomic hydrogen on silicon dioxide (SiO2) and organosilicate glass (OSG) substrates. TBTDET partially

48 reacts with O atoms in both SiO2 and OSG substrates by 300 K to form strong Ta-O interfacial bonds. Subsequent bombardment with atomic hydrogen at 500K results in formation of a stoichiometric TaN phase on top of the Ta-O interfacial phase (“interphase”). Subsequent depositions of the precursor on the reacted layer on SiO2 and OSG, followed by atomic hydrogen bombardment, result in increased TaN formation. The data summarized here indicate that

TBTDET is a potential precursor for an atomic hydrogen driven ALD/CVD process.

3.2. Experiment Details

Experiments were carried out in a combined UHV analysis/dual magnetron sputter deposition system, which has been described previously in chapter 2. For the charging shift of apparent photoelectron energies on insulating samples, XPS spectra were calibrated by referencing the peak from adventitious carbon at 284.5 eV [18]. Relative film composition was derived from XPS intensities according to Eq. (2.1). The film thickness d was determined by the attenuation of the Si(2p) substrate signal (photoelectron inelastic mean free path (IMFP) for

Si(2p) electrons (λ) = 31.3 Ǻ), according to Eq. (2.2).

Ta precursor exposures were carried out in the PVD/CVD chamber (with a base pressure of 1 x 10-8 torr after bake out), followed by sample transfer to and XPS analysis in the ultra high vacuum (UHV) chamber. TBTDET (Schumacher) was used as Ta precursor. The liquid precursor was contained in a bubbler heated to 343K. TBTDET was condensed onto a clean SiO2 or OSG sample at 120K, annealed to 300K, and then exposed to hydrogen (H/H2) flux. Atomic H was generated within the UHV chamber by passing H2 through a TC-50 thermal gas cracker

(Oxford Applied Research), which consisted of a heated tungsten tube and external cooling

49 system. In this study, atomic hydrogen was generated at a constant power of 55 W and caused a

–7 slight rise in the sample temperature to 500K. Hydrogen pressure was 2*10 torr during H2/H bombardment. According to manufacturer specifications, such operating conditions should result in ~50% dissociation of H2. That significant H2 dissociation occurred was demonstrated by the fact that results obtained using the thermal gas cracker differed sharply from those observed for exposure to pure H2 at 500K, which had no significant effect.

2 The 1cm samples consisted of either a 4000Å SiO2 on Si substrate, or commercially prepared OSG film grown on Si. Adventitious carbon was removed by heating the sample to

1000 K for SiO2 and 500K for OSG in the UHV chamber. Sample cleanliness was verified by

XPS.

3.3. Results

3.3.1. TaN Formation on SiO2

TBTDET precursor was condensed on SiO2 at 120 K in the PVD/CVD chamber. After the dose, the sample was transferred to the UHV chamber for XPS analysis at 120K. The sample was then annealed to 300K in UHV, followed by exposure to an H/H2 flux for 90 mins

(involving an incidental temperature increase to 500K). XPS analyses were obtained after each treatment. The Si(2p) spectrum (Fig. 3.2a) of a clean SiO2 substrate is well fit by a single component with a FWHM of 2.1eV at binding energy of 103.2eV. The corresponding oxygen

(O) (1s) spectrum (Fig. 3.2b) is similarly well fit with a single component (FWHM of 2.1 eV) at binding energy of 532.1 eV [19]. After exposure to TBTDET precursor at 120 K, the absence of

50 Si(2p) and O(1s) intensity from the substrate (Fig. 3.2a and 3.2b) indicates that precursor dosing at 120 K results in a thick multilayer. The adsorbed multilayer on SiO2 exhibits significant carbon (C) (1s) intensity at ~285 eV (Fig. 3.2e), N(1s) intensity at ~398eV (Fig. 2d) and negligible Ta(4f) intensity (Fig. 3.2c). The XPS spectra acquired after annealing to 300K are shown in Fig. 3.2. The Ta(4f) spectra (Fig. 3.2c) include both the Ta(4f7/2) and the Ta(4f5/2) photoelectron lines, which are present in a 4:3 intensity ratio, with a 1.9 eV separation [18]. The binding energies reported in this paper refer to the Ta(4f7/2) photoelectron line. By 300K, considerable precursor interaction with SiO2 substrate has occurred and resulted in Ta-O bond formation, characterized by Ta(4f7/2) peaks at 24.7 eV and 25.8 eV for Ta suboxides [4, 20, 21], and the corresponding O(1s) peak at 529.8 eV [18]. The existence of the nitrogen (N) (1s) 398 eV feature (Fig. 3.2d) and C(1s) 285 eV peak (Fig. 3.2e) at 300K indicates that some precursor species still remained in the film. The total average film thickness determined by Eq. (2.2), was

101Ǻ. Interaction with atomic hydrogen resulted in the formation of a new Ta(4f7/2) feature at

23.2 eV binding energy (top of Fig. 3.2c), which indicates the formation of Ta nitride [4]. The formation of Ta nitride is corroborated by the emergence of a new N(1s) feature at 396.6 eV (top of Fig. 3.2d) [4]. The Ta(4p3/2) feature overlaps this portion of the N(1s) spectrum, with a broad

Ta(4p3/2) peak at binding energy of 403-404 eV [1, 22]. According to the intensities of the Ta(4f) and N(1s) features at 23.2 eV and 396.6 eV, Ta:N atomic ratio, determined by Eq. (2.1), is 1:1, indicating a stoichiometric phase: TaN. The atomic hydrogen flux induced more Ta/oxide reactions as evidenced by enhanced Ta(4f7/2) and O(1s) intensities at 24.7 eV and 529.8 eV, respectively (sub-oxide) and the appearance of new Ta(4f7/2) feature at 26.5 eV (Ta2O5) [18].

The formation of a new Si(2p) feature at 102.1 eV indicates that SiO2 is partially reduced as the deposited Ta is oxidized. The atomic hydrogen flux induced the shifting of C(1s) feature to

exposure exposure Ta-O exposure

)

) )

s s

t t

i i

n n

U U

U 300K

300K 300K

y y

r r

r annealing

a a

a annealing annealing

r r

t t

i i

b b

r r

A A

(

( (

TBTDET TBTDET

y y

t t

t TBTDET

i i

i dose dose s dose dose s dose

s dose

n n n

e e e

t t t

n n n

I I

I SiO

2 SiO2 SiO2

S S S

P P P

X X X

108 104 100 96 536532 528 524 36 32 28 24 20 16 Binding Energy (eV)

(d) N(1s) (e) C(1s) Ta(4p3/2) H2/H Reaction of

TaN H2/H Adventitious

) )

exposure ) C and O

) ) )

s s

s exposure C

t t t

s s s

i i i

t t t

i i i

n n n

U U U

U U

U Precursor

y y

y Precursor

r r r

y y y

r r r

a a

300K a 300K

r r r

a a a

t t t

r r r

i i i

t t t

i i

i annealing annealing

b b b

r r r

b b b

r r r

A A A

( ( (

A A A

( ( (

y y y

t t t

y y y

i i i

t t t

i i i

s s

TBTDET s TBTDET

s s s

n n n

e e

dose e dose t t t

e e e

t t t

n n n

I I I

n n n

I I I

S S S

P P P

X X X

X X X SiO2 SiO2

412408 404 400 396 392 304300 296 292 288284 280 Binding Energy (eV) Binding Energy (eV)

Fig. 3.2 XPS spectral changes upon reaction with TBTDET precursor. (Bottom traces) SiO2 sample before precursor exposure; (bottom middle traces) precursor adsorption at 120 K; (top middle traces) after annealing to 300K; (top traces) after H/H2 exposure. (a) Si(2p); (b) O(1s); (c)

Ta(4f); (d) N(1s) and (e) C(1s).

higher binding energy at 295.5 eV (top of Fig. 3.2e), suggesting reactions with electron- withdrawing groups, such as oxygen [18, 23]. The binding energies of all C(1s) peaks (Fig. 3.2e) are above 284 eV, indicating no Ta carbide in the film. The later would yield C(1s) spectra near

281.9 eV [18, 24].

Formation of TaN upon exposure to H/H2 at 500 K might arguably be due to the rise in temperature to 500K. In order to explore this possibility, a sequential experiment was carried out as previously described, but with annealing to 500K in the absence of atomic hydrogen. Ta(4f) and N(1s) XPS spectra (Fig. 3.3a and 3.3b respectively) show that annealing the adsorbed precursor to 500 K without H/H2 induced more TaOx formation at binding energy of 24.7 eV and negligible N change, indicating no TaN formation. Subsequent exposure to H/H2 at 500K, however, immediately resulted in a significant Ta(4f7/2) feature at 23.2 eV and N(1s) intensity at

396.6 eV. It was reported that Ta nitride was formed by metal organic precursors and H2 only at high temperature above ~650K [25]. The changes observed for Ta(4f) (Fig. 3.3a) and N(1s) (Fig.

3.3b) are therefore due specifically to interaction with the atomic hydrogen flux, rather than the rise in temperature or H2 flux.

The evolution of Ta(4f) and N(1s) spectra upon repeated precursor physisorption/warmup/H/H2 exposure cycles are displayed in Fig. 3.4. The increase in relative intensities of the Ta(4f) feature at 23.2 eV and N(1s) peak at 396.6 eV corresponding to TaN [4] is evident. Total average film thickness determined by Eq. (2.2), is 36Ǻ after the first cycle (Fig.

3.4a). After the second cycle, no Si(2p) intensity could be observed, indicating that the total average thickness was >> 31.3 Ǻ - more like > ~100 Ǻ (assuming a Si(2p) photoelectron mean free path of ~31.3 Ǻ). The ratio of Ta intensity in TaN to total Ta intensity (TaTaN/TaTotal)

53 increases from 0.09±0.01 for 36Ǻ film (first cycle) to 0.32±0.02 for the thick film (second cycle), indicating the initial formation of Ta oxide at the overlayer/SiO2 interface, followed by enhanced TaN during the second cycle.

(a) TaOx Ta(4f)(b) N(1s) Ta O H2/H 2 5 TaN H2/H exposure, exposure, TaN Ta(4p ) 500K 500K 3/2

500K 500K Precursor

n annealing annealing

u t

o n

t 300K 300K

t annealing annealing

5 0

2 0

7 3632 28 24 20412408 404400 396 392 388 Binding Energy (eV) Binding Energy (eV)

Fig. 3.3 XPS spectra change after annealing without and with subsequent H2/H flux exposure on

SiO2 sample. (a) Ta(4f) and (b) N(1s).

(b) N(1s) Ta(4p3/2) TaN Second Cycle

t First Cycle

t ~36Ă

7 412408 404 400 396392 388 Ta(4f) (a) Ta-N TaOx Ta-N/Ta total Ta2O5 =0.32 Second Cycle

o Ta-N/Ta total

First Cycle

0 =0.09 0 ~36Ă

2 2 3632 28 24 20 16 Binding Energy (eV)

Fig. 3.4 XPS spectra change with cycles of TBTDET dose and H2/H flux exposure on SiO2 sample. (a) Ta(4f) and (b) N(1s).

Ta(4f) spectra acquired with the sample aligned normal to the analyzer axis (normal emission) and at 60º (from the surface normal - grazing emission) are shown in Fig. 3.5a and

3.5b for the thick film, respectively. The ratios of TaTaN/TaTotal are 0.32±0.02 in normal incidence and 0.46±0.02 in grazing incidence, respectively. The increase in the relative intensity of TaN at

55 grazing incidence [Fig. 3.5b] also indicates that TaN is associated with the surface region rather than the film/substrate interface.

TaOx Ta-N Ta2O5 (b) Ta-N/Ta Grazing total =0.46 Emission

o (a)

t Ta-N/Ta

o Normal

t total =0.32

0 Emission

2 2 3632 28 24 20 16 Binding Energy (eV)

Fig. 3.5 Ta(4f) XPS spectra after TBTDET dose and H2/H flux exposure on SiO2. (a) Normal incidence and (b) Grazing incidence.

3.3.2. TaN Formation on OSG

Typical Si(2p), O(1s) and C(1s) XPS spectra for an OSG sample are shown in Fig. 3.6, along with the evolution of the Ta(4f) spectrum upon reaction with TBTDET. Since XPS on an insulating OSG sample can result in sample charging that shifts the binding energies, spectra

56 were calibrated by referencing the peak from adventitious carbon at 284.5 eV [18]. The Si(2p) spectrum (Fig. 3.6a) is well fit with a FWHM of 2.1 eV at binding energies of 103.2 eV and

102.1 eV, respectively. The peak at 103.2 eV can be assigned to the Si-O bond in SiO2, while the feature at 102.1 eV indicates a different chemical environment, which can be assigned to the O-

Si-C bond formation [17, 18, 26]. The O(1s) spectrum (Fig. 3.6b) can be fit with two features

(FWHM = 2.1 eV) at binding energies of 532.1 eV and 531.1 eV, consistent with O-Si in SiO2 and O-Si-C bonding environments, respectively [18, 27]. In the C(1s) spectrum (Fig. 3.6e), the peak at binding energy of 283.5 eV is consistent with a C-Si-O environment, and the feature at

284.5 eV is assigned to adventitious carbon [18, 28]. The XPS data indicate that this OSG sample is an O-Si-C material with substantial amounts of SiO2 and a stoichiometry of SiC0.72O1.08

(Eq. (2.1)). As with SiO2, the TBTDET precursor was condensed on OSG at 120K, annealed to

300K, and exposed to H/H2 flux. After exposure to TBTDET precursor at 120K and annealing to

300K, considerable precursor interaction with OSG substrate occurred and resulted in Ta suboxide formation with Ta(4f7/2) features at 24.7 eV and 25.8 eV (Fig. 3.6c), and the corresponding O(1s) peaks at 529.8 eV (Fig. 3.6b) [4, 18, 20, 21]. The existence of the N(1s)

398 eV feature (Fig. 3.6d) and C(1s) 285 eV peak (Fig. 3.6e) at 300K indicates that some precursor still remained in the film. Interaction with atomic hydrogen results in the formation of a Ta(4f7/2) feature at 23.2 eV binding energy (top trace, Fig. 3.6c) and N(1s) feature at 396.6 eV

(top trace, Fig. 3.6d), with a Ta:N atomic ratio of 1:1 (Eq. (2.1)), indicating the formation of a stoichiometric TaN phase on the OSG substrate [4]. Ta2O5 and additional Ta suboxide were formed with atomic hydrogen exposure, characterized by the appearance of a Ta(4f) peak at 26.5 eV and increase of Ta(4f7/2) intensity at 24.7 eV, respectively (Fig. 3.6d) [18]. The atomic hydrogen flux induced the shifting of C(1s) feature to higher binding energy at 295.5 eV (top of

57 Fig. 3.6e), suggesting reactions with electron-withdrawing groups, such as oxygen [18, 23]. The binding energy of C(1s) (Fig. 3.6e) is above 284 eV, indicating no Ta carbide in the film on OSG sample [18, 24].

Further evidence of the effects of H/H2 bombardment on precursor/substrate interactions is displayed in figure 3.7. The data show changes in sample charging as a function of various stages in precursor dose/annealing/reaction. Charging is frequently observed for photoemission studies of insulating samples, due to the ejection of photoelectrons without compensating charge flow from a conductive substrate layer. This net excess of charge usually leads to a uniform shift of apparent photoelectron energies, as measured against the known energies of standard transitions (in this case substrate C(1s) and Si(2p) transitions, as shown). Note that the application of an H/H2 flux eliminates any significant charging, coincident with the formation of a TaN overlayer. The elimination of sample charging is evidence that the TaN surface coverage is sufficient to alter the electron emission of the sample so as to reduce the net charge imbalance.

This is in turn evidence that, at this point, the reacted overlayer covers a large proportion of the surface area.

58 (a) Si(2p)(b) O(1s)(c) Ta(4f)

TaOx

) )

) O-Si-C

s s s

t t

t TaN

i i

i SiO Ta2O5

n n n SiO 2

2 Ta-O

U U U

y y

y H /H H /H

r r

r H /H C-Si-O 2 2

a a

a 2

r r

r exposure exposure

t t t

i i

i exposure

b b b

r r r

A A A

( ( (

300K 300K 300K

y y y

t t t

i i

i annealing annealing annealing

s s s

n n n

e e e

t t t

n n n

I I I

S S

S Clean Clean Clean P P P

X X X OSG OSG OSG

108 104 100 96 536532 528 524 36 32 28 24 20 16

Binding Energy (eV)

) ) )

) )

(d) ) (e) C(1s)

s s s

t t t

t t

N(1s) t

i i i

n n n

U U U

y y y

y y

y H /H

r r r

r r

H2/H r 2

a a a

a a

Ta(4p3/2) TaN a C-C

r r r

r r

r exposure

t t

t exposure

t t t

i i i

i i

i Reaction of C-Si-O

b b b

r r r

A A A

A A C and O

( ( (

y y y

t t t

i i i

i i

Precursor i

s s Precursor

s s s

n n n

e e e

t t t

t t

300K t 300K

n n n

I I I

annealing annealing

S S S

P P P

X X X

X X X Clean Clean OSG OSG 420412 404 396 388308300 292 284 276 Binding Energy (eV) Binding Energy (eV)

Fig. 3.6 XPS spectral changes upon reaction with TBTDET precursor. (Bottom traces) OSG sample before precursor exposure; (middle traces) precursor adsorption at 120 K followed by annealing to 300K; (top traces) after H/H2 exposure. (a) Si(2p); (b) O(1s); (c) Ta(4f); (d) N(1s) and (e) C(1s).

6 H2/H

Charge Shift (eV) 2

0 01234567 Process Stages

Fig. 3.7 XPS charge shift during various process stages of TaN formation, stage 1: OSG sample as received; 2: cool down to 120K; 3: TBTDET dose at 120K; 4: anneal to 300K; 5: hydrogen flux exposure for 90 min; 6: hydrogen flux exposure for 270 min.

The evolution of Ta and N core level spectra during sequential exposures to TBTDET at

120 K, followed by annealing to 300 K and H/H2 bombardment, are shown in figure. 3.8. The increase in relative intensities of the Ta(4f) at 23.2 eV and N(1s) at 396.6 eV features corresponding to TaN is evident. This suggests that a cyclic ALD or starved CVD process could be used to form a TaN interfacial layer on an OSG dielectric substrate. The numbers to the left of the Ta(4f) feature indicate the estimated total thickness (Eq. (2.2)) of the Ta-containing overlayer (By the end of the third cycle, a substrate Si feature is no longer observed.).

(b) N(1s) Ta(4p3/2)

TaN Third Cycle

Second Cycle ~79Ă

First Cycle ~37Ă 5500 total counts 420412 404 396 388 (a) TaOx Ta(4f) TaN Ta2O5 Third Cycle

Second Cycle ~79Ă

First Cycle ~37Ă 21000 total counts 3632 28 24 20 16 Binding Energy (eV)

Fig. 3.8 XPS spectra change with cycles of TBTDET dose and H2/H flux exposure on OSG sample. (a) Ta(4f) and (b) N(1s).

61 3.4. Discussion

TBTDET condensed at 120K, annealed in UHV to 300K, and then exposed to an H/H2 flux at 500 K results in the formation of TaN on both OSG and SiO2 substrates, with Ta(4f7/2) feature at binding energy 23.2 eV, N(1s) feature at 396.6 eV, and Ta:N atomic ratio of 1:1. No

TaN features were observed when similar experiments were carried out in the absence of atomic hydrogen. TaN formation after subsequent atomic hydrogen exposure at 500K, indicates that these features are formed only under atomic hydrogen flux. Temperature for TaN formation by

TBTDET CVD deposition without atomic hydrogen was reported to be ~720 K – 920 K [12, 13].

Such a high temperature limits its application involving low κ materials, as such materials often degrade at temperature above ~600 K – 650 K. In this study, low-temperature of 500K TaN

CVD deposition on OSG with TBTDET and atomic hydrogen was developed. The intensities of the Ta(4f) and N(1s) at 23.2 eV and 396.6 eV result in a atomic ratio of 1:1, indicating the formation of TaN. TaN is an excellent diffusion barrier due to its low resistivity [3, 29].

Alternatively, attempt to grow TaN by CVD methods using Ta(NMe2)5 and ammonia resulted in high resistivity Ta3N5, which hampers its usefulness as a diffusion barrier for advanced metallization in IC applications [30].

By annealing to 300K, considerable TBTDET precursor interaction with SiO2 and OSG substrate occurred and resulted in Ta-O bond formation, characterized by Ta(4f7/2) peaks at 24.7 and 25.8 eV, O(1s) at 529.8 eV for Ta suboxide. The atomic hydrogen flux induced new Ta2O5 and more Ta suboxide formation. Comparison of the relative intensity of Ta(4f7/2) at 23.2 eV at normal incidence and grazing incidence on SiO2 (Fig. 3.5) indicates that TaN is associated with the outer surface of the film, while the Ta oxide is associated with the interface. The formation of

62 Ta-O interfacial bonds strongly suggests good adhesion between the adlayer and the substrate.

The shifting of C(1s) feature to higher binding energies upon reaction with atomic hydrogen suggests reactions with electron-withdrawing groups, such as oxygen. This aspect of the reaction is not yet understood, but it is notable that an atomic hydrogen flux at low temperatures does not readily remove organic species from the surface/interface region.

Ta(4f) XPS spectra resulting from a single TBTDET physisorption/300 K anneal/H/H2 cycle with ~36Ǻ Ta film thickness on SiO2 and OSG substrates are shown in figures. 3.9a and

3.9b, respectively. The TaTaN/TaTotal intensity ratios are 0.09±0.01 on SiO2 and 0.36±0.02 on

OSG. These data indicate more facile Ta-O bond formation on SiO2 than on OSG. The formation of Ta oxide by reaction of Ta from the precursor and O from the substrate is also corroborated by the fact that SiO2 is reduced, resulting in a Si(2p) substrate feature at 102.1 eV

(Fig. 3.2a) as the deposited Ta is oxidized at SiO2 substrate after atomic hydrogen exposure.

(b) TaOx Ta O 2 5 TaN ~37Ă Ta-N/Ta On OSG total =0.36

(a) ~36Ă On SiO Ta-N/Ta 2 total =0.09

13000 total counts 3632 28 24 20 16 Binding Energy (eV)

Fig. 3.9 Ta(4f) XPS spectra after TBTDET dose and H2/H flux exposure. (a) on SiO2 and (b) on

OSG.

63 Many factors are not understood, including the effects of H bombardment on ligand/substrate interactions, precursor interaction with conductive surfaces (e.g., Cu), and changes in reaction pathways for precursor deliveries at 300 K, rather than by physisorption/annealing. The effects of higher hydrogen doses, exposure to NH3/NH2 fluxes, etc., have not been examined. Nonetheless, these data demonstrate the potential for a

TBTDET/H process to yield conformal, ultra thin TaN barrier on low dielectric substrates.

3.5. Conclusions

TBTDET partially reacts with O atoms in both SiO2 and OSG substrate by 300 K to form strong Ta-O interfacial bonds, characterized by the O(1s) feature at binding energy 529.8 eV,

Ta(4f7/2) at 24.7 eV and 25.8 eV. Reaction with atomic hydrogen at 500 K results in stoichiometric TaN formation with the Ta(4f7/2) feature at binding energy 23.2 eV and N(1s) at

396.6 eV, leading to a TaN phase bonded to the substrate by Ta-O interactions. Ta is oxidized as

SiO2 is reduced, with the decrease of Si(2p) binding energy at 103.2 eV in SiO2 and appearance of new feature at 102.1 eV in Si suboxide. The increase in the relative intensity of TaN on SiO2 at grazing incidence compared to normal incidence indicates that TaN is associated with the surface region rather than the overlayer/substrate interface. Annealing of TBTDET on SiO2 to

500K without atomic hydrogen results in no TaN formation, indicating that the atomic hydrogen induces the TaN formation. Sequential exposures of TBTDET at 120K, followed by annealing to

300K and exposure to H/H2 on the reacted layer on both SiO2 and OSG result in a gradual increase in TaN formation. These data indicate precursor reaction with SiO2 and OSG at low temperature (e.g. 300K) and TaN formation upon reaction with H/H2 flux at temperature < 500K.

64 The data therefore indicate that a low temperature TaN deposition process is possible on SiO2,

OSG or similar surfaces. The ready formation of Ta-O interfacial bonds strongly suggests good adhesion between the adlayer and the SiO2 or OSG substrates. The data summarized here indicate that TBTDET is a potential precursor for an atomic hydrogen driven ALD/CVD process.

3.6 Chapter References

[1] Kuo, Y.-L., Huang, J.-J., Lin, S.-T., Lee, C. and Lee, W.-H., Mater. Chem. Phys. 80 (2003)

690.

[2] Kim, H., Kellock, A. J. and Rossnagel, S. M., J. Appl. Phys. 92 (2002) 7080.

[3] Olowolafe, J. O., Mogab, C. J., Gregory, R. B. and Kottke, M., J. Appl. Phys. 72 (1992)

4099.

[4] Badrinarayanan, S. and Sinha, S., J. Appl. Phys. 69 (1991) 1141.

[5] Baba, K., Hatada, R., Udoh, K. and Yasuda, K., Nucl. Instrum. Methods in Phys. Res. B

127/128 (1997) 841.

[6] Lee, Y. K., Latt, K. M., Jaehyung, K. and Lee, K., Mater. Sci. Semicond. Processing 3

(2000) 179.

[7] Angelkort, C., Berendes, A., Lewalter, H., Bock, W. and Kolbesen, B. O., Thin Solid Films

437 (2003) 108.

[8] Wang, Q., Ekerdt, J. G., Gay, D., Sun, Y.-M. and White, J. M., Appl. Phys. Lett. 84 (2004)

1380.

[9] Ugolini, D., Kowalczyk, S. P. and McFeely, F. R., J. Appl. Phys. 72 (1992) 4912.

[10] Ugolini, D., Powalczyk, S. P. and McFeely, F. R., J. Appl. Phys. 70 (1991) 3899.

[11] Yokoyama, N., Hinode, K. and Homma, Y., J. Electrochem. Soc. 136 (1989) 882.

65 [12] Tsai, M. H., Sun, S. C., Chiu, H. T., Tsai, C. E. and Chuang, S. H., Appl. Phys. Lett. 67

(1995) 1128.

[13] Tsai, M. H., Sun, S. C., Lee, C. P., Chiu, H. T., Tsai, C. E., Chuang, S. H. and Wu, S. C.,

Thin Solid Films 270 (1995) 531.

[14] Park, J.-S., Lee, M.-J., Lee, C.-S. and Kang, S.-W., Electrochem. Solid-State Lett. 4 (2001)

C17.

[15] Eisenbraun, E., Straten, O. V. D., Zhu, Y., Dovidenko, K. and Kaloyeros, A., Proc. 2001

IITC (2001) 207.

[16] Ohshita, Y., Ogura, A., Hoshino, A., Hiiro, S. and Machida, H., J. Cryst. Growth (2000)

604.

[17] Tong, J., Martini, D., Magtoto, N. and Kelber, J., J. Vac. Sci. and Technol. B21 (2003) 293.

[18] Moulder, J. F., Stickle, W. F., Sobol, P. E., Bomben, K. D., Chastain, J. and King, R. C.,

Handbook of X-Ray Photoelectron Spectroscopy, Physical Electronics, Eden Prairie, MN.

[19] Barr, T. L., J. Vac. Sci. Technol. A 9 (1991) 1793.

[20] Atnassova, E. and Spassov, D., Appl. Surf. Sci. 135 (1998) 76.

[21] Nefedov, V. I., Firsov, M. N. and Shaplygin, I. S., J. Electron Spectrosc. Relat. Phenom. 26

(1982) 65.

[22] Engbrecht, E. R., Sun, Y.-M., Smith, S., Pfiefer, K., Bennett, J., White, J. M. and Ekerdt, J.

G., Thin Solid Films 418 (2002) 145.

[23] http://srdata.nist.gov/xps/Bind_e_spec_query.asp, retrieved the information on July 25,

2004.

[24] Kim, K. S., Jang, Y. C., Kim, H. J., Quan, Y.-C., Choi, J., Jung, D. and Lee, N.-E., Thin

Solid Films 377 (2000) 122.

66 [25] Machida, H., Hoshino, A., Suzuki, T., Ogura, A. and Ohshita, Y., J. Cryst. Growth 237-239

(2002) 586.

[26] Fisher, I., Kaplan, W. D., Eizenberg, M., Nault, M. and Weidman, T., Mat. Res. Soc. Symp.

Proc. 716 (2002) B7.20.1.

[27] Contarini, S., Howlett, S. P., Rizzo, C. and Deangelis, B. A., Appl. Surf. Sci. 51 (1991) 177.

[28] Chourasia, A. R., Surf. Sci. Spectra 8 (2001) 45.

[29] Holloway, K., Fryer, P. M., Cabral, C., Jr., Harper, J. M. E. and Bailey, P. J., J. Appl. Phys.

71 (1992) 5433.

[30] Fix, R. M., Gordon, R. G. and Hoffman, D. M., Chem. Mater. 5 (1993) 614.

67 CHAPTER 4

RUTHENIUM SPUTTER DEPOSITION ON ORGANOSILICATE GLASS AND ON

PARYLENE: AN XPS STUDY OF INTERFACIAL CHEMISTRY, NUCLEATION AND

GROWTH [1]*

4.1. Introduction

Ruthenium (Ru) is of growing interest as a diffusion barrier for copper (Cu) in low-k integration, primarily because the semi-noble nature of the metal allows for facile electrodeposition of Cu films, while the high melting point and immiscibility with Cu are favorable for microelectronics applications [2]. In diffusion barrier applications, the adhesion of

Ru films to dielectric substrates is also important. The semi-noble nature of Ru is detrimental for adhesion and conformal growth on dielectric substrates, and one would intuitively expect that Ru would exhibit much less interfacial interaction with a silicon(Si)-oxygen(O)-carbon(C) low k dielectric than would, e.g., tantalum (Ta) [3]. X-ray photoelectron spectroscopy (XPS) has been demonstrated [3-5] to be an excellent method for characterizing the formation of metal-substrate chemical bonds (or lack of same) during gradual metal deposition under controlled conditions.

In such experiments, it is usually helpful to transfer the sample from the deposition chamber to the analysis chamber without exposure to atmosphere. This report documents results of XPS core level spectra for sputter-deposited Ru on organosilicate glass (OSG) and on parylene, at

* Reproduced with permission from [Zhao, X., Magtoto, N. P., and Kelber, J. A., Mat. Res. Soc.

68 coverages up to several monolayers. OSG is a leading low-k candidate for replacing SiO2 as an interlayer dielectric, while parylene is a candidate for pore sealing of porous ultra-low k materials [6]. The XPS results confirm generally weak Ru-substrate interactions, and this is corroborated by results for a Scotch® tape (3M, Inc., St. Paul, Minnesota, www.3m.com) test of an electrolessly deposited Cu film on Ru sputter-deposited onto OSG, showing failure at the

Ru/OSG interface.

4.2. Experiment Detail

Experiments were carried out in an ultrahigh vacuum (UHV) system attached to a dual magnetron sputter deposition system. This system has been described previously in chapter 2.

Since XPS on insulating samples can result in sample charging that shifts the apparent photoelectron energies, spectra were calibrated by referencing the peak from adventitious carbon at 284.5 eV.

An elemental Ru sputter target was used, with target/sample distance adjusted so that a minimum sputter deposition rate of 0.004 Å-sec-1 was observed, as determined by the attenuation of core-level XPS intensity (Eq. (2.2)) during Ru deposition on clean Cu foil. Sample transfer between deposition and analysis chambers occurred under UHV (< 10-8 torr). The ~ 1 cm2 samples consisted of either commercially prepared OSG deposited on Si, or 1000Å thick parylene deposited on ~20Å silicon dioxide (SiO2) on a p-type Si(100) substrate.

4.3. Results

XPS core-level spectra are displayed in figure 4.1 for the OSG sample after annealing

69 in UHV at 500 K for 30 minutes, which removed some adventitious carbon. This particular type of OSG has been described in chapter 3. The main effect of annealing in UHV is to have removed a significant amount of adventitious carbon from the sample, resulting in a post-anneal stoichiometry of SiC0.72O1.08 (Eq. (2.1)).

XPS spectra of samples containing carbon and ruthenium are complicated by the overlap between the C(1s) spectrum and the Ru(3d3/2) component of the Ru(3d) spectrum, as shown in figure 4.2. For this reason, Ru coverage was determined from intensity of the Ru(3d5/2) feature.

Since the relative intensities of the Ru(3d3/2)/Ru(3d5/2) features are well known [7], the estimation of the C(1s) signal intensity by subtraction of the determined Ru(3d3/2) intensity is straightforward, as shown (Fig. 4.2).

C(1s) O(1S) Si(2p) O-Si-C(- C-C(-284.5 eV) SiO2(-532.1 SiO (-103.2 eV) C-Si-O(- eV) 531.1 eV) 2 C-Si-O(- 283.5 eV) 102.1 eV)

After Annealing

ss tt SiC0.72O1.08

s sss

t ttt

n nn

u uu

o ooo

lll

c ccc

l lll

a aaa

t ttt Before

o ooo

t ttt

0 000 Annealing

0 555

5 555

11 3 333 SiC0.86O1.11 288 286 284 282 280 536 534 532 530 528 106 104 102 100 Binding Energy (eV)

Fig. 4.1 Core level XPS spectra for the OSG film before (bottom) and after (top) annealing at

500 K for 30 min. in UHV.

The Ru(3d5/2) binding energy of 280.0 eV is indicative of Ru in a metallic or zerovalent state [8], and this binding energy was observed independent of Ru coverage, indicating negligible charge transfer between the Ru and the substrate. The variation of

Ru(3d) intensity with deposition time on OSG is compared to the behavior on parylene in Fig.

4.3 (left), normalized to the substrate C(1s) intensity.

C(1s) + Ru(3d ) 5000 3/2

Ru(3d5/2) Intensity (counts) Intensity (counts) 0

290Binding Energy (eV) 278

Fig. 4.2 Ru(3d)/C(1s) region, showing the partial overlap of the Ru(3d) and C(1s) core level features. Spectrum corresponds to about 30 min Ru deposition (approximately 3 monolayers).

Ru(3d3/2) component indicated by dashed line was determined from the intensity and binding energy of the Ru(3d5/2) component.

(a) PVD Ru on OSG (b) PVD Ru on Parylene 0.7 0.9 0.6 ~2.5Å Ru 0.7 ~2.5ÅRu 0.5 0.4 0.5 0.3 0.3 0.2 )/C(1s) XPS Intensity)/C(1s) Ratio XPS 5/2 0.1 0.1 0

Ru(3d 0 010203040010203040 Ru Deposition Time(min)

Fig. 4.3 The evolution of the Ru(3d5/2)/C(1s) intensity ratio as function of Ru deposition vs time at 300 K. (Left) Deposition of Ru on OSG. The film thickness at 10 min corresponds to 1 monolayer. (Right) Deposition of Ru on parylene.

The data in figure 4.3 (left) indicate a linear increase in Ru intensity on OSG up to 10 minutes deposition time, followed by a change in slope. Such behavior is indicative of initial conformal (2D) growth in the first layer, followed by a transition to 3D growth with the formation of the second layer— SK growth [9]. Although such behavior—when observed for Cu deposition— generally correlates with significant chemical interaction and charge exchange between the metal overlayer and dielectric substrate [5, 10, 11], no evidence for Ru oxidation was observed in the XPS spectra.

In contrast to Ru/OSG, similar data for Ru deposition on parylene shows different behavior (Fig. 4.3, right). There is no sharp change in slope, as for deposition on OSG, and the region where a gradual change is observed occurs well before an average thickness of 1 monolayer is achieved. The data in Fig. 4.3 therefore suggest a much less pronounced tendency

72 for initial conformal growth on parylene than on OSG. A comparison of these results in turn suggests a stronger interfacial interaction between Ru and OSG than between Ru and parylene.

In order to examine the suitability of sputter-deposited Ru films for Cu integration, a

Ru/OSG sample was prepared for Cu electroless deposition. A sample was prepared by 60 minutes sputter deposition of Ru onto an OSG substrate at 300 K—corresponding to an average thickness of 6 Ru monolayers. The sample was removed from UHV and placed in a formaldehyde-based electroless plating solution (6g/l copper sulfate, 32 g/l ethylenediaminetetraacetic acid (EDTA), 4g/l formaldehyde and sodium hydroxide to adjust the pH of the solution to 11). After 60 minutes, the sample was emersed from the solution, and air- dried. A shiny Cu film was readily visible to the eye. The sample was then subjected to a simple

Scotch Tape test in order to provide a rough, qualitative assessment of film substrate adhesion.

The Scotch Tape test resulted in the loss of all visible Cu from the sample. XPS spectra of the sample before Cu deposition (but after air exposure) and after Cu deposition and Scotch Tape test, are shown in Fig. 4.4.

C(1s)/Ru(3d) O(1s)

C-Si-O(- Ru-O(- SiO2/OH( O-Si-C(- 283.5eV) 281.4eV) -532.1eV) 531.1eV) After Cu and O-Ru(- Ru(-280eV) Tape Test 529.7eV)

Before Cu Intensity (arb. units) Intensity (arb. units) 290 285 280 275 536 532 528 Binding Energy (eV) Binding Energy (eV)

Fig. 4.4 XPS C(1s)/Ru(3d) spectra (left) and O(1s) spectra (right).

73 The XPS spectra taken after air exposure but before Cu electroless deposition (bottom traces) clearly show the presence of oxidized Ru, as evidenced by a broadening of the Ru(3d5/2) feature (figure 4.4, bottom, left) which is well fit by a component at 281.4 eV—characteristic of oxidized Ru [7, 8]. The presence of oxidized Ru is corroborated by the presence of a component near 529.7 eV in the O(1s) spectrum [7](Fig. 4.4, bottom, right). After the Scotch tape test, the

Ru(3d5/2) intensity is reduced by 73% (Fig. 4.4, top, left). This is accompanied by a corresponding loss in intensity of the O(1s) component near 529.7 eV (Fig. 4.4, top, right).

Cu(2p) spectra (not shown) indicated the presence of only trace amounts of Cu after the Scotch tape test.

4.4. Discussion

The XPS data shown here indicate that sputter deposition of Ru on OSG at 300 K results in the growth of a conformal interface (SK growth). In contrast, Ru deposition on parylene shows less tendency towards SK behavior, indicative of a weaker interfacial interaction.

During deposition on OSG, there is no evidence for Ru oxidation, which would be evidenced by a broadening of the Ru(3d5/2) component to higher binding energies. Such a broadening is only observed after exposure of the sample to air (Fig. 4.4). Possible XPS evidence for the formation of a Ru carbide interfacial layer is obscured by the overlap of Ru (3d) and C(1s) features in the region of 283-281 eV (Fig. 4.2)—the region in which metal carbide features are often observed

[7]. The absence of interfacial carbide formation can be deduced from the fact that the

Ru(3d)/C(1s) intensity ratio (Fig. 4.3, left) increases linearly with Ru coverage up to the completion of the first monolayer. Similar Ru/Si data (not shown) similarly strongly suggests an absence of interfacial silicide formation. The weak chemical interaction between Ru and the

74 substrate is consistent with the results of the Scotch tape test on a Cu(electroless)/Ru/OSG sample indicating failure mainly at the Ru/OSG interface.

The lack of interfacial chemical reaction between sputter-deposited Ru and OSG or parylene is in contrast to what is observed for Ta. Ta sputter deposited onto a Si-O-C substrate, where formation of interfacial Ta oxide and Ta carbide is observed [3]. Similarly, Ta deposited on parylene results in the formation of interfacial carbide [12]. This would suggest that the use of

Ru on OSG or parylene sealed dielectric materials should be accompanied by the use of a Ta

“glue” layer between the Ru and the dielectric in order to increase interfacial adhesion. Such a scheme, of course, replaces the deposition of a Cu seed on Ta with the deposition of a Ru layer, which can then be used for direct electroless or electrodeposition of Cu. An alternative approach, which would avoid the necessity for a glue layer, would be to hydroxylate the substrate prior to Ru deposition. Such an approach has been shown to enhance the chemical interaction of Cu with a variety of substrates [10, 11]. Other seminoble metals may exhibit similar effects [13].

4.5 Conclusions

XPS has been used to characterize the chemical interactions of sputter-deposited Ru with OSG and parylene at 300 K. The results indicate the lack of interfacial oxide, carbide or silicide formation on OSG, and the lack of interfacial carbide formation on parylene. Ru exhibits

SK growth (conformal growth of the first layer) on OSG, but this tendency is much less clear-cut on parylene. The formation of an electroless Cu layer on a Ru/OSG sample, followed by a

Scotch Tape test and subsequent XPS analysis, revealed failure at the Ru/OSG interface, consistent with the weak interfacial chemical interactions observed by XPS. The results indicate

75 that the use of Ru as a diffusion barrier for Cu/low-k integration requires either the use of a

“glue” layer (e.g., Ta) between the Ru and dielectric substrate, or possibly some other chemical modification, such as hydroxylation of the substrate.

4.6. Chapter References

[1] Zhao, X., Magtoto, N. P., and Kelber, J. A., Material Research Society (MRS)

Proceedings 812 (2004) F2.5.

[2] Chyan, O., Arunagirl, T. N. and Ponnuswamy, T., J. Electrochem. Soc. 150 (2003) C347.

[3] Tong, J., Martini, D., Magtoto, N. and Kelber, J., J. Vac. Sci. Technol. B21 (2003) 293.

[4] Shepherd, K. and Kelber, J., Appl. Surf. Sci. 151 (1999) 287.

[5] Zhao, X., Leavy, M., Magtoto, N. P. and Kelber, J. A., Appl. Phys. Lett. 79 (2001) 3479.

[6] Lu, T.-M., Rensselaer Polytechnic Institute, private communication.

[7] Moulder, J. F., Stickle, W. F., Sobol, P. E., Bomben, K. D., Chastain, J. and King, R. C.,

Handbook of X-Ray Photoelectron Spectroscopy, Physical Electronics, Eden Prairie,

Minnesota, 1995.

[8] Madhavaram, H., Idriss, H., Wendt, S., Kim, Y. D., Knapp, M., Over, H., Asman, J.,

Loffler, E. and Muhler, M., J. Catalysis 202 (2001) 296.

[9] Argile, S. and Rhead, G. E., Surf. Sci. Repts. 10 (1989) 277.

[10] Niu, C., Magtoto, N. P., Kelber, J. A., Jennison, D. R. and Bogicevic, A., Surf. Sci. 465

(2000) 163.

[11] Pritchett, M., Magtoto, N. P., Tong, J., Kelber, J. A. and Zhao, X., Thin Solid Films 440

(2003) 100.

[12] Yang, G. R. and Lu, T.-M., Appl. Phys. Lett. 72 (1998) 1846.

76 [13] Baumer, M. and Freund, H.-J., Prog. Sur. Sci. 61 (1999) 127.

77 BIBLIOGRAPHY

http://nepp.nasa.gov/index_nasa.cfm/934, retrieved the information on July 25, 2004.

ESCA Tools, Surface/Interface, Mountain View, CA. http://srdata.nist.gov/xps/Bind_e_spec_query.asp, retrieved the information on July 25, 2004.

Alers, G. B., Werder, D. J., Chabal, Y., Lu, H. C., Gusev, E. P., Gustafsson, T. and Urdahl, R. S.,

Appl. Phys. Lett. 73 (1998) 1517.

Angelkort, C., Berendes, A., Lewalter, H., Bock, W. and Kolbesen, B. O., Thin Solid Films 437

(2003) 108.

Angyal, M. S., Shacham-Diamand, Y., Reid, J. S. and Nicollet, M.-A., Appl. Phys. Lett. 67

(1995) 2152.

Argile, S. and Rhead, G. E., Surf. Sci. Repts. 10 (1989) 277.

Atnassova, E. and Spassov, D., Appl. Surf. Sci. 135 (1998) 76.

Baba, K., Hatada, R., Udoh, K. and Yasuda, K., Nucl. Instrum. Methods in Phys. Res. B 127/128

(1997) 841.

Badrinarayanan, S. and Sinha, S., J. Appl. Phys. 69 (1991) 1141.

Barr, T. L., J. Vac. Sci. Technol. A 9 (1991) 1793.

Baumer, M. and Freund, H.-J., Prog. Surf. Sci. 61 (1999) 127.

Beyers, R., J. Appl. Phys. 56 (1984) 147.

Bohr, M. T., Solid State Technol 9 (1996) 105.

Chang, C. and Uh, C., Appl. Phys. Lett. 57 (1990) 617.

Chen, L., Magtoto, N. P., Ekstrom, B. and Kelber, J. A., Thin Solid Films 376 (2000) 115.

78 Cheng, Y. L., Wang, Y. L., Wu, Y. L., Liu, C. P., Liu, C. W., Lan, J. K., O'Neil, M. L., Ay, C.

and Feng, M. S., Thin Solid Films 447-448 (2004) 681.

Chiang, C., Tzeng, S., Raghavan, G., Villasol, R., Bai, G., Bohr, M., Fujimoto, H. and Fraser, D.

B., Proc. VMIC (1994) 414.

Chyan, O., Arunagirl, T. N. and Ponnuswamy, T., J. Electrochem. Soc. 150 (2003) C347.

Contarini, S., Howlett, S. P., Rizzo, C. and Deangelis, B. A., Appl. Surf. Sci. 51 (1991) 177.

Eisenbraun, E., Straten, O. V. D., Zhu, Y., Dovidenko, K. and Kaloyeros, A., Proc. 2001 IITC

(2001) 207.

Ekstrom, B. M., Lee, S., Magtoto, N. and Kelber, J. A., Appl. Surf. Sci. 171 (2001) 275.

Engbrecht, E. R., Sun, Y.-M., Smith, S., Pfiefer, K., Bennett, J., White, J. M. and Ekerdt, J. G.,

Thin Solid Films 418 (2002) 145.

Fisher, I., Kaplan, W. D., Eizenberg, M., Nault, M. and Weidman, T., Mat. Res. Soc. Symp.

Proc. 716 (2002) B.7.20.1.

Fix, R. M., Gordon, R. G. and Hoffman, D. M., Chem. Mater. 5 (1993) 614.

Ganesan, P. G. and Eizenberg, M., Mat. Sci. Eng. 103 (2003) 213.

Garenstroom, S. W. and Winograd, N., J. Chem. Phys. 67 (1977) 3500.

Garza, M., Liu, J., Magtoto, N. P. and Kelber, J. A., Appl. Surf. Sci. 222 (2004) 253.

Gill, P., Miller, M. and Nguyen, B. Y., Microelectronic Eng. 56 (2001) 169.

Gottlieb, U., Laborde, O., Giauque, P. H., Nicolet, M.-A. and Madar, R., Microlectronic Eng. 60

(2002) 107.

Grill, A., Diamond Related Mater. 10 (2001) 234.

Guittet, M. J., Crocombette, J. P. and Salim, M., Phys. Rev. B 63 (2001) 125117.

79 Holloway, K., Fryer, P. M., Cabral, C., Harper, J. M., Bailey, P. J. and Kelleher, K. H., J. Appl.

Phys. 71 (1992) 5433.

Huheey, J. E., Keiter, E. A. and Keiter, R. L., Inorganic Chemistry, Principles of Structure and

Reactivity, Fourth Ed., Harper Collins, New York (1993) 114.

Jeng, S.-P., Havemann, R. H. and Chang, M.-C., Mat. Res. Soc. Symp. Proc. 337 (1994) 25.

Jiang, Q. T., Faust, R., Lam, H. and Mucha, J., Proceedings of the 1999 International

Interconnect Technology Conference (IEEE), New York, USA (2000) 125.

Kaloyeros, A. E. and Eisenbraun, E., Annu. Rev. Mater. Sci 30 (2000) 363.

Kim, K. S., Jang, Y. C., Kim, H. J., Quan, Y.-C., Choi, J., Jung, D. and Lee, N.-E., Thin Solid

Films 377 (2000) 122.

Kim, H., Kellock, A. J. and Rossnagel, S. M., J. Appl. Phys. 92 (2002) 7080.

Klein, J. C., Proctor, A., Hercules, D. M. and Black, J. F., Anal. Chem. 55 (1983) 2205.

Klein, T. M., Niu, D., Epling, W. S., Li, W., Maher, D. M., Hobbs, C. C., Hedge, R. I., Baumvol,

I. J. R. and Parsons, G. N., Appl. Phys. Lett. 75 (1999) 4001.

Kowalczky, S. P., Pollak, R. A., Mcfeely, F. R., Ley, L. and Shirley, D. A., Phys. Rev. B 8

(1973) 2387.

Kuo, Y.-L., Huang, J.-J., Lin, S.-T., Lee, C. and Lee, W.-H., Mater. Chem. Phys. 80 (2003) 690.

Laurila, T., Zeng, K., Kivilahti, J. K., Molarius, J. and Suni, I., Superficies y Vacio 9 (1999) 199.

Lee, W. W. and Ho, P. S., Mater. Res. Soc. Bull. 22 (1997) 19.

Lee, Y. K., Latt, K. M., Jaehyung, K. and Lee, K., Mater. Sci. Semicond. Processing 3 (2000)

179.

Lee, W.-H., Lin, J.-C. and Lee, C., Mater. Chem. Phys. 68 (2001) 266.

Li, J., Shacham-Diamand, Y. and Mayer, J. W., Mater. Sci. Rep. 9 (1992) 1.

80 Lu, T.-M., Rensselaer Polytechnic Institute, private communication.

Lu, Z. H., McCaffery, J. P., Brar, B., Wilk, G. D., Wallace, R. M., Feldman, L. C. and Tay, S. P.,

Appl. Phys. Lett. 71 (1997) 2764.

Machida, H., Hoshino, A., Suzuki, T., Ogura, A. and Ohshita, Y., J. Cryst. Growth 237-239

(2002) 586.

Madhavaram, H., Idriss, H., Wendt, S., Kim, Y. D., Knapp, M., Over, H., Asman, J., Loffler, E.

and Muhler, M., J. Catalysis 202 (2001) 296.

Maex, K., Baklanov, M. R., Shamiryan, D., Lacopi, F., Brongersma, S. H. and Yanovitskaya, Z.

S., J. Appl. Phys. 93 (2003) 8793.

Maier, G., Prog. Polym. Sci. 26 (2001) 3.

Massalski, T. B., Binary Phase Diagram. Westerville, OH: American Society For Metals (1990).

Mcbrayer, J. D., Swanson, R. M. and Sigmon, T. W., J. Electrochem. Soc. 133 (1986) 1242.

Mekki, A., Holland, D., McConville, C. F. and Salim, M., J. Non-Cryst. Solids 208 (1996) 267.

Miller, M. L. and Linton, R. W., Anal. Chem. 57 (1985) 2314.

Min, K., Chun, K. and Kim, K., J. Vac. Sci. Technol. B 14 (1996) 3263.

Moulder, J. F., Sticle, W. F., Sobol, P. E. and Bomben, K. D., Handbook of X-ray Photoelectron

Spectroscopy, Physical Electronics, Inc., Eden Prairie, Minnesota, 1995.

Murarka, S. P., Gutmann, R. J., Kaloyeros, A. E. and Lanford, W. A., Thin Solid Films 236

(1993) 257.

Murarka, S. P. and Hymes, S. W., Crit. Rev. Solid State Mater. Sci. 20 (1995) 87.

Murarka, S. P., Mater. Sci. Eng. R 19 (1997) 87.

Nefedov, V. I., Firsov, M. N. and Shaplygin, I. S., J. Electron Spectrosc. Relat. Phenom. 26

(1982) 65.

81 Nicolet, M.-A., Suni, I. and Finetti, M., Solid State Technol. 26 (1983) 129.

Niu, C., Shepherd, K., Martini, D., Tong, J., Kelber, J. A., Jennison, D. R. and Bogicevic, A.,

Surf. Sci. 465 (2000) 163.

Nuesca, G. M. and Kelber, J. A., Thin Solid Films 262 (1995) 224.

Ohshita, Y., Ogura, A., Hoshino, A., Hiiro, S. and Machida, H., J. Cryst. Growth (2000) 604.

Ohuchi, C. F. S., French, R. H. and Kasowski, R. V., J. Appl. Phys. 62 (1987) 2286.

Olowolafe, J. O., Mogab, C. J., Gregory, R. B. and Kottke, M., J. Appl. Phys. 72 (1992) 4099.

Park, J.-S., Lee, M.-J., Lee, C.-S. and Kang, S.-W., Electrochem. Solid-State Lett. 4 (2001) C17.

Peden, C. H. F., Kidd, K. B. and Shinn, N. D., J. Vac. Sci. Technol. A9 (1991) 1518.

Powell, C. J., Jablonski, A., Tilinin, I. S., Tanuma, S. and Penn., D. R., J. Electron Spectrosc.

Relat. Phenom. 98 (1999) 1.

Pritchett, M., Magtoto, N. P., Tong, J., Kelber, J. A. and Zhao, X., Thin Solid Films 440 (2003)

100.

Seah, M. P., in Auger and X-ray Photoelectron Spectroscopy, D. Briggs and M. P. Seah (New

York, 1990), Vol. 1, 201 & 245.

Shacham-Diamond, Y. and Dubin, V. M., Microelctronic Eng. 33 (1997) 47.

Shepherd, K. and Kelber, J., Appl. Surf. Sci. 151 (1999) 287.

Shepherd, K., Niu, C., Martini, D. and Kelber, J. A., Appl. Surf. Sci. 158 (2000) 1.

Singer, P., Semicond. Int. 5 (1996) 88.

Slaugther, J. M., Weber, W., Guntherodt, G. and Falco, C. M., MRS Bull. 17 (1992) 39.

Stampanoi, M., Vaterlans, A., Aeschlimann, M. and Pescia, D., J. Appl. Phys. 64 (1988) 5321.

Strausser, Y. E., Scheibner, E. J. and Johannssen, J. S., Thin Solid Films 52 (1978) 203.

Ting, C. Y. and Wittmer, M., Thin Solid Films 96 (1982) 327.

82 Tompkins, H. G. and Tray, C., J. Electrochem. Soc. 136 (1989) 2331.

Tong, J., Martini, D., Magtoto, N. and Kelber, J., J. Vac. Sci. and Technol. B21 (2003) 293.

Torres, J., Appl. Surf. Sci. 91 (1995) 112.

Treichel, H., Ruhl, G., Ansmann, P., Wurl, R., Muller, C. and Dietlmeier, M., Microelectron.

Eng. 40 (1998) 1.

Treichel, H. and Goonetilleke, C., Adv. Engin. Mater. 3 (2001) 461.

Tsai, M. H., Sun, S. C., Chiu, H. T., Tsai, C. E. and Chuang, S. H., Appl. Phys. Lett. 67 (1995)

1128.

Tsai, M. H., Sun, S. C., Lee, C. P., Chiu, H. T., Tsai, C. E., Chuang, S. H. and Wu, S. C., Thin

Solid Films 270 (1995) 531.

Ugolini, D., Powalczyk, S. P. and McFeely, F. R., J. Appl. Phys. 70 (1991) 3899.

Ugolini, D., Kowalczyk, S. P. and McFeely, F. R., J. Appl. Phys. 72 (1992) 4912.

Wagner, C. D., Gale, L. H. and Raymond, R. H., Anal. Chem. 151 (1979) 466.

Wang, S. Q., Mater. Res. Soc. Bull. 19 (1994) 30.

Wang, Q., Ekerdt, J. G., Gay, D., Sun, Y.-M. and White, J. M., Appl. Phys. Lett. 84 (2004) 1380.

Wendt, H., Cerva, H., Lehmann, V. and Pamler, W., J. Appl. Phys. 65 (1989) 2402.

Wilk, G. D. and Wallace, R. M., Appl. Phys. Lett. 74 (1999) 2854.

Wilk, G. D. and Wallace, R. M., Appl. Phys. Lett. 76 (2000) 112.

Wu, Y. Y., Kohn, A. and Eizenberg, M., J. Appl. Phys. 95 (2004) 6167.

Yang, G. R. and Lu, T.-M., Appl. Phys. Lett. 72 (1998) 1846.

Yokoyama, N., Hinode, K. and Homma, Y., J. Electrochem. Soc. 136 (1989) 882.

83 Zeng, Y., Russell, S. W., McKerrow, A. J., Chen, L. and Alford, T. L., J. Vac. Sci. Technol. B 18

(1999) 221.

Zhang, L., Persaud, R. and Madey, T. E., Phys. Rev. B 56 (1997) 549.

Zhao, X., Leavy, M., Magtoto, N. P. and Kelber, J. A., Appl. Phys. Lett. 79 (2001) 3479.

Zhao, X., Magtoto, N. P., and Kelber, J. A., Mat. Res. Soc. Symp. (MRS) Proc. 812 (2004) F2.5.

Zhao, X., Magtoto, N. P., Leavy, M. and Kelber, J. A., Thin Solid Films 415 (2002) 308.