Work Function and Process Integration

WORK FUNCTION AND PROCESS INTEGRATION
ISSUES OF METAL GATE MATERIALS
IN CMOS TECHNOLOGY
REN CHI
NATIONAL UNIVERSITY OF SINGAPORE
2006
WORK FUNCTION AND PROCESS INTEGRATION
ISSUES OF METAL GATE MATERIALS
IN CMOS TECHNOLOGY
REN CHI
B. Sci. (Peking University, P. R. China) 2002
A THESIS SUBMITTED
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING
NATIONAL UNIVERSITY OF SINGAPORE
OCTOBER 2006
_____________________________________________________________________
ACKNOWLEGEMENTS
First of all, I would like to express my sincere thanks to my advisors, Prof.
Chan Siu Hung and Prof. Kwong Dim-Lee, who provided me with invaluable
guidance, encouragement, knowledge, freedom and all kinds of support during my
graduate study at NUS. I am extremely grateful to Prof. Chan not only for his
patience and painstaking efforts in helping me in my research but also for his kindness
and understanding personally, which has accompanied me over the past four years.
He is not only an experienced advisor for me but also an elder who makes me feel
peaceful and blessed. I also greatly appreciate Prof. Kwong from the bottom of my
heart for his knowledge, expertise and foresight in the field of semiconductor
technology, which has helped me to avoid many detours in my research work. I do
believe that I will be immeasurably benefited from his wisdom and professional
advice throughout my career and my life. I would also like to thank Prof. Kwong for
all the opportunities provided in developing my potential and personality, especially
the opportunity to join the Institute of Microelectronics, Singapore to work with and
learn from so many experts in a much wider stage. My best wishes will be with Prof.
Chan and Prof. Kwong always.
I would also like to greatly acknowledge Prof. Li Ming-Fu, Dr. Yeo Yee-Chia
from NUS and Prof. Kang Jinfeng from Peking University for their valuable
suggestions and inspirational discussions which had been indispensable for my
research work. My special acknowledgement goes to Dr. Yu Hongyu, who is my
senior in the lab previously and currently with IMEC at Belgium, for his self-giving
help in most of the technical problems that I had encountered in the first two years of
my research.
i
I also owe the opportunity to collaborate with so many talented graduate
students in Silicon Nano Device Lab at NUS. Many thanks to Dr. Hou Yongtian, Dr.
Chen Jinghao, Dr. Yu Xiongfei, Mr. Whang Sung Jin, Mr. Wang Xinpeng, Mr. Shen
Chen, Mr. Hwang Wan Sik, Mr. Liow Tsung-Yang, Mr. Lim Eu-Jin, Mr. Faizhal Bin
Bakar, and Mr. Peng Jianwei for their useful discussions and kind assistances during
the course of my research, as well as the friendships that will be cherished always. I
would also like to extend my best appreciation to all other SNDL teaching staff,
graduate students, and technical staff for the good academic environment created.
A significant part of my research was performed in Institute of
Microelectronics (IME), Singapore. Many of my thanks also go to the managers and
technical staff in the Semiconductor Process Technologies (SPT) lab of IME. I would
like to appreciate Dr. Balasubramanian Narayanan, Dr. Lo Guo-Qiang, Dr. Rakesh
Kumar, and Dr. Feng Han-Hua for all the supports during my stay at IME. I also
must acknowledge Dr. Alastair David Trigg for the help in AES analysis, Dr. Tung
Chih-Hang for the help in TEM characterization, and Dr. Loh Wei-Yip, Dr. Agarwal
Ajay, Dr. Lakshimi Kanta Bera, Dr. Yu Ming-Bin, and Dr. Subramaniam Balakumar
for their knowledge and experiences which had helped me so much. My gratitude
also goes to the excellent team of the technical staff in the IME cleanroom for their
skillful and responsible work. Without these, I would not have gained so much during
the course of my doctoral research.
I also need to thank Dr. Pan Jishen in Institute of Materials Research and
Engineering (IMRE), Singapore, for the help in XPS analysis, and Dr. Thomas
Osipowicz in the Department of Physics, NUS, for the help in RBS analysis.
Last but not least, to my family, especially the love of my life, Zhang Li, for
their love and enduring supports.
ii
TABLE OF CONTENTS
Acknowledgements ......................................................................................................... i
Table of Contents ........................................................................................................ iii
Summary .................................................................................................................... viii
List of Tables ................................................................................................................. x
List of Figures ............................................................................................................... xi
List of Symbols ........................................................................................................ xviii
List of Abbreviations ................................................................................................... xx
Chapter 1. Introduction ...................................................................................... 1
1.1. Overview ............................................................................................................ 1
1.2. MOSFET Scaling: Challenges and Opportunities ............................................. 2
1.2.1. Leakages in Deep-Submicrometer MOSFET ............................................. 4
1.2.2. Vertical Scaling of MOSFET Gate Stack ................................................... 6
1.2.3. Innovations in Device Structures .............................................................. 10
1.2.4. Mobility Enhancement for Performance Gain .......................................... 11
1.3. Summary .......................................................................................................... 13
References .................................................................................................................... 15
Chapter 2. Developments in Metal Gate Materials for CMOS
Technology .............................................................................................................. 22
2.1. Limitations of Poly-Si Electrode ..................................................................... 22
2.1.1. Poly-Si Depletion Effect ........................................................................... 23
iii
2.1.2. Dopant Penetration Effect ......................................................................... 24
2.1.3. Gate Electrode Resistivity......................................................................... 25
2.1.4. Compatibility with High-κ Dielectrics...................................................... 27
2.2. Post Polysilicon Era: Metal Gate Technology ................................................. 30
2.2.1. Historical Perspective of Metal Gate Electrodes ...................................... 30
2.2.2. Considerations for Metal Gate Candidates ............................................... 31
2.2.2.1. Work Function Requirement.............................................................. 31
2.2.2.2. Thermal Stability Considerations ...................................................... 33
2.2.2.3. Process Integration Issues .................................................................. 34
2.2.2.4. Co-optimization of Metal Gate/High-κ Gate Stack ........................... 36
2.2.3. Research Status of Metal Gate Technology .............................................. 37
2.2.3.1. Direct Metal Gates ............................................................................. 38
2.2.3.2. Binary Metal Alloys ........................................................................... 40
2.2.3.3. Fully-Silicided (FUSI) Metal Gates ................................................... 40
2.3. Challenges in Metal Gate Technology ............................................................. 43
2.3.1. Understanding of the Metal-Dielectric Interface ...................................... 43
2.3.2. Developing Appropriate Metal Gate Materials......................................... 44
2.3.3. Dual Metal Gate Integration Issues........................................................... 44
2.4. Research Scope and Major Achievements in this thesis .................................. 45
References .................................................................................................................... 48
Chapter 3. The Metal-Dielectric Interface and Its Impact on the
Effective Work Function of Metal Gates...................................................... 56
3.1. Introduction ...................................................................................................... 56
3.2. Theoretical Backgrounds ................................................................................. 58
iv
3.2.1. Work Function of Metal Materials ........................................................... 58
3.2.2. Definition of Effective Work Function ..................................................... 58
3.2.3. Factors Affecting the Work Function of Metals ....................................... 60
3.2.4. Fermi-Level Pinning: Schottky Model and Bardeen Model ..................... 61
3.2.5. Metal Induced Gap States (MIGS) Theory and Its Limitations ................ 62
3.2.6. Work Function Measurement Techniques ................................................ 65
3.3. Experimental .................................................................................................... 67
3.4. Results and Discussions ................................................................................... 68
3.4.1. Work Function Thermal Instability of TaN .............................................. 68
3.4.2. General Trends in the Process Dependentce of Φm,eff on SiO2 and high-κ
Dielectrics ............................................................................................... 74
3.4.3. Model: Fermi Level Pinning Induced by Extrinsic States ........................ 77
3.4.4. Investigation of Hf-Si Bond Induced Extrinsic States .............................. 79
3.5. Conclusion ....................................................................................................... 85
References .................................................................................................................... 87
Chapter 4. Lanthanide-Incorporated Metal Nitrides Electrodes for
NMOS Applications ............................................................................................. 91
4.1. Introduction ...................................................................................................... 91
4.2. Experimental .................................................................................................... 93
4.3. Material Characteristics of Lanthanide-MNx ................................................... 96
4.3.1. Composition Analysis ............................................................................... 96
4.3.2. Auger Electron Spectroscopy (AES) Study .............................................. 98
4.3.3. X-ray Photoelectron Spectroscopy (XPS) Study .................................... 100
4.3.4. X-ray Diffraction (XRD) Study .............................................................. 103
v
4.3.5. Resistivity ............................................................................................... 105
4.4. Work Function Tunability.............................................................................. 108
4.5. Thermal Stability Study ................................................................................. 113
4.6. MOSFET Characteristics ............................................................................... 121
4.7. Conclusion ..................................................................................................... 128
References .................................................................................................................. 129
Chapter 5. Process Integration for Dual Metal Gate Electrodes...... 132
5.1. Introduction .................................................................................................... 132
5.2. A Gate-First Dual Metal Gate Integration Scheme by High-Temperature Metal
Intermixing Technique ........................................................................................... 140
5.2.1. Motivation ............................................................................................... 140
5.2.2. Process Integration Flow and Device Fabrication .................................. 141
5.2.3. Feasibility Study of the High-Temperature Intermixing Method ........... 143
5.2.4. Compatibility with High-κ Dielectrics.................................................... 147
5.2.5. Dual Work Function Metal Gate Integration using InM ....................... 151
5.2.6. Summary ................................................................................................ 155
5.3. A Gate-Last Dual Metal Gate Integration Process Employing a Novel HfN
Replacement Gate .................................................................................................. 156
5.3.1. Motivation ............................................................................................... 156
5.3.2. Proposed Integration Flow and Device Fabrication................................ 157
5.3.3. Results and Discussions .......................................................................... 158
5.3.4. Summary ................................................................................................. 164
5.4. Conclusion ..................................................................................................... 164
References .................................................................................................................. 166
vi
Chapter 6. Conclusion...................................................................................... 169
5.1. Summary ........................................................................................................ 169
5.1.1. Understanding the Metal-Dielectric Interface ........................................ 169
5.1.2. Lanthanide-Incorporated Metal Nitride Gate Electrodes ........................ 170
5.1.3. Process Integration of Dual Metal Gates ................................................ 171
5.2. Suggestions for Future Work ......................................................................... 173
References .................................................................................................................. 176
Appendix
List of Publications .................................................................................................... 177
vii
SUMMARY
Rapid advances in CMOS technology have led to aggressive scaling of the
MOSFET gate stack. Conventional poly-Si/SiO2 gate stack is approaching some
practical limits, and novel metal gate materials and high-κ dielectrics may need to be
introduced into IC industry as will novel process integration technologies. Immense
challenges arise in material engineering and process integration of novel metal gate
electrodes. This thesis attempts to address some of these challenges.
The metal-dielectric interface is important since it directly affects the effective
work function of metal gates. The influence of the metal-dielectric interface on the
effective work function has been investigated systematically in this thesis. It is found
that the creation of extrinsic states at the metal-dielectric interface, which appears to
be thermodynamically driven, could be the major cause for the instability of metal
gate effective work function during the high-temperature annealing process. The
chemical bond configurations at the metal-dielectric interface could be correlated with
the creation of extrinsic states. In general, the Hf-Si bond tends to create extrinsic
states upon annealing while Hf-Hf or Si-Si bonds’ effect is less pronounced. A model
considering the impact of extrinsic states has also been proposed to qualitatively
explain the dependence of metal effective work function on the annealing process.
One of the most urgent issues for metal gate technology is to find a way to
tune the work function of metal gates for CMOS applications. We demonstrate, for
the first time, that lanthanide elements can be very useful in modulating the work
function of refractory metal-nitride gate electrodes, which provides a new way for
metal gate work function engineering. In this work, lanthanide elements with very
low work function are incorporated into metal-nitride materials to get the best trade-
off between thermal stability and low work function. By varying the lanthanide
viii
concentration in lanthanide-incorporated metal-nitrides, a work function value of
4.2~4.3 eV can be obtained even after a 1000 oC RTA treatment. This is promising
for NMOS devices using a gate-first bulk-Si CMOS process. The good thermal
stability has been attributed to the high nitrogen concentration in these lanthanide-
incorporated metal-nitrides, therefore the N concentration needs to be carefully
engineered in process. Good transistor characteristics have also been demonstrated
using these novel metal gate materials.
Dual metal gate integration issues for advanced CMOS devices are also
discussed in this thesis. A novel dual metal gate integration process using a high-
temperature metal intermixing technique is first demonstrated for gate-first CMOS
flow. In this process, a TaN buffer layer is used to protect the gate dielectric during
the selective metal etching process. The work function of the TaN buffer layer can be
modulated for CMOS by a subsequent metal intermixing process at high-temperature,
which is compatible with the conventional gate-first process flow. By using this
integration scheme, dual work function of 4.15 and 4.72 eV has been achieved in
TaN/Tb/TaN (NMOS) and TaN/Ti/HfN (PMOS) metal stacks, respectively.
Another dual metal gate integration process proposed in this thesis is a gate-
last replacement gate process employing HfN as a novel dummy gate electrode. In
this process, a high-quality HfN/HfO2 gate stack with HfO2 EOT less than 1 nm is
first fabricated using a gate-first process. The dummy HfN gate can then be
selectively removed from HfO2 so that other metal gate candidates with suitable work
functions for bulk-Si CMOS can be integrated. In a prototype demonstration, large
work function difference for about 0.8 eV has been achieved by using Ta and Ni to
replace HfN for NMOS and PMOS devices, respectively, with no degradation in the
EOT, gate leakage, and TDDB characteristics of the ultra-thin HfO2 gate dielectric.
ix
LIST OF TABLES
Table 1.1 Technology roadmap for the scaling of dielectrics thickness 7

in next ten years.
Table 2.1 Specifications for the scaling of gate electrode, derived from 26
ITRS-2005.
Table 3.1 Experimental splits of different laminated stacks consisting 80

Hf (N) and Si layers on slanted SiO2 and HfO2 dielectrics.
Table 3.2 VFB shifts for TaN/HfN/Si and TaN/Si/HfN stacks on SiO2 84
and HfO2 after RTA at 1000oC for 5 sec, observed from the
lower parts of the C-V curves shown in Fig. 3.10 & Fig. 3.11.
Table 4.1 Work functions of some lanthanide metals. 92
Table 4.2 Experiment splits and the compositions for the Lanthanide- 94
MNx films.
Table 4.3 Experiment splits and the compositions for Ta0.9Tb0.1Ny with 94
different N2 flow rates during reactive sputtering deposition.
Table 4.4 Work function and barrier height of lanthanide-incorporated 112

TaN on SiO2 as a function of rapid thermal anneal (RTA)
temperatures.
x
LIST OF FIGURES
Fig. 1.1 CPU transistor counts from 1970s to present, showing the 2
device scaling according to Moore’s Law; © Intel corp.
Fig. 1.2 Historical scaling trends of supply voltage (VDD), threshold 3

voltage (Vth) and gate-oxide thickness (tox) vs. channel length
(Lg) for CMOS logic technologies, showing the different
scaling factors for supply voltage and device dimension.
Fig. 1.3 Schematic cross section of MOSFET showing the major 4

leakage current paths. I1 for direct tunneling through gate oxide;
I2 for subthreshold leakage; I3 for BTBT; I4 for GIDL; and I5 for
punchthrough.
Fig. 1.4 Gate leakage current density of some high-κ dielectrics as a 9

function of EOT, compared with the gate leakage specifications
at 100oC for high-performance (HP), low-operating-power
(LOP), and low-standby-power (LSTP) applications according
to ITRS 2004.
Fig. 2.1 (a) The energy band diagram of an NMOS device showing the 23
poly-Si gate depletion effect; (b) Equivalent circuit for the gate
stack of MOSFET. C denotes the total gate capacitance which
determines the inversion charge density Qi in the channel, Cpoly,
Cox, CSi represent the capacitance from the poly depletion, gate
oxide, and substrate, respectively. CSi is further broken up into
a depletion charge capacitance Cd and inversion-layer
capacitance Ci. CET represents the capacitance equivalent
thickness of the MOS gate stack, and ψs is the surface potential.
Fig. 2.2 Work function of some metal elements collected from 32

experiments.
Fig. 2.3 Work function modulation by various mechanisms in some 42

NiSi-based-silicide metal gates; data from [75], [77], [80]-[84].
Fig. 3.1 Band diagram of a MOS structure in flat-band condition (a) 59

without interface dipoles and (b) with interface dipoles at
metal-dielectric interface.
Fig. 3.2 Schematic energy band diagram (left) and the characteristics of 63
the gap states (right) for metal gate on dielectrics. The
character of MIGS becomes more acceptor- (donor-) like
toward the Ec (Ev), as indicated by the solid (dashed) line.
Fig. 3.3 Plots of VFB versus EOT of (a) TaN/SiO2 and (b) TaN/HfO2 69
xi
devices before and after 1000°C RTA treatment, from which
the effective work function of TaN can be extracted.
Fig. 3.4 The comparison of (a) EOT (with three different gate-oxide 71
thickness) and (b) effective work function of TaN as a function
of RTA temperature with and without HfN capping layer on top
of the TaN/SiO2 stack
Fig. 3.5 (a) Gate leakage measurement of HfN/TaN/SiO2 devices and 72

(b) barrier height extraction by F-N tunneling analysis before
and after 1000oC RTA treatment.
Fig. 3.6 The variation of metal gate work function Φm with the 74
annealing temperature on SiO2 dielectric.
Fig. 3.7 Work function of metal gates on HfO2 before and after high- 76
temperature annealing. HfNx-1 and HfNx-2 denotes HfNx with
different N concentration.
Fig. 3.8 Schematic energy band diagram for a metal gate on a dielectric, 78
showing the mechanism of Fermi-level pinning by extrinsic
states. (a) When EF,m is above the pinning level, (b) When the
EF,m is below the extrinsic pinning level. The conduction band
edge and the valence band edge of the dielectric are denoted by
Ec,d and Ev,d, respectively.
Fig. 3.9 Work function of the TaN/Hf/Si laminated stack on SiO2 and 81
HfO2 after annealing at different conditions: as-deposited,
420oC FGA and RTA at 1000oC for 5 sec followed by FGA.
Fig. 3.10 C-V measurements of the TaN/HfN/Si laminated stack on (a) 83

SiO2 and (b) HfO2/SiO2 dielectrics after different annealing.
Fig. 3.11 C-V measurements of the TaN/Si/HfN laminated stack on (a) 83

SiO2 and (b) HfO2/SiO2 dielectrics after different annealing.
Fig. 3.12 XTEM image of a TaN/HfN/Si stack on SiO2 after 1000oC 85

RTA for 5 sec.
Fig. 4.1 Illustration of the idea to modulate the work function of metal 93
nitrides by incorporating lanthanide elements for n-MOSFET
applications.
Fig. 4.2 Relationship between Tb concentration in Ta1-xTbxNy and the 96

sputtering power applied on the Tb target during co-sputtering
deposition.
Fig. 4.3 RBS spectrum of Ta0.92Tb0.08N1.0 film, where the concentration 97

of each species are determined from the simulation by XRUMP
[9].
xii
Fig. 4.4 AES depth profiling for TaN/Ta0.94Tb0.06Ny/SiO2 (a-b) and 99
TaN/Ta0.95Er0.05Ny/SiO2 (c-d) gate stacks before and after
1000 oC RTA in N2 ambient.
Fig. 4.5 The XPS spectra of the (a) N 1s, (b) Ta 4f, and (c) Tb 4d region 101
for the as-deposited Ta1-xTbxNy films with different Tb
concentrations: (1) TaN; (2) Ta0.97Tb0.03Ny; (3) Ta0.94Tb0.06Ny;
(4) Ta0.9Tb0.1Ny; (5) Ta0.87Tb0.13Ny.
Fig. 4.6 The XPS core level spectra in the N 1s region for the as- 102
deposited (a) Ta1-xErxNy, (b) Ta1-xYbxNy, and (c) Hf1-xTbxNy &
films with different lanthanide concentrations. 103
Fig. 4.7 XRD spectrums of Ta1-xTbxNy materials with different Tb 104

concentrations before and after 1000 oC RTA anneal for 20 sec,
compared with that of TaN.
Fig. 4.8 XRD spectrums of Ta0.9Tb0.1Ny films with different N 105

concentrations measured in the following conditions: as-
deposited, after 900 oC RTA and after 1000 oC RTA.
Fig. 4.9 Resistivity of Ta0.9Tb0.1Ny films as a function of N2/Ar flow rate 106
ratio during the sputtering, with and without RTA performed.
Fig. 4.10 Resistivity of lanthanide-incorporated TaN materials as a 107

function of lanthanide type and concentration before and after
RTA treatments in N2 ambient.
Fig. 4.11 High-frequency C-V characteristics (100 kHz) of Ta1-xTbxNy 108

gated MOS capacitors with different Tb concentration in
Ta1-xTbxNy on SiO2. The measurements are taken after a 420oC
forming gas anneal.
Fig. 4.12 High-frequency C-V characteristics (100 kHz) of MOS- 109

capacitors with Ta0.94Tb0.06Ny gate electrode after 420oC FGA
and after 1000oC RTA. The lines show the simulated C-V
curves which takes quantum mechanical effect into account.
Fig. 4.13 VFB vs. EOT plots of Ta0.94Tb0.06Ny/SiO2, Ta0.95Er0.05Ny/SiO2, 110

and Hf0.8Tb0.2Ny/SiO2 gate stacks (a) after 420oC FGA and (b)
after 1000oC RTA treatment, compared with that of TaN and
HfN.
Fig. 4.14 Work function values of some MNx and lanthanide-MNx gate 111
electrodes as a function of lanthanide type and concentrations
under different annealing conditions, showing the tunability of
MNx work functions by incorporating lanthanide.
Fig. 4.15 C-V characteristics of TaN and Ta0.9Tb0.1Ny metal gates on 113
ALD HfAlO dielectrics after FGA at 420oC for 30 min. and
xiii
RTA at 1000oC for 5 sec.
Fig. 4.16 XTEM images of Ta0.94Tb0.06Ny/SiO2 gate stack on (100) Si 114
substrate after 420oC FGA for 30 min, 900oC RTA for 30 sec
and 1000oC RTA for 30 sec.
Fig. 4.17 EOT variation of the Ta1-xTbxNy/SiO2 gate stacks before and 115
after 1000oC RTA for 20 sec, as a function of Tb concentrations
in Ta1-xTbxNy.
Fig. 4.18 Gate leakage characteristics of Ta0.94Tb0.06Ny/SiO2 gate stack 116

with PMA performed at different temperatures.
Fig. 4.19 Gate leakage characteristics of Ta0.95Er0.05Ny/SiO2 gate stack 117

with PMA performed at different temperatures.
Fig. 4.20 TDDB characteristics of Ta0.94Tb0.06Ny/SiO2 gate stack 117

(SiO2~3.2 nm) after PMA at different temperatures, measured
under negative constant voltage stress (CVS) at room
temperature.
Fig. 4.21 EOT variation as a function of annealing temperature for 118

Ta0.9Tb0.1Ny/SiO2 gate stacks with different N2 flow rates
during the deposition of Ta0.9Tb0.1Ny. Two groups of oxides
with initial thickness of ~3.3 nm and ~ 5.8 nm were
investigated.
Fig. 4.22 Typical I-V characteristics of Ta0.9Tb0.1Ny gated MOS 119

capacitors with different N2 flows during metal gate deposition,
measured after FGA at 420oC for 30 min.
Fig. 4.23 Work function of Ta0.9Tb0.1Ny gate electrodes with different N 120
concentrations as a function of annealing temperature.
Fig. 4.24 XTEM image of TaN/Ta0.9Tb0.1Ny/SiO2 gate stack with 122

Ta0.9Tb0.1Ny thickness of ~ 40 Å on SiO2 dielectric.
Fig. 4.25 C-V characteristics of Ta0.9Tb0.1Ny/SiO2 gate stacks with 123

different Ta0.9Tb0.1Ny thickness. The measurement was taken
on the samples with 1000oC RTA for 10 sec performed.
Fig. 4.26 (a) VFB versus EOT plots of Ta0.9Tb0.1Ny/SiO2 gate stacks with 123
different Ta0.9Tb0.1Ny thickness after 900oC RTA for 20 sec. &
(b) VFB versus EOT plots of Ta0.9Tb0.1Ny/SiO2 gate stacks with 124
different Ta0.9Tb0.1Ny thickness after 1000oC RTA for 10 sec.
Fig. 4.27 Process flow of the damascene process used to pattern the TaN/ 125
Ta0.9Tb0.1Ny metal gate stack in MOSFET fabrication.
Fig. 4.28 Typical high-frequency C-V measurement of Ta0.9Tb0.1Ny/SiO2 126

gated n-MOSFET.
xiv
Fig. 4.29 IDS ~ VDS characteristics of Ta0.9Tb0.1Ny/SiO2 gated n-MOSFET, 126
with the substrate doping concentration of NA = 5 × 1015 cm-3.
Fig. 4.30 IDS ~ VGS characteristics of Ta0.9Tb0.1Ny/SiO2 gated n-MOSFET, 127

with the substrate doping concentration of NA = 5 × 1015 cm-3.
Fig. 4.31 Effective electron mobility in Ta0.9Tb0.1Ny/SiO2 gated n- 127

MOSFET.
Fig. 5.1 Process flow of dual metal gate integration by direct etching 133
method. PR denotes photoresist, HM denotes hard-mask, and
HK denotes high-κ dielectric.
Fig. 5.2 Process flow of the dual metal gate/dual high-κ integration 135
scheme. (a) Metal-A/HK-A deposition; (b) Metal-A/HK-A
selective etching from one side of CMOS; (c) Metal-B/HK-B
deposition; (d) hard-mask-B deposition and patterning; (e)
Metal-B/HK-B selective removal; (f) hard-mask removal, thick
poly-Si top-up, and gate patterning.
Fig. 5.3 Process flow of the dual metal gate integration via metal inter- 136
diffusion.
Fig. 5.4 Process flow of the FUSI process. (a) CMOS fabrication 137
conventionally; (b) oxide re-flow and planarization by CMP; (c)
hard-mask stripping followed by ion-implantation or poly-Si
etch-back; (d1) deposition of a same metal, e.g. Ni, for both
NMOS and PMOS; (d2) deposition of different metals for
NMOS and PMOS, respectively (in parallel with step (d1)); (e)
silicidation and unreacted metal stripping.
Fig. 5.5 Process flow of the replacement gate process. (a) CMOS 138
fabrication with poly-Si as dummy gate; (b) oxide re-flow and
planarization; (c) dummy poly-Si & SiO2 removal; (d) filling
the groove with new high-κ and metal gate; (e) metal CMP to
pattern the metal gate; (f) dual metal gate CMOS formation by
repeat steps (c)-(e).
Fig. 5.6 Dual metal gate integration process flow by high-temperature 141
metal intermixing technique: (1) TaN buffer layer deposition;
(2) P-type metal gate stack (e.g TaN/Ti/HfN) formation
followed by selective etching; (3) N-type metal gate stack
formation (e.g TaN/Tb/TaN) and capping layer deposition; (4)
gate etching, S/D implantation, and dopant activation annealing
(also for intermixing).
Fig. 5.7 XTEM images of the TaN/Tb/TaN stack on SiO2 (a) as- 143
deposited and (b) after 1000oC RTA in N2 ambient for 5 sec.
xv
Fig. 5.8 (a), (b) STEM and (c), (d) EDX depth profiles of the 144
TaN/Tb/TaN stack on SiO2 as-deposited and after 1000oC RTA
in N2 ambient for 5 sec. In (a)& (b), the dark layer in the
sandwich structure denotes Tb element.
Fig. 5.9 Work function versus annealing temperature for different 146
TaN/Metal stacks. Thickness of TaN or TaNx (less N%) bottom
layers are about 2.0~2.5 nm, and that of Tb or Ir are about 2.5
nm. The N2 gas flow rate during deposition of thin TaN layer is
5 sccm, while that for TaNx (less N%) is 4 sccm. All samples
are capped with thick TaN film of ~100 nm.
Fig. 5.10 Gate leakage characteristics of TaN/Tb stack after different 146
RTA treatments. The corresponding WF of the sample is
denoted by open circle in Fig. 5.9.
Fig. 5.11 C-V characteristics of n-MOSFETs with TaN, TaN/Tb/TaN and 148
co-sputtered Ta0.9Tb0.1Ny metal gates on HfTaON after RTA at
1000oC for 1 sec.
Fig. 5.12 IDS-VGS characteristics of n-MOSFETs with TaN, TaN/Tb/TaN 149

(InM), and co-sputtered Ta0.9Tb0.1Ny gates on HfTaON high-κ
dielectric.
Fig. 5.13 Vth distribution of n-MOSFETs with TaN, TaN/Tb/TaN (InM) 150
and co-sputtered Ta0.9Tb0.1Ny gates on HfTaON/HfO2
dielectric. (W/L=320 μm / 5 μm)
Fig. 5.14 Effective electron mobility in n-MOSFETs with TaN, 150

TaN/Tb/TaN (InM) and co-sputtered Ta0.9Tb0.1Ny metal gates
on HfTaON dielectric, measured by split C-V measurement.
(W/L = 200 μm / 20 μm)
Fig. 5.15 C-V characteristics of TaN/Ti/HfN metal stack with and without 152
Tb on top, where the HfN thickness is (a) ~15 Å and (b)
~100 Å.
Fig. 5.16 AFM images of TaN (~ 2 nm) deposited on the bare-Si wafer 152
before and after wet etching in DHF (1:200) for 30 sec.
Fig. 5.17 XTEM images of as-deposited dual metal gate stacks on a 153
single wafer: TaN/Tb/TaN (left) for NMOS and TaN/Ti/HfN
(right) for PMOS on SiO2.
Fig. 5.18 (a) C-V and (b) I-V characteristics of TaN/Tb/TaN (N-type) and 153
TaN/Ti/HfN (P-type) metal gate stacks on SiO2 in as-deposited
condition.
Fig. 5.19 C-V characteristics of TaN/Tb/TaN (NMOS) and TaN/Ti/HfN 154

(PMOS) stacks on SiO2 before and after 1000oC RTA for 1 sec.
xvi
Fig. 5.20 Work Function extraction for TaN/Tb/TaN (NMOS) and 155
TaN/Ti/HfN (PMOS) metal stacks on SiO2 after metal
intermixing process.
Fig. 5.21 Proposed replacement gate process using HfN as dummy gate: 157
(a) CMOS fabrication using TaN/HfN/HfO2 as the gate stacks;
(b) high selective etching of TaN and HfN by wet chemicals;
(c) new metal gate deposition and CMP planarization; (d) dual
metal gate integration by repeating steps (b)-(c).
Fig. 5.22 Etching properties of the HfN/HfO2 gate stack (open triangle 159
symbol) and the HfO2 (solid symbol) film after 1000oC RTA
process in diluted HF solution (1:100). The etch rate of HfN is
determined by surface profiler, and the remaining HfO2
thickness is measured by ellipsometer.
Fig. 5.23 AFM images of HfO2 with different process history: as- 160
deposited HfO2 film, HfO2 after 1000 °C RTA anneal, and
HfO2 in HfN/HfO2 stack with HfN removed by DHF solution
after 1000 °C RTA anneal.
Fig. 5.24 C-V and I-V (inset) characteristics of the “control” HfN/HfO2 161
devices and “re-deposited” HfN/HfO2 devices with HfO2
EOT~0.83 nm. The C-V curves were measured at 100 kHz and
1 MHz on devices with an area of 50×50 μm2.
Fig. 5.25 High frequency C-V curves of the HfN/HfO2 “control” devices 162
and “re-deposited” Ta/HfO2, Ni/HfO2 devices. The inset
compares the C-V curves measured from the “re-deposited”
Ni/HfO2 devices with ultra-thin HfO2 (EOT~0.9 nm) and that of
a “control” HfN/HfO2 device. All the C-V curves were
measured at 100 kHz on devices with an area of 50×50 μm2.
Fig. 5.26 Comparison of TDDB and gate leakage (inset) characteristics 163
between the “control” HfN/HfO2 devices and “re-deposited”
HfN/HfO2, Ta/HfO2, Ni/HfO2 devices. For the TDDB study,
CCS with a current density of ~8 A/cm2 was performed on
devices with an area of 100×100 μm2 at room temperature.
xvii
LIST OF SYMBOLS
Cd depletion-layer capacitance
Ci inversion-layer capacitance
Cox gate-oxide capacitance
Cpoly poly-depletion capacitance
Csi Si substrate capacitance
Ec conduction band
EF,m Fermi level of metal
Eox electric field in gate oxide
Ev valence band
ћ reduced Planck constant
JFN F-N tunnelling current density
K Boltzmann constant
Lg gate length
m body-effect coefficient
m* effective electron mass
Nb substrate doping concentration
q elementary charge
Qd depletion charge density
Qi inversion charge density
Qox equivalent oxide charge density at oxide/Si interface
S Schottky pinning parameter
Tgate gate electrode thickness
tox gate-oxide thickness
xviii
VDD supply voltage
VFB flatband voltage
Vth threshold voltage
γ electronic specific heat coefficient
ε permittivity
ε0 permittivity in vacuum
μeff effective carrier mobility
ρ resistivity
effective density of states at Fermi-level
κ permittivity (or dielectric constant)
φb Schottky barrier height
φox barrier height between gate electrode and gate oxide
ΦCNL,d charge neutrality level of dielectric
Φm metal work function
ΦMS work function difference between gate electrode and substrate
Φm,eff effective work function
Φm,vac metal work function in vacuum
Φs semiconductor work function
ΦTaN/HfO2 effective work function of TaN in TaN/HfO2 stack
ΦTaN/SiO2 effective work function of TaN in TaN/SiO2 stack
χs electron affinity in semiconductor
xix
LIST OF ABBREVIATIONS
AES Auger electron spectroscopy
AFM atomic force microscopy
ALCVD atomic-layer chemical vapor deposition
ALD atomic-layer deposition
BEoL back-end of line
BTBT band-to-band tunnelling
BTI bias-temperature-instability
CCS constant current stress
CES constant-field scaling
CET capacitance equivalent thickness
CMOS complementary metal-oxide-semiconductor
C-V capacitance-voltage
CVD chemical vapor deposition
CVS constant-voltage scaling
constant-voltage stress
DG double-gate
DHF diluted hydrofluoric (acid)
DIBL drain-induced barrier lowering
DOF depth-of-focus
ECNL charge-neutrality level
EDX energy dispersive X-ray
EELS Electron energy loss spectroscopy
EOT equivalent oxide thickness
xx
EWF effective work function
FLP Fermi-level pinning
FUSI fully-silicided (metal gate)
F-N Fowler-Nordheim (tunnelling)
FGA forming-gas annealing
GIDL gate-induced-drain leakage
GOI gate oxide integrity
HFCV high-frequency C-V
HK high-κ (dielectric)
HM hard-mask
HOT hybride-orientation-technology
HP high-performance
HRTEM high-resolution transmission electron microscopy
IC integrated circuits
I/I ion implantation
InM (metal) intermixing
IPE internal photoemission
ITRS International Technology Roadmap for Semiconductors
I-V current-voltage
Lanthanide-MNx lanthanide-incorporated metal nitride
LOP low-operation-power
LSTP low-standby-power
MG metal gate
MIGS metal-induced gap state
MNx (refractory) metal nitride
xxi
MOCVD metal-organic chemical vapor deposition
MOSFET metal-oxide-semiconductor field-effect transistor
PC phase-controlled (silicide)
PDA post-deposition-annealing
PMA post-metal-annealing
PMD post-metal-dielectric
PR photoresist
PVD physical vapor deposition
RBS Rutherford backscattering spectrometry
RF radio-frequency
RSF relative sensitivity factor
RTA rapid thermal annealing
SBH Schottky barrier height
SC-1 standard cleaning-1 (NH4OH+H2O2+H2O) solution
S/D source/drain
SIIS silicidation induced impurity segregation
SS subthreshold swing
SSDOI strained-Si directly on insulator
SSOI strained-Si on insulator
STI shallow trench isolation
UPS ultraviolet photoemission spectroscopy
UTBSOI ultra-thin-body silicon-on-insulator
UV ultraviolet
Vo oxygen vacancy
WF work function
xxii
XPS X-ray photoelectron spectroscopy
XRD X-ray diffraction
XTEM cross-sectional transmission electron microscope
xxiii
Ch 1: Introduction
Chapter 1
Introduction
1.1 Overview
Since the invention of the first integrated circuit (IC) in 1958, the
semiconductor industry has undergone unprecedented growth through the latter half
of the 20th century. Today, silicon-based IC products have infiltrated every corner of
our daily life. Driven by the demand for IC chips with higher speed, greater
functionality, and lower cost, the physical dimensions of metal-oxide-semiconductor
field-effect transistor (MOSFET), the basic element in IC chips, have been scaled
down continuously over the past 40 years so that more transistors can be integrated on
a single chip. In 1965, Gordon Moore of Intel predicted the trend of MOSFET
scaling, which is popularly known as Moore’s Law: the number of transistors on a
chip doubles about every two years, as shown in Fig. 1.1 [1]-[2]. This trend has been
made possible by the advancements in semiconductor process technology from ~ 8
μm in 1972 to the current 65 nm technology. As a result, cost per function has
decreased at an average rate of ~ 25-30% per year per function [3]. According to the
prediction of the latest 2005 International Technology Roadmap for Semiconductors
(ITRS), in the year of 2015, the physical gate length (Lg) for high-performance logic
applications will shrink down to 10 nm [3], which is about 10,000 times smaller than
the diameter of a hair!
1
Ch 1: Introduction
I billion transistors by 2007

9
10 Itanium® II CPU
CPU Transistor Counts Itanium® CPU
8
10 Pentium® 4 CPU
7 Pentium® III CPU

10 Pentium® CPU Pentium® II CPU
TM
6 486 CPU
10 TM
386 CPU
5
10 286
8086
4
10 8080
8008
3 4004
10
1970 1980 1990 2000 2010
Calendar Year
Fig. 1.1 CPU transistor counts from 1970s to present, showing the device scaling
according to Moore’s Law; © Intel corp. [2].
1.2 MOSFET Scaling: Challenges and Opportunities
Several scaling rules were proposed to guide the scaling of MOSFETs, such as
constant-field scaling (CES), constant-voltage scaling (CVS), and generalized scaling
[4]-[6]. In practice, the generalized scaling rule was followed in the modern
complementary metal-oxide-semiconductor (CMOS) technology. The principle of the
generalized scaling is to scale the electric field and the physical dimensions (both
lateral and vertical) of MOSFET by different factors, α and κ, respectively [6]. Under
this protocol, the supply voltage (VDD) typically scales slower than the channel length,
which leads to the increase of electric field by a factor of α, as well as the increase of
power density by a factor of α2 to α3 [6]. Fig. 1.2 illustrates the scaling of VDD,
2
Ch 1: Introduction
threshold voltage (Vth), and gate-oxide thickness (tox) as a function of channel length
[7], showing the different scaling factors of VDD compared with that of Lg. As a result,
the power dissipation of chips became higher and higher.
10 1000
VDD
Gate Oxide Thickness (nm)

1 100
VDD / Vth (V)
Vth
0.1 10
tox
~ Lg
0.01 1
0.01 0.1 1
MOSFET Channel Length (μm)
Fig. 1.2 Historical scaling trends of supply voltage (VDD), threshold voltage (Vth) and
gate-oxide thickness (tox) vs. channel length (Lg) for CMOS logic technologies,
showing the different scaling factors for supply voltage and device dimension [7].
The power-performance trade-off has become the major road-block for the
continuous scaling of CMOS into deep-submicron regimes. In order to squeeze the
maximum performance gain from the continuous scaling while maintaining the power
consumption at an acceptable level, innovative device structures and new materials
have been explored extensively in recent years [8].
3
Ch 1: Introduction
1.2.1 Leakages in Deep-Submicrometer MOSFET
High leakage current is becoming a most serious issue for aggressively scaled
MOSFETs in the sub-50-nm regime, and is very likely to be the show-stopper for the
MOSFET scaling eventually. The major leakage components in short-channel
devices are 1) tunneling current through the thin gate oxide; 2) subthreshold leakage
between source and drain; 3) band-to-band tunneling (BTBT) current through drain-
well junction; 4) gate-induced-drain leakage (GIDL), and 5) punchthrough leakage [9].
Fig. 1.3 is the schematic cross section of a MOSFET, illustrating various leakage
components.
Fig. 1.3 Schematic cross section of MOSFET showing the major leakage current
paths. I1 for direct tunneling through gate oxide; I2 for subthreshold leakage; I3 for
BTBT; I4 for GIDL; and I5 for punchthrough.
The leakage problems in MOSFET rise with the decreasing distance between
the four terminals in the vertical and horizontal directions. As the source and drain
terminals approach each other and the distance becomes comparable with the MOS
depletion width in the channel region, the conventional 1-D field pattern for long-
channel devices, where the electric field in the channel region is controlled by the gate
4
Ch 1: Introduction
electrode only, will no longer be valid. Instead, the source and drain fields penetrate
deeply into the middle of the channel, which lowers the potential barrier between the
source and drain and causes a substantial increase of the subthreshold current. This is
referred to as drain-induced barrier lowering (DIBL) effect [10]. The DIBL effect
will be further amplified when a high drain voltage is applied, leading to the decrease
of Vth and dramatic increase of the subthreshold current by diffusion.
The subthreshold leakage current (Isub) is considered as one of the major
contributors to the “off-state” power dissipation (or passive power), as described by
the following equation [11]:
qVth
Poff = WtotVDD I off = WtotVDD I 0 exp(− ) (1-1)
mKT
where Wtot is the total turn-off device width, Ioff is the average off-current per device
width, I0 is the extrapolated current per width at threshold voltage, and m is body-
effect factor typically ranging from 1.1 to 1.4 [11]. For the modern CMOS circuits,
the passive power can even exceed the active switching power eventually [12].
In order to regulate the Isub to a tolerable level so that the lateral scaling of
MOSFET can be implemented, the influence of drain electric field to the channel
region must be minimized so that the “gate control” can overwhelm that of the drain.
There are several approaches to achieve this goal in device design: increase the gate
capacitance (Cox), reduce the source/drain (S/D) junction depth, or engineer the
doping profiles in the channel region. Increasing channel doping level is a typical
way to control Isub. In order to minimize the side-effects associated with the high
channel doping, including the high electric-field, high BTBT leakage, large
subthreshold swing (SS) and low mobility, non-uniform channel engineering
techniques such as retrograde channel doping [13] and halo implantation [14] have
been introduced into the MOSFET fabrication. Moreover, increasing Cox and/or
5
Ch 1: Introduction
reducing S/D junction depths can also help to control the short-channel effects, but the
trade-off is the increased gate leakage and S/D series resistance. Today, transistor
with 10 nm Lg has been demonstrated with manageable subthreshold current [15].
1.2.2 Vertical Scaling of MOSFET Gate Stack
In addition to the lateral scaling, the gate stack of MOSFET also needs to be
scaled down to provide a better gate control to the channel and improve the drive
capability of a MOSFET. The saturation drive current of a MOSFET can be depicted
by the following equation:
W (Vg − Vth )
2
ε 0ε SiO2 W (Vg − Vth ) 2
I ds , sat = μeff Cox = μeff (1-2)
L 2m CET L 2m
where Cox is the gate capacitance, μeff is the effective carrier mobility, εSiO2 is the
permittivity of gate oxide, CET is the capacitance equivalent thickness of the gate
stack which includes the contributions from the poly-depletion and quantum
mechanical effects, and m is the body-effect coefficient [11]. From this equation, it is
clear that the drive current can be improved by reducing the gate oxide thickness.
Table 1.1 summarizes the requirements for today’s and tomorrow’s gate dielectric in
different applications according to ITRS 2004, showing the aggressive scaling of gate
oxide thickness.
The outstanding properties of SiO2 have enabled the vertical scaling of Si-
based MOSFET for several decades. SiO2 possess many ideal dielectric properties,
such as amorphous structure, thermodynamical and electrical stability, wide band gap
of ~ 9 eV, smooth interface with Si, and so on. However, when the physical thickness
of SiO2 shrinks to less than about 3 nm, the direct tunneling current [16]-[17] through
the thin SiO2 will become significant and rise exponentially as the thickness of SiO2
6
Ch 1: Introduction
decreases. This has become one of the major issues for MOSFET scaling. Although
incorporating nitrogen into SiO2 to form SiON can slightly reduce the gate leakage,
the situation is still getting worse and worse as the gate oxide thickness shrinks
towards the sub-1-nm regime [18]-[19]. On the other hand, from the material point of
view, the minimum thickness needed for SiO2 to maintain its bulk properties (i.e.
band-gap) is about 7 Å [20], which is too thick for the requirement of high-end
applications after year 2010 according to Table 1.1. Therefore, an alternative way out
of the quandary needs to be carved out.
Table 1.1 Technology roadmap for the scaling of dielectrics thickness in next ten
years [3].
Year of Production 2004 2007 2010 2013 2016
Technology node hp90 hp65 hp45 hp32 hp22
Physical gate length for MPU

37 25 18 13 9
(nm)
Physical gate length for low-
53 32 22 16 11
operating-power (nm)
Physical gate length for low-
65 37 25 18 13
standby-power (nm)
EOT for MPU (nm) 1.2 0.9 0.7 0.6 0.5
EOT for low-operating-power

1.5 1.2 0.9 0.8 0.7
(nm)
EOT for low-standby-power
2.1 1.6 1.3 1.1 1
(nm)
Gate leakage at 100oC for high
460 920 1,833 7,692 18,556
performance (A/cm2)
Gate leakage at 100oC for low-
1.9 5.2 10.6 20.8 90.9
operating-power (A/cm2)
Gate leakage at 100oC for low-
4.6 21.6 80 150 254
standby-power (mA/cm2)
7
Ch 1: Introduction
Using high-permittivity (κ) dielectrics to replace SiO2 or SiON would be a
potential solution to enable the further scaling of the gate stack in MOSFET [21]-[22].
The advantage of high-κ gate dielectrics rather than SiO2 is to provide a physically
thicker film for leakage current reduction while improving the gate capacitance by
higher permittivity, as described in equation 1-3:
ε SiO
EOT = 2
T (1-3)
ε high −κ high −κ , Phy
where EOT is the equivalent oxide thickness of the high-κ dielectric, εSiO2 and εhigh-κ
are the permittivity of SiO2 and the high-κ dielectric, respectively, and Thigh-κ,Phy is the
physical thickness of the high-κ film.
The candidate high-κ materials should have suitable permittivity (κ≈ 15-25),
large barrier height for both electron and hole, high crystallization temperature, good
thermal stability and good interface quality with Si substrate and gate electrodes, and
high carrier mobility for both electrons and holes [21]. Among various high-κ
materials investigated, Hf-based high-κ dielectrics have drawn considerable attention
due to their appropriate κ values and relatively high barrier heights for both electrons
and holes. High-κ materials such as HfO2 [23]-[24], HfAlO [25], HfSiO [26]-[27],
HfON [28], and HfSiON [29]-[30] have been extensively studied. Fig. 1.4 shows the
scalability of some higk-κ materials compared with the ITRS requirements. It clearly
shows that the gate leakage reduction by 2 to 4 orders compared to SiO2 can be
achieved using high-κ dielectrics.
However, there are still some challenges for high-κ dielectric to replace SiO2.
The first one is mobility degradation. Coulomb scattering due to the pre-existing and
trapped charges in high-κ [31]-[32] and remote phonon scattering associated with the
ionic properties of the “soft” metal-oxygen bonds in high-κ films [33]-[34] have been
8
Ch 1: Introduction
Gate Leakage Density (A/cm )

hp
2
3 HP Poly/HfO2 [23]
10
Poly/HfSiON [29]
1 Poly/HfSiON [30]
10
LOP
-1
10
LSTP
-3
10
SiO2 [29]
-5
10 TiN/HfO2 [23]
-7
HfN/HfO2 [24]
10
0 5 10 15 20 25 30 35
EOT (Å)
Fig. 1.4 Gate leakage current density of some high-κ dielectrics as a function of EOT,
compared with the gate leakage specifications at 100oC for high-performance (HP),
low-operating-power (LOP), and low-standby-power (LSTP) applications according
to ITRS 2004 [3].
proposed to account for the electron mobility degradation in high-κ stacks. Among
various high-κ dielectric candidates, HfSiOx shows most promising mobility
characteristics [35], but the permittivity is relatively low. Some methods have been
demonstrated to improve the mobility characteristics, such as improving the
microstructure of high-κ films [36] or inserting a SiOx(N) layer under the high-κ [37]-
[38]. The second issue for high-κ dielectric is the Fermi-level pinning (FLP) problem
at the high-κ/poly-Si interface [39]-[40], which causes a high Vth for MOSFETs,
especially for p-MOSFET. This problem may exclude the use of poly-Si/high-κ gate
stack in high-performance applications where low Vth of around +/-0.25 V are
required to maintain enough gate overdrive. Hf-Si bond-induced interface dipole [39]
and/or oxygen vacancy induced charge transfer across the poly-Si/high-κ interface
9
Ch 1: Introduction
[41]-[42] have been proposed to explain the FLP phenomenon. Some techniques such
as F- implantation could be useful in mitigating the FLP problem [43]. Finally, some
reliability issues, such as charge trapping and bias-temperature-instability (BTI) in
high-κ dielectric [44]-[45], are still not well understood. Therefore, scientific
understanding and technology development of high-κ materials are still expected
before its implementation in CMOS technology.
Poly-Si depletion effect is another factor affecting the device scaling in the
vertical direction [46]. The poly-depletion layer accounts for an additional thickness
of about 4 Å to the capacitance equivalent thickness (CET) of the gate stack and
results in a significant loss of gate control. This problem is particularly serious when
the gate oxide scales to the sub-1-nm regime. Metal gate technology can be used to
eliminate the poly-Si depletion effect and hence improve the device performance.
Meanwhile, metal gate electrodes can also address some other concerns associated
with the poly-Si gate electrodes. These make the metal gate technology one of the
hottest research areas in recent years. The detailed backgrounds about metal gate
technology will be discussed in Chapter 2.
1.2.3 Innovations in Device Structures
In order to manage the short-channel effects in the aggressively scaled devices,
many novel device structures have been proposed and investigated, including ultra-
thin-body silicon-on-insulator (UTBSOI), double-gate (DG), FinFET, triple-gate, Ω-
gate FET, nanowire FET etc. [47]-[53]. In these device structures, the gate to channel
potential coupling can be greatly improved by their special device geometry
compared with planar bulk-Si CMOS, so that the short-channel characteristics can be
controlled. Consequently, the intrinsic silicon channel can be adopted, which enables
10
Ch 1: Introduction
lower channel electric field, lower BTBT leakage, sharper subthreshold slope and
better carrier mobility to be achievable. These advantages make them very attractive
as potential technology options for the future high-performance applications.
However, there are still many challenges for these novel device structures.
One of these challenges is threshold voltage adjustment. Due to the small amount of
depletion charges and the intrinsic Si channel used, the desirable gate work function
for these devices should be close to the mid-gap of Si [3]. Conventional poly-Si gate
will not work properly in this situation and novel metal gate electrodes with mid-gap
work functions are required [54]. Secondly, the high source/drain series resistance
caused by the thin silicon body used in these 3-D structures is another concern which
may affect the overall performance of these novel FETs. Thirdly, the carrier transport
characteristics in the ultra-thin Si channels will be very sensitive to the Si-body
thickness TSi [55]-[57], rendering a requirement for very strict process control of TSi.
The manufacturing tolerance for TSi would be added up with the tolerance in defining
the gate length, resulting in even smaller process windows and higher manufacturing
cost in fabricating these novel structures compared with the conventional planar
devices. Therefore these novel structures may only be used for some kernel parts in
MPU or ASIC chips.
1.2.4 Mobility Enhancement for Performance Gain
Another way to improve the performance of MOSFET is to enhance the
carrier mobility, which may provide a key to escape from the power/speed box in
device scaling [58]. Basically there are three avenues to achieving the enhanced
mobility for MOSFET: inducing strain to the channel region, utilizing the high
mobility surface orientation, or employing new channel materials with high mobility
11
Ch 1: Introduction
and high saturation velocity. Sometimes these techniques are combined to get the
utmost gain in performance.
There are two groups of technologies to introduce strain into the channel of
MOSFET. One is the local strain or called process induced strain technologies,
including the strain from shallow trench isolation (STI), Si3N4 stress liners, silicide
induced strain, and embedded SiGe or SiC stressors in the source/drain region, etc
[62]-[67]. These techniques are based on the conventional bulk-Si CMOS process
and thus have the advantages like low-cost and easy integration. Some of these
techniques have already been adopted in the latest 65-nm CMOS technology for the
mass production. The other group of technologies is global-strain technologies, in
which the strain is induced from substrate, such as strained-Si on relaxed-SiGe,
strained-Si on insulator (SSOI), strained-Si directly on insulator (SSDOI), and so on
[68]-[71]. One of the concerns for the global-strain techniques is the difficulty to
optimize the n-MOSFET and p-MOSFET individually. Another concern is the cost
issue because strained-Si substrates with very low defect level are required.
The surface orientation and channel direction (current flow direction) can also
affect the carrier mobility in MOSFET. In conventional bulk-Si CMOS technology,
CZ-Si with (100) surface orientation are commonly used. M. Yang et al. developed a
novel hybride-orientation-technology (HOT) to integrate (100) and (110) surface
orientation on a same wafer to get ideal mobility for n-MOSFET and p-MOSFET,
respectively [72]. Recent progress demonstrates that the HOT technology can also be
integrated with other uniaxial local strain technologies for more performance
enhancement [73].
High-mobility and high-saturation-velocity semiconductors, such as SiGe, Ge,
InP, and GaAs, have also attracted considerable attentions as the possible candidates
12
Ch 1: Introduction
for channel materials [74]-[76]. Compared with Si, the physical properties of these
materials are still not very well understood yet. Many process issues also need to be
addressed, including the dielectric/channel interface engineering and the S/D junction
formation [77]. Moreover, integration of these materials into the conventional Si-
based CMOS process flow would be another challenge. More comprehensive study
on these materials will be appreciated in the future.
1.3 Summary
In summary, immense challenges arise as the Si-based CMOS technology
enters the sub-50-nm node, accompanied by the great opportunities for novel
materials and integration technologies. CMOS scaling will no longer be driven by
photolithography alone; rather it will be driven by the innovations in developing
advanced materials/structures and the ability to integrate these novel materials and
structures into CMOS fabrication processes. Reliable and cost-effective solutions are
expected to unlash the power-performance deadlock. Among the various technology
choices discussed above, metal gate and high-κ technologies are among the most
urgent technologies required by the semiconductor industry to keep step with Moore’s
law in the future ten years [3].
The work in this thesis will be focused on understanding and developing the
advanced metal gate technology, which is particular important for the scaling of
MOSFET in the vertical direction. According to ITRS 2005, metal gate technology
has been considered as one of the performance booster for CMOS in 32-nm
technology node and beyond. The overall objective of this thesis is to gain insights
into the major issues and to address some of the critical challenges in implementing
the metal gate electrodes in the future CMOS platform. The background information
13
Ch 1: Introduction
and recent developments of the metal gate technology will be introduced in Chapter 2,
where the major issues and challenges will be highlighted. According to the different
challenges and the efforts to address these problems, the work in this thesis will be
divided into three aspects. In Chapter 3, the role of metal-dielectric interface in
affecting the effective work function of metal gates will be investigated systematically.
After that, a novel approach to modify the work function of metal nitride materials
will be presented in Chapter 4, which would be valuable for the applications in
NMOS devices using the conventional gate-first process flow. The dual metal gate
integration issues will be discussed in Chapter 5, where two novel integration
schemes will be proposed and demonstrated in order to address some of the
challenges faced in the process integration of dual metal gates. Finally, the major
results achieved in this thesis will be summarized in Chapter 6, as well as some
suggestions on future research work.
14
Ch 1: Introduction
References
1. G. E. Moore, "Progress in digital integrated electronics," in IEDM Tech. Dig., pp.

11-13, 1975.
2. From Intel corp.; http://www.intel.com/technology/silicon/mooreslaw/index.htm
3. International Technology Roadmap of Semiconductors (ITRS), Semiconductor

Industry Association (SIA), SanJose, CA 95110. Available: http://public.itrs.net/
4. R. H. Dennard, F. H. Gaensslen, H. N. Yu, V. L. Rideout, E. Bassous, and A. R.

Leblanc, “Design of ion-implanted MOSFETs with very small physical
dimensions,” IEEE J. Solid-State Circuits, vol. SC-9, p. 256, 1974.
5. G. Baccarani, M.R. Wordeman, and R. H. Dennard, “Generalized scaling theory

and its application to ¼ micrometer MOSFET design,” IEEE Tran. Electron. Dev.,
vol. 31, pp. 452, 1984.
6. Y. Taur and T. H. Ning, Fundamentals of Modern VLSI Devices, Cambridge, U.K.,

Cambridge Univ. Press, 1998, ch. 4, pp. 167-170.
7. Y. Taur, “CMOS design near the limit of scaling,” IBM J. Res. & Dev., vol. 46, pp.
213-222, 2002.
8. H.-S. P. Wong, “Beyond the conventional transistors,” IBM J. Res. & Dev., vol.
46, pp. 133-168, 2002.
9. K. Roy, S. Mukhopadhyay, and H. Mahmoodi-Meimand, “Leakage current

mechanisms and leakage reduction techniques in deep-submicrometer CMOS
circuits,” Proc. IEEE, vol. 91, pp. 305-327, 2003.
12. E. J. Nowak, “Maintaining the benefits of CMOS scaling when scaling bogs
down,” IBM J. Res. & Dev., vol. 46, pp. 169-180, 2002.
13. Indranil De and C. M. Osburn, “Impact of super-steep-retrograde channel doping

profiles on the performance of scaled devices,” IEEE Tran. Electron. Dev., vol. 46,
pp. 1711-1717, 1999.
14. S. Thompson, P. Packan, and M. Bohr, “MOS scaling: transistor challenges for
the 21st century,” Intel Tech. Journal, vol. 2, no.3, 1998.
15. H. Wakabayashi, T. Ezaki, M. Hane, T. Ikezawa, T. Sakamoto, H. Kawaura, S.

Yamagami, N. Ikarashi, K. Takeuchi, T. Yamamoto, and T. Mogami, “Transport
15
Ch 1: Introduction
properties of sub-10-nm planar-bulk-CMOS devices,” in IEDM Tech. Dig., pp.

429-432, 2004.
17. N. Yang, W. K. Henson, J. R. Hauser, and J. J. Wortman, “Modeling study of

ultrathin gate oxides using direct tunneling current and capacitance-voltage
measurements in MOS devices,” IEEE Tran. Electron. Dev., vol. 46, pp. 1464-
1471, 1999.
18. X. Guo, and T. P. Ma, “Tunneling leakage current in oxynitride: dependence on

oxygen/nitrogen content,” IEEE Electron Device Lett., vol. 19, no. 6, pp. 207-209,
1998.
19. S. Inaba, T. Shimizu, S. Mori, K. Sekine, K. Saki, H. Suto, H. Fukui, M.

Nagamine, M. Fujiwara, T. Yamamoto, M. Takayanagi, I. Mizushima, K. Okano,
S. Matsuda, H. Oyamatsu, Y. Tsunashima, S. Yamada, Y. Toyoshima, and H.
Ishiuchi, "Device performance of sub-50 nm CMOS with ultra-thin plasma
nitrided gate dielectrics," in IEDM Tech. Dig., pp. 651-654, 2002.
20. D. A. Muller, T. Sorsch, S. Moccio, F. H. Baumann, K. Evans-Lutterodt, and G.

Timp, “The electronic structure at the atomic scale of ultrathin gate oxides,”
Nature (London) vol. 399, pp.758-761, 1999.
21. G. D. Wilk, R. M. Wallace, and J. M. Anthony, “High-k gate dielectrics: Current

status and materials properties considerations,” J. Appl. Phys., vol. 89, pp. 5243-
5275, 2001.
22. Y.-C. Yeo, T.-J. King, and C. Hu, “MOSFET gate leakage modeling and selection
guide for alternative gate dielectrics based on leakage considerations,” IEEE Tran.
Electron. Dev., vol. 50, pp. 1027-1035, 2003.
23. W. Tsai, L. Ragnarsson, L. Pantisano, P. J. Chen, B. Onsia, T. Schram, E. Cartier,

A. Kerber, E. Young, M. Caymax, S. De Gendt, and M. Heyns, “Performance
comparison of sub-1-nm sputtered TiN/HfO2 nMOS and pMOSFETs,” in IEDM
Tech. Dig. pp. 311-314, 2003.
24. H. Y. Yu, J. F. Kang, J. D. Chen, C. Ren, Y. T. Hou, S. J. Whang, M. Li, D. S. H.

Chan, K. L. Bera, C. H. Tung, A. Du, and D. Kwong, “Thermally robust high
quality HfN/HfO2 gate stack for advanced CMOS devices,” in IEDM Tech. Dig.,
pp. 99-102, 2003.
25. S. G. Park, Y. K. Lee, S. B. Kang, H. S. Jung, S. J. Doh, J.-H. Lee, J. H. Choi, G.

H. Kim, G. H. Choi, U. I. Chung, and J. T. Moon, “Performance improvement of
MOSFET with HfO2-Al2O3 laminate gate dielectric and CVD-TaN metal gate
deposited by TAIMATA,” in IEDM Tech. Dig., pp. 327-330, 2003.
16
Ch 1: Introduction
26. G. D. Wilk and R. M. Wallace, “Electrical properties of hafnium silicate gate

dielectrics deposited directly on silicon,” Appl. Phys. Lett., vol. 74, pp. 2854-2856,
1999.
27. M. A. Quevedo-Lopez, M. El-Bouanani, B. E. Gnade, R. M. Wallace, M. R.

Visokay, M. Douglas, M. J. Bevan, and L. Colombo, “Interdiffusion studies for
HfSixOy and ZrSixOy on Si,” J. Appl. Phys., vol. 92, pp. 3540-3550, 2002.
28. C. S. Kang, H.-J. Cho, K. Onishi, R. Choi, R. Nieh, S. Goplan, S. Krishnan, and J.
C. Lee, “Improved thermal stability and device performance of ultra-thin
(EOT<10Å) gate dielectric MOSFETs by using hafnium oxynitride (HfOxNy),” in
VLSI Tech. Dig., pp. 146-147, 2002.
29. A.L.P. Rotondaro, M.R. Visokay, J.J. Chambers, A. Shanware, R. Khamankar, H.

Bu, R.T. Laaksonen, L.Tsung, M. Douglas, R. Kuan, M.J. Bevan, T. Grider, J.
McPherson, and L. Colombo, “Advanced CMOS transistors with a novel HfSiON
gate dielectric,” in VLSI Tech. Dig., pp. 148-149, 2002.
30. M. Koike, T. Ino, Y. Kamimuta, M. Koyama, Y. Kamata, M. Suzuki, Y. Mitani, A.

Nishiyama, and Y. Tsunashima, “Effect of Hf-N bond on properties of thermally
stable amorphous HfSiON and applicability of this material to sub-50 nm
technology node LSIs, ” in IEDM Tech. Dig., pp. 107-110, 2003.
31. M. Hiratani, S. Saito, Y. Shimamoto, K. Torii, “Effective electron mobility

reduced by remote charge scattering in high-kappa gate stacks” Jpn. J. Appl. Phys.
vol. 41, pp.4521-4522, 2002.
32. F. Lime, K. Oshima, M. Casse, G. Ghibaudo, S. Cristoloveanu, B. Guillaumot, H.

Iwai, “Carrier mobility in advanced CMOS devices with metal gate and HfO2
gate dielectric,” Solid-State Electronics, vol. 47, pp. 1617-1621, 2003.
33. M. Fischetti, D. Neumayer, and E. Cartier, “Effective electron mobility in Si

inversion layers in MOS systems with a high-k insulator: The role of remote
phonon scattering,” J. Appl. Phys., vol. 90, pp. 4587-4608, 2001.
34. W. J. Zhu and T. P. Ma, “Temperature dependence of channel mobility in HfO2-

gated NMOSFETs,” IEEE Electron Device Lett., vol. 25, pp. 89-91, 2004.
35. Y. Kim, H. J. Lim, H. Jung, J. Lee, J. Park, S. K. Han, J. H. Lee, S. Doh, J. P. Kim,
N. I. Lee, Y. Chung, H. Y. Kim, N. K. Lee, S. Ramanathan, T. Seidel, M.
Boleslawski, G. Irvine, B. Kim, H. Lee, and H. Kang, “Characteristics of ALD
HfSiOx using new Si precursors for gate dielectric applications,” in IEDM Tech.
Dig., pp. 511-514, 2004.
36. X. Yu, C. Zhu, X. P. Wang, M. F. Li, A. Chin, A. Y. Du, W. D. Wang, and D.-L.
Kwong, “High mobility and excellent electrical stability of MOSFETs using a
novel HfTaO gate dielectric,” in Symp. VLSI Tech. Dig., pp. 110-111, 2004.
37. A. Morioka, H. Watanabe, M. Miyamura, T. Tatsumi, M. Saitoh, T. Ogura, T.

Iwamoto, T. Ikarashi, Y. Saito, Y. Okada, H. Watanabe, Y. Mochiduki, and T.
17
Ch 1: Introduction
Mogami, “High mobility MISFET with low trapped charge in HfSiO films,” in
Symp. VLSI Tech. Dig., pp. 165-166, 2003.
38. H. Ota, N. Yasuda, T. Yasuda, Y. Morita, N. Miyata, K. Tominaga, M.

Kadoshima, S. Migita, T. Nabatame, and A. Toriumi, “Study on oxynitride buffer
layers in HfO2 metal-insulator-semiconductor structures for improving metal-
insulator-semiconductor field-effect transistor performance,” Jpn. J. Appl. Phys.,
vol. 44, pp.1698-1703, 2005.
39. C. Hobbs, L. Fonseca, V. Dhandapani, S. Samavedam, B. Taylor, J. Grant, L. Dip,

D. Triyoso, R. Hegde, G. Gilmer, R. Garcia, D. Roan, L. Lovejoy, R. Rai, L.
Hebert, H. Tseng, B. White, and P. Tobin, “Fermi level pinning at the
polySi/metal oxide interface”, in Symp. VLSI Tech. Dig., pp. 9-10, 2003.
40. C. W. Yang, Y. K. Fang, C. H. Chen, S. F. Chen, C. Y. Lin, and C. S. Lin, M. F.

Wang, Y. M. Lin, T. H. Hou, C. H. Chen, L. G. Yao, S. C. Chen, and M. S. Liang,
“Effect of polycrystalline-silicon gate types on the opposite flatband voltage shift
in n-type and p-type metal–oxide–semiconductor field-effect transistors for high-
k-HfO2 dielectric,” Appl. Phys. Lett. vol. 83, pp. 308-310, 2003.
41. K. Shiraishi, K. Yamada, K. Torii, Y. Akasaka, K. Nakajima, M. Kohno, T.

Chikyo, H. Kitajima, and T. Arikado, “Physics in Fermi level pinning at the
polySi/Hf-based high-k oxide interface,” in Symp. VLSI Tech. Dig., pp. 108-109,
2004.
42. H. Takeuchi, H. Y. Wong, D. Ha, and T. King, “Impact of oxygen vacancies on

high-k gate stack engineering,” in IEDM Tech. Dig., pp. 829 - 832, 2004.
43. M. Inoue, S. Tsujikawa, M. Mizutani, K. Nomura, T. Hayashi, K. Shiga, J.

Yugami, J. Tsuchimoto, Y. Ohno, and M. Yoneda, “Fluorine incorporation into
HfSiON dielectric for Vth Control and its impact on reliability for Poly-Si gate
pFET,” in IEDM Tech. Dig., pp. 413-416, 2005.
44. D. K. Schroder and J. A. Babcock, “Negative bias temperature instability: Road to

cross in deep submicron silicon semiconductor manufacturing,” J. Appl. Phys., vol.
94, pp. 1-18, 2003.
45. B. H. Lee, C. D. Young, R. Choi, J. H. Sim, G. Bersuker, C. Y. Kang, R. Harris, G.

A. Brown, K. Matthews, S. C. Song, N. Moumen, J. Barnett, P. Lysaght, K. S.
Choi, H. C. Wen, C. Huffman, H. Alshareef, P. Majhi, S. Gopalan, J. Peterson, P.
Kirsh, H. Li, J. Gutt, M. Gardner, H. R. Huff, P. Zeitzoff, R. W. Murto, L. Larson,
and C. Ramiller, “Intrinsic characteristics of high-k devices and implications of
fast transient charging effects (FTCE),” in IEDM Tech. Dig., pp. 859-862, 2004.
47. H.-S. Wong, D. Frank, Y. Taur, and J. Stork, “Design and performance
considerations for sub-0.1 μm double-gate SOI MOSFETs,” in IEDM Tech. Dig.,
pp. 747, 1994.
18
Ch 1: Introduction
48. B. Doris, M. Ieong, T. Kanarsky, Y. Zhang, R. A. Roy, O. Dokumaci, Z. Ren, F.

Jamin, L. Shi, W. Natzle, H. Huang, J. Mezzapelle, A. Mocuta, S. Womack, M.
Gribelyuk, E. C. Jones, R. J. Miller, H. P. Wong, and W. Haensch, “Extreme
scaling with ultra-thin Si channel MOSFETs,” in IEDM Tech. Dig., pp. 267-270,
2002.
49. D. Hisamoto, W.-C. Lee; J. Kedzierski, E. Anderson, H. Takeuchi, K. Asano, T.-J.

King, J. Bokor, and C. Hu, “A folded-channel MOSFET for deep-sub-tenth
micron era,” in IEDM Tech. Dig., pp. 1032-1034, 1998.
50. Y. Choi, N. Lindert, P. Xuan, S. Tang, D. Ha, E. Anderson, T. King, J. Bokor, and
C. Hu, “Sub-20nm CMOS FinFET technologies,” in IEDM Tech. Dig., pp. 421-
424, 2001.
51. R. Chau et al., “Advanced depleted-substrate transistors: single-gate, double-gate

and tri-gate,” in Proc. Int. Conf. on Solid State Dev. and Mat.(SSDM), pp. 68-69,
2002.
52. J. Kavalieros, B. Doyle, S. Datta, G. Dewey, M. Doczy, B. Jin, D. Lionberger, M.

Metz, W. Rachmady, M. Radosavljevic, U. Shah, N. Zelick and R. Chau, “Tri-
gate transistor architecture with high-k gate dielectrics, metal gates and strain
engineering,” in Symp. VLSI Tech. Dig., pp. 62-63, 2006.
53. F. Yang, D. Lee, H. Chen, C. Chang, S. Liu, C. Huang, T. Chung, H. Chen, C.

Huang, Y. Liu, C. Wu, C. Chen, S. Chen, Y. Chen, Y. Chen, C. Chen, B. Chan, P.
Hsu, J. Shieh, H. Tao, Y. Yeo, Y. Li, J. Lee, P. Chen, M. Liang, and C. Hu, “5nm-
gate nanowire FinFET,” in Symp. VLSI Tech. Dig., pp. 196-197, 2004.
54. J. Kedzierski, E. Nowak, T. Kanarsky, Y. Zhang, D. Boyd, R. Carruthers, C.

Cabral, R. Amos, C. Lavoie, R. Roy, J. Newbury, E. Sullivan, J. Benedict, P.
Saunders, K. Wong, D. Canaperi, M. Krishnan, K. Lee, B. A. Rainey, D. Fried, P.
Cottrell, H. P. Wong, M. Ieong, and W. Haensch, “Metal-gate FinFET and fully-
depleted SOI devices using total gate silicidation,” in IEDM Tech. Dig., pp. 247 -
250, 2002.
55. K. Uchida, H. Watanabe, A. Kinoshita, J. Koga, T. Numata, and S. Takagi,

“Experimental study on carrier transport mechanism in ultrathin-body SOI n-and
p-MOSFETs with SOI thickness less than 5 nm,” in IEDM Tech. Dig., pp. 47-50,
2002.
56. S. Takagi, J. Koga, and A. Toriumi, “Subband structure engineering for

performance enhancement of Si MOSFETs,” in IEDM Tech. Dig., pp. 219-222,
1997.
57. T. Low, M. F. Li, W. J. Fan, S. T. Ng, Y. Yeo, C. Zhu, A. Chin, L. Chan, and D. L.
Kwong, “Impact of surface roughness on silicon and germanium ultra-thin-body
MOSFETs,” in IEDM Tech. Dig., pp. 151-154, 2004.
19
Ch 1: Introduction
58. Chenming Hu, “Device Challenges and opportunities,” in Symp. VLSI Tech. Dig.,
pp. 4-5, 2004.
59. S. Takagi, J. L. Hoyt, J. J. Welser, and J. F. Gibbons, “Comparative study of

phonon-limited mobility of two-dimensional electrons in strained and unstrained
Si metal–oxide–semiconductor field-effect transistors,” J. Appl. Phys., vol. 80, pp.
1567-1577, 1996.
60. K. Rim, J. L. Hoyt, and J. F. Gibbons, “Fabrication and analysis of deep

submicron strained-Si NMOSFETs,” IEEE Tran. Electron. Dev., vol. 47, pp.
1406-1415, 2000.
61. S. E. Thompson, M. Armstrong, C. Auth, M. Alavi, M. Buehler, R. Chau, S. Cea,

T. Ghani, G. Glass, T. Hoffman, C.-H. Jan, C. Kenyon, J. Klaus, K. Kuhn, Z. Ma,
B. Mcintyre, K. Mistry, A. Murthy, B. Obradovic, R. Nagisetty, P. Nguyen, S.
Sivakumar, R. Shaheed, L. Shifren, B. Tufts, S. Tyagi, M. Bohr, and Youssef El-
Mansy, “A 90-nm logic technology featuring strained-silicon,” IEEE Tran.
Electron. Dev., vol. 51, pp. 1790-1797, 2004.
62. S. Tiwari, M. Fischetti, P. Mooney, and J. Welser, “Hole mobility improvement in

silicon-on-insulator and bulk silicon transistors using local strain,” in IEDM Tech.
Dig., pp. 939-941, 1997.
63. A. Shimizu, K. Hachimine, N. Ohki, H. Ohta, M. Koguchi, Y. Nonaka, H. Sato,

and F. Ootsuka, “Local mechanical-stress control (LMC): A new technique for
CMOS-performance enhancement,” in IEDM Tech. Dig., pp. 433-436, 2001.
64. H. S. Yang, R. Malik, S. Narasimha, Y. Li, R. Divakaruni, et al., “Dual stress liner
for high performance sub-45nm gate length SOI CMOS manufacturing,” in IEDM
Tech. Dig., pp. 1075-1078, 2004.
65. A. Steegen, M. Stucchi, A. Lauwers, and K. Maex, “Silicide induced pattern

density and orientation dependent transconductance in MOS transistors,” in IEDM
Tech. Dig., pp. 497-500, 1999.
66. T. Ghani, M. Armstrong, C. Auth, M. Bost, P. Charvat, G. Glass, T. Hoffmann, K.

Johnson, C. Kenyon, J. Klaus, B. McIntyre, K. Mistry, A. Murthy, J. Sandford, M.
Silberstein, S. Sivakumar, P. Smith, K. Zawadzki, S. Thompson, and M. Bohr, “A
90nm high volume manufacturing logic technology featuring novel 45nm gate
length strained silicon CMOS transistors,” in IEDM Tech. Dig., pp. 978-980, 2003.
67. K. W. Ang, K. J. Chui, V. Bliznetsov, A. Du, N. Balasubramanian, M. F. Li, G.

Samudra, and Y. Yeo, “Enhanced performance in 50 nm N-MOSFETs with
silicon-carbon source/drain regions,” in IEDM Tech. Dig., pp. 1069-1071, 2004.
68. K. Rim, J. Chu, H. Chen, K.A. Jenkins, T. Kanarsky, K. Lee, A. Mocuta, H. Zhu,
R. Roy, J. Newbury, J. Ott, K. Petrarca, P. Mooney, D. Lacey, S. Koester, K.
Chan, D. Boyd, M. Ieong, and H.-S. Wong, “Characteristics and device design of
sub-100 nm strained Si N- and PMOSFETs,” in Symp. VLSI Tech. Dig., pp. 98-99,
2002.
20
Ch 1: Introduction
69. T. Mizuno, S. Takagi, N. Sugiyama, J. Koga, T. Tezuka, K. Usuda, T.

Hatakeyama, A. Kurobe, and A. Toriumi, “High performance strained-Si p-
MOSFETs on SiGe-on-insulator substrates fabricated by SIMOX technology,” in
IEDM Tech. Dig., pp. 934-936, 1999.
70. T. Tezuka, N. Sugiyama, T. Mizuno and S. Takagi, “High-performance strained-

Si-on-insulator MOSFETs by novel fabrication processes utilizing Ge-
condensation technique,” in Symp. VLSI Tech. Dig., pp. 96-97, 2002.
71. K. Rim, K. Chan, L. Shi, D. Boyd, J. Ott, N. Klymko, F. Cardone, L. Tai, S.

Koester, M. Cobb, D. Canaperi, B. To, E. Duch, I. Babich, R. Carruthers, P.
Saunders, G. Walker, Y. Zhang, M. Steen, and M. Ieong, “Fabrication and
mobility characteristics of ultra-thin strained Si directly on insulator (SSDOI)
MOSFETs,” in IEDM Tech. Dig., pp. 47-52, 2003.
72. M. Yang, M. Ieong, L. Shi, K. Chan, V. Chan, A. Chou, E. Gusev, K. Jenkins, D.

Boyd, Y. Ninomiya, D. Pendleton, Y. Surpris, D. Heenan, J. Ott, K. Guarini, C.
D'Emic, M. Cobb, P. Mooney, B. To, N. Rovedo, J. Benedict, R. Mo, and H. Ng,
“High performance CMOS fabricated on hybrid substrate with different crystal
orientations,” in IEDM Tech. Dig., pp. 453-456, 2003.
73. C. D. Sheraw, M. Yang, D. M. Fried, G. Costrini, T. Kanarsky, W-H. Lee, V.

Chan, M. V. Fischetti, J. Holt, L. Black, M. Naeem, S. Panda, L. Economikos, J.
Groschopf, A. Kapur, Y. Li, R. T. Mo, A. Bonnoit, D. Degraw, S. Luning, D.
Chidambarrao, X. Wang, A. Bryant, D. Brown, C-Y. Sung, P. Agnello, M. Ieong,
S-F. Huang, X. Chen, and M. Khare, “Dual stress liner enhancement in Hybrid
Orientation Technology,” in Symp. VLSI Tech. Dig., pp. 12-13, 2005.
74. H. Shang, H. Okorn-Schmidt, K. K. Chan, M. Copel, J. A. Ott, P. M. Kozlowski,

S. E. Steen, S. A. Cordes, H. P. Wong, E. C. Jones, and W. E. Haensch, “High
mobility p-channel germanium MOSFETs with a thin Ge oxynitride gate
dielectric,” in IEDM Tech. Dig., pp. 441-444, 2002.
75. H. Shang, J. O. Chu, S. Bedell, E. P. Gusev, P. Jamison, Y. Zhang, J. A. Ott, M.

Copel, D. Sadana, K. W. Guarini, and M. Leong, “Selectively formed high
mobility strained Ge PMOSFETs for high performance CMOS,” in IEDM Tech.
Dig., pp. 157-160, 2004.
76. C. H. Huang, M. Y. Yang, Albert Chin, W. J. Chen, C. X. Zhu, B. J. Cho, M.-F.

Li, and D. L. Kwong, “Very low defects and high performance Ge-on-insulator p-
MOSFETs with Al2O3 gate dielectrics,” in Symp. VLSI Tech. Dig., pp. 119-120,
2003.
77. N. Wu, Q. Zhang, C. Zhu, D. S. H. Chan, M.-F. Li, N. Balasubramanian, A. Chin,

and D. L. Kwong, “Alternative surface passivation on germanium for metal-
oxide-semiconductor applications with high-k gate dielectric,” Appl. Phys. Lett.
vol. 85, pp. 4127-4129, 2004.
21
Ch 2: Developments in Metal Gate Materials for CMOS Technology
Chapter 2
Developments in Metal Gate Materials

for CMOS Technology
2.1 Limitations of Poly-Si Gate Electrode
Before discussing the metal gate technology, we first briefly review the
advantages and limitations of the conventional polycrystalline-Si (poly-Si) gate
electrode in today’s CMOS technology.
Poly-Si has been used as the gate electrode material in MOSFETs since 1970s.
One advantage of the poly-Si gate is its excellent thermal stability with SiO2, which
enables the self-aligned gate-first process to be implemented in IC manufacturing.
Another inherent advantage of poly-Si is the easy adjustment of its work function by
dopant implantation, which enables the dual work function poly-Si gate process to be
realized. These merits make the poly-Si a superior gate electrode material for the
gate-first CMOS process.
However, as the transistors scales into the high-performance (HP)-45-nm
technology node and beyond, some fundamental limits of poly-Si become more and
more serious and tends to retard the advancing of CMOS performance, as will be
discussed below.
22
2.1.1 Poly-Si Depletion Effect
The poly-Si depletion effect happens when a MOSFET is working in the
inversion region [1]. Fig. 2.1 illustrates the poly-Si depletion effect in a NMOS
device with n+ poly-Si gate. When a positive bias is applied on the n+ poly-Si gate, a
depletion layer with a finite thickness will be formed in the poly-Si gate side near the
poly-Si/oxide interface, leading to a non-negligible band bending and voltage drop in
the poly-Si gate. As a result, the actual voltage applied on the gate capacitor becomes
smaller, such that less inversion charges will be formed in the channel side. This
causes a loss of gate control and a reduction of MOSFET drive current. The result is
equivalent to introducing a capacitor Cpoly in series with Cox.
1 1 1 1 CET
= + + =
CEff C poly Cox CSi ε 0ε SiO 2
(a) (b)
Fig. 2.1 (a) The energy band diagram of an NMOS device showing the poly-Si gate
depletion effect; (b) Equivalent circuit for the gate stack of MOSFET. CEff denotes
the total gate capacitance which determines the inversion charge density Qi in the
channel, Cpoly, Cox, CSi represent the capacitance from the poly-depletion, gate oxide,
and substrate, respectively. CSi is further broken up into a depletion-layer capacitance
Cd and inversion-layer capacitance Ci. CET represents the capacitance equivalent
thickness of the MOS gate stack, and ψs is the surface potential.
23
The equivalent circuit diagram is shown in Fig. 2.1 (b). The existence of the
poly-Si depletion capacitance CPoly is equivalent to using a capacitor with thicker gate
oxide. Assuming a heavily doped poly-Si with active doping concentration of 1.5 ×
1020 cm-3 (actually this could be difficult for p+ poly-Si with boron), channel doping
concentration of 5 × 1018 cm-3 [2], and inversion charge of 1 × 1013 cm-2, the poly-Si
depletion layer thickness will be around 1.2 nm, which is equivalent to 0.4-nm SiO2 in
terms of the capacitance. Obviously, this none-scalable additive component makes all
other efforts in the vertical scaling of MOSFET less pronounced, since the gate oxide
thickness has been reduced to about 1.2 nm in today’s technology.
The equivalent thickness due to poly-depletion can be reduced by increasing
the doping concentration in poly-Si, but this may aggravate the dopant penetration
problem, which will be discussed in the following section. Moreover, the active
dopant concentration will finally be restricted by the solid solubility of dopants in
poly-Si. In order to eliminate the poly-depletion effect ultimately, the poly-Si gate
needs to be replaced by metal gate electrodes in which the electron density is high
enough to make the gate depletion layer negligible.
2.1.2 Dopant Penetration Effect
As discussed in Chapter 1, the vertical scaling of MOSFET gate stack requires
continuous reduction of gate oxide thickness below 1 nm. On the other hand, the
doping in poly-Si is becoming very heavy (more than 1020 cm-3) in order to minimize
the poly-depletion effect and reduce the sheet resistance of poly-Si. As a result,
dopant penetration from the heavily doped poly-Si gate into the channel would be a
severe issue, especially for boron in p+ poly-Si gate because boron can easily
segregate into SiO2. The dopant penetration effect will affect the threshold voltage of
24
transistors and raise some reliability concerns. It was suggested that using high-κ
dielectrics with physical thickness larger than SiO2 may help to address this problem.
However, it has been reported that dopant penetration also happens for some high-κ
materials like HfO2 [3], implying that this issue still need to be addressed for high-κ
dielectrics.
One way to suppress the dopant penetration problem is to incorporate nitrogen
into gate oxide. It has been shown that nitrogen in SiON [4] and high-κ HfSiON [5]
dielectrics can effectively block the boron from penetrating into the channel region,
which was suggested to be due to the particular Si-O-N bonding lattice formed in
silicon nitride and oxynitride [6]. However, it has been found that a large amount of
nitrogen near the dielectric/Si interface would degrade the channel mobility and thus
the drive current [7]. Therefore, a sandwich nitrogen profile would be advantagous,
and precise nitrogen profile control will be the key to optimize the SiON dielectric
further [8]. But this would be more and more difficult as the gate oxide thickness
scales into the sub-1-nm regime.
Therefore, using a dopant-free metal gate electrode to replace heavily doped
poly-Si gate would be a potential solution to avoid the dopant penetration problem
ultimately.
2.1.3 Gate Electrode Resistivity
Another important parameter for a gate electrode is the resistivity. The
significance of the gate electrode resistivity is that it directly determines the gate RC
delay of digital circuits, particularly in radio-frequency (RF) applications [9].
Typically the sheet resistance of a gate electrode needs to be kept below 5 Ω/
according to ITRS roadmap. However, as the channel length Lg becomes smaller and
25
smaller, the gate electrode thickness (Tgate) should be scaled down with Lg in order to
maintain roughly a constant aspect ratio (Tgate/Lg ≈2) [2]. On the other hand, the
contact silicide thickness also has to be scaled down because the maximum silicon
consumption in the salicide process should not exceed half of the S/D junction depth
Xj [2]. As a result, it is necessary to reduce the resistivity of the poly-Si layer steadily
to meet the sheet resistance specification for the gate electrodes. These relationships
are summarized in Table 2.1.
Table 2.1 Specifications for the scaling of gate electrode, derived from ITRS-2005 [2].
Year 2004 2006 2008 2010 2012
Technology Node /
90 70 57 45 36
DRAM half pitch (nm)
Physical gate length for
37 28 22 18 14
MPU/ASIC (nm)
Average gate electrode sheet
<5 <5 <5 <5 <5
resistance (Ω/ )
Contact silicide thickness
20 19 15 12 9
(nm)
Gate electrode thickness (nm)
74 56 46 36 28
(poly-Si or metal gate)
Contact silicide sheet
7.9 8.6 10.5 13.5 17.3
resistance (for NiSi) (Ω/ )
Poly-Si sheet resistance (Ω/ ) 13.6 11.9 9.5 7.9 7.0
Poly-Si resistivity (μΩ-cm) 100.8 66.9 43.9 28.6 19.7
The way to reduce the resistivity of poly-Si is to increase the doping
concentration. However, the active doping concentration in poly-Si would be limited
by the solid solubility of the dopants in Si. For the commonly used dopants such as B,
26
P, and As, the solid solubility at 1100oC are ~ 5×1020 cm-3, ~ 1.2×1021 cm-3, and ~
2×1021 cm-3, respectively [10]. Therefore, the minimum resistivity values for poly-Si
with B, P, and As dopants would be about 268 μΩ-cm, 74 μΩ-cm, and 58 μΩ-cm,
respectively. Obviously, these values are not adequate to meet the specifications
summarized in Table 2.1. This suggests that the choice of gate materials need to be
reconsidered and metal gates with lower resistivity than poly-Si could be useful to
solve this problem.
2.1.4 Compatibility with high-κ dielectrics
As discussed in Chapter 1, a high-κ dielectric would probably be finally
required to breakthrough the scaling limits of SiO2 and SiON dielectrics in terms of
gate leakage. Beside the considerations from the high-κ dielectrics themselves,
however, the interface quality between poly-Si and some high-κ materials is also a big
challenge for the implementation of high-κ with conventional n+/p+ poly-Si electrodes.
First and foremost, the Fermi-level pinning (FLP) phenomenon will lead to an
undesirable shift of the flat-band voltage (VFB) when n+ and p+ poly-Si gate electrodes
are in contact with many Hf-based high-κ dielectrics. This is particularly severe for p-
MOSFET (ΔVFB ≈ 0.5-0.6 V) [11]-[13], and causes asymmetric high threshold
voltages for n- and p-MOSFETs, making them difficult to be used for circuit design
[14]. The origin of the FLP problem has been attributed to diverse effects. Dopant
penetration from poly-Si gate into the dielectric [11] or the formation of HfB2 at the
interface between p-type poly-Si and HfO2 [15] were proposed to explain the high Vth
in p-MOSFET. However, some later studies showed that the high Vth in p-MOSFET
had been established during the poly-Si deposition, irrespective of the dopant-type
and the activation process [16]. Electron energy loss spectroscopy (EELS)
27
measurement also showed no evidence of HfB2 formation at the poly-Si/HfO2
interface [17]. C. W. Yang et al. suggested that the FLP effect can be attributed to the
formation of acceptor- and donor-like interface states, but the origin of the interface
states was not identified [14]. C. Hobbs et al. proposed a Hf-Si bond induced dipole
theory to explain the observed FLP phenomenon at the poly-Si/HfO2 interface [18]-
[19]. According to this theory, electrons can transfer from Hf to Si and create an
interface dipole layer, which “pins” the effective work function (EWF) of p+ poly-Si
at a position near the conduction band (Ec) of Si. The formation of the dipole layer at
the p+-poly-Si/HfO2 interface was then evidenced by a potential imaging method
using atomic force microscopy (AFM), and the “pinning” position was found to be
about 0.4 eV below Ec [20]. Recently, K. Shiraishi et al. proposed another model in
which the VFB shift for p+ poly-Si on Hf-based dielectric was attributed to the oxygen
vacancy (Vo) promoted charge transfer across the interface [17]. In this model, the
interactions between poly-Si and HfO2 will cause oxygen transport from the ionic
high-κ material into poly-Si, leaving an oxygen vacancy Vo near the poly-Si/high-κ
interface and two weakly bonded electrons in the high-κ film. These electrons can
easily transfer into the conduction band of Si and leave positively charged Vo+ in the
dielectric side, resulting in the formation of dipole layer across the poly-Si/high-κ
interface and consequent shift of VFB. However, it is difficult to explain the opposite
shift of VFB for n+ and p+ poly-Si gates using this model.
Apart from the disagreements in understanding the origin of the FLP effect, it
is also a practical challenge to minimize this effect in the device fabrication process
and to obtain reasonably low Vth for p-MOSFET. M. Koyama et al. demonstrated a
method to reduce the FLP effect by carefully engineering the gradient of Hf in
HfSiON film, where a Hf-ratio below 10% near the poly-Si/HfSiON interface is
28
required to achieve 0.64 V VFB difference between n+ and p+ poly-Si [21]. But the
disadvantage of this method is the over-all dielectric constant of the high-κ dielectric
will be sacrificed. Another way to reduce the high Vth for p-FET is to cap the Hf-
based high-κ by a thin AlOx layer in the p-FET region [22]-[23]. However, this will
lead to a different equivalent oxide thickness (EOT) for n-FET and p-FET, as well as
immense challenges in process integration. Finally, adding Ge into the poly-Si gate
or using n+ Ge as gate electrode was proposed as a potential solution to minimize FLP
due to the smaller possibility to form Vo at the Ge/high-κ interface [24], but the
process integration will be a potential issue.
The thermodynamic instability between poly-Si and many high-κ materials
will be another concern for the applications of poly-Si with high-κ dielectrics. Poly-
Si was found to be reactive with some of the high-κ materials, e.g. ZrO2 and HfO2,
which leads to excess leakage current and poor charge trapping properties in the high-
κ films. Inserting an Al2O3, Si3N4 or amorphous-Si layer between poly-Si and HfO2
had been demonstrated to improve the thermodynamic stability and lower the gate
leakage density of poly-Si/HfO2 stack by several orders [13], [25]-[26]. Incorporating
nitrogen into the gate dielectric to form HfON or HfSiON is another approach to
improve the thermal/electrical stability of poly-Si/high-κ gate stacks [27]-[29], [5].
In summary, the conventional poly-Si gate has been an ideal gate electrode
material in the SiO2 era, whereas many issues arise when the gate stack of MOSFET
scales into the sub-1-nm regime. Metal gate technology is believed to be a promising
technology solution because it not only eliminates the poly-Si depletion and dopant
penetration effects, but also greatly reduces the gate electrode resistivity. Therefore
the metal gate technology has been extensively studied in recent years.
29
2.2 Post Polysilicon Era: Metal Gate Technology
2.2.1 Historical Perspective of Metal Gate Electrodes
The technology evolution always advances in a cyclical way. In the early era
of MOSFET fabrication, aluminium (Al) metal was selected as the gate electrode
material of MOS devices until 1968. At that time, a low-thermal-budget gate-last
process, in which the source/drain was formed prior to the gate deposition and
patterning, was used for MOSFET fabrication with an Al gate electrode. However,
this scheme makes it quite difficult to align the gate with the channel accurately,
which limits the development of more complex logic circuits. A self-aligned gate-
first process was then proposed by R. W. Bower in 1966. However, no one could
make the idea work since the melting point of Al is only about 660oC. This is not
adequate for the dopant activation in Si processing, which is typically higher than
900oC. Fortunately, F. Faggin in Fairchild® invented a silicon gate process and
adopted this technology into their Fairchild 3708 chip firstly in 1968, which shows
improved performance over its Al gate counterpart. From then on, the poly-Si gate
technology as well as the self-aligned process has been developed into a standard
process for MOSFET fabrication.
Today, poly-Si electrodes have encountered many challenges and are expected
to be replaced by metal gates in the sub-50-nm regime. However, the biggest
challenge is the difficulty to identify the appropriate metal(s). Many issues in
material selection and process integration need to be thought out before we reenter the
metal gate era.
30
2.2.2 Considerations for Metal Gate Candidates
2.2.2.1 Work Function Requirements
One of the most important parameters for metal gate candidates is the work
function (WF) because it directly affects the threshold voltage of a MOSFET. The
threshold voltage of a MOSFET is typically given by the following expression [30]:
Qd 4ε Si qN bφB
Vth = VFB + 2φB + = VFB + 2φB + (2-1)
Cox Cox
where φB is the difference between the Fermi-level and the intrinsic level in Si, εSi is
the permittivity of Si, Qd is the total depletion charge in the channel region, Nb is the
doping concentration of the Si substrate (for uniform channel doping), Cox is the gate
oxide capacitance per unit area, and VFB is the flat band voltage across the MOS stack.
VFB can be further given by the following equation:
Qox Q
VFB = Φ MS − = (Φ M − Φ S ) − ox (2-2)
Cox Cox
where ΦMS denotes the work function difference between the metal gate and silicon
substrate, Qox represents the equivalent oxide charge density at the oxide/Si interface.
It is therefore clear that Vth is jointly controlled by Nb and ΦM for a given technology
(for a given Cox and Qox). For sub-50-nm bulk-Si devices, the optimal work function
values required for NMOS and PMOS should be about 4.05~4.25 eV and 4.97~5.17
eV, respectively [31]. In other words, the Fermi level of metal gates should be within
0.2 eV from the band-edges of Si. For the fully-depleted and multi-gate devices, e.g.
FDSOI or FinFET, the Vth will be determined by the gate work function only since the
channel region is almost intrinsic; accordingly metal gates with work functions of ±
0.15 eV from the midgap position of Si will be best served for high-performance
applications (± 0.1 eV for low-standby-power (LSTP) applications) [2], [32]. Recent
31
study shows that when the body thickness of UTBSOI devices shrinks to less than 5
nm, band-edge work function will again be required due to the carrier quantization
effect [33].
Fig. 2.2 summarizes the vacuum work function values of various metal
materials in nature. It shows that metals such as Ta, Hf, Zr, In, Al, and Nb could be
useful for NMOS, while Ni, Ir, Re, Rh, and Pt could be used in PMOS, and W, Cr,
and Mo could be possible metal gate candidates for fully-depleted or multi-gated
devices. Note that these inferences are only from the vacuum work function point of
view, and may not be valid or practical in a real MOS system. Actually, the effective
work function of metal gates in MOSFET will be affected by many factors, such as
crystallinity of metals, metal-dielectric interface, metal deposition techniques, and
even the thermal process used in the device fabrication. These effects will be
discussed in Chapter 3.
6.0
Pt(5.65)
5.5 Ir,P+Poly(5.3)
PMOS Ni(5.15)
Work Function (eV)
Au(5.1)
5.0 Os(4.83) Rh(5.0)
Re(4.95)
Mo(4.7)
Cr(4.55) Ru(4.8)
4.5 NMOS Ta,Cd,n+Poly(4.2)Ti,Ag,Nb(4.3)

Hg(4.52)
Sn(4.38)
W(4.6)
Fe(4.31)
Sb(4.08) Al(4.25)
4.0 Hf(3.9)
In(4.1)
Zr(4.05)
Be(3.92)
Mg(3.64) Mn(3.83)
3.5 La(3.5)
Tb(3.1)
3.0
Yb(2.6)
2.5
Fig. 2.2 Work function of some metal elements collected from experiments [34].
32
2.2.2.2 Thermal Stability Considerations
Besides the consideration for proper work function, another concern for metal
gate candidates is the thermal stability in process. The thermal stability requirements
for the metal candidates will vary with the different process sequences used in
fabricating the MOSFET. In a gate-first process, the metal gates are formed prior to
the source/drain implantation and dopant activation. This implies that the metal gate
material and the metal-dielectric interface should be robust enough to sustain a high
thermal budget used for the dopant activation, which is typically above 1000oC.
The most crucial thermal stability concern for the metal gates is the reaction or
inter-diffusion between metal gate and the underlying gate dielectric. The interface
reaction is thermodynamically driven, and likely to happen at the interface where the
atoms have large differences in electronegativity and radius [35]. Once the reaction
happens, a shift in the device properties with annealing conditions will be expected, in
terms of gate leakage, CET, threshold voltage, gate oxide charge density and gate
oxide reliability. It has been found that many metals with low work function, like Ta,
Hf, Ti, etc, tend to react with the gate dielectric at high temperature, leading to a
reduction of EOT and an increase in gate leakage [36]-[37]. On the other hand, some
high work function metals such as Pt, Ir, and Ni tend to diffuse or penetrate through
the gate oxide during the high temperature annealing process. The penetration of
metals may introduce serious degradations to the gate oxide integrity (GOI), therefore
it needs to be carefully managed.
In addition to the metal reaction/diffusion, some other thermal stability issues
such as microstructure change, stress generation, and oxygen penetration at high-
temperature should also be carefully avoided. Phase change or grain growth may
affect the work function of metal gates and roughen the gate-dielectric interface,
33
leading to a change of Vth and channel mobility upon annealing [38]-[39]. From this
point of view, an amorphous metal gate material would be desirable. The process
related stress generation is another concern for some metal gate candidates. Due to
the extreme high temperature and the ultra-fast temperature ramp up/down rate (200-
250oC/s in spike annealing case) used in a state-of-the-art CMOS manufacturing
process, the thermal induced stress would be a serious problem for metal gates,
especially for those which have very different expansion properties with Si and the
underlying dielectric. The stress in the MOS stack will result in some adherence
problems, and even cause the metal film to crack or peel after annealing. In the light
of this, it is necessary to choose a gate material with a thermal expansion coefficient
close to that of bulk Si. Moreover, some metals also show poor barrier properties
with respect to the diffusion of oxygen at high temperature [40]. This can lead to
growth of an interfacial layer under the high-κ dielectric due to the penetration of
oxygen residues or moisture from the gas ambient during annealing. This effect
makes it challenging to scale the EOT down to the sub-1-nm regime.
2.2.2.3 Process Integration Issues
Process integration is another tough challenge for the implementation of metal
gate electrode in CMOS fabrication. The considerations for the process integration
are closely tied up with the selection of metal gate candidates, which forms a negative
feedback loop until the best solution arrives.
The process integration issues include the deposition techniques, etching and
post-etching cleaning issues, and the dual metal gate integration process. For the dual
metal gate integration, a simple, reliable, and cost-effective scheme to integrate dual
metal gates on the same wafer would be desirable. Basically there are three groups of
34
integration schemes: gate-first, gate-last, and fully-silicided (FUSI) gate processes.
The gate-first process is a relatively cost-effective and reliable process, but it requires
excellent thermal stability for both n+ and p+ metal gate candidates, as well as refined
metal etching and post-etching cleaning techniques [41]-[42]. The gate-last process
has the lowest thermal budget requirement and hence imposes fewer demands on the
selection of metal gate materials, but the process is relatively complex [43]-[44]. The
FUSI gate process is most compatible with the conventional poly-Si CMOS
technology [45]-[46]. More details on the dual metal gate integration schemes will be
discussed in Chapter 5.
The metal gate deposition techniques can affect the properties of metal gate
electrode in many aspects, such as film morphology, resistivity, work function,
thermal stability, and even the gate stack reliability [47]-[50]. The most commonly
used method to deposit metal gate is the physical vapor deposition (PVD) technique,
including sputtering, evaporation, etc [51]. One advantage of the sputtering method is
that it allows easy control of the film composition in metal alloys, metal nitrides,
metal carbides, etc. However, the charged ions used in the sputtering process may
induce plasma damage problems [52]-[53]. The evaporation method can avoid the
plasma damage problem, but it is not as flexible as sputtering in modifying the
composition of metal gate candidates. For all the PVD deposition techniques, a
fundamental limit is the step coverage issue in high aspect ratio structures [51], which
may limit the applications of PVD techniques in 3-D device structures (FinFET, Ω-
gate, nanowire, etc). However, this problem can be addressed by using chemical
vapor deposition (CVD) methods. The CVD methods not only have the advantages
such as good step coverage and low damage to the dielectric, but they provide a
number of variables, viz. temperature, pressure and gas flow which could be useful to
35
control the film microstructure [51]. Among many kinds of CVD techniques, metal-
organic chemical vapor deposition (MOCVD) and atomic-layer chemical vapor
deposition (ALCVD) are two most important forms. MOCVD have the merits of high
deposition rate, good uniformity control, etc. However, it will introduce some
impurities like carbon into the films due to the use of organic precursors. Recently,
ALCVD (or ALD) has drawn considerable attention and has been utilized to deposit
many high-κ and metal gate films. The major advantage of the ALD technique is that
it provides the ability to control the atoms layer by layer such that the film
concentration and even the interface chemistry can be well engineered. This
advantage makes ALD a powerful tool in scientific research. However, one concern
for the ALD technique is its relative long lead time and hence low throughput due to
the need to purge the ambient in every deposition cycle.
Etching and post-etching cleaning of metal gate electrodes is another practical
issue for the integration of metal gate in CMOS process. To achieve high selectivity,
vertical profile, and small feature size in the metal gate etching process, the selection
of masks (photoresist or hard mask), etchant and the post-etching cleaning process
need to be optimized systematically according to the properties of the specific metal
gate materials. These practical issues can even affect the selection of metal gate
candidates.
2.2.2.4 Co-optimization of Metal Gate/High-κ Gate Stack
According to ITRS 2005, it is most likely that metal gate technology will be
introduced into production together with high-κ dielectric. Therefore the co-
optimization of metal gate/high-κ gate stack would be of significance.
36
The first issue that needs to be considered is the effective work function of
metal gate on high-κ dielectric. It has been found that the effective work functions of
some metal gates measured on high-κ dielectrics deviate from the work function
values measured in vacuum and exhibits large dependence on the underlying
dielectric, which makes the work function tuning and threshold voltage adjustment on
high-κ dielectric more complicated. The dependence of the metal work function on
the permittivity of the gate dielectric has been explained by Yeo et al using a metal-
induced gap state (MIGS) theory [54]. Recently, our study indicates that the extrinsic
states relating to the interface chemistry and the thermal process will also affect the
effective work function of metal gates [55]-[56]. This topic will be further discussed
in detail in Chapter 3.
The impact of the metal gate to the carrier mobility in MOSFET has also
drawn considerable attention recently. It has been observed that the electron mobility
in TiN/high-κ stacks is higher than that of poly-Si/high-κ gate stacks, and this was
attributed to the screening of remote phonon scattering by metal gate [57]. Some
recent research by Akasaka et al. reveals that the chemistry of the metal gate electrode
can also affect the electron mobility [58]. Since mobility degradation is one of the
most crucial issues for high-κ gate dielectrics, co-optimization of metal/high-κ stack
to achieve mobility close to that of the poly-Si/SiO2 stacks will become an
indispensable criterion in evaluating potential metal gate electrodes.
2.2.3 Research Status of Metal Gate Technology
According to the different stratagems used to achieve dual work function for
CMOS applications (referring to Chapter 5 for more details), metal gate candidates
can be grouped into several categories, such as direct metal gates, binary metal alloys,
37
and FUSI metal gates. Here the direct metal gates refer to those metal candidates
whose work function is directly determined by the metal material itself after
deposition, without a following process step to modify its work function intentionally.
These metal gate candidates are typically used in a gate-first process. For the binary
(or ternary) metal alloys, the work function are not directly determined by the as-
deposited layered metal stacks; instead, a subsequential annealing process will be
carried out to form new alloys and achieve the final work function. Depending on the
thermal budgets of the alloying process, these metal gates can either be used in a gate-
first integration process or gate-last process. As for FUSI metal gates, the work
function is determined after the silicidation of poly-Si gate. For all these metal gate
solutions, remarkable progress has been achieved in recent research.
2.2.3.1 Direct Metal Gates
The direct metal gates include some pure metals, metal nitrides, metal carbides
and conductive metal oxides. The development of these direct metal gates is mainly
targeted at applications using a gate-first integration scheme, therefore the thermal
stability is a pre-requisite. From this point of view, metal nitrides and metal carbides,
such as TaN, TiN, HfN, WN, and TaC, would be of priority due to their excellent
thermal stability compared with many pure metal materials [40],[50]. In addition,
these metal gate candidates also tend to form a stable interface with high-κ dielectric,
rendering excellent EOT scalability to be achievable. To date, most of the reported
metal gate/high-κ stacks fabricated using the gate-first approach with sub-1-nm EOT
are based on the metal nitride gate electrodes [40],[50].
However, the work functions of most metal nitride materials are close to the
mid-gap position of Si. In order to modulate the work function of metal nitride
38
materials towards the band-edge of Si, several approaches have been explored.
Adding Si [59]-[60] and Al [61] into metal nitrides to form ternary compound have
been demonstrated to be useful for NMOS and PMOS, respectively, but the work
function tuning range is still not ideal. In our work, we investigated a novel approach
to incorporate lanthanide into metal nitrides to reduce the work function of TaN or
HfN, which shows promising characteristics for NMOS (to be discussed in Chapter 4).
Besides these methods, another way to tune the work function of metal nitride
materials is to modify the nitrogen concentration, which can be achieved either by
tuning the N2 flow rate during deposition [62] or by a post-deposition N implantation
process [63]-[64]. The later approach is quite attractive from the process integration
point of view. However, the work function can only be tuned in the range of 4.5 ~
4.95 eV, which is not adequate for bulk-Si CMOS [65].
In terms of the process integration of direct metal gates for dual metal gate
application, the most commonly studied approach is to selectively remove the first-
deposited metal layer in one side (NMOS or PMOS), followed by the deposition of
the second metal material. Many metal gate combinations have been demonstrated
using this approach, such as Ti/Mo, TiN/TaSiN, TaSiN/Ru, etc. [66]-[67], [41]. Note
that the etching of the first-deposited metal layer from the underlying gate dielectric
may introduce some reliability concern if the process is not optimized. In our work,
we proposed a new integration scheme which can avoid the etching of metal film
from the dielectric and hence avoid the exposure of the gate dielectric during process.
This novel process will be discussed in Chapter. 5.
39
2.2.3.2 Binary Metal Alloys
Binary metal alloys are another group of metal gate candidates which have
been studied from the very beginning of metal gate research. Polishchuk et al. first
proposed a Ti-Ni inter-diffusion process to achieve dual work function values for n-
MOSFET and p-MOSFET [68]. The advantage of this approach is that there is no
need to expose the dielectric to any etching process. Instead, the work function tuning
is realized by a metal inter-diffusion process. This approach had been further studied
by many researchers using other material combinations, such as Ru-Ta alloy [69], Ta-
Pt alloy [70], Ti-Pt alloy [71], and Hf-Mo alloy [72] etc. Wide range work function
modulation (more than 1 eV) and precise work function control could be achieved
using this approach. However, an inherent concern of this method is the thermal
stability, which is typically limited by the low work function metals used in the alloys.
As a result, the alloying temperature is typically around 400oC to 600oC. This greatly
limits the applications of these metal candidates in the gate-first CMOS process.
However, they are still promising metal gate solutions for a gate-last integration
process where only the thermal budget for back-end of line (BEoL) processes is
required.
2.2.3.3 Fully-Silicided (FUSI) Metal Gates
All the metal gate solutions mentioned above need notable modification to the
existing poly-Si based CMOS process, which implies high cost and high risk for the
technology migration. For the semiconductor industry, a simple approach with little
modification to the current CMOS process flow would be more attractive in the near-
term. Therefore, the fully-silicided metal gate technology has drawn considerable
attention and steady progresses have been achieved in recent year. Tavel et al. first
40
demonstrated a FUSI CoSi2 process with low sheet resistance (~2Ω/ ) [45]. However,
the work function of CoSi2 is about 4.6 eV, rendering the Vth of bulk-Si MOSFET too
high to be acceptable. Min et al. reported that the work function of FUSI NiSi can be
modulated from about 4.6 eV to 5.0 eV by the dopants in poly-Si [46], which makes
the FUSI gate process more advantageous than any other dual metal gate solutions.
The silicidation induced impurity segregation (SIIS) and pile-up at the silicide-
dielectric interface was believed to be the main mechanism for the work function
modulation in NiSi [73]-[75]. Various impurities have been investigated in tuning the
work function of NiSi, as summarized in Fig. 2.3 [76]. Applications of mid-gap NiSi
in FinFET were also demonstrated successfully with good electrical characteristics
[77].
However, the work function tuning of NiSi through the dopant segregation
approach seems no longer efficient on high-κ dielectric [78]-[79]. This was again
attributed to the Fermi-level pinning effect induced by Hf-Si bonds at the NiSi/high-κ
interface, similar with that in poly-Si/high-κ interface [79]. In order to solve this
problem, phase-controlled (PC) silicide technology was developed [78]. The Ni3Si
phase seems to be useful for p-MOSFET with the high work function of 4.8 eV on
HfSiON dielectric, while NiSi2 phase would be used for n-MOSFET with the low
work function of 4.4 eV [78]. However, the work function tuning range (< 0.4 eV) is
still not large enough for planar bulk-Si devices. Therefore, the PC-silicide process
will be potentially useful only for those novel 3-D devices featuring intrinsic channel
and high-κ dielectric. In addition, controlling the phases of silicide in process is a
challenging task, particularly for MOSFETs with a very small gate length [80].
In order to achieve a larger work function window in FUSI metal gates and
address the dopant fluctuation problem associated with the SIIS approach, some
41
ternary silicides were proposed. The basic idea is to introduce the third element into
the NiSix system to (i) reduce the number of Hf-Si bonds at NiSix/high-κ interface and
(ii) set the work function. Many material systems have been studied, such as Ni1-
xYbxSiy, Ni1-xTaxSiy, or Ni1-xAlxSiy for NMOS and Ni1-xPtxSiy for PMOS [81]-[83], as
summarized in Fig. 2.3.
In some extreme cases such as substituted Al (SA) gate or partial silicided
PtxSi (x >10) processes, the silicon is almost replaced by metals. The work function
will thus be determined by that of the metals at the metal-dielectric interface [84]-[85].
Dopant Effect Phase Control Ternary Silicide Others

5.2 Ev
5.1 PC-NiSix on HfAlON
NiSi on SiO2 Ni(M)Si on HfSiO
or HfO(N) PMOS
5.0 on HfSiON
Work Function (eV)
4.9 NiSi Ni3Si

NiYbxSi
4.8 Control (on SiON)
NiTaxSi
Pt-rich
4.7 Sb As P PtxSi
4.6 Ei
B Al NiSi
4.5
i
t)S
i(P
4.4
-N
Al
4.3 NiSi2 NiPtxSi

4.2 SA-Al NMOS
NiAlxSi
4.1
4.0
Ec
Fig. 2.3 Work function modulation by various mechanisms in some NiSi-based-

silicide metal gates; data from [76], [78], [81]-[85].
In summary, though many successes have been achieved in developing
potential metal gate materials for next generation transistors, the right materials and
the appropriate integration flows have not been identified yet. Fundamental
understanding on the work function tuning mechanisms is still needed, as well as the
solutions which can reach every aspects of the matter.
42
2.3 Challenges in Metal Gate Technology
2.3.1 Understanding of the Metal-Dielectric Interface
The interface between the gate electrode and gate dielectric is of importance
since it determines the effective work function of the metal gate electrode as well as
many other characteristics of the MOSFET. However, the knowledge on this crucial
interface is still insufficient to guide us in developing the metal-gated MOSFET. The
impacts of many factors to the properties of metal-dielectric interface still need to be
identified, including the metal crystallinity, interface chemistry, dangling bonds and
defects, oxygen vacancy in dielectrics, interface reactions, metal and dopant
penetrations, and so on. Simple models depicting the general behavior of this
interface will be appreciated.
Among these issues, one of the most critical challenges is to understand the
role of the metal-dielectric interface in determining the effective work function of
metals. Although Yeo et al. proposed the interface dipole theory to explain the
dependence of the effective work function of metal gates on the underlying dielectric,
it is still not clear whether it can be used to explain the process dependence of metal
work function during high-temperature annealing, which finally determines the Vth of
the MOSFET. Therefore, the factors contributing to the thermal instability of metal
work functions need to be identified and the mechanisms need to be understood.
Moreover, engineering of the material system at the metal-dielectric interface is also
of technical importance for work function tuning of metal gates.
43
2.3.2 Developing Appropriate Metal Gate Materials
As we discussed above, there are many criteria in the material selection and
process integration of novel metal gate candidates. The materials fulfilling all the
requirements have not been identified yet, and novel metal gate candidates with
promising properties are still expected.
For the direct metal gate candidates, one of the biggest difficulties is how to
get the best trade-off between the thermal stability concern and the low work function
requirement for NMOS applications. Developing a thermally stable N-type metal
gate candidate and finding a simple approach to integrate it into a CMOS process will
be of significance from this point of view. For P-type metal candidates, the thermal
stability issue will be less pronounced. However, co-optimization of P-type metal
gates with high-κ dielectrics will be a challenging issue because it has been observed
that the work function of many P-type materials will decrease significantly on high-κ
dielectrics, probably due to the oxygen vacancies induced interface dipoles [86]. A
proper method to minimize the oxygen vacancies and obtain a high work function
above 5.0 eV would be desired.
2.3.3 Dual Metal Gate Integration Issues
Besides the difficulties in developing the individual metal gate materials for
both n- and p-MOSFETs, the integration of dual metal gate in a CMOS process flow
is always a challenge for metal gate technology. Many factors need to be considered
when integrating dual metal gates, such as process complexity, cost, GOI degradation,
process controllability, EOT scalability, and so on. These considerations will in turn
affect the selection of metal materials.
44
For the gate-first CMOS integration flow, one of the major concerns for the
conventional direct-etching integration scheme is the exposure of gate dielectrics to
the etching environment during the metal selective etching process, as we discussed
above in Section 2.2.3.1. A proper integration flow to avoid this concern is needed.
For the gate-last process, there have been some solutions for the dual metal gate
integration process, thanks to the low thermal budget used. However, the co-
optimization of the metal-gate/high-κ stacks should be emphasized in this case,
regarding the EOT scalability, carrier mobility, and device reliability. For the FUSI
metal gate process, integration of different silicide materials and the precise control of
the dopants, phases, and stress in the silicidation process will still be the key issues.
2.4 Research Scope and Major Achievements in This Thesis
The overall objective of this work is to: (i) study the role of the metal-
dielectric interface in determining the effective work function of metal gate electrodes,
and (ii) develop appropriate metal materials as well as proper integration schemes for
the implementation of metal gates in next generation transistors.
For the metal-dielectric interface study, this work will be focused on
understanding the impact of dielectric types and different thermal treatments to the
effective work functions of metal gates. The behavior of metal effective work
function as a function of annealing temperature was investigated in several
combinations of metals and dielectrics. It was observed that the work function
thermal instability problem is more serious in metal/SiO2 systems than in metal/HfO2
stacks. A metal-dielectric interface model that takes the role of extrinsic states into
account was proposed to qualitatively explain the work function thermal instability.
The generation of extrinsic states seems to be thermodynamically driven, and
45
becomes more pronounced as the annealing temperature increases. The interface
chemistry plays an important role in determining the generation of extrinsic states.
These results may provide some useful understanding of the properties of metal-
dielectric interface and be of practical significance in engineering the metal-dielectric
interface for future study. These contents will be discussed in Chapter 3.
Based on the above understanding, the key to modulating the metal gate work
function is to engineer the metal-dielectric interface with proper materials. In Chapter
4, a novel approach to modifying the work function of metal nitride gate electrodes by
incorporating lanthanide elements are investigated systematically for the first time.
Thanks to the very low work function of lanthanide elements (Tb, Er, Yb, etc.), the
work function of metal nitride gate electrodes (TaN, HfN, etc.) can be tuned gradually
from mid-gap position down to ~ 4.2 eV by incorporating lanthanide. Good thermal
stability can also be achieved simultaneously due to the very low lanthanide
concentration and the enhanced nitrogen levels in these materials. These results
suggest that lanthanide-incorporated metal nitrides could be a promising metal gate
candidate for n-MOSFETs.
Moreover, in order to address some of the challenges associated with the
existing dual metal gate integration processes, several potential dual metal gate
integration flows will be proposed and discussed in Chapter 5. The first one is a gate-
first integration process using a novel high-temperature metal intermixing technique
to achieve dual work function. Unlike the conventional dual metal gate integration
scheme where selective etching of the metal material from the gate dielectric is
required, the new integration scheme avoids the exposure of the gate dielectric during
the metal etching process by using an ultra-thin TaN buffer layer on top of the gate
dielectric. The work function of the TaN buffer layer can then be modulated by the
46
intermixing of TaN with other metals in a high-temperature annealing process, which
is compatible with the source/drain dopant activation process in the gate-first CMOS
integration flow. The second integration scheme discussed in Chapter 5 is a gate-last
process utilizing HfN as the dummy gate material. The aim of this approach is to
achieve large work function difference and optimal high-κ properties simultaneously
for the integration of dual metal gates with high-κ materials. The use of the HfN
dummy gate enables high quality high-κ dielectric with EOT of less than 1 nm to be
achieved, as well as the work function difference of ~ 0.8 eV by using Ta and Ni as
the electrodes for NMOS and PMOS, respectively. These attempts could be of
practical values for the community in developing the dual metal gate solutions for
future CMOS technology. Note that the intention in this part of study is not to solve
all the potential issues from the material selection to the process integration in
nanometer level, but to propose the possible technology approaches and discuss their
advantages and limitations.
47
References
1. Y. Taur and T. H. Ning, Fundamentals of Modern VLSI Devices, Cambridge,

U.K., Cambridge Univ. Press, 1998, ch. 2, pp. 74-78.
2. International Technology Roadmap of Semiconductors (ITRS), 2005 edition,

Semiconductor Industry Association (SIA), SanJose, CA 95110. Available:
http://public.itrs.net/
3. K. Onishi, L. Kang, R. Choi, E. Dharmarajan, S. Gopalan, Y. Jeon, C. S. Kang, B.

H. Lee, R. Nieh, and J. C. Lee, “Dopant penetration effects on polysilicon gate
HfO2 MOSFET’s,” in Symp. VLSI Tech. Dig., pp. 131-132, 2001.
4. T. Morimoto, H. S. Momose, Y. Ozawa, K. Yamabe, and H. Iwai, “Effect of

boron penetration and resultant limitations in ultra thin pure-oxide and nitrided-
oxide gate flims,” in IEDM Tech. Dig., pp. 429-432, 1990.
5. M. Koyama, A. Kaneko, T. Ino, M. Koike, Y. Kamata, R. Iijima, Y. Kamimuta,

A. Takashima, M. Suzuki, C. Hongo, S. Inumiya, M. Takayanagi, and A.
Nishiyama, “Effects of nitrogen in HfSiON gate dielectric on the electrical and
thermal characteristics,” in IEDM Tech. Dig., pp. 849-852, 2002.
6. K. A. Ellis and R. A. Buhrman, “Time-dependent diffusivity of boron in silicon

oxide and oxynitride,” Appl. Phys. Lett. vol. 74, pp. 967-969, 1999.
7. H. Yang and G. Lucovsky, “Integration of ultrathin (1.6~2.0 nm) RPECVD

oxynitride gate dielectrics into dual poly-Si gate submicron CMOSFETs,” in
IEDM Tech. Dig., pp. 245-248, 1999.
8. D. Ishikawa, S. Sakai, K. Katsuyama, and A. Hiraiwa, “Nitride-sandwiched-oxide

gate insulator for low power CMOS,” in IEDM Tech. Dig., pp. 869-872, 2002.

10. S. M. Sze, Physics of Semiconductor Devices, 2nd Edition, John Wiley & Son,
1981, ch. 2, p. 69.
11. L. Kang, K. Onishi, Y. Jeon, B. H. Lee, C. Kang, W. Qi, R. Nieh, S. Gopalan, R.

Choi, and J. C. Lee, “MOSFET Devices with polysilicon on single-layer HfO2
high-K dielectrics,” in IEDM Tech. Dig., pp. 35-38, 2000.
12. S.Pidin, Y.Morisaki, Y.Sugita, T.Aoyama, K.Irino, T.Nakamura, and T.Sugii,

“Low standby power CMOS with HfO2 gate oxide for 100-nm generation,” in
13. V. S. Kaushik, E. Rohr, S. De Gendt, A. Delabie, S. Van. Elshocht, M. Claes, X.

Shi, Y. Shimamoto, L.-A. Ragnarsson, T. Witters, Y. Manabe, M. Heyns,
48
“Effects of interactions between HfO2 and poly-Si on MOSCAP and MOSFET

electrical behavior,” Proc. Int. Workshop Gate Insulator (IWGI), pp. 62-63, 2003.

k HfO2 dielectric,” Appl. Phys. Lett. vol. 83, pp. 308-310, 2003.
15. T. Aoyama et al., Int. Workshop on Gate Insulator (IWGI), pp. 175-176, 2003.
16. E. Cartier, V. Narayanan, E. P. Gusev, P. Jamison, B. Linder, M. Steen, K. K.

Chan, M. Frank, N. Bojarczuk, M. Copel, S. A. Cohen, S. Zafar, A. Callegari, M.
Gribelyuk*, M. P. Chudzik, C. Cabral Jr., R. Carruthers , C. D’Emic, J. Newbury,
D. Lacey, S. Guha and R. Jammy, “Systematic study of pFET Vt with Hf-based
gate stacks with poly-Si and FUSI gates,” in Symp. VLSI Tech. Dig., pp. 44-45,
2004.

2004.

D. Triyoso, R. Hegde, D. Gilmer, R. Garcia, D. Roan, L. Lovejoy, R. Rai, L.
polySi/metal oxide interface,” in Symp. VLSI Tech. Dig., pp. 9-10, 2003.
19. C. Hobbs, L. Fonseca, A. Knizhnik, V. Dhandapani, S. Samavedam, W. Taylor, J.

Grant, L. Dip, D. Triyoso, R. Hegde, D. Gilmer, R. Garcia, D. Roan, M. Lovejoy,
R. Rai, E. Hebert, H. Tseng, S. Anderson, B. White, and P. Tobin, “Fermi-level
pinning at the polysilicon/metal oxide interface,” IEEE Tran. Electron. Dev., vol.
51, pp. 971-984, 2004.
20. R. Ludeke, V. Narayanan, E. P. Gusev, E. Cartier, and S. J. Chey, “Potential

imaging of Si/HfO2/polycrystalline silicon gate stacks: Evidence for an oxide
dipole,” Appl. Phys. Lett., vol. 86, 122901, 2005.
21. M. Koyama, Y. Kamimuta, T. Ino, A. Kaneko, S. Inumiya, K. Eguchi, M.

Takayanagi, and A. Nishiyama, “Careful examination on the asymmetric Vfb shift
problem for poly-Si/HfSiON gate stack and its solution by the Hf concentration
control in the dielectric near the poly-Si interface with small EOT expense,” in
IEDM Tech. Dig., pp. 499-502, 2004.
22. H.-S. Jung, J.-H. Lee, S. K. Han, Y.-S. Kim, H. J. Lim, M. J. Kim, S. J. Doh, M.
Y. Yu, N.-I. Lee, H.-L. Lee1, T.-S. Jeon, H.-J. Cho, S. B. Kang, S. Y. Kim, I. S.
Park, D. Kim, H. S. Baik, Y. S. Chung, “A highly manufacturable MIPS (metal
inserted poly-Si stack) technology with novel threshold voltage control,” in Symp.
VLSI Tech. Dig., pp. 232-233, 2005.
49
23. H.-J. Li and M. Gardner, “Dual high-k gate dielectric with poly gate electrode:
HfSiON on nMOS and Al2O3 capping layer on pMOS,” IEEE Electron Device
Lett., vol. 26, pp. 441-443, 2005.
24. H. Takeuchi, H. Y. Wong, D. Ha, and T. King, “Impact of oxygen vacancies on

high-k gate stack engineering,” in IEDM Tech. Dig., pp. 829-832, 2004.
25. D. C. Gilmer, R. Hegde, R. Cotton, R. Garcia, V. Dhandapani, D. Triyoso, D.

Roan, A. Franke, R. Rai, L. Prabhu, C. Hobbs, J. M. Grant, L. La, S. Samavedam,
B. Taylor, H. Tseng, and P. Tobin, “Compatibility of polycrystalline silicon gate
deposition with HfO2 and Al2O3/HfO2 gate dielectrics,” Appl. Phys. Lett., vol. 81,
pp. 1288-1290, 2002.
26. Y. Morisaki, T. Aoyama, Y. Sugita, K. Irino, T. Sugii, and T. Nakamura, “Ultra-

thin (Teffinv=1.7 nm) poly-Si-gated SiN/HfO2/SiON high-k stack dielectrics with
high thermal stability (1050oC),” in IEDM. Tech. Dig., pp. 861-864, 2002.
27. R. M. Wallace, R. A. Stoltz, G. D. Wilk, “Zirconium and/or hafnium oxynitride

gate dielectric,” US Patent 6,013,553, Jan, 2000.
28. R. M. Wallace, R. A. Stoltz, G. D. Wilk, “Zirconium and/or hafnium silicon-

oxynitride gate dielectric,” US Patent 6,020,243, Feb, 2000.
29. C. H. Choi, S. J. Rhee, T. S. Jeon, N. Lu, J. H. Sim, R. Clark, M. Niwa, and D. L.

Kwong, "Thermally stable CVD HfOxNy advanced gate dielectrics with poly-Si
gate electrode," in IEDM Tech. Dig., pp. 857-860, 2002.

31. I. De, D. Johri, A. Srivastava, and C. M. Osburn, “Impact of gate workfunction

on device performance at the 50 nm technology node,” Solid State Electronics,
vol. 44, no. 6, pp. 1077-1080, Jun. 2000.
32. B. Cheng, B. Maiti, S. Samavedam, J. Grant, B. Taylor, P. Tobin, and J. Mogab,

“Metal gates for advanced sub-80-nm SOI CMOS technology,” in IEEE Intl. SOI
Symp. Proc., pp. 91-92, 2001.
33. T. Low, M. F. Li, W. J. Fan, S. T. Ng, Y. Yeo, C. Zhu, A. Chin, L. Chan, and D.
L. Kwong, “Impact of surface roughness on silicon and germanium ultra-thin-
body MOSFETs,” in IEDM Tech. Dig., pp. 151-154, 2004.
34. H. B. Michaelson, “The work function of the elements and its periodicity,” J.
Appl. Phys., vol. 48, pp. 4729-4733, 1977.
35. J. K. Schaeffer, S. B. Samavedam, D. C. Gilmer, V. Dhandapani, P. J. Tobin, J.

Mogab, B.-Y. Nguyen, B. E. White, Jr., S. Dakshina-Murthy, R. S. Rai, Z.-X.
Jiang, R. Martin, M. V. Raymond, M. Zavala, L. B. La, J. A. Smith, R. Garcia, D.
Roan, M. Kottke, and R. B. Gregory, “Physical and electrical properties of metal
50
gate electrodes on HfO2 gate dielectrics,” J. Vac. Sci. Technol. B, vol. 21, no. 1,
pp. 11-17, Jan./Feb. 2003.
36. V. Misra, G. P. Heuss, and H. Zhong, “Use of metal-oxide-semiconductor

capacitors to detect interactions of Hf and Zr gate electrodes with SiO2 and
ZrO2,” Appl. Phys. Lett., vol. 78, no. 26, pp. 4166-4168, 2001.
37. Takeo Ushiki, Kunihiro Kawai, Ichiro Ohshima, and Tadahiro Ohmi, “Chemical
reaction concerns of gate metal with gate dielectric in Ta gate MOS devices: an
effect of self-sealing barrier configuration interposed between Ta and SiO2,”
IEEE Trans. Electron Devices, vol. 47, no.11, pp. 2201-2207, Nov. 2000.
38. S. H. Bae, W. P. Bai, H. C. Wen, S. Mathew, L. K. Bera, N. Balasubramanian, N.

Yamada, M. F. Li, and D. L. Kwong, “Laminated metal gate electrode with
tunable work function for advanced CMOS,” in Symp. VLSI Tech. Dig., pp. 188-
189, 2004.
39. M.-F. Wang, T.-Y. Huang, Y.-C. Kao, H.-C. Lin, and C.-Y. Chang, “Impact of
thermal stability on the characteristics of complementary metal oxide
semiconductor transistors with TiN metal gate,” Jpn. J. Appl. Phys., vol. 41, pp.
546-551, 2002.
40. H. Y. Yu, J. F. Kang, J. D. Chen, C. Ren, Y. T. Hou, S. J. Whang, M.-F. Li, D. S.

H. Chan, K. L. Bera, C. H. Tung, A. Du, and D.-L. Kwong, “Thermally robust
high quality HfN/HfO2 gate stack for advanced CMOS devices,” in IEDM Tech.
Dig., pp. 99-103, 2003.
41. S. B. Samavedam, L. B. La, J. Smith, S. Dakshina-Murthy, E. Luckowski, J.

Schaeffer, M. Zavala, R. Martin, V. Dhandapani, D. Triyoso, H. H. Tseng, P. J.
Tobin, D. C. Gilmer, C. Hobbs, W. J. Taylor, J. M. Grant, R. I. Hegde, J. Mogab,
C. Thomas, P. Abramowitz, M. Moosa, J. Conner, J. Jiang, V. Arunachalam, M.
Sadd, B.-Y. Nguyen, and B. White, “Dual-metal gate CMOS with HfO2 gate
dielectric,” in IEDM Tech. Dig., pp. 433-436,2002.
42. Z. B. Zhang, S. C. Song, C. Huffman, J. Barnett, N. Moumen, H. Alshareef, P.

Majhi, M. Hussain, M.S. Akbar, J.H. Sim, S.H. Bae, B. Sassman, and B.H. Lee,
“Integration of dual metal gate CMOS with TaSiN (NMOS) and Ru (PMOS) gate
electrodes on HfO2 gate dielectric,” in VLSI Tech. Dig., pp. 50-51. 2005.
43. I. Polishchuk, P. Ranade, T.-J. King, and C. Hu, “Dual work function metal gate
CMOS technology using metal interdiffusion,” IEEE Electron Device Lett., vol.
22, pp. 444-446, Sep. 2001.
44. C. Ren, H. Y. Yu, J. F. Kang, X. P. Wang, H. H. H. Ma, Yee-Chia Yeo, D. S. H.

Chan, M.-F. Li, and D.-L. Kwong, “A dual-metal gate integration process for
CMOS with sub-1 nm EOT HfO2 by using HfN replacement gate,” IEEE
Electron Device Lett., vol. 25, pp. 580-582, Aug. 2004.
45. B. Tavel, T. Skotnicki, G. Pares, N. Carrière, M. Rivoire, F. Leverd, C. Julien, J.

Torres, and R. Pantel, "Totally silicided (CoSi2) polysilicon: A novel approach to
51
very low-resistive gate (~2Ω/ ) without metal CMP nor etching," in IEDM Tech.
Dig., pp. 825-828, 2001.
46. M. Qin, V. M. C. Poon, and S. C. Ho, “Investigation of polycrystalline nickel

silicide films as a gate material,” J. Electrochem. Soc., vol. 148, no. 5, pp. G271-
G274, 2001.
47. Y. H. Kim, C. H. Lee, T. S. Jeon, W. P. Bai, C. H. Choi, S. J. Lee, L. Xinjian, R.

Clarks, D. Roberts, and D. L. Kwong, "High quality CVD TaN gate electrode for
sub-100nm MOS devices," in IEDM Tech. Dig., pp. 667-670, 2001.
48. J. C. Hooker, R.J.P Lander, F.N. Cubaynes, T. Schram, F. Roozeboom, J. van Zijl,
M. Maas, F.C. van den Heuvel, E.P. Naburgh, J.G.M. van Berkum, Y. Tamming,
T. Dao, K. Henson, M. Schaekers, A. van Ammel, Z. Tokei, M. Demand, C.
Dachs, “Tantalum-based gate electrode metals for advanced CMOS
applications,” Proc. 207th ECS meeting, 632, Quebec, Canada, 2005.
49. D.-G. Park, K.-Y. Lim, H.-J. Cho, T.-H. Cha, J.-J. Kim, J.-K. Ko, I.-S. Yeo, and J.
W. Park, “Novel damage-free direct metal gate process using atomic layer
deposition,” in VLSI Tech. Dig., pp. 65-66. 2001.
50. D.-G. Park, Z.J. Luo, N. Edleman, W. Zhu, P. Nguyen, K. Wong, C. Cabral1, P.
Jamison1, B.H. Lee, A. Chou, M. Chudzik, J. Bruley, O. Gluschenkov, P.
Ronsheim, A. Chakravarti, R. Mitchell, V. Ku, H. Kim1, E. Duch1, P.
Kozlowski1, C.D’Emic1, V. Narayanan1, A. Steegen, R. Wise, R. Jammy1, R.
Rengarajan, H. Ng, A. Sekiguchi, and C.H. Wann, “Thermally robust dual-work
function ALD-MNx MOSFETs using conventional CMOS process flow,” in
51. S. Wolf and R. N. Tauber, Silicon Processing for the VLSI Era, Volumn 1, 2nd
Edition, California, Lattice Press, 2000, ch. 11& ch. 6.
52. T. Ushiki , M.-C. Yu, K. Kawai, T. Shinohara, K. Ino , M. Morita and T. Ohmi
“Gate oxide reliability concerns in gate-metal sputtering deposition process: an
effect of low-energy large-mass bombardment”, Microelectronics Reliability, vol.
39, pp. 327-332, 1999.
53. T. Yamada, M. Moriwaki, Y. Harada, S. Fujii, and K. Eriguchi, “Effects of the

sputtering deposition process of metal gate electrode on the gate dielectric
characteristics,” Microelectronics Reliability, vol. 41, pp. 697-704, 2001.
54. Y.-C. Yeo, T.-J. King, and C. Hu, “Metal-dielectric band alignment and its
implications for metal gate complementary metal-oxide-semiconductor
technology,” J. Appl. Phys., vol. 92, no. 12, pp. 7266-7271, Dec. 2002.
55. C. Ren, H. Y. Yu, J. F. Kang, Y. T. Hou, M.-F. Li, W. D. Wang, D. S. H. Chan,

and D.-L. Kwong, “Fermi-level pinning induced thermal instability in the
effective work function of TaN in TaN/SiO2 gate stack,” IEEE Electron Device
Lett., vol. 25, pp. 123-125, Mar. 2004.
52
56. H. Y. Yu, C. Ren, Y.-C. Yeo, J. F. Kang, X. P. Wang, H. H. H. Ma, M.-F. Li, D.
S. H. Chan, and D.-L. Kwong, “Fermi pinning induced thermal instability of
metal gate work functions,” IEEE Electron Device Lett., vol. 25, pp. 337-339,
May 2004.
57. R. Chau, S. Datta, M. Doczy, B. Doyle, J. Kavalieros, and M. Metz, “High-

k/metal–gate stack and its MOSFET characteristics,” IEEE Electron Device Lett.,
vol. 25, pp. 408-410, Jun. 2004.
58. Y. Akasaka, K. MiyagawaA, T. Sasaki,K. Shiraishi, S. Kamiyama, O. Ogawa, F.

Ootsuka and Y. Nara, “Impact of electrode-side chemical structures on electron
mobility in metal/HfO2 MISFETs with sub-1nm EOT,” in VLSI Tech. Dig., pp.
228-229. 2005.
59. Y.-S. Suh, G. Heuss, H. Zhong, S.-N. Hong, and V. Misra, “Electrical
characteristics of TaSixNy gate electrodes for dual gate Si-CMOS devices,” in
VLSI Tech. Dig., pp. 47-48, 2001.
60. H. C. Wen, H. N. Alshareef, H. Luan, K. Choi, P. Lysaght, H. R. Harris, C.

Huffman, G. A. Brown, G. Bersuker, P. Zeitzoff, H. Huff, P. Majhi, and B. H.
Lee, “Systematic investigation of amorphous transition-metal-silicon-nitride
electrodes for metal gate CMOS applications,” in Symp. VLSI Tech. Dig., pp. 46-
47, 2005.
61. T.-H. Cha, D.-G. Park, T.-K. Kim, S.-A. Jang, I.-S. Yeo, J.-S. Roh, and J. W.
Park, “Work function and thermal stability of Ti1-xAlxNy for dual metal gate
electrodes,” Appl. Phys. Lett., vol. 81, pp. 4192-4194, 2002.
62. C. S. Kang, H.-J. Cho, Y. H. Kim, R. Choi, K. Onishi, A. Shahriar, and J. C. Lee,
“Characterization of resistivity and work function of sputtered-TaN film for gate
electrode applications,” J. Vac. Sci. Technol. B, vol. 21, pp. 2026-2028, 2003.
63. H. Wakabayashi, Y. Saito, K. Takeuchi, T. Mogami, and T. Kunio, “A dual-metal

gate CMOS technology using nitrogen-concentration-controlled TiNx film,” IEEE
Trans. Electron Devices, vol. 48, pp. 2363-2369, 2001.
64. P. Ranade, Y.-C. Yeo, Q. Lu, H. Takeuchi, T.-J. King, C. Hu, “Molybdenum as a
gate electrode for deep sub-micron CMOS technology”, Mat. Res. Soc. Symp., vol.
611, pp.C3.2.1-C3.2.6, 2000.
65. P. Ranade, Y. Choi, D. Ha, A. Agarwal, M. Ameen, and T. King, “Tunable work
function molybdenum gate technology for FDSOI-CMOS,” in IEDM Tech. Dig.,
pp. 363-366, 2002.
66. Y.-C. Yeo, Q. Lu, P. Ranade, H. Takeuchi, K. J. Yang, I. Polishchuk, T.-J. King,
C. Hu, S. C. Song, H. F. Luan, and D.-L. Kwong, “Dual-metal gate CMOS
technology with ultra-thin silicon nitride gate dielectric,” IEEE Electron Device
Lett., vol. 22, no. 5, pp. 227-229, May 2001.
53

electrodes on HfO2 gate dielectric,” in VLSI Tech. Dig., pp. 50-51, 2005.
22, pp. 444-446, Sep. 2001.
69. J. Lee, H. Zhong, Y. Suh, G. Heuss, J. Gurganus, B. Chen, and V. Misra,

“Tunable work function dual metal gate technology for bulk and non-bulk
CMOS,” in IEDM Tech. Dig., pp. 359-362, 2002.
70. B.-Y. Tsui, and C.-F. Huang, “Wide range work function modulation of binary
alloys for MOSFET application,” IEEE Electron Device Lett., vol. 24, pp. 153-
155, Mar. 2003.
71. I. S. Jeon, J. Lee, P. Zhao, P. Sivasubramani, T. Oh, H. J. Kim, D. Cha, J. Huang,

M. J. Kim, B. E. Gnade, J. Kim, and R. M. Wallace, “A novel methodology on
tuning work function of metal gate using stacking bi-metal layers,” in IEDM Tech.
Dig., pp. 303-306, 2004.
72. T.-L. Li, C.-H. Hu, W.-L. Ho, H. C.-H. Wang, and C.-Y. Chang, “Continuous and
precise work function adjustment for integratable dual metal gate CMOS
technology using Hf–Mo binary alloys,” IEEE Trans. Electron Devices, vol. 52,
pp. 1172-1179, 2005.
73. W. P. Maszara, Z. Krivokapic, P. King, J. Goo, and M. Lin, “Transistors with

dual work function metal gates by single full silicidation (FUSI) of polysilicon
gates,” in IEDM Tech. Dig., pp. 367-370, 2002.
74. J. H. Sim, H. C. Wen, J. P. Lu, and D. L. Kwong, “Dual work function metal
gates using full nickel silicidation of doped poly-Si,” IEEE Electron Device Lett.,
vol. 24, pp. 631-633, Oct. 2003.
75. J. Kedzierski, D. Boyd, C. Cabral, Jr., P. Ronsheim, S. Zafar, P. M. Kozlowski, J.

A. Ott, and M. Ieong, “Threshold voltage control in NiSi-gated MOSFETs
through SIIS,” IEEE Trans. Electron Devices, vol. 52, pp. 39-46, 2005.
76. C. Cabral, Jr., J. Kedzierski, B. Linder, S. Zafar, V. Narayanan, S. Fang, A.

Steegen, P. Kozlowski, R. Carruthers , and R. Jammy, “Dual workfunction fully
silicided metal gates,” in VLSI Tech. Dig., pp. 184-185, 2004.
77. J. Kedzierski, E. Nowak, T. Kanarsky, Y. Zhang, D. Boyd, R. Carruthers, C.

Cabral, R. Amos, C. Lavoie, R. Roy, J. Newbury, E. Sullivan, J. Benedict, P.
Saunders, K. Wong, D. Canaperi, M. Krishnan, K. Lee, B. A. Rainey, D. Fried, P.
Cottrell, H. P. Wong, M. Ieong, and W. Haensch, “Metal-gate FinFET and fully-
depleted SOI devices using total gate silicidation,” in IEDM Tech. Dig., pp. 247-
250, 2002.
54
78. K. Takahashi, K. Manabe, T. Ikarashi, N. Ikarashi, T. Hase, T. Yoshihara, H.

Watanabe, T. Tatsumi, and Y. Mochizuki, “Dual workfunction Ni-
silicide/HfSiON gate stacks by phase-controlled full-silicidation (PC-FUSI)
technique for 45nm-node LSTP and LOP devices,” in IEDM Tech. Dig., pp. 91-
94, 2004.
79. M. Terai, K. Takahashi, K. Manabe, T. Hase, T. Ogura, M. Saitoh, T. Iwamoto, T.

Tatsumi, and H. Watanabe, “Highly reliable HfSiON CMOSFET with phase
controlled NiSi (NFET) and Ni3Si (PFET) FUSI gate electrode,” in VLSI Tech.
Dig., pp. 68-69, 2005.
80. J. A. Kittl, A. Veloso, A. Lauwers, K. G. Anil, C. Demeurisse, S. Kubicek, M.

Niwa, M. J. H. van Dal, O. Richard, M. A. Pawlak, M. Jurczak, C. Vrancken, T.
Chiarella, S. Brus, K. Maex and S. Biesemans, “Scalability of Ni FUSI gate
process: phase and Vt control to 30 nm gate lengths,” in VLSI Tech. Dig., pp. 72-
73, 2005.
81. H. Y. Yu, J. D. Chen, M. F. Li, S. J. Lee, D. L. Kwong, M. van Dal, J.A. Kittl, A.
Lauwers, E. Augendre, S. Kubicek, C. Zhao, H. Bender, B. Brijs, L. Geenen, A.
Benedetti, P. Absil, M. Jurczak, and S. Biesemans, “Modulation of the Ni FUSI
workfunction by Yb doping: from midgap to n-type band-edge,” in IEDM Tech.
Dig., pp. 630-633, 2005.
82. N. Biswas, S. Novak, B. Chen, D. Lichtenwalner, M. Ozturk and V.Misra,

“NixTa1-xSi and NixPt1-xSi ternary alloys for work function tuning on SiO2,
HfSiOx and HfO2 dielectrics,” in IEDM Tech. Dig., pp. 650-653, 2005.
83. Y. H. Kim, C. Cabral. Jr, E. P. Gusev, R. Carruthers, L. Gignac, M. Gribelyuk, E.

Cartier, S. Zafar, M. Copel. V. Narayanan, J. Newbury, B. Price, J. Acevedo, P.
Jamison, B. Linder, W. Natzle, J. Cai, R. Jammy, and M. Ieong, “Systematic
study of workfunction engineering and scavenging effect using NiSi alloy FUSI
metal gates with advanced gate stacks,” in IEDM Tech. Dig., pp. 642-645, 2005.
84. C. S. Park, B. J. Cho, L. J. Tang, and D. Kwong, “Substituted aluminum metal

gate on high-K dielectric for low work-function and Fermi-level pinning free,” in
IEDM Tech. Dig., pp. 299-302, 2004.
85. T. Nabatame, M. Kadoshima, K. Iwamoto, N. Mise, S. Migita, M. Ohno, H. Ota,

N. Yasuda, A. Ogawa, K. Tominaga, H. Satake, and A. Toriumi, “Partial silicides
technology for tunable work function electrodes on high-k gate dielectrics-Fermi
level pinning controlled PtSix for HfOx(N) pMOSFET,” in IEDM Tech. Dig., pp.
83-86, 2004.
86. E. Cartier, F. R. McFeely, V. Narayanan, P. Jamison, B. P. Linder, M. Copel, V.

K. Paruchuri, V.S. Basker, R. Haight, D. Lim, R. Carruthers, T. Shaw, M. Steen,
J. Sleight, J. Rubino, H. Deligianni, S. Guha, R. Jammy, G. Shahidi, “Role of
oxygen vacancies in VFB/Vt stability of pFET metals on HfO2,” in VLSI Tech.
Dig., pp. 230-231, 2005.
55
Ch 3: The Metal-Dielectric Interface and Its Impact on the EWF of Metal Gates
Chapter 3
The Metal-Dielectric Interface and Its

Impact on the Effective Work Function
of Metal Gates
3.1 Introduction
As discussed in Chapter 2, metal gate is more advantageous than the
conventional poly-Si gate due to the elimination of the gate-depletion and dopant
penetration problems, therefore it is seen as a performance booster for future CMOS
transistors. However, one of the major challenges for the metal gate technology is
how to get desirable work function (Φm) values in MOSFETs. Although the work
function of a metal material in vacuum can be precisely measured as a physical
parameter of the metal material itself, the effective work function (Φm,eff, EWF) of the
metal gate in a real MOS structure will be affected by many other factors such as the
crystallinity, deposition techniques, impurities, and most importantly, the underlying
gate dielectrics [1]-[2]. Moreover, integrating metal gates in a gate-first CMOS
fabrication process requires the gate-stack to undergo a high-temperature source/drain
(S/D) dopant activation annealing, which may lead to a change in the microstructures
of the metal gates or some interactions at the metal-dielectric interface, and hence
affects the effective work function. All these effects make the engineering of the
56
metal gate effective work function in MOSFET a complex and challenging task.
From a scientific perspective, identifying the major factors which affects the effective
work function of metal gates and understanding the mechanisms determining how
these factors contribute to the effective work functions would be of significance.
The dependence of the metal gate effective work function on the gate
dielectric materials was explained by Yeo et al. [1]-[2]. An interface dipole theory
was proposed where the intrinsic states or metal-induced gap states (MIGS) are
believed to be the major factor in determining the effective work function of metals
on high-κ materials. However, it is not clear whether this model can be applied to
explain the instability of the metal gate work function during the thermal annealing
processes used in the CMOS fabrication, which finally determines the transistor
threshold voltage.
In this chapter, we examined the dependence of the metal gate effective work
function on the underlying gate dielectric materials as well as the process temperature.
It was found that the creation of extrinsic states, which is typically related to the
interface defects arising from the interface reactions, could be the major reason
responsible for the thermal instability of the metal gate effective work functions.
Different from the intrinsic states, the creation of these extrinsic states are
thermodynamically driven and can result in a dipole layer at the metal-dielectric
interface upon annealing, which modifies the Φm,eff of metal gates. A model
considering the contribution of these extrinsic states is used to qualitatively explain
the phenomenon of the process-induced work function instability. This insight could
be useful in the understanding and engineering of this critical metal-dielectric
interface.
57
3.2 Theoretical Backgrounds
3.2.1 Work Function of Metal Materials
The work function of a solid (either a metal or a semiconductor) in vacuum
(Φm,vac) is defined as the energy difference between the Fermi level in the solid and
the vacuum level. For a semiconductor, the work function is a statistical concept and
stands for the weighted average of the amount of energy required to raise an electron
from the valence band and the conduction band to the vacuum level, respectively [3].
There are two parts for the work function calculation using quantum
mechanical effect [3]: (1) A volume contribution which stems from the energy of an
electron due to the periodic potential of the crystal and interaction of the electron with
other electrons; (2) A surface contribution which is due to a possible surface dipole.
As the electron charge distribution around the atoms at the solid surface is not
symmetrically disposed around the nucleus, the center of the positive and the negative
charge will not coincide, leading to a surface dipole. Therefore, any change in the
surface electron charge distribution of a solid will result in the modification of the
surface dipole layer, and hence the work function value of the solid.
3.2.2 Definition of Effective Work Function
The effective work function (Φm,eff) of a metal gate in MOSFET represents the
effective value of metal work function from the device operation point of view. Φm,eff
directly determines the flat band (VFB) and threshold voltage (Vth) of a MOSFET, and
hence is of importance for device design. Although Φm,vac is a physical parameter of
the metal material itself, the effectiveness of the metal gate electrode in determining
VFB and Vth of MOSFET may be affected by many other factors, e.g. the dipole layer
58
at the metal-dielectric interface. Fig. 3.1 illustrates the impact of interface dipoles in
modulating the effective work function of metal gates. As shown in Fig. 3.1 (a), the
VFB is defined as the gate voltage required to flatten the energy bands in metal,
dielectric and silicon, and is equal to the Fermi level difference between the metal and
silicon if we assume that the oxide charge is negligible. However, the interface dipole
layer will introduce an additional electric field at the metal-dielectric interface, and
hence modify the alignment of the metal Fermi level to gate dielectric, as shown in
Fig. 3.1 (b). As a result, the VFB will be modified, as well as the Vth of transistors.
Fig. 3.1 Band diagram of a MOS structure in flat-band condition (a) without interface
dipoles and (b) with interface dipoles at metal-dielectric interface.
From the device operation point of view, this result is equivalent to using a
new metal material with a modified work function Φm,eff as the gate electrode, as
illustrated in Fig. 3.1 (b). It is clear that the effective work function consists of both
the contributions from the metal itself and that from the metal-dielectric interface, or
other factors if any. Form the device engineering point of view, the effective work
function is more important than the work function value measured in vacuum.
59
To avoid the possible confusion between these two concepts, the metal work
function measured in vacuum will be deliberately denoted by Φm,vac, while the
effective work function from a device perspective will be expressed as Φm,eff in this
chapter.
3.2.3 Factors Affecting the Work Function of Metals
In order to study the behavior of the Φm,eff in a MOS structure, we need to first
understand the factors which may affect the effective work function of metals. These
factors can be grouped into two categories: bulk contribution and interface
contribution.
Let’s start with the factors which affect the bulk properties on the metal work
function. First, the different crystal orientations of a metal material will result in
different work functions [4]-[5]. For Al, the work function difference between the
different crystal orientations could be about 0.2 eV [4]; while that for Mo was
reported to be about 0.5 eV [5]. The reason could be attributed to the different surface
atom densities on difference facets [4]. Second, the microstructure of the metal film,
e.g. grain size and film texture, may affect its work function. Generally, the film with
an amorphous structure tends to have a low work function [6], while the poly-
crystalline structures with lots of grains may have a relatively high work function [7].
Third, the impurities in the metal film can also have impact, and this effect could be
more pronounced if the impurities pile up near the metal-dielectric interface. Pan et
al. reported that impurities like P, C, and Si can modify the work function of TaN or
TiN metal gates [8]. In addition, impurities like O and N are also believed to be
capable of affecting the metal work function.
60
Besides these factors, the metal-dielectric interface contribution is also
important for the Φm,eff of metal gates in a MOS stack. The contribution of the
interface is realized through the interface dipoles, which is typically associated with
the energy states or defects near the metal-dielectric interface. The creation of the
interface states/defects can even “pin” the Fermi-level of metals at a certain position,
which is commonly cited as the “Fermi-level pinning” (FLP) effect. However, the
origin of these interface states/defects, as well as their material and process
dependence, are still not well understood. This will be the main objective of this
chapter.
3.2.4 Fermi Level Pinning: Schottky Model and Bardeen Model
The Fermi-level pinning effect is initially studied in the metal-semiconductor
contact. A Schottky Model is first proposed to describe the ideal metal-
semiconductor contact [9]. In this model, there is no charge transfer across the metal-
semiconductor interface, and the Schottky Barrier Height (SBH, φb ) for electron is
determined only by Φm,vac and the electron affinity of the semiconductor χs. Therefore
there is,
φb = Φ m ,Vac − χ s (3-1)
However, it was experimentally observed that the above equation is not
generally obeyed. Bardeen then proposed a Surface State model to explain this
observation [10]. Bardeen’s model assumes that there exists high density of surface
states at the semiconductor surface (1 per surface atom), as well as a gap δ of atomic-
scale dimension between the metal and semiconductor. The filling of these surface
states results in a charged dipole layer across the small gap δ, which “pins” the metal
Fermi level to a certain energy level. However, later it was further pointed out by
61
Heine [11] that the high density of surface states did not exist in the fundamental gap
for most metal-semiconductor interfaces. Instead, the metal-induced gap states
(MIGS) or simply intrinsic states could be responsible for the FLP phenomenon
observed.
3.2.5 Metal-Induced Gap States (MIGS) Theory and Its Limitations
When the conduction band of the metal overlaps the band gap of
semiconductor, the wave functions of the electrons in the metal will decay into the
band gap of the semiconductor, resulting in states in the forbidden gap of the
semiconductor. These states are known as intrinsic states or called metal-induced gap
states (MIGS). These MIGS can exist not only in the metal-semiconductor interface,
but also in a metal-dielectric interface [1]-[2], [12].
Since the intrinsic states are split off from the valence and the conduction
band, their character varies across the gap from mostly donor-type close to the top of
the valence band Ev to mostly acceptor-type close to the bottom of the conduction
band Ec [13]. Charge transfer generally happens between the metal and these intrinsic
states in the dielectric. Filling the acceptor-type state gives an excess negative charge,
while leaving the donor-type state empty gives an excess positive charge. These
excess charges create a dipole at the interface, and this dipole layer tends to drive the
band lineup toward a position that would give zero dipole charge, which is called
charge neutrality level ECNL [14]. Fig. 3.2 illustrates the characteristics of the intrinsic
states at a metal-dielectric interface and the mechanism to create dipoles across the
interface. As shown in Fig. 3.2, the energy level at which the dominant character of
the interface states changes from acceptor-like to donor-like is defined as the charge-
neutrality level (ECNL) [14]. In other words, ECNL is the balanced point of the weights
62
of the conduction-band and valence-band density of states. High-density states near
the valence (conduction) band edge tend to drive the ECNL towards the conduction
(valence) band edge.
Fig. 3.2 Schematic energy band diagram (left) and the characteristics of the gap states
(right) for metal gate on dielectrics. The character of MIGS becomes more acceptor-
(donor-) like toward the Ec (Ev), as indicated by the solid (dashed) line [2].
Fig. 3.2 depicts the case where the metal Fermi level (EF,m) is above the
dielectrics ECNL. Electrons from the metal tend to transfer across the interface to fill
the MIGS where the energy is below EF,m. Consequently, negative charges are
created in the dielectrics side, and a dipole layer is then formed at the interface. EF,m
would be driven towards the ECNL by this interface dipole, resulting in the Φm,eff shift.
According to the MIGS theory, the relationship between Φm,eff and Φm,vac is given by
the following equation [2]:
Φ m,eff = Φ CNL ,d + S (Φ m ,vac − Φ CNL ,d ) (3-2)
63
where S is the Schottky pinning parameter [12], describing the dielectric screening.
In other words, S determines the strength of the FLP effect. W. Mönch found that the
parameter of S empirically obeys the following equation [13]:
1
S= (3-3)
1 + 0.1(ε ∞ − 1) 2
where ε ∞ stands for the electronic part of the dielectric constant. The smaller the S
parameter for a material, the more effective this material is to pin the metal Fermi
level. When S is equal to zero, EF,m would be fully pinned to the ECNL, and this is the
“Bardeen limit”. When S is equal to 1, there is no pinning of EF,m, and this is called
the “Schottky limit”.
According to this MIGS theory, the Fermi-level pinning effect on high-κ
dielectric will be dominated by the S parameter as well as the ECNL of the gate
dielectric materials [2]. On a SiO2 dielectric, the FLP effect would be negligible due
to the large S value of 0.95. On the other hand, significant FLP would be expected on
HfO2 or ZrO2 because the S values are as small as 0.52 [2]. Such a model has been
successful in fitting the experimental results in many metal-dielectric systems [2].
However, it is assumed in the intrinsic states or MIGS model that the metal-
semiconductor or metal-dielectric interface is perfect without any physical/chemical
defects. This may not be the case in real Schottky contact or MOS structures. W. E.
Spicer et al. proposed an Interface Defect Model to explain the FLP phenomenon of
III-V materials [15], where the extrinsic defects caused by the deficit of some atoms
(e.g. As in GaAs) are believed responsible for the FLP. Nevertheless, there has been
no research on the impact of the extrinsic states on the Φm,eff of metal gates in MOS
structure, as well as the material and process dependence. This will be investigated
and discussed in this chapter.
64
3.2.6 Work Function Measurement Techniques
Before starting the discussions on the metal gate work function, we will first
briefly review the techniques to measure the Φm,vac and Φm,eff of metal gates. To
measure the vacuum work function of metals, the commonly used technique is the
photoemission method. Since the work function of a metal is determined by the
valence band electrons instead of the core level electrons, ultraviolet (UV) light with
hv of 5~100 eV is typically used to excite the photoelectrons from valence band,
instead of the high-energy x-ray which can lead to too high a kinetic energy for the
excited photoelectrons from the valence band. Therefore this technique is also called
ultraviolet photoemission spectroscopy (UPS). From an analysis of the kinetic energy
and angular distribution of the photoelectrons, information on the electronic structure
of the material under investigation, including the Φm,vac, can be extracted. Note that
the UPS measurement is surface sensitive so that the absorbates on the surface need to
be cleaned properly.
To get the Φm,eff of a metal material, an internal photoemission (IPE) technique
has been developed and applied by Afanas’ev et al [16]. In this technique, an MOS
structure with very thin transparent metal layer is required to enable the photon to
reach the metal-dielectric interface. A gate bias will be applied on the metal electrode
so that the leakage current through the dielectric can be monitored. The electrons in
metal will be excited by the photons and give rise to a significant current when the
photon energy hv is large enough for electrons to overcome the potential barrier at the
metal-dielectric interface. Therefore, the barrier height can be determined from the
threshold of the IPE current, so that the Φm,eff can be deduced. The disadvantage of
this technique is that it requires special samples with transparent metal layers.
65
A straightforward approach to determine the Φm,eff of metal gate electrode in
MOS structure is to measure the capacitance-voltage (C-V) characteristics. The
following equation describes the VFB of metal gate in a MOS structure:
Qox Qox
VFB = (Φ m − Φ s ) − = (Φ m − Φ s ) − EOT (3-4)
Cox ε oε SiO 2
where Qox is the equivalent oxide charge per unit area, Cox is the oxide capacitance,
and ΦS is the work function of Si substrate. By plotting the VFB versus the EOT
determined from the C-V measurements, the work function difference between the
metal and the silicon substrate can be obtained from the intercept on the y-axis. Φm,eff
can therefore be calculated. This method provides an easy and convenient way to
extract the effective work function of metal gates from the real MOS devices. Note
that one of the assumptions for Eq. 3-3 is that the oxide charges should be located at
the dielectric-Si interface and the amount of Qox does not change with the dielectric
thickness. This assumption is reasonable for the conventional SiO2/Si stack where the
charge in bulk-SiO2 is negligible. However, this assumption may not be valid in the
high-κ dielectrics due to the relatively high Qox level in the bulk high-κ film as well as
the existence of an interfacial layer under high-κ, which makes the distribution of
oxide charge in high-κ stack more complex. Therefore some corrections may be
needed to improve the accuracy of this approach [17].
Another method to determine the Φm,eff of metal gate is to extract the barrier
height between metal and dielectric by analyzing the Fowler-Nordheim (F-N)
tunneling current through the dielectric, as depicted in Eq. 3-4:
J FN q3 4 2 m * φox 3 / 2 1 (3-5)
= exp( − )
Eox 2 16π 2 hφox 3h q Eox
where JFN is the F-N tunneling current density, Eox is the electric field in the oxide,
φox is the barrier height between metal and oxide, and m* is the effective electron
66
mass in oxide. By plotting ln (JFN/Eox2) versus 1/Eox, a straight line should be
obtained so that the barrier height φox can be extracted from the slope of this plot.
However, this method requires a prior knowledge of the effective mass in the
dielectric, which may limit its applications on some high-κ dielectrics where the
effective mass has not been very well determined.
3.3 Experimental
The metal-oxide-semiconductor capacitors were fabricated using p-Si (100)
substrates (B, 6x1015/cm-3). After the definition of the active area with 4000Å field
oxide, and a standard DHF-last RCA pre-gate clean process, either thermally grown
SiO2 or MOCVD HfO2 with different thicknesses was grown. Capacitors with TaN,
TiN, HfN, TaTi, TaTiN and WN metal gates were fabricated. The various metal gate
electrodes were deposited by dc sputtering. HfN was used as the capping layer on
TaN, TaTi, and TaTiN gate electrodes to minimize oxygen diffusion through the gate
stack during high-temperature post process, and hence minimize the EOT variation
induced by oxygen diffusion. To study the thermal stability of Φm, the capacitors
were annealed by rapid thermal annealing (RTA) in N2 gas at 800-1000 ºC for ~20 s.
The C-V characteristics were measured on large area (100x100 μm2) MOS
capacitors with an HP 4285A LCR meter at a high frequency (100 kHz). EOT and
the flat-band voltage VFB were determined through simulation using a model which
takes the quantum mechanical effect into account [18]. The current-voltage
measurements were performed using an HP 4156A semiconductor parameter analyzer.
Values of metal gate work function Φm,eff were extracted from plots of VFB versus the
gate dielectric EOT.
67
3.4 Results and Discussions
3.4.1 Work Function Thermal Instability of TaN
In order to identify the major factors which influence the Φm,eff during process,
we first compare the effective work function of TaN in TaN/SiO2 and TaN/HfO2 gate
stacks (ΦTaN/SiO2 and ΦTaN/HfO2) before and after RTA treatments. As shown in Fig.
3.3, the VFB vs. EOT plots exhibit linear relationship with almost identical slopes
before and after RTA treatments. The fixed oxide charge density extracted from Fig.
3.3 is about +1.6 × 1011 cm-2 for TaN/SiO2 and -1.38 × 1012cm-2 for TaN/HfO2 devices,
and both are independent of RTA temperature. These results indicate that the VFB
shift after RTA is caused by the effective work function change instead of the increase
in the fixed oxide charge. This allows accurate extraction of ΦTaN/SiO2 and ΦTaN/HfO2
from VFB versus EOT plots. The values of ΦTaN/SiO2 and ΦTaN/HfO2 before and after
RTA are noted in Fig. 3.3.
Comparing Fig. 3.3 (a) and Fig. 3.3 (b), it is interesting to note that for low
temperature (420°C forming gas) annealed devices, ΦTaN/SiO2 (4.4eV) is quite close to
ΦTaN/HfO2 (4.34eV) (also consistent with published data [19]-[20]), indicating that the
interface structure and composition of TaN/SiO2 and TaN/HfO2 are very similar and
the impact of MIGS is not very significant. However, the amount of work function
change before and after 1000 °C RTA anneal is considerably different for the
TaN/SiO2 and TaN/HfO2 systems. The increase of ΦTaN/SiO2 (from 4.4eV to 4.7eV) is
much larger than that of ΦTaN/HfO2 (from 4.34eV to 4.41eV), indicating that the
effective work function during RTA is strongly influenced by the interaction between
metal and the underlying dielectric. As we discussed in Section 3.2.3, the structural
change, such as the crystallization effect, may lead to a change of the metal gate work
68
function during the thermal annealing process. However, this may not be able to
explain the large difference observed between ΔΦTaN/SiO2 and ΔΦTaN/HfO2 because both
0.2
o
TaN on SiO2 420 C FG/30min
o
0.0 880 C RTA/20sec
Φ1000 C=4.70 eV
o
o
1000 C RTA/20sec
-0.2
VFB (V)
-0.4
-0.6
Φ420 C=4.40 eV
o
a)
-0.8
0 1 2 3 4 5 6
EOT (nm)
0.2
o
TaN on HfO2 420 C FG/30min
0.0 o
1000 C RTA/20sec
-0.2
Φ1000 C=4.41 eV
VFB (V)
-0.4
-0.6
Φ420 C=4.34 eV
VFB=ΦMS-Qox/Cox
o
-0.8
b)
-1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
EOT (nm)
Fig. 3.3 Plots of VFB versus EOT of (a) TaN/SiO2 and (b) TaN/HfO2 devices before
and after 1000 °C RTA treatment, from which the effective work function of TaN can
be extracted.
69
TaN/SiO2 and TaN/HfO2 devices received the same RTA and hence the same
crystallization effects. Interfacial reaction between the metal electrode and the gate
dielectric and the formation of an interfacial layer (metal silicide or high-κ) could be
another reason accounting for the phenomenon observed. The interface reaction
between the metal gate and gate dielectric has been studied in metal gates such as Hf,
Ta, and TaSixNy [21]-[23]. One way to monitor the interfacial reaction is to check the
EOT variation before and after high-temperature annealing. However, it is important
to recognize that oxygen in the RTA ambient can diffuse through the gate stack and
oxidize the Si substrate at the dielectric/Si interface, causing EOT to increase during
RTA.
In order to eliminate the oxygen diffusion effects, we capped TaN with an HfN
layer (50nm) to block the O2 diffusion during RTA because HfN has been
demonstrated to be an excellent O2 diffusion barrier [24]. Fig. 3.4 shows both the
EOT and work function of TaN devices as a function of RTA temperature with and
without HfN capping layer. With the HfN capping layer, the EOT stability of
TaN/SiO2 stack is improved significantly even after 1000 °C RTA anneal. This
clearly indicates that the EOT increase observed in TaN/SiO2 stack after 1000 °C
RTA is mainly caused by the O2 diffusion and not due to the interfacial layer
formation between TaN and SiO2 during RTA, which would have affected EOT. In
other words, the reaction between TaN and SiO2 could be very slight and results in no
detectable EOT change. However, even with the HfN capping layer, the effective
work function still changes significantly after 1000 °C RTA, as shown in Fig. 3.4 (b).
The gate leakage measurements of HfN/TaN/SiO2 gate stack under various
RTA conditions are depicted in Fig. 3.5 (a). The leakage current decreases with the
increasing RTA temperature, which is attributed to the increase of the effective work
70
function of TaN. The barrier height between TaN and SiO2 is extracted before and
after RTA treatment using the F-N tunneling ayalysis, as shown in Fig. 3.5 (b). The
increase of the barrier height after RTA can be well correlated with the work function
increase obtained from the C-V measurement.
8
7 Solid: TaN / SiO2
Open: HfN / TaN / SiO2
6
EOT (nm)
5
4
3
2
(a)
1
400 500 600 700 800 900 1000
ο
Annealing Temperature ( C)
5.2
5.0 TaN on SiO2

Work Function (eV)
HfN/TaN on SiO2
4.8
4.6
4.4
4.2
(b)
4.0
400 600 800 1000
ο
Fig. 3.4 The comparison of (a) EOT (with three different gate-oxide thickness) and (b)
effective work function of TaN as a function of RTA temperature with and without
HfN capping layer on top of the TaN/SiO2 stack.
71
-3
10
HfN/TaN / SiO2 / p-Si
Current Dencity (A/cm )
-4
2
10
-5
10
-6 o
10 420 C FGA
o
800 C RTA
-7 o
10 1000 C RTA
-8
10
tox=7nm
-9
10
-9 -8 -7 -6 -5 -4 -3
VG-VFB (V)
(a)
-10
ΦB=3.44eV
Ln(J/Eox )
-15
2
o
: 420 C FGA
o
: 1000 C RTA
-20
ΦB=3.73eV
-25
0.06 0.08 0.10 0.12 0.14
1/Eox (cm/MV)
(b)
Fig. 3.5 (a) Gate leakage measurement of HfN/TaN/SiO2 devices and (b) barrier
height extraction by F-N tunneling analysis before and after 1000 oC RTA treatment.
72
Another mechanism which would affect the effective work function of TaN
during the high-temperature annealing is the reaction between TaN and SiO2 that only
occur within an “interface” distance and thus have no impact on EOT. Comparing the
changes in ΦTaN/SiO2 and ΦTaN/HfO2 caused by RTA, it suggests that the Fermi level of
TaN at the TaN/SiO2 interface is pinned from 4.4eV at 420°C to 4.7eV at 1000°C.
However, this effect is not seen at the TaN/HfO2 interface, as shown by Fig. 3.3 (b)
where ΦTaN/HfO2 remains almost a constant, independent of the RTA temperature. We
believe that the Ta(N)-Si bonds at the TaN/SiO2 interface are responsible for the
ΦTaN/SiO2 increase during the high-temperature RTA, similar to the pinning effect
induced by Hf-Si bonds in poly-Si/HfO2 devices [25]. At low temperatures (420°C),
there are few Ta(N)-Si bonds due to the lack of activation, so the pinning effect is
weak and ΦTaN/SiO2 is very close in value to ΦTaN/HfO2. As annealing temperature
increases, the formation of Ta(N)-Si bonds are more likely to occur than that of Ta-Hf
bonds due to the larger electronegativity difference between Ta and Si atoms
compared to that between Ta and Hf atoms (Pauling’s electronegativity: Ta~1.5,
Si~1.9, and Hf~1.3). This may result in more interface defects or extrinsic states at
the TaN/SiO2 interface. Therefore, more charge transfer across the TaN/SiO2 interface
and stronger extrinsic pinning effect could be expected in TaN/SiO2 stack. The lack
of Φm,eff variation during RTA process in TaN/HfO2 stake could be attributed to the
chemical similarities between metal Ta(N) and metal oxide HfO2, and hence the small
number of extrinsic states created at the TaN/HfO2 interface.
73
3.4.2 General Trends in the Process Dependence of Φm,eff on SiO2 and
High-κ Dielectrics
Besides TaN, the thermal instability issue regarding the Φm,eff value of many
other metal gates are also studied in this work. To reveal the general trends of the
Φm,eff’s behaviors during process, the Φm,eff of many metals on SiO2 are plotted in
Fig.3.6 as a function of the RTA temperature. The experimental data from the
literature [26]-[29] are also included in this figure.
5.2 a
TiN
Work Function Φm (eV)
5.0 Ta-Pt
d
4.8
WN
HfN
4.6
TaTiN
4.4 TaN
b
Ta
4.2 c
TaN
TaTi
4.0
400 600 800 1000
o
a b
taken from Ref. [26]; taken from Ref. [27];
c d
taken from Ref. [28]; taken from Ref. [29].
Fig. 3.6 The variation of metal gate work function Φm with the annealing temperature
on SiO2 dielectric.
From the above figure, a noteworthy trend is that the work function of some
metals like TiN, Ta, TaTi, TaN, TaPt, and TaTiN converge towards the mid-gap
position of Si as the annealing temperature increases. This suggests that the metal
74
gate Fermi levels are pinned at about 4.7-4.8 eV after high-temperature treatment.
According to the theory of intrinsic states (MIGS), the FLP effect will be negligible
for SiO2 dielectric due to the high S parameter of SiO2 [2]. Therefore, the observed
change in Φm,eff with increasing temperature is likely to be due to the creation of a
high density of extrinsic states upon annealing. The creation of these extrinsic states
seems to be correlated with the interface reactions between metal gates and SiO2 since
the annealing temperature directly determines the magnitude of the Φm,eff variation in
process. Like the case of TaN that has been discussed earlier in this chapter, the
extrinsic states could be associated with some bonding defects between metals and Si.
Some other possible source of the extrinsic states could be oxygen-related vacancies
[30], nitrogen-related vacancies, or the penetrations of metal atoms into dielectrics.
Given the chemical similarity of Ti, Hf, and Ta, it is plausible that extrinsic states
with similar characteristics are formed between SiO2 and these metals. This may
explain why the Φm,eff tends to converge towards a similar energy position around the
mid-gap of Si.
However, a different trend has been obtained on high-κ dielectric. In Fig. 3.7,
the Φm,eff values of various metals on HfO2 dielectric before high-temperature anneal
are plotted on the horizontal axis, while that after high-temperature anneal are plotted
on the vertical axis. If the work function of a metal gate does not change appreciably
upon annealing at high temperature, it contributes a data point on the solid invariant
line, otherwise the data point will deviate from the line. It is observed in Fig. 3.7 that
the Φm,eff of many metals on HfO2 dielectric does not change significantly upon
annealing. Although the work function of some metals such as TaN or TaSiN on
HfO2 departs slightly from the invariant line, the deviation is quite small compared
with that on SiO2. This suggests that the RTA process temperature has less
75
significant impact on the metal-HfO2 interface than that on the metal-SiO2 interface.
This phenomenon is possibly due to two reasons. Firstly, the intrinsic states at the
metal-HfO2 interface has already “pinned” the Φm,eff strongly, and the role of extrinsic
states is comparatively diminished. Secondly, it could be possible that the creation of
extrinsic states or interfacial bonding defects upon thermal annealing is less
significant for the metal-HfO2 interface compared with the metal-SiO2 interface. The
likelihood of extrinsic states generation could be related to the chemical constituents
in the gate dielectric and the gate electrode. For example, given the similarity in
atomic radii and electronegativity of metal atoms in the metal gate investigated in this
work and the metal atoms (the hafnium) in the HfO2 gate dielectric, it is plausible that
chemical reactions are less likely to occur at the metal-HfO2 interface [31].
5.2
Φm after high-temperature anneal (eV)
Annealing Temp. Φ m
nt
o ir a
5.0 900 C va
o In
1000 C
4.8
TaSi* TaN
4.6 WN* HfN
HfNx-1
4.4 HfNx-2
4.2 TaN*
TaSiN* Metal - HfO2

4.0
4.0 4.2 4.4 4.6 4.8 5.0 5.2
Φm before high-temperature anneal (eV)
* taken from Ref. [31].

Other data points are measured from this work.
Fig. 3.7 Work function of metal gates on HfO2 before and after high-temperature
annealing. HfNx-1 and HfNx-2 denotes HfNx with different N concentration.
76
These results on SiO2 and HfO2 suggest that the creation of extrinsic states in
process and the associated Fermi-level pinning problem could be an important
consideration in tuning the effective work function of metal gates in MOSFET. The
creation of extrinsic states is determined by the thermal/chemical stability of the
metal-dielectric interface. Elevation of the annealing temperature probably increases
the density of extrinsic states and hence the effectiveness in pinning the Fermi level of
the metal gates.
3.4.3 Model: Fermi Level Pinning Induced by Extrinsic States
A metal-dielectric interface model considering the contribution of the extrinsic
states to the Φm,eff is hence proposed to qualitatively explain the thermal instability of
metal gate work function on SiO2, as shown in Fig. 3.8. The creation of the extrinsic
states is thermodynamically driven, and the states will have a certain energy level in
the band gap of the gate dielectric, which is determined by the properties of both the
gate electrode and the gate dielectric materials. Fig. 3.8 (a) illustrates the case where
the metal Fermi level Ef,m is above the energy level of the extrinsic states, and thus the
extrinsic states at the pinning location (at an energy Φpin,ex below the vacuum level)
are filled with electrons from the metal. This creates an interface dipole that is
charged negatively in the dielectric side, driving Ef,m towards the pinning position.
Conversely, for the case where the metal Fermi level Ef,m is below the energy level of
the extrinsic states, as shown in Fig. 3.8 (b), the existing electrons at the pinning level
tend to redistribute toward the metal side, resulting in an interface dipole that is
charged positively in the dielectric side. As a result, the Ef,m will move up towards the
pinning position. This explains why the Φm,eff of various metals with different Φm,vac
values tends to converge towards a similar position, as depicted in Fig. 3.6. Note that
77
the Φm,eff variation will be less pronounced if Ef,m is close to the pinning level of the
extrinsic states, like the situation of HfN and WN on SiO2 as illustrated in Fig. 3.6.
(a)
(b)
Fig. 3.8 Schematic energy band diagram for a metal gate on a dielectric, showing the
mechanism of Fermi-level pinning by extrinsic states. (a) When EF,m is above the
pinning level, (b) When the EF,m is below the extrinsic pinning level. The conduction
band edge and the valence band edge of the dielectric are denoted by Ec,d and Ev,d,
respectively.
78
The source of the extrinsic states could be related to the interfacial bonding
defects between the gate electrode and the gate dielectrics, or some vacancies (e.g.
oxygen vacancy) near the interface [30]. In order to avoid or minimize the creation of
extrinsic states in device design and processing, the source of the extrinsic states need
to be identified. However, it is difficult for the conventional physical analytical tool
such as XPS, AES, or SIMS to investigate this problem due to their limited resolution.
Therefore we will discuss some preliminary study using electrical characterization.
3.4.4 Investigation of Hf-Si Bond Induced Extrinsic States
As discussed in the above sections, the creation of the extrinsic states could be
responsible for the thermal instability of metal gate work function on SiO2. The
existence of extrinsic states can also be found in the poly-Si/HfO2 interface, as
described in [32]-[33]. However, there has been no consensus on the origin of the
extrinsic states at the gate-dielectric interface. The oxygen vacancy induced interface
dipoles was proposed by K. Shiraishi et al. to explain the flatband voltage (VFB) shift
of p+ poly-Si gate on HfO2 [34], and this mechanism has been found to be as
important for many p-type metal candidates [30], [36]. However, this theory has
difficulty in explaining the symmetric shift of Φm,eff for n+ and p+ poly-Si electrodes
[33]. On the other hand, some researchers believe that the Hf-Si bonds at the poly-
Si/HfO2 or FUSI/high-κ interfaces are responsible for the FLP effects observed in
these gate stacks [33], [37]. However, it is still not clear how the Hf-Si bond affects
the Φm,eff and whether the direction of the Hf-Si bond at the interface is a factor or not.
To investigate the origin of the extrinsic states at the gate-dielectric interface
and to study their properties, we investigated the behavior of Φm,eff in novel laminated
metal gate structures consisting of ultra-thin Si and Hf (N) layers, where the bond
79
configurations at the gate-dielectric interface can be deliberately varied by using
different laminated layers on SiO2 and HfO2 dielectrics.
Capacitors with different laminated structures on SiO2 and HfO2 dielectrics
were fabricated, as summarized in Table 3.1. For the splits with high-κ dielectric,
HfO2 with a thickness of 10 Å was deposited by MOCVD on top of the slanted SiO2
to minimize the impact of oxide fixed charge associated with thick high-κ films. Gate
electrodes comprising different laminated stacks were then deposited. From these
splits, different interface bond configurations including Si-Si, Hf-Si, Si-Hf, and Hf-Hf
bonds can be achieved. After gate patterning, RTA was then performed at 1000oC for
5 s in some splits, followed by a final forming gas anneal (FGA) at 420 oC for 30 min.
Table 3.1 Experimental splits of different laminated stacks consisting Hf (N) and Si
layers on slanted SiO2 and HfO2 dielectrics.
No. Gate Dielectric Laminated Stacks Thickness (Å)

1A SiO2 TaN /Hf /Si 1000 /10 /10
2A SiO2 TaN /HfN /Si 1000 /40 /10
3A SiO2 TaN /Si /HfN 1000 /10 /10
1B SiO2 + HfO2 (10 Å) TaN /Hf /Si 1000 /10 /10
2B SiO2 + HfO2 (10 Å) TaN /HfN /Si 1000 /40 /10
3B SiO2 + HfO2 (10 Å) TaN /Si /HfN 1000 /10 /10
The behavior of Φm,eff in TaN/Hf/Si stack was first investigated on SiO2 and
SiO2/HfO2 dielectrics before and after 1000 oC RTA. As shown in Fig. 3.9, Φm,eff of
TaN/Hf/Si stack on SiO2 and HfO2 are about 4.15 and 4.2 eV, respectively, in the as-
deposited condition, and the values become ~ 4.6 eV and ~ 4.62 eV after RTA
treatment at 1000 oC. The variation of Φm,eff upon annealing could be attributed to
two factors: “bulk” effect and/or “interface” effect. The “bulk” effect is typically
related to the microstructure or composition change of the metal gate, while the
80
“interface” effect represents the contribution of a dipole layer possibly formed at the
metal-dielectric interface. The dipole formation at the metal-dielectric interface could
be arising from both intrinsic states and extrinsic states, and the later one is typically
associated with the thermal process used in device fabrication. Since the Φm,eff of
TaN/Hf/Si stack in the as-deposited condition shows only small difference on SiO2
and HfO2 dielectrics, it can thus be inferred that the contribution from the intrinsic
states to the Φm,eff is not a major factor in our case [2]. On the other hand, the
contribution of extrinsic states associated with the interface defects could become
more pronounced after RTA annealing, leading to the Φm,eff instability problem after
RTA treatments [38]. However, it should be noted that a silicidation process between
the laminated ultra-thin Si and Hf layers is also likely to happen during RTA,
resulting in the formation of HfSix material and a change of the bulk work function.
Therefore, it is difficult to distinguish the “interface” effect with the “bulk” effect in
this case.
4.8
Effective Work Function (eV)
4.7 TaN/Hf/Si laminated Stack (1A & 1B)

on SiO2
4.6
on SiO2/HfO2
4.5
4.4
4.3
4.2
4.1
4.0
o
As-Dep FGA 1000 C RTA
+ FGA
Annealing Conditions
Fig. 3.9 Work function of the TaN/Hf/Si laminated stack on SiO2 and HfO2 after
annealing at different conditions: as-deposited, 420 oC FGA and RTA at 1000 oC for 5
sec followed by FGA.
81
In order to separate the contributions of “interface” and “bulk-material” to the
variation of Φm,eff during the RTA process, the Hf layer is replaced by HfN in the
laminated structures so that the silicidation process can be minimized during the RTA
process. As shown in Fig. 3.10 and Fig. 3.11, the variation in the Φm,eff of these metal
stacks is observed to be dependent to the different dielectrics, suggesting that the
“interface contribution” become dominant. Interestingly, a “kink” can be observed in
some of the C-V measurements after RTA annealing, which seems to be a kind of
“non-uniformity” problem since both the smooth C-V curves and the “kinked” curves
can be obtained from the same sample, as shown in Fig. 3.10 (b) & Fig. 3.11 (b).
Similar “kinked” C-V curves have also been reported in Ref. [35] and are attributed to
the different localized work function values in the gate electrode comprising different
metal materials. In our case, a localized crystallization effect among the ultra-thin
laminated layers could have happened during the high-temperature process, so that the
grain growth can break the continuity of the well-ordered bottom layer. The
crystallized grains may have different compositions compared with the well-ordered
stacks, leading to the different localized Φm,eff values as well as the “kinks” in the C-V
measurements. The morphology of the gate-dielectric interface has been examined by
XTEM, as illustrated in Fig. 3.12. The result confirms that the localized
crystallization effects are visible after RTA and some of the grains can breakthrough
the bottom layer and come into physical contact with the dielectric, which is
consistent with the explanation above. Based on this understanding, therefore, it can
be inferred that the upper part of the “kinked” C-V curve possibly represents the Φm,eff
of the localized newly-grown grains, while the lower part represents that of the well-
ordered laminated regions.
82
Sample 2A Sample 2B
Gate Capacitance (fF/μm )
2
10 10
TaN/HfN/Si/SiO2 TaN/HfN/Si/HfO2
8 8
∼ 0.5 V ∼ 0.6 V
6 6
As-Deposited
As-Deposited
4 FGA 4
FGA < 0.1 V
o & ∼ 0.3 V
RTA 1000 C 5s o
2 RTA 1000 C 5s 2
0 (a) (b) 0
-2 -1 0 1 2 -2 -1 0 1 2
Gate Voltage (V) Gate Voltage (V)
Fig. 3.10 C-V measurements of the TaN/HfN/Si laminated stack on (a) SiO2 and (b)
HfO2/SiO2 dielectrics after different annealing.
10 Sample 3A Sample 3B
2
10
TaN/Si/HfN/SiO TaN/Si/HfN/HfO2
8 2 8
∼ 0.35 V
6 6
As-Deposited
4 FGA ∼ 0.45 V FGA 4
o & < 0.1 V
RTA 1000 C 5s o
2 RTA 1000 C 5s 2
0 (a) (b) 0
-2 -1 0 1 2 -2 -1 0 1 2
Gate Voltage (V)
Gate Voltage (V)
Fig. 3.11 C-V measurements of the TaN/Si/HfN laminated stack on (a) SiO2 and (b)
HfO2/SiO2 dielectrics after different annealing.
Despite the “kinks” observed in the C-V measurements, the VFB shift observed
from the lower part of the C-V curves in Fig. 3.10 and Fig. 3.11 shows a clear
dependence on different gate dielectrics and the different bond configurations at the
interface, as summarized in Table 3.2. First, the magnitude of the VFB shift after RTA
is found to be considerably different for the devices with SiO2 and HfO2 dielectrics,
83
suggesting that the bond configuration at the gate-dielectric interface contributes to
the VFB shift. Secondly, as depicted in Table 3.2, the Hf-Si bonds at the gate-
dielectric interface always lead to significant VFB shift after RTA, while the Si-Si or
Hf-Hf bond configuration results in negligible VFB shift. This implies that the creation
of the extrinsic states, which lead to the dipole formation and VFB shifts, is determined
by the chemical properties of elements at the gate-dielectric interface. A metal-metal
or Si-Si atoms terminated interface is not likely to induce interface defects during the
high-temperature treatment so that the creation of extrinsic states is negligible,
enabling better Φm,eff stability to be achievable compared with the Hf-Si terminated
interface. Thirdly, irrespective of the direction of Hf-Si (or Si-Hf) bonds across the
interface, the VFB and Φm,eff always shift in a same direction (comparing devices 2B
and 3A), implying that the dipole layer formation is mainly due to the creation of
some energy states rather than the charge transfer from one kind of atom to another
[33]. Finally, these results are difficult to be explained by the oxygen vacancy theory,
where a negative VFB shift should be expected for device with HfO2 dielectric.
Therefore it is believed that the Si-metal bond related interface defects should be the
major source for the extrinsic states at the gate-dielectric interface.
Table 3.2 VFB shifts for TaN/HfN/Si and TaN/Si/HfN stacks on SiO2 and HfO2 after
RTA at 1000 oC for 5 sec, observed from the lower parts of the C-V curves shown in
Fig. 3.10 & Fig. 3.11.
SiO2 HfO2
bonds @ bonds @
No. ΔVFB (V) No. ΔVFB (V)
interface interface
TaN /HfN /Si 2A < 0.1 Si-Si 2B ~ 0.3 Si-Hf
TaN /Si /HfN 3A ~ 0.45 Hf-Si 3B < 0.1 Hf-Hf
84
Fig. 3.12 XTEM image of a TaN/HfN/Si stack on SiO2 after 1000 oC RTA for 5 sec.
In short summary, by investigating the behavior of the WF in laminated metal
gate structures consisting of Si and Hf (N) layers, it was found that the presence of a
Hf-Si bond is correlated with the creation of extrinsic states which can lead to the
EWF instability during high-temperature annealing process, while Hf-Hf or Si-Si
bonds do not. The creation of the bond-related interface states as well as the dipole
layer formation is irrespective of the direction of Hf-Si bond across the interface.
This understanding could be useful in engineering the gate-dielectric interface in
MOSFETs.
3.5 Conclusion
The dependence of metal gate work function on annealing temperature was
investigated in devices with several combinations of metal gate and gate dielectric
85
materials. It is identified that the extrinsic states created at the metal-dielectric
interface is the major factor responsible for the instability of metal gate effective work
function during high-temperature process. The creation of extrinsic states and the
resulting Fermi level pinning problem appears to be thermodynamically driven, and
becomes more pronounced when the annealing temperature is higher. A metal-
dielectric interface model that takes the role of extrinsic states into account is
proposed to qualitatively explain the work function instability phenomenon. In
general, the generation of extrinsic states upon annealing can be correlated with the
chemical bond configurations at the metal-dielectric interface. The presence of a
metal-Si bond at the gate-dielectric interface can lead to the Φm,eff instability problem
during high-temperature annealing process, while metal-metal or Si-Si bonds do not.
Moreover, the creation of the bond-related interface states as well as the dipole layer
formation is irrespective of the direction of the bonds across the interface. These
insights regarding the metal-dielectric interface could be useful for work function
tuning and interface engineering of the metal gate electrodes in future MOS devices.
86
References
1. Y.-C. Yeo, P. Ranade, T.-J. King, and C. Hu, “Effects of high-k gate dielectric
materials on metal and silicon gate workfunctions”, IEEE Electron Device Lett.,
vol. 23, pp. 342-344, 2002.
2. Y.-C. Yeo, T.-J. King, and C. Hu, “Metal-dielectric band alignment and its
implications for metal gate complementary metal-oxide-semiconductor
technology”, J. Appl. Phys., vol. 92, pp. 7266-7271, 2002.
3. E. H. Rhoderick, and R. H. Williams, “Metal-Semiconductor Contacts,” 2nd

Edition, Clarendon Press, Oxford, 1988.
4. R M Eastmenti and C H B Mee, “Work function measurements on (100), (110)

and (111) surfaces of aluminium,” J. Phys. F: Metal Phys., vol. 3, pp. 1728-1745,
1973.
5. S. Berg, P. O. Gartland and B. J. Slagsvold, “Photoelectric work function of a

molybdenum single crystal for the (100), (110), (111), (112), (114), and (332)
faces,” Surface Science, vol. 43, pp275-292, 1974.
6. P. Ranade, Y. Choi, D. Ha, A. Agarwal, M. Ameen, and T. King, “Tunable work

function molybdenum gate technology for FDSOI-CMOS,” in IEDM Tech. Dig.,
pp. 363-366, 2002.
7. S. H. Bae, W. P. Bai, H. C. Wen, S. Mathew, L. K. Bera, N. Balasubramanian, N.

Yamada, M. F. Li, and D. L. Kwong, “Laminated metal gate electrode with
tunable work function for advanced CMOS,” in Symp. VLSI Tech. Dig., pp. 188-
189, 2004.
8. J. Pan, C. Woo, M.-V. Ngo, P. Besser, J. Pellerin, Q. Xiang, and M.-R. Lin, “A
low-temperature metal-doping technique for engineering the gate electrode of
replacement metal gate CMOS transistors,” IEEE Electron Device Lett., vol. 24,
pp. 547-549, Sep. 2003.
9. W. Schottky, Phys. Z. vol. 41, pp. 570, 1940.
10. J. Bardeen, “Surface states and rectification at a metal semiconductor contact,”

Phys. Rev., vol. 71, pp. 717-727, 1947.
11. V. Heine, “Theory of Surface States,” Phys. Rev, vol. 138, pp. A1689-A1696,
1965.
12. John Robertson, “Band offsets of wide-band-gap oxides and implications for
future electronic devices,” J. Vac. Sci. Technol. B, vol. 18, pp.1785-1791, 2000.
13. W. Mönch, “Role of virtual gap states and defects in metal-semiconductor

contacts,” Phys. Rev. Lett., vol. 58, pp. 1260-1263, 1987.
87
14. J. Tersoff, “Theory of semiconductor heterojuction: The role of quantum

dipoles,” Physical Review B, vol. 30, No. 8, pp. 4874, 1984.
15. W. E. Spicer, P. W. Chye, P. R. Skeath, C. Y. Su and I. Lindau, “New and unified

model for Schottky barrier and III-V insulator interface states formation,” J. Vac.
Sci. Technol., vol. 16, pp.1422-1433, 1979.
16. V K Adamchuk and V. V. Afanas’ev, “Internal photoemission spectroscopy of

semiconductor-insulator interface,” Pro. In Surf. Sci., vol. 41, pp. 111-211, 1992.
17. R. Jha, J. Gurganos, Y. H. Kim, R. Choi, J. Lee, and V. Misra, “A capacitance-

based methodology for work function extraction of metals on high-k,” IEEE
Elecron Device Lett., vol. 25, pp. 420-423, 2004.
18. Keven Yang, Ya-Chin King, and Chenming Hu, “Quantum effect in oxide
thickness determination from capacitance measurement,” in Symp. VLSI Tech.
Dig., pp. 77-78, 1999.
19. D.-G. Park, T.-H. Cha, K.-Y. Lim, H.-J. Cho, T.-K. Kim, S.-A. Jang, Y.-S. Suh,
V. Misra, I.-S. Yeo, J.-S. Roh, J. W. Park, and H.-K. Yoon, “Robust Ternary
Metal Gate Electrodes for Dual Gate CMOS Devices,” in IEDM Tech. Dig., pp.
671-674, 2001.
20. J. C. Hooker, R. J. P. Lander, Z. M. Rittersma, T. Schram, G. S. Lujan, J. van Zijl,

E. van den Heuvel and F. Roozeboom, “Materials and Electrical Characterization
of Metal Gate Electrodes on High-k Dielectrics for Advanced CMOS
Technologies,” in Proc. Int. Conf. Solid State Devices and Materials, pp. 174-175,
2002.
21. V. Misra, G. P. Heuss, and Huicai Zhong, “Use of metal-oxide-semiconductor

capacitors to detect interactions of Hf and Zr gate electrodes with SiO2 and ZrO2,”
Appl. Phys. Lett., vol. 78, no. 26, pp. 4166-4168, 2001.
22. T. Ushiki, K. Kawai, I. Ohshima, and T. Ohmi, “Chemical Reaction Concerns of

Gate Metal with Gate Dielectric in Ta Gate MOS Devices: An Effect of Self-
Sealing Barrier Configuration Interposed Between Ta and SiO2,” IEEE Trans.
Electron Devices, vol. 47, no.11, pp. 2201-2207, Nov. 2000.
23. Y.-S. Suh, G. P. Heuss, and V. Misra, “Electrical characteristics of

TaSixNy/SiO2/Si structures by Fowler-Nordheim current analysis,” Appl. Phys.
Lett., vol. 80, no. 8, pp. 1403-1405, 2002.
24. H. Y. Yu, H. F. Lim, J. H. Chen, M. F. Li, C. X. Zhu, D.-L. Kwong, C. H. Tung,

K. L. Bera, and C. J. Leo, “Robust HfN Metal Gate Electrode for Advanced MOS
Devices Application,” in Symp. VLSI Tech. Dig., pp. 151-152, 2003.

88
Hebert, H. Tseng, B. White and P. Tobin, “Fermi Level Pinning at the

PolySi/Metal Oxide Interface,” in Symp. VLSI Tech. Dig., pp. 9-10, 2003.
26. J. Westlinder, T. Schram, L. Pantisano, E. Cartier, A. Kerber, G. S. Lujan, J.

Olsson, and G. Groeseneken, “On the thermal stability of atomic layer deposited
TiN as gate electrode in MOS devices”, IEEE Elecron Device Lett., vol. 24, pp.
550-552, 2003.

CMOS”, in Tech. Dig. IEDM, 2002, pp. 359-362.
28. C. S. Kang, H. J. Cho, Y. H. Kim, R. Choi, K. Onishi, A. Shahriar, and J. C. Lee,

electrode applications”, J. Vac. Sci. Technol. B, vol. 21, pp. 2026-2028, 2003.
29. B. Y. Tsui and C. F. Huang, “Wide range work function modulation of binary
alloys for MOSFET application”, IEEE Elecron Device Lett., Vol. 24, pp. 153-
155, 2003.

Dig., pp. 230-231, 2005.
31. J. K. Schaeffer, S. B. Samavedam, D. C. Gilmer, V. Dhandapani, P. J. Tobin, J.

Mogab, B. Y. Nguyen, B. E. White, Jr. S. Dakshina-Murthy, R. S. Rai, Z. X.
Jiang, R. Martin, M. V. Raymond, M. Zavala, L. B. La, J. A. Smith, R. Garcia, D.
Roan, M. Koeeke, and R. B. Gregory, “Physical and electrical properties of metal
gate electrodes on HfO2 gate dielectrics”, J. Vac. Sci. Technol. B, vol. 21, pp. 11-
17, 2003.

k HfO2 dielectric,” Appl. Phys. Lett. vol. 83, pp. 308-310, 2003.

polySi/metal oxide interface,” in Symp. VLSI Tech. Dig., pp. 9-10, 2003.

2004.
89

Dig., pp. 303-306, 2004.
36. J. K. Schaeffer, C. Capasso, L. R. C. Fonseca, S. Samavedam, D. C. Gilmer, Y.

Liang, S. Kalpat, B. Adetutu, H. Tseng, Y. Shiho, A. Demkov, R. Hegde, W. J.
Taylor, R. Gregory, J. Jiang, E. Luckowski, M. V. Raymond, K. Moore, D.
Triyoso, D. Roan, B. E. White Jr., and P. J. Tobin, “Challenges for the integration
of metal gate electrodes,” in IEDM Tech. Dig., pp. 287 - 290, 2004.
37. M. Terai, K. Takahashi, K. Manabe, T. Hase, T. Ogura, M. Saitoh, T. Iwamoto, T.

Tatsumi, and H. Watanabe, “Highly reliable HfSiON CMOSFET with phase
controlled NiSi (NFET) and Ni3Si (PFET) FUSI gate electrode,” in VLSI Tech.
Dig., pp. 68-69, 2005.
May 2004.
90
Ch 4: Lanthanide-Incorporated Metal Nitrides for NMOS Applications
Chapter 4
Lanthanide-Incorporated Metal Nitrides

for NMOS Applications
4.1 Introduction
Based on the understanding in Chapter 3, the metal-dielectric interface is a
critical factor in determining the effective work function (EWF) of metal gates (MG).
Since the nature of the extrinsic interface states tends to drive the effective work
function of metals towards the mid-gap position of Si, adopting materials with work
function (WF) values far below the conduction band Ec or far above the valence band
Ev of Si as the gate electrodes would be a potential solution to achieve the required
band-edge work function values after the rigorous thermal process.
As we know, most lanthanide series elements, including lanthanum (La),
gadolinium (Gd), terbium (Tb), dysprosium (Dy), erbium (Er), ytterbium (Yb) etc.,
exhibit work function values of less than 4.0 eV, as summarized in Table 4.1.
However, these metal materials could be quite reactive in the presence of oxygen or
moisture so that they are difficult to handle and integrate into the gate-first CMOS
process flow. On the other hand, refractory metal nitrides (MNx), such as tantalum
nitride (TaN), hafnium nitride (HfN) and titanium nitride (TiN), have been widely
studied due to their relatively good thermal stability compared with other metal gate
91
candidates [1]-[3]. In addition, the metal nitride gate electrodes also show superior
compatibility with high-κ gate dielectrics, enabling excellent EOT scalability of below
1 nm to be achievable [2]-[4]. These advantages make the metal nitrides a most
important category among all the metal gates being developed. However, the
limitation of MNx materials is their mid-gap work functions, which leads to the Vth of
the device too high to be acceptable for plenary bulk-Si devices. Therefore, it is
intuitional to think of the possibilities of combining the advantages of lanthanide and
MNx metal gates together.
Table 4.1 Work functions of some lanthanide metals.
Metals La Gd Tb Dy Er Yb
Work
3.5a 3.17a 3.1a 3.25b 3.25b 2.6c
Function (eV)
a
taken from Ref. [5];
b
taken from Ref. [6];
c
taken from Ref. [7].
In this work, we systematically investigated a novel approach to tune the work
function of MNx gate electrodes (TaN, HfN, etc) by incorporating lanthanide series
elements (Tb, Er, and Yb) into MNx films, as illustrated in Fig. 4.1. The results show
that the WF of lanthanide-incorporated metal nitrides (lanthanide-MNx) can be
successfully modulated continuously from mid-gap values down to 4.2 ~4.3 eV with
good thermal stability achieved even after rapid thermal annealing (RTA) treatments
at 1000 oC. The deposition methods, material and electrical properties, and the
MOSFET characteristics of the novel lanthanide-MNx metal gates will be presented
and discussed in this chapter.
92
Fig. 4.1 Illustration of the idea to modulate the work function of metal nitrides by
incorporating lanthanide series elements for n-MOSFET applications.
4.2 Experimental
Lanthanide-MNx films were deposited by reactive co-sputtering of refractory
metals (Ta, Hf, etc) with lanthanide metals (Tb, Er, Yb, etc) in an Ar + N2 ambient
using a PVD facility. The background pressure in the process chamber is about 3 ×
10-7 Torr or less. For the deposition of refractory metal nitride TaN and HfN, a dc
power of 450 W was applied on a 3-inch plenary Ta or Hf target with purity of
99.99 %, and the N2/Ar flow rate was kept at 5/25 sccm. For the deposition of
lanthanide-MNx, a relatively long burn-in step (a dummy sputtering of lanthanides
with a dc power of 150 W in pure Ar ambient for 10~15 min) was performed on the
lanthanide targets prior to the co-sputtering process in order to remove the possible
surface moisture or oxides on the lanthanide targets. During the co-sputtering, the
power applied on the Ta (or Hf) target was kept at a constant value of 450 W, whereas
that on the lanthanide Tb, Er, or Yb targets was varied from 70 W to 200W to
modulate the lanthanide concentration in the films (Table 4.2). The N2/Ar flow in
these experiments was kept at 5/25 sccm unless otherwise stated. In some
experimental splits, the N2/Ar flow rate ratio was varied intentionally from around
93
3/25 to 8/25 in sccm to evaluate the impact of N concentration on the properties of
lanthanide-MNx. The deposition conditions and the compositions of the films
investigated are summarized in Table 4.2 and Table 4.3.
Table 4.2 Experimental splits and the compositions for the Lanthanide-MNx films.
Power Power on
N N2/Ar Composition Composition
MNx Lan. on Ta or Lantha.
o. (sccm) (by XPS) (by RBS)
Hf (watt) (watt)
1 N.A. 450 0 5/25 TaN Ta1.05N0.95
2 450 70 5/25 Ta0.97Tb0.03Ny N.A.
3 450 100 5/25 Ta0.94Tb0.06Ny N.A.

Tb
4 450 150 5/25 Ta0.9Tb0.1Ny Ta0.92Tb0.08N1.0
TaN
5 450 200 5/25 Ta0.87Tb0.13Ny N.A.
6 450 100 5/25 Ta0.97Er0.03Ny N.A.

Er
7 450 150 5/25 Ta0.95Er0.05Ny N.A.
8 Yb 450 60 5/25 Ta0.97Yb0.03Ny N.A.
9 N.A. 450 0 5/25 HfN N.A.
10 HfN 450 100 5/25 Hf0.89Tb0.11Ny N.A.

Tb
11 450 150 5/25 Hf0.8Tb0.2Ny N.A.
Table 4.3 Experimental splits and the compositions for Ta0.9Tb0.1Ny with different N2
flow rates during reactive sputtering deposition.
Power on Power on
Lanthanide- N2/Ar N/(Ta+Tb)
No. Ta or Hf Lantha.
MNx (sccm) (by RBS)
(watt) (watt)
1 Ta0.9Tb0.1Ny 450 150 3/25 0.96
2 Ta0.9Tb0.1Ny 450 150 5/25 1.0
3 Ta0.9Tb0.1Ny 450 150 8/25 1.06
94
MOS capacitors and n-MOSFETs with lanthanide-MNx gates and SiO2 or
HfAlO gate dielectrics were fabricated on p-Si (100) substrates (6-9 Ω·cm). For MOS
capacitor fabrication, Ta1-xTbxNy, Ta1-xErxNy, Ta1-xYbxNy and Hf1-xTbxNy metal gates
with thicknesses of about 500 Å were deposited using the method described above,
followed by the deposition of an in-situ TaN capping layer with a thickness of 1000 Å
on top of all the lanthanide-MNx films in order to reduce the sheet resistance of the
whole gate electrode stacks and prevent the oxidation of lanthanide-MNx surface.
After gate patterning, some of the capacitors were subjected to RTA treatment from
800°C to 1000°C in N2 ambient to evaluate the thermal stability. For n-MOSFETs
fabrication, a damascene gate process was used to pattern the Ta0.9Tb0.1Ny/TaN stack,
which will be discussed in Section 4.6. The thickness of Ta0.9Tb0.1Ny was reduced to
100 Å to minimize the difficulties in gate patterning. The As implantation with a dose
of 4 × 1015 was performed followed by a RTA at 1000 oC for 5 sec to form the
source/drain (S/D) regions of the MOSFET. NiSi salicide was adopted to reduce the
S/D parasitic resistance. Finally, all the samples were subjected to a backside Al
metallization and a forming gas annealing (FGA) at 420 oC for 30 min.
Capacitance-voltage (C-V) and current-voltage (I-V) characteristics were
measured using a HP4284A LCR meter and HP4156A semiconductor parameter
analyzer, respectively, on MOS capacitors with area of 100 × 100 μm2. EOT and flat-
band voltage (VFB) were obtained by fitting the C-V measurements with the theoretical
C-V curves, which take the quantum mechanical effect into account. Auger electron
spectroscopy (AES), X-ray photoelectron spectroscopy (XPS), Rutherford
backscattering spectrometry (RBS), X-ray diffraction (XRD) and high-resolution
transmission electron microscopy (HRTEM) were also performed on blanket films
deposited on thin (2~3 nm) SiO2 dielectric for material characterizations.
95
4.3 Material Characteristics of Lanthanide-MNx
4.3.1 Composition Analysis
The compositions of the films investigated were analyzed by various
techniques including XPS, AES, and RBS. For Ta1-xTbxNy, the Tb/ (Ta+Tb) ratios are
determined by XPS analysis according to the Ta 4f and Tb 4d peaks, as summarized
in Table 4.2. It is found that the Tb concentration in Ta1-xTbxNy exhibits a linear
dependence on the sputtering power applied on the Tb target, as depicted in Fig. 4.2.
This suggests good controllability of lanthanide concentration by adjusting the
sputtering power. The Tb/ (Ta+Tb) ratios in some Ta1-xTbxNy films were also
measured by the RBS technique, as indicated in Table 4.2. For the most frequently
used deposition condition, the Tb/(Ta+Tb) fraction determined by RBS is about 8 %,
which is quite consistent with the value of 10% measured by XPS after taking the
possible experiment errors induced by different characterization techniques into
account.
16
14
12
Tb/(Ta+Tb) (%)
10
8
6
4
2
0
0 50 100 150 200 250
Tb Sputtering Power (watt)
Fig. 4.2 Relationship between Tb concentration in Ta1-xTbxNy and the sputtering

power applied on the Tb target during co-sputtering deposition.
96
However, the nitrogen concentration is relatively difficult to determine very
accurately. For XPS analysis, the N 1s peak partially overlaps with Ta 4p3 peak,
which implies that curve fitting of the measured spectrum is necessary to differentiate
the contributions from the two elements. However, it should be noted that the Ta/N
ratio determined from Ta 4p3 and N 1s peak should be same with that determined
from the Ta 4f and N 1s peak theoretically. In practice, unfortunately, the best curve
fitting may not fulfill the later constraint very well, resulting in some uncertainty in
determining the N concentration. For the RBS measurement, the channel of N
spectrum is masked by the background signal from the Si substrate because the
atomic number of N is smaller than that of Si [8]. Moreover, the RBS yield of the N
element is relatively low because of the small scattering cross section of N atoms,
leading to the low signal detected from N, as shown in Fig. 4.3. As a result, accurate
determination of N concentration from RBS analysis could also be difficult.
Energy (MeV)
0.5 1.0 1.5
4
2.5x10
tan1
Simulation of Ta-Tb-N/Si
______ ‘Tb’ Subplot
2.0 Energy (MeV)
0.6 0.8 1.0
5000
tan1
Simulation of Ta-Tb-N/Si
4000
1.5
Counts
3000
Counts
2000
1.0
1000
N
0
0.5 60 80 100 120 140 160
Channel
N Si Ta
0.0
100 150 200 250 300
Channel
Fig. 4.3 RBS spectrum of Ta0.92Tb0.08N1.0 film, where the concentration of each
species are determined from the simulation by XRUMP [9].
97
Although it is difficult to predict the N concentration very accurately, the
general trend of N concentration in the different experimental splits can still be
observed. For the TaN control sample, the N % determined from XPS, RBS, and
AES analysis is about 38%, 47%, and 50%, respectively. The RBS and AES
measurements give quite close values for N content. On the other hand, for the
Ta0.9Tb0.1Ny film, the measured N % values are about 43%, 50%, and 55%,
respectively, by XPS, RBS, and AES. Therefore, it can be concluded that the N
concentration in Ta0.9Tb0.1Ny is higher than that in TaN. Indeed, an increase of N
concentration can be observed while increasing the lanthanide concentration in the
lanthanide-MNx films.
In the following discussion, the compositions determined by XPS will be used
as a standard metrology. The actual N fraction in lanthanide-MNx will not be
explicitly labeled to avoid any possible confusion. Instead, where needed, the N2/Ar
flow rates during the sputtering deposition will be used as an indication of the relative
level of N content. The default N2/Ar flow rates will be 5/25 sccm unless otherwise
stated.
4.3.2 Auger Electron Spectroscopy (AES) Study
The depth profiles of the TaN/lanthanide-MNx gate stacks were studied by
AES analysis. Fig. 4.4 presents the AES profiles of the TaN/Ta0.94Tb0.06Ny/SiO2 and
TaN/Ta0.95Er0.05Ny/SiO2 gate stacks before and after 1000 °C RTA. The spectra and
the relative sensitivity factor (RSF) for Ta, N, O, and Si elements are obtained from
the standard reference book [10], while that of Tb, Er, and Yb elements are measured
from the pure metal samples deposited in this experiment. The RSF for these
lanthanide elements are relatively low, rendering the noise level of Tb or Er to be
98
relatively higher compared with O. From the depth profiles shown in Fig. 4.4, it can
be observed firstly that the Ta, Tb (Er), and N profiles appear to be very uniform
across the TaN and Ta0.94Tb0.06Ny (Ta0.95Er0.05Ny) films, without notable changes
before and after the 1000 °C RTA. Secondly, at the interface between the undoped
TaN and the Ta0.94Tb0.06Ny (Ta0.95Er0.05Ny), there is a clear change of the N percentage.
The nitrogen concentration in Ta0.94Tb0.06Ny (or Ta0.95Er0.05Ny) is observed to be
remarkably higher than that in the TaN film, and keeps stable even after 1000 oC RTA
treatment. Note that the N concentrations in other lanthanide-MNx films also tend to
increase with the lanthanide concentration, as judged from both AES and XPS studies.
100 100
a) as-dep b) 1000 oC RTA
80 TaN TaTb xN y TaN TaTb xN y 80
Concentration at %
Si
Si
60 60
N N
40 40
Ta Ta
O O
20 Tb
20
Tb
0 0
0 100 200 300 400 500 0 100 200 300 400 500
Sputtering Tim e (Sec)
100 100
c) as-dep d) 1000oC RTA
80 TaN TaErxN y TaN TaErxN y 80
Concentration at %
Si Si
60 N N 60
40 40
Ta Ta
O O
20 Er Er 20
0 0
0 100 200 300 400 500 0 100 200 300 400 500 600
Sputtering Time (Sec)
Fig. 4.4 AES depth profiling for TaN/Ta0.94Tb0.06Ny/SiO2 (a-b) and

TaN/Ta0.95Er0.05Ny/SiO2 (c-d) gate stacks before and after 1000 oC RTA in N2 ambient.
99
Since the N2/Ar flow rates, the process pressure and the sputtering power
applied on the Ta target were all kept constant during the deposition of Ta0.94Tb0.06Ny
(or Ta0.95Er0.05Ny) and TaN films, this abrupt change of nitrogen concentration implies
that the presence of Tb (Er) makes it possible to incorporate more N into the
Ta1−xTbxNy (or Ta1−xErxNy) films. The low work functions of lanthanide series
elements and the relatively large electronegativity differences to nitrogen could be
responsible for the affinity of lanthanide elements to nitrogen. The enhanced N
concentration could be of importance to the properties of lanthanide-MNx, especially
its thermal stability.
4.3.3 X-ray Photoelectron Spectroscopy (XPS) Study
XPS analysis was also performed to study the binding energy and composition
of the lanthanide-MNx films investigated in this experiment. Fig. 4.5 shows the core
level spectra of N 1s, Ta 4f, and Tb 4d regions for the as-deposited Ta1−xTbxNy films
with different Tb/(Ta+Tb) ratios. The ratio of Tb/(Ta+Tb) was obtained by XPS
analysis from the Ta 4f and Tb 4d spectra, and summarized in Table 4.2. As shown in
Fig. 4.5 (a), the binding energy of the N 1s peak shifts gradually from 498.7 eV to
about 487.5 eV as the Tb concentration increases, while that of the Ta 4p3/2 peak
remains almost unchanged. This implies that N may form new bonds with the
incorporated Tb, either in the form of Tb-N or Ta-N-Tb compound, as the Tb
concentration in Ta1−xTbxNy increases. In Fig. 4.5 (b), the Ta 4f7/2 peak in TaN
shows a binding energy of 23.0 eV, which is higher than the value of 21.9 eV for pure
Ta, suggesting the existence of Ta-N bonds [11]. As the Tb concentration increases,
the Ta 4f peak shows only a 0.3-eV-shift which could be attributed to the formation of
100
Ta-N-Tb compound. The signals from the Tb 4d peaks in Fig. 4.5 (c) appear noisy
due to the low Tb content.
Ta 4p3 N 1s Ta 4f5 Ta 4f7 Tb 4d

(5)
Intensity (a.u.)
(5)
(4) (5)
(4)
(3) (4)
(3) (3)
(2)
(2)
(1) (2)
(1)
(a) (b) (c)
408 404 400 396 30 28 26 24 22 20 165 160 155 150 145 140
Binding Energy (eV)
Fig. 4.5 The XPS spectra of the (a) N 1s, (b) Ta 4f, and (c) Tb 4d region for the as-
deposited Ta1-xTbxNy films with different Tb concentrations: (1) TaN; (2)
Ta0.97Tb0.03Ny; (3) Ta0.94Tb0.06Ny; (4) Ta0.9Tb0.1Ny; (5) Ta0.87Tb0.13Ny.
The similar shift of N 1s peak as a function of nitrogen concentration can also
be observed in Ta1−xErxNy, Ta1−xYbxNy and Hf1−xTbxNy materials, as depicted in Fig.
4.6. These results suggest that the nitrogen may form some complex compounds with
lanthanide, either in the form of lanthanide-N or Ta (Hf)-N-lanthanide, as the
lanthanide elements are introduced into the binary Ta (Hf)-N system. We believe that
the formation of these compounds could be helpful for improving the stability of the
lanthanide-MNx since the lanthanide metals themselves are quite reactive. The role of
nitrogen in lanthanide-MNx will be further discussed in Section 4.5.
101
Ta 4p3 N 1s
Ta0.95Er0.05Ny
Intensity (a.u.)
Ta0.97Er0.03Ny
TaN
415 410 405 400 395

Binding Energy (eV)
(a)
Ta 4p3 N 1s
Ta0.95Yb0.05Ny
Intensity (a.u.)
Ta0.97Yb0.03Ny
TaN
415 410 405 400 395

Binding Energy (eV)
(b)
102
N 1s
Hf0.43Tb0.11N0.46
Intensity (a.u.)
Hf0.48Tb0.06N0.46
HfN
402 400 398 396 394

Binding Energy (eV)
(c)
Fig. 4.6 The XPS core level spectra in the N 1s region for the as-deposited
(a) Ta1-xErxNy, (b) Ta1-xYbxNy, and (c) Hf1-xTbxNy films with different lanthanide
concentrations.
4.3.4 X-ray Diffraction (XRD) Study
The film morphology of lanthanide-MNx was studied by XRD analysis. Fig.
4.7 shows the XRD spectra of Ta1-xTbxNy films with different Tb contents, compared
with that of TaN. We can observe that TaN exhibits crystalline structure even in the
as-deposited condition. Moreover, a peak shift can be clearly observed in the spectra
of TaN after 1000 oC RTA treatment, suggesting the phase of TaN films is not stable
and some re-crystallization or condensation effect happens during the high-
temperature annealing process. On the other hand, a broad diffraction peak can be
found in the XRD spectrum of Ta1-xTbxNy when the Tb/(Ta+Tb) ratio is higher than
10 %. This indicates that the crystallization of TaN can be effectively suppressed by
the Tb additive, and very fine grains or amorphous morphology can be achieved.
103
More importantly, the amorphous structure can be maintained up to 1000oC, which is
adequate for the S/D dopant activation in gate-first process.
: As-Deposited
o
: 1000 C 20sec Ta0.87Tb0.13Ny
Ta0.9Tb0.1Ny
Intensity (a.u.)
Ta0.94Tb0.06Ny
Ta0.97Tb0.03Ny
(TaN)
TaN (111) TaN (002) TaN (220)
25 30 35 40 45 50 55 60 65 70
2θ (Degree)
Fig. 4.7 XRD spectrums of Ta1-xTbxNy materials with different Tb concentrations

before and after 1000 oC RTA anneal for 20 sec, compared with that of TaN.
For a metal gate electrode, amorphous morphology as well as good phase
stability against the high thermal budget used in the CMOS process will be desirable
since it will result in less stress and uniformity problems for the large-scale IC
fabrication [12]. The phase stability of Ta1-xTbxNy was also investigated as a
function of N concentration, as shown in Fig. 4.8. Ta0.9Tb0.1Ny gate electrodes with
different N content were prepared by varying the N2/Ar flow rates (N2/Ar = 3/25, 5/25,
and 8/25 sccm) during sputtering, as summarized in Table 4.3. According to the XRD
spectra, the crystallization of TaN can be significantly suppressed by adding Tb into
TaN. However, it is observed in the sample with the highest-N-split (N2/Ar = 8/25)
that a peak at TaN (200) position tends to increase after the high temperature RTA
104
process, implying that some re-crystallization may happen at high temperature.
Therefore the N concentration needs to be well engineered to get a stable amorphous
morphology in Ta1-xTbxNy.
As-dep
a) TaN b) o
900 C
o
1000 C
Intensity (cps)
Ta0.9Tb0.1Ny
N2/Ar = 3/25
c) Ta0.9Tb0.1Ny d) Ta0.9Tb0.1Ny
N2/Ar = 5/25 N2/Ar = 8/25
30 35 40 45 50 55 60 65 30 35 40 45 50 55 60 65
2θ (degree)
Fig. 4.8 XRD spectrums of Ta0.9Tb0.1Ny films with different N concentrations

measured in the following conditions: as-deposited, after 900 oC RTA and after 1000
o
C RTA.
4.3.5 Resistivity
The resistivity of lanthanide-MNx was investigated as a function of N and
lanthanide contents, respectively. We first discuss the dependence of the resistivity
on the N concentration in lanthanide-MNx. Fig. 4.9 compares the resistivity of the
Ta0.9Tb0.1Ny films with different nitrogen flow rates during deposition (see Table 4.3).
When the nitrogen flow increases from 3 sccm to 8 sccm, the resistivity of
Ta0.9Tb0.1Ny increases from 416 μΩ·cm to 965 μΩ·cm for the as-deposited films. An
increase of resistivity with N concentration can also be observed in the transition MNx
105
materials such as TaN, TiN and HfN [13]. Microstructure change and the formation
of some poorly conducting phases in the MNx film could be one of the reasons [14].
Besides this, the formation of a Tb-N compound, which probably behaves as a narrow
gap insulator [15], may also account for the increased resistivity as a function of N
content in the case of Ta1-xTbxNy. The higher resistivity values after the high
temperature (900 oC, 1000 oC) RTA process could be correlated with the oxidation of
the metal surface during the RTA process. This is due to the existence of oxygen
residues in the process chamber, and can be suppressed by capping the lanthanide-
MNx by TaN or other electrodes which are difficult to be oxidized.
o
420 c FGA
Volume Resistivity (μΩ•cm)
4
Ta0.9Tb0.1Ny o
900 c 20s
10 o
1000 c 5s
3
10
2
10
25
25
25
3/
5/
8/
N2/Ar Flow Ratio (sccm)
Fig. 4.9 Resistivity of Ta0.9Tb0.1Ny films as a function of N2/Ar flow rate ratio during
the sputtering, with and without RTA performed.
The dependence of the film resistivity on the lanthanide concentration in
lanthanide-MNx materials was also studied. Fig. 4.10 summarizes dependence of the
106
resistivity of lanthanide-TaN films on the lanthanide type and concentration. As
shown in Fig. 4.10, the resistivity of lanthanide-TaN is higher than that of TaN and
tends to increase with the lanthanide concentration. This could be correlated with the
N concentration in the lanthanide-TaN films which is higher than that in TaN, as
observed from the AES analysis in Section 4.3.2. Again, an increase of the resistivity
with annealing temperature can be observed in Fig. 4.10, which could be correlated
with the oxygen traces in the N2 ambient during the RTA process. A capping layer
with very low resistivity and high immunity to oxidation would be helpful in
achieving the low sheet resistance required by ITRS.
2000
Volume Resistivity (μΩ⋅cm)
As-Dep
1600 o
900 C RTA
1200 o
1000 C RTA
800
400
N
N
y
N
y
N
N
y
N
y
y
N
y
y
N
03
05
03
13
06
3
1
Ta
0.
Tb
0.
0.
0.
0.
0.
Yb
0.
Er
Er
Tb
Tb
b
97 T
95
97
97
87
94
0.
Ta
0.
0.
Ta
Ta
0.
0.
0.
0.
Ta
Ta
Ta
Ta
Fig. 4.10 Resistivity of lanthanide-incorporated TaN materials as a function of

lanthanide type and concentration before and after RTA treatments in N2 ambient.
107
4.4 Work Function Tunability
The work function tunability of lanthanide-MNx was studied by varying the
lanthanide content in lanthanide-MNx. Fig. 4.11 shows the high-frequency C-V
(HFCV) measurement of the Ta1-xTbxNy gated MOS capacitors with varied Tb
concentrations in Ta1-xTbxNy (corresponding to Table 4.2). Compared with that of
TaN, a parallel shift in the C-V curves of Ta1-xTbxNy/SiO2 MOS capacitors is observed,
indicating continuous work function tuning by varying the Tb concentrations. Fig.
4.12 compares the C-V measurement of Ta0.94Tb0.06Ny/SiO2 MOS capacitors after 420
o
C FGA for 30 min and 1000 oC RTA for 20 sec. The C-V measurements fit well
with the simulated C-V, implying good interface quality. A shift in VFB for about +
0.15 V is observed after 1000 oC RTA treatment, possibly due to the Fermi-level
pinning effect similar with that in TaN/SiO2 stack as discussed in Chapter 3 [16].
12
Capacitance Density (fF/μm )
2
TaN
10 Ta0.97Tb0.03Ny
Ta0.9Tb0.1Ny
8
Ta0.87Tb0.13Ny
6 Tb%
4
TaN
2
0
-2.0 -1.5 -1.0 -0.5 0.0 0.5
Gate Voltage (V)
Fig. 4.11 High-frequency C-V characteristics (100 kHz) of Ta1-xTbxNy gated MOS
capacitors with different Tb concentration in Ta1-xTbxNy on SiO2. The measurements
are taken after a 420oC forming gas anneal.
108
10
Capacitance Density (fF/um )

2
o
420 C FGA
o
1000 C RTA for 30 sec
8
4
simulated
Ta0.94Tb0.06Ny HFCV
2
on SiO2
0 EOT ~3.2 nm
-2.0 -1.5 -1.0 -0.5 0.0 0.5

Gate Voltage (V)
Fig. 4.12 High-frequency C-V characteristics (100 kHz) of MOS-capacitors with

Ta0.94Tb0.06Ny gate electrode after 420 oC FGA and after 1000 oC RTA. The lines
show the simulated C-V curves which takes quantum mechanical effect into account.
The work function of all the experimental splits investigated in this work were
extracted from the VFB versus EOT plots which exclude the contribution of oxide
fixed charge to the VFB of MOS capacitor such that accurate WF can be obtained. Fig.
4.13 compares the VFB versus EOT plots of Ta0.94Tb0.06Ny, Ta0.95Er0.05Ny, and
Hf0.8Tb0.2Ny metal gates with that of TaN and HfN on SiO2 dielectric after 420 oC
FGA and 1000 oC RTA processes. The VFB vs. EOT plots for these gate stacks show
similar slopes after different post-metal-annealing (PMA) treatments, indicating the
gate oxide quality is not affected much by the PMA processes. It is observed clearly
that the WF of MNx is lowered by adding lanthanide into MNx metal gates. The WF
values of Ta0.94Tb0.06Ny, Ta0.95Er0.05Ny, and Hf0.8Tb0.2Ny metal gates are determined to
be 4.08 eV, 4.17 eV, and 4.23 eV after 420 oC FGA, and they increase to 4.23 eV, 4.3
eV, and 4.31 eV, respectively, after 1000 oC RTA treatment. The WF increase after
109
the high temperature anneal process could be due the Fermi pinning effect resulting
from some extrinsic defect states generated during the high temperature anneal
process at the metal gate/SiO2 interface [17].
(a)
-0.2 SiO2, FGA
HfN, 4.65 eV
Flat Band Voltage VFB (V)
w/o Lanthanide
-0.4 TaN, 4.46 eV
-0.6 HfTbN, 4.23 eV with Lanthanide
-0.8
TaErN, 4.17 eV
-1.0 TaTbN, 4.08 eV
0 1 2 3 4 5 6 7 8 9 10
EOT (nm)
(b)
o
HfN, 4.71 eV SiO2, 1000 C RTA
Flat Band Voltage VFB (V)
-0.2
w/o Lanthanide
-0.4 TaN, 4.65 eV
HfTbN, 4.31 eV with Lanthanide

-0.6
-0.8 TaErN, 4.30 eV

TaTbN, 4.23 eV
-1.0
0 1 2 3 4 5 6 7 8 9 10
EOT (nm)
Fig. 4.13 VFB vs. EOT plots of Ta0.94Tb0.06Ny/SiO2, Ta0.95Er0.05Ny/SiO2, and

Hf0.8Tb0.2Ny/SiO2 gate stacks (a) after 420 oC FGA and (b) after 1000 oC RTA
treatment, compared with that of TaN and HfN.
110
Fig. 4.14 summarizes the WF values of some lanthanide-MNx as functions of
lanthanide type and concentration after different PMA treatments. It shows clearly
that WF of MNx, such as TaN and HfN, can be tuned continuously from mid-gap
values down to 4.2~4.3 eV by incorporating a small amount of lanthanide elements
into these MNx. It is believed that the presence of lanthanide at the metal gate-
dielectric interface is responsible for drawing the effective work function of MNx
down to a lower value, since lanthanide elements generally possess very low WF
values (see Table 4.1). Note that the metal gate work function of 4.2~4.3 eV is
achievable even after 1000 oC RTA treatment, which is promising as N-type metal
gate candidates in the gate-first bulk-Si CMOS process. Further tuning of the work
function is still possible by increasing the concentration of lanthanide elements.
o o
Effective Work Functions (eV)
5.0 420 C 800 C 1. TaN

o o 2. Ta0.97Tb0.03Ny
900 C 1000 C
3. Ta0.94Tb0.06Ny
4.8
4. Ta0.9Tb0.1Ny
4.6 5. Ta0.87Tb0.13Ny
6. Ta0.97Er0.03Ny
4.4 7. Ta0.95Er0.05Ny
8. Ta0.97Yb0.03Ny
4.2 9. HfN
10. Hf0.89Tb0.11Ny
4.0 11. Hf0.8Tb0.2Ny
1 2 3 4 5 6 7 8 9 1011
Fig. 4.14 Work function values of some MNx and lanthanide-MNx gate electrodes as
a function of lanthanide type and concentrations under different annealing conditions,
showing the tunability of MNx work functions by incorporating lanthanide.
111
Table 4.4 summarizes the barrier heights measured from several lanthanide-
MNx/SiO2 stacks as well as the corresponding work function values. The work-
function values are obtained from VFB versus EOT plots. The barrier height values are
extracted by Fowler-Nordheim current analysis on capacitors with ~6 nm SiO2
dielectrics. The effective mass of 0.4 (mox/m0) was chosen for the tunneling electrons
[18]. Most barrier height values correlate well with the extracted work function
values, suggesting the work function of MNx has indeed been modulated by the
lanthanide.
Table 4.4 Work function and barrier height of lanthanide-incorporated TaN on SiO2 as a
function of RTA temperatures.
Work function Barrier height

Gate (eV) (eV)
electrode 1000 1000
420 oC 800 oC 900 oC o 420oC 800 oC 900 oC o
C C
TaN 4.4 4.5 4.7 3.44 3.5 3.73
Ta0.94Tb0.06Ny 4.08 4.15 4.23 3.12 3.25 3.36
Ta0.95Er0.05Ny 4.17 4.28 4.34 4.30 3.14 3.23 3.32 3.29
Ta0.97Yb0.03Ny 4.44 4.34 4.36 4.37 3.40 3.25 3.41 3.05
The work function tunability of lanthanide-MNx on high-κ dielectric was also
investigated. Fig. 4.15 shows the C-V characteristics of Ta0.9Tb0.1Ny gated capacitors
on HfAlO dielectric before and after 1000 oC RTA process. The C-V measurements
fit well with the C-V curves produced by simulation, suggesting good interface quality
between HfAlO and Si substrate. Compared with the TaN/HfAlO stack, the
Ta0.9Tb0.1Ny/HfAlO gate stack shows a negative shift in VFB by 0.2~0.3 V irrespective
of the PMA temperature, which could be attributed to the lower WF of Ta0.9Tb0.1Ny
compared with that of TaN. Considering the EWF value of TaN on HfAlO (~ 4.6 eV
112
on HfAlO in our previous work; ~ 4.5 ± 0.1 eV on ALD HfAlO in Ref. [19]), the
EWF of Ta0.9Tb0.1Ny on HfAlO could be around ~ 4.3 eV. The Qf difference between
the two gate stacks, if any, is not likely to take the major role on the VFB difference
observed since the high-κ films used in the two gate stacks are identical. Further
study will be needed to study the work function tunability of lanthanide-MNx on high-
κ dielectrics in more detail.
o
Capacitance Density (fF/μm )
420 C FGA
2
15 o
1000 C 5 sec.
10 HfAlO, 55Å
TaN
5 Ta0.9Tb0.1Ny
Symbols: Measured
0 Lines: Simulated
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Gate Voltage (V)
Fig. 4.15 C-V characteristics of TaN and Ta0.9Tb0.1Ny metal gates on ALD HfAlO
dielectrics after FGA at 420 oC for 30 min. and RTA at 1000 oC for 5 sec.
4.5 Thermal Stability Study
Besides appropriate work function values, another important consideration in
choosing a suitable metal gate candidate is the thermal stability of the metal material
against the high-thermal budget used in the gate-first CMOS process. Therefore,
another effort in this chapter is to study the thermal stability of the lanthanide-MNx
metal gates and its dependence.
113
Fig. 4.16 shows the XTEM pictures of a Ta0.94Tb0.06Ny/SiO2 gate stack after
420 oC FGA, 900 oC RTA and 1000 oC RTA. No significant change has been
observed in the thickness of SiO2, indicating the good stability of Ta0.94Tb0.06Ny on
SiO2. This is also consistent with the C-V measurements shown in Fig. 4.12 where
the EOT of the Ta0.94Tb0.06Ny/SiO2 stack does not show much change with and
without 1000 oC RTA process.
Fig. 4.16 XTEM images of Ta0.94Tb0.06Ny/SiO2 gate stack on (100) Si substrate after
420 oC FGA for 30 min, 900 oC RTA for 30 sec and 1000 oC RTA for 30 sec.
Fig. 4.17 compares the EOT stability of the Ta1-xTbxNy/SiO2 gate stacks with
various Tb content after a 1000 oC RTA treatment. As shown in Fig. 4.17, TaN/SiO2
shows a significant increase in EOT for about 0.6 nm after 1000 oC RTA. This
increase is believed due to the poor oxygen diffusion barrier properties of TaN,
allowing the oxygen residues to penetrate through the film. However, the EOT
variation observed in the Ta1-xTbxNy/SiO2 stacks are much smaller than that in the
TaN/SiO2 stack, and it tends to decrease with the concentration of Tb in Ta1-xTbxNy.
Similar results can also be observed in other lanthanide-MNx-gated MOS devices.
114
The improved EOT stability in lanthanide-MNx-gated MOS capacitors implies that
lanthanide-MNx is a good diffusion barrier to O2, which could be related to the
amorphous structure of the Ta1-xTbxNy film as discussed in the section 4.3.4. An EOT
reduction by 1~2 Å was also observed in Fig. 4.17 when the Tb concentration
becomes high. This could be due to some reactions of the excess Tb with SiO2 at the
Ta1-xTbxNy/SiO2 interface, which can be improved by increasing the N content in the
film. The role of N to the thermal/chemical stability of lanthanide-MNx will be
discussed later in this section.
0.8
Ta1-xTbxNy on SiO2
Δ EOT during RTA (nm)
0.6
Initial EOT ~2.7 nm before RTA
0.4
0.2
0.0
-0.2
o
RTA at 1000 C 20 s
-0.4
N
N
y
y
N
N
N
y
Ta
03
13
06
1
0.
Tb
0.
0.
0.
Tb
Tb
Tb
9
97
87
0.
94
Ta
0.
0.
0.
Ta
Ta
Ta
Fig. 4.17 EOT variation of the Ta1-xTbxNy/SiO2 gate stacks before and after 1000 oC
RTA for 20 sec, as a function of Tb concentrations in Ta1-xTbxNy.
Fig. 4.18 and Fig. 4.19 present the gate leakage characteristics of the
Ta0.94Tb0.06Ny/SiO2 and Ta0.95Er0.05Ny/SiO2 stacks after thermal treatments at different
temperatures. The I-V characteristic does not show significant change after annealing
115
at different temperatures, suggesting good thermal and chemical stability of the
Ta0.94Tb0.06Ny/SiO2 (or Ta0.95Er0.05Ny/SiO2) interface. The TDDB reliability
characteristics of lanthanide-MNx metal gates after different RTA treatments were
also investigated. The Ta0.94Tb0.06Ny/SiO2 capacitors with different PMA performed
were stressed under a constant negative bias to examine the potential impact of high
temperature annealing on the reliability of the Ta0.94Tb0.06Ny /SiO2 gate stack, as
shown in Fig. 4.20. No significant reliability degradation is observed even after RTA
at 1000 oC for 30 secs, demonstrating the excellent thermal stability of Ta0.94Tb0.06Ny.
The appropriate N content in the lanthanide-MNx metal gates, which can suppress the
reaction of lanthanide with the underlying SiO2, is believed to be responsible for the
thermal stability of lanthanide-MNx.
-2
Leakage Current Density (A/cm )
10
2
o
420 C FGA
o
900 C RTA 30 sec
-4
10 o
1000 C RTA 30 sec
-6
10
10
-8 Ta0.94Tb0.06Ny
on SiO2 (~3.2 nm)

-10
10
-3 -2 -1 0
VG-VFB (V)
Fig. 4.18 Gate leakage characteristics of Ta0.94Tb0.06Ny/SiO2 gate stack with PMA
performed at different temperatures.
116
-3
Leakage Current Density (A/cm )

10
2
-4
Ta0.95Er0.05Ny
10
on SiO2 (~ 3.5 nm)
-5
10
-6
10
o
420 c FGA
o
-7 800 c RTA
10 o
900 c RTA
o
-8 1000 c RTA
10
-3 -2 -1 0
VG-VFB (V)
Fig. 4.19 Gate leakage characteristics of Ta0.95Er0.05Ny/SiO2 gate stack with PMA
performed at different temperatures.
o
420 C FGA
1 o
900 C RTA 30 sec
o
1000 C RTA 30 sec
0
ln(-ln(1-F))
-1 Ta0.94Tb0.06Ny/SiO2
SiO2 ~ 3.2 nm
CVS stress
-2 Gate Injection
Eox~10.5 MV/cm
-3 0 1 2 3
10 10 10 10
Time to Breakdown (s)
Fig. 4.20 TDDB characteristics of Ta0.94Tb0.06Ny/SiO2 gate stack (SiO2~3.2 nm) after
PMA at different temperatures, measured under negative constant voltage stress (CVS)
at room temperature.
117
In order to study the role of N in the thermal stability of lanthanide-MNx
materials, we varied the N2/Ar flow rates during the deposition of Ta0.9Tb0.1Ny (N2/Ar
= 3/25, 5/25, and 8/25 sccm) on purpose. Fig. 4.21 compares the EOT stability
among Ta0.9Tb0.1Ny gates with different N2 flow rates during deposition. It is
observed that Ta0.9Tb0.1Ny with low N concentration (N2/Ar = 3/25) shows poor EOT
stability against the high thermal budget used in the process. This is attributed to
some reactions between Tb and/or Ta metals with the underlying SiO2 dielectrics.
Fortunately, this kind of interface reaction can be suppressed by increasing the
nitrogen concentration in lanthanide-MNx. As shown in Fig. 4.21, the EOT of
Ta0.9Tb0.1Ny (N2/Ar = 8/25)/SiO2 gate stack keeps almost constant even after 1000 oC
RTA treatment, implying a stable interface between Ta0.9Tb0.1Ny (N2/Ar = 8/25) and
SiO2.
8
N2/Ar = 3/25
7 N2/Ar = 5/25
N2/Ar = 8/25
6
EOT (nm)
5
4
3
Ta0.9Tb0.1Ny/SiO2
2 No Yield
420 900 1000
o
RTA Temperature ( C)
Fig. 4.21 EOT variation as a function of annealing temperature for Ta0.9Tb0.1Ny/SiO2

gate stacks with different N2 flow rates during the deposition of Ta0.9Tb0.1Ny. Two
groups of oxides with initial thickness of ~3.3 nm and ~ 5.8 nm were investigated.
118
The higher N concentration can also contribute to the leakage current
characteristics in Ta0.9Tb0.1Ny/SiO2 stack, as depicted in Fig. 4.22. Ta0.9Tb0.1Ny with
lower N2 flow rate (N2/Ar = 3/25) and consequently lower N concentration exhibits
higher gate leakage than the other two splits. These results suggest that the N
concentration in lanthanide-MNx could be responsible for the chemical-thermal
stability of the lanthanide-MNx/SiO2 interface.
-3
10
N2/Ar = 3/25
-4
Gate Leakage (A/cm )
10 N2/Ar = 5/25
2
-5 N2/Ar = 8/25
10
-6
10
-7
10 Ta0.9Tb0.1Ny/SiO2
o
-8
10 Tox~3.3 nm 420 C FGA
-3 -2 -1 0
Vg-VFB (V)
Fig. 4.22 Typical I-V characteristics of Ta0.9Tb0.1Ny gated MOS capacitors with
different N2 flows during metal gate deposition, measured after FGA at 420 oC for 30
min.
If the N concentration is too high, however, it will lead to high work function
values for lanthanide-MNx. Fig. 4.23 compares the WF of Ta0.9Tb0.1Ny metal gates
with different N concentrations as a function of annealing temperature. The result
shows that Ta0.9Tb0.1Ny with N2/Ar = 5/25 exhibits a low WF of ~ 4.2 eV after 1000
o
C RTA process. However, the split with highest nitrogen flow (N2/Ar = 8/25) shows
119
higher WF (4.4 ± 0.05 eV) than the others. The reason for the dependence of the
metal work function on the N concentration is still not very clear, but a possible
reason could be the reduction of Tb atoms near the metal-dielectric interface due to
the increased N concentrations. A more plausible explanation could be that the
number of metallic Tb species, which contributes to the low WF, decreases when the
N concentration in Ta0.9Tb0.1Ny increases due to the formation of Tb-N compounds,
which may possess insulator properties [15].
4.6
Ta0.9Tb0.1Ny/SiO2
Work Function ΦM (eV)
4.5
N2/Ar = 8/25
4.4
4.3
4.2
4.1
N2/Ar = 3/25 N2/Ar = 5/25
4.0
420 900 1000
o
RTA Temperature ( C)
Fig. 4.23 Work function of Ta0.9Tb0.1Ny gate electrodes with different N

concentrations as a function of annealing temperature.
In summary, the thermal stability of lanthanide-MNx is found to be related
with the N content. Therefore carefully engineering of the N content according to the
lanthanide additives in the lanthanide-MNx metal gate could be important in the
process to achieve a best trade-off between the good thermal stability and a proper
work function value desired.
120
4.6 MOSFET Characteristics
A further effort of this chapter is to discuss some process integration issues of
the lanthanide-MNx metal gate in CMOS process. As mentioned before, lanthanide-
MNx has higher resistivity than the MNx metal gates. This leads to high sheet
resistance of the gate electrode stack and is not desirable. One way to address this
problem is to minimize the thickness of lanthanide-MNx in the metal gate stack. An
additional advantage to use a thin lanthanide-MNx layer is that this minimizes the
difficulties in patterning these ternary metal nitride films since the lanthanide and the
refractory metal species (e.g. Tb and Ta) may possess different etching properties
(etching rate, selectivity, byproduct properties, etc.).
However, the thickness of a lanthanide-MNx layer must be large enough to
determine the WF, otherwise the capping layer material may affect the WF of the gate
stack [20]. To evaluate the dependence of work function on Ta0.9Tb0.1Ny thickness,
Ta0.9Tb0.1Ny-gated MOS capacitors with different Ta0.9Tb0.1Ny thickness (~ 40 Å, 100
Å, 150 Å and 200 Å) were fabricated. Fig. 4.24 shows the cross-section image of the
gate stack consisting of TaN capping layer, 40 Å Ta0.9Tb0.1Ny and SiO2 dielectric.
The C-V characteristics of these devices are measured after 1000oC PMA and plotted
in Fig. 4.25. It is observed that the device with ~ 40 Å Ta0.9Tb0.1Ny shows different
VFB compared with the other samples.
The WF values of Ta0.9Tb0.1Ny with different thickness are extracted in Fig.
4.26. It is found that Ta0.9Tb0.1Ny with a thickness of ~ 40 Å exhibits higher WF
value than the other splits with thicker Ta0.9Tb0.1Ny of 100 ~ 200 Å. It is also
interesting to note that the WF difference tends to increase with the annealing
temperature. As shown in Fig. 4.26, the WF difference for Ta0.9Tb0.1Ny/TaN stack
with varying Ta0.9Tb0.1Ny thicknesses is only about 0.08 eV if the RTA is performed
121
at 900 oC, but it becomes 0.18 eV when the RTA is performed at 1000 oC for 10 sec.
This implies that the different WF for Ta0.9Tb0.1Ny/TaN stacks with varied
Ta0.9Tb0.1Ny thicknesses could be attributed to some interaction between the
Ta0.9Tb0.1Ny and the capping TaN layer during the high-temperature annealing process,
causing the effective work function of Ta0.9Tb0.1Ny/TaN metal gate stack to be shifted.
The interaction becomes more severe when the annealing temperature becomes higher.
The behaviors of EWF in the stacked thin metal layers are quite interesting and have
been reported in many studies [20]-[22]. Carrier redistribution in thin continuous
metal stacks [20] and/or inter-diffusion of the metal species in bi-layer structures [22]
have been proposed to explain the dependence of EWF on the metal layer thicknesses
in the stacked metal structures. The mechanisms in our case, however, are still not
very clear and need to be further investigated from a scientific perspective. To
guarantee the low EWF value, Ta0.9Tb0.1Ny with thickness of 100 Å is chosen for the
MOSFET fabrication in this work.
Fig. 4.24 XTEM image of TaN/Ta0.9Tb0.1Ny/SiO2 gate stack with Ta0.9Tb0.1Ny

thickness of ~ 40 Å on SiO2 dielectric.
122
14
TaTbN-40 Å

2
12 TaTbN-100 Å
10 TaTbN-150 Å
TaTbN-200 Å
8
6 o
RTA 1000 C 10 s
4
SiO2
2
EOT~ 2.25 nm
0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Gate Voltage (V)
Fig. 4.25 C-V characteristics of Ta0.9Tb0.1Ny/SiO2 gate stacks with different

Ta0.9Tb0.1Ny thickness. The measurement was taken on the samples with 1000 oC
RTA for 10 sec performed.
-0.2
TaTbN-40 Å TaTbN-100 Å
-0.4
VFB (V)
ΦM=4.3 eV
-0.6
-0.8 ΦM=4.22 eV
(a) RTA 900oC 20 s
-1.0
0 1 2 3 4 5 6 7 8 9 10
EOT (nm)
Fig. 4.26 (a) VFB versus EOT plots of Ta0.9Tb0.1Ny/SiO2 gate stacks with different
Ta0.9Tb0.1Ny thickness after 900 oC RTA for 20 sec.
123
0.0
-0.2 TaTbN-150 Å TaTbN-200 Å
ΦM=4.43 eV
-0.4
VFB (V)
-0.6
-0.8 ΦM=4.25 eV
(b) RTA 1000oC 10 s
-1.0
0 1 2 3 4 5 6 7 8 9 10
EOT (nm)
Fig. 4.26 (b) VFB versus EOT plots of Ta0.9Tb0.1Ny/SiO2 gate stacks with different
Ta0.9Tb0.1Ny thickness after 1000 oC RTA for 10 sec.
In the MOSFET fabrication, a damascene-gate process was used to pattern the
gate electrode in order to avoid the S/D recess problem associated with the non-
optimized plasma etching process, as illustrated in Fig. 4.27. First, a thick passivation
oxide layer of ~1500 Å was grown on p-Si substrate, followed by the opening of the
gate area on the passivation oxide. Second, the thermal oxide with thickness of about
30 Å was grown as the gate dielectric, followed by the deposition of 100-Å-thick
Ta0.9Tb0.1Ny and an in-situ deposited TaN capping layer of about 2000 Å. Third, a
CMP process was used to polish away the metals on top of the passivation oxide.
Finally, the passivation oxide was removed by diluted hydrofluoric (DHF) acid
solution, followed by a standard process to form spacer and S/D region. Note that
such a process is only a temporary solution when the plasma etching process for these
novel metal gate materials has not been well developed. Further development of a dry
etching solution will be required in future study.
124
Fig. 4.27 Process flow of the damascene process used to pattern the TaN/
Ta0.9Tb0.1Ny metal gate stack in MOSFET fabrication.
Fig. 4.28 to Fig. 4.31 present the characteristics of n-MOSFET with
Ta0.9Tb0.1Ny/SiO2 gate stack. The source/drain activation anneal was performed at
1000oC for 5 sec in N2 ambient. A typical C-V curve measured from the n-MOSFET
is depicted in Fig. 4.28, with no gate-depletion observed. Fig. 4.29 and Fig. 30
present the IDS ~ VDS and IDS ~ VGS characteristics of n-MOSFET with Ta0.9Tb0.1Ny
metal gate. The threshold voltage is only about 0.07 V thanks to the low WF of
Ta0.9Tb0.1Ny (4.25 ± 0.05 eV). It was reported that N can diffuse out from TaNx
during the high-temperature process when the N concentration is high [23], resulting
in the degradation of the SiO2/Si interface. However, an excellent subthreshold swing
(SS) of 68 mV/dec has been obtained from Ta0.9Tb0.1Ny/SiO2 gated n-MOSFET,
suggesting that the interface quality may not be affected in the Ta0.9Tb0.1Ny-gated
devices. This could be attributed to the presence of Tb which is more capable to
combine with N, as discussed in section 4.3.2 and 4.3.3. The effective electron
mobility in the Ta0.9Tb0.1Ny/SiO2 stack has also been investigated, as shown in Fig.
4.31. No mobility degradation has been observed, suggesting good interface quality.
125
These results demonstrate that Ta0.9Tb0.1Ny could be potential N-type metal gate
candidates for next-generation transistors. More studies on the characteristics of
lanthanide-MNx with high-κ dielectrics will be needed in the future work.
10 Ta0.9Tb0.1Ny / SiO2
2
EOT = 2.96 nm
8 VFB = -0.78V
2 Measured
Simulation
0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Gate Voltage (V)
Fig. 4.28 Typical high-frequency C-V measurement of Ta0.9Tb0.1Ny/SiO2 gated n-

MOSFET.
Ta0.9Tb0.1Ny / SiO2 VG = 1.2 V

Drain Current IDS (μA/μm)
300 W/L= 50 μm / 0.6 μm

250 Vth = 0.07 V VG = 1.0 V
200
VG = 0.8 V
150
VG = 0.6 V
100
VG = 0.4 V
50
0
0.0 0.5 1.0 1.5 2.0
Drain Voltage VDS (V)
Fig. 4.29 IDS ~ VDS characteristics of Ta0.9Tb0.1Ny/SiO2 gated n-MOSFET, with the
substrate doping concentration of NA = 5 × 1015 cm-3.
126
-2
10
Ta0.9Tb0.1Ny / SiO2 VD = 1.0 V
Drain Current IDS (A/μm)

-4
10
-6 VD = 0.1 V
10
-8
10 SS = 68 mV/dec.
-10 Vth = 0.07 V
10
-12
W/L= 50 μm / 0.6 μm
10
EOT=2.96 nm
-14
10
-1 0 1 2
Gate Voltage VGS (V)
Fig. 4.30 IDS ~ VGS characteristics of Ta0.9Tb0.1Ny/SiO2 gated n-MOSFET, with the
substrate doping concentration of NA = 5 × 1015 cm-3.
500
Electron Mobility (cm / V⋅s)
Universal
400
2
300
200
100 Ta0.9Tb0.1Ny / SiO2
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Effective Field (MV/cm)
Fig. 4.31 Effective electron mobility in Ta0.9Tb0.1Ny/SiO2 gated n-MOSFET.
127
4.7 Conclusion
In summary, for the first time, we successfully demonstrated a novel method
to tune the work function of MNx gate electrodes by incorporating lanthanide-series
elements. The material and electrical properties of lanthanide-incorporated metal
nitrides have been systematically studied. The results show that the work function of
the lanthanide-MNx metal gates can be continuously tuned by varying the lanthanide
concentration, and a work function value of 4.2~4.3 eV can be obtained even after a
1000 oC RTA treatment, which is attractive for the applications in n-MOSFET using a
gate-first process. Good thermal stability with respect to EOT, leakage current and
TDDB characteristics have also been achieved up to 1000oC in the TaTbN/SiO2
stacks. The nitrogen concentration is identified as an important parameter to the
properties of lanthanide-MNx, especially for the thermal stability; therefore the N
concentration needs to be carefully engineered in lanthanide-MNx. MOSFETs with
Ta0.9Tb0.1Ny as the metal gate electrode have also been successfully demonstrated,
with excellent transistor characteristics reported. All these results indicate that
lanthanide-MNx can be a promising N-type metal gate candidate for the next
generation CMOS technology.
128
References
1. Y. H. Kim, C. H. Lee, T. S. Jeon, W. P. Bai, C. H. Choi, S. J. Lee, L. Xinjian, R.

Clarks, D. Roberts, and D. L. Kwong, “High quality CVD TaN gate electrode for
sub-100 nm MOS devices,” in IEDM Tech. Dig., pp. 667-670, 2001.
2. D.-G. Park, Z.J. Luo, N. Edleman, W. Zhu, P. Nguyen, K. Wong, C. Cabral, P.

Jamison, B.H. Lee, A. Chou, M. Chudzik, J. Bruley, O. Gluschenkov, P.
Ronsheim, A. Chakravarti, R. Mitchell, V. Ku, H. Kim, E. Duch, P. Kozlowski, C.
D’Emic, V. Narayanan, A. Steegen, R. Wise, R. Jammy, R. Rengarajan, H. Ng, A.
Sekiguchi, and C.H. Wann, “Thermally robust dual-work function ALD-MNx
MOSFETs using conventional CMOS process flow,” in VLSI Tech. Dig., pp. 186-
187, 2004.

Dig., pp. 99-103, 2003.
4. W. Tsai, L.-Å Ragnarsson, L. Pantisano, P. J. Chen, B. Onsia, T. Schram, E.

Cartier, A. Kerber, E. Young, M. Caymax, S. De Gendt, and M. Heyns,
“Performance comparison of sub 1 nm sputtered TiN/HfO2 nMOS and
pMOSFETs,” in IEDM Tech. Dig., pp. 311-314, 2003.
5. H. B. Michaelson, “The work function of the elements and its periodicity,” J.

Appl. Phys., vol. 48, pp. 4729-4733, 1977.
6. T. Durakiewicz, S. Halas, A. Arko, J. J. Joyce, and D. P. Moore, “Electronic

work-function calculations of polycrystalline metal surfaces revisited,” Phys. Rev.
B, vol. 64, 045101, Jun. 2001.
7. M. K. Fung, S. W. Tong, S. L. Lai, S. N. Bao, and C. S. Lee, W. W. Wu, M.

Inbasekaran, and J. J. O’Brien, S. Y. Liu, and S. T. Lee, “Role of ytterbium and
ytterbium/cesium fluoride on the chemistry of poly (9,9-dioctylfluorene-co-
benzothiadiazole) as investigated by photoemission spectroscopy,” J. Appl. Phys.,
vol. 94, pp. 2686-2694, Aug. 2003.
8. Dieter K. Schroder, Semiconductor material and device characterization, 2nd

Edition, John Wiley & Sons, Inc, New York, 1998, ch. 10, pp. 689-697.
9. Refer to www.genplot.com for XRUMP simulator.
10. T. Sekine, Y. Nagasawa, M. Kudoh, Y. Sakai, A. S. Parkes, J. D. Geller, A.

Mogami, and K. Hirata, “Handbook of Auger Electron Spectroscopy,” JEOL Inc.,
Tokyo,1982.
11. K. Sasaki, A. Noya, and T. Umezawa, “Stoichiometry of Ta-N film and its
application for diffusion barrier in the Al3Ta/Ta-N/Si contact system,” Jpn. J.
Appl. Phys., vol. 29, pp. 1043-1047, 1990.
129
12. M.-F. Wang, T.-Y. Huang, Y.-C. Kao, H.-C. Lin, and C.-Y. Chang, “Impact of
thermal stability on the characteristics of complementary metal oxide
semiconductor transistors with TiN metal gate,” Jpn. J. Appl. Phys., vol. 41, pp.
546-551, Feb. 2002.
13. C. S. Kang, H.-J. Cho, Y. H. Kim, R. Choi, K. Onishi, A. Shahriar, and J. C. Lee,
electrode applications,” J. Vac. Sci. Technol. B, vol. 21, pp. 2026-2028, Sep./Oct.
2003.
14. X. Sun, E. Kolawa, J.-S. Chen, J. S. Reid, and M.-A. Nicolet, “Properties of
reactively sputter-deposited Ta-N thin films,” Thin Solid Film, vol. 236, pp. 347-
351, 1993.
15. C. M. Aerts, P. Strange, M. Horne, W. M. Temmerman, Z. Szotek, and A. Svane,

“Half-metallic to insulating behavior of rare-earth nitrides,” Phys. Rev. B, vol. 69,
045115, 2004.

Lett., vol. 25, pp. 123-125, Mar. 2004.
May 2004.
18. M. Lenzlinger and E. H. Snow, “Fowler-Nordheim tunneling into thermally

grown SiO2,” J. Appl. Phys., vol. 40, pp. 278-283, 1969.
19. S. G. Park, Y. K. Lee, S. B. Kang, H. S. Jung, S. J. Doh, J.-H. Lee, J. H. Choi, G.

H. Kim, G. H. Choi, U. I. Chung, and J. T. Moon, “Performance improvement of
MOSFET with HfO2-Al2O3 laminate gate dielectric and CVD-TaN metal gate
deposited by TAIMATA,” in IEDM Tech. Dig., pp. 327-330, 2003.

Dig., pp. 303-306, 2004.
21. H.N. Alshareef, K. Choi, H.C. Wen, H. R. Harris, H. Luan, P. Lysaght, P. Majhi,
and B.-H. Lee, “Gate work function modification using ultra-thin metal
interlayers,” Proc. ECS 207th meeting, 630, Quebec, Canada, 2005.
22. C. H. Lu, G. M. T. Wong, M. D. Deal, W. Tsai, P. Majhi, C. O. Chui, M. R.

Visokay, J. J. Chambers, L. Colombo, B. M. Clemens, and Y. Nishi,
“Characteristics and mechanism of tunable work function gate electrodes using a
130
bilayer metal structure on SiO2 and HfO2,” IEEE Electron Device Lett., vol. 26,
pp. 445-447, July 2005.
23. Y.-S. Suh, G. P. Heuss, J.-H. Lee, and Veena Misra, “Effect of the composition
on the electrical properties of TaSixNy metal gate electrodes,” IEEE Electron
Device Lett., vol. 24, pp. 439-441, July 2003.
131
Ch 5: Process Integration of Dual Metal Gate Electrodes
Chapter 5
Process Integration of Dual Metal Gate

Electrodes
5.1 Introduction
Integration of dual metal gate electrodes into the CMOS process is another
major challenge for the development of metal gate technology. The criteria for an
ideal dual metal gate integration process include: (i) achieving dual work function
(WF) values for NMOS and PMOS respectively; (ii) with no process-induced
damages or reliability concerns to the gate oxide; and (iii) less process complexity.
Moreover, good scalability is also required for the ideal dual metal gate integration
process for the implementation of metal gates in nanometer CMOS fabrication.
Depending on the different integration schemes used, the issues and considerations
regarding the metal gate integration would be different.
Generally, the schemes to integrate the dual work function metal gates onto a
single wafer can be reduced to three categories: gate-first approach, gate-last approach,
and FUSI gate integration process. Among the three categories of integration
schemes, the gate-first approach is most challenging and many issues like metal
selection, metal etching and cleaning need to be considered systematically. The gate-
last integration schemes show less difficulty in material selection, but the problem is
the process complexity and cost. The FUSI process requires least change to the
132
conventional CMOS process, but the work function tuning issues on high-κ dielectric
need to be addressed.
In this chapter, we will first review the existing dual metal gate integration
processes ever reported and discuss the problems and concerns related to these
processes. This will facilitate the following discussions on the motivation and
innovations of two proposed integration schemes in this work.
The first dual metal gate CMOS integration process was reported by Lu et al.
by a direct-etching method, where Ti and Mo were used as the gate electrodes for
NMOS and PMOS, respectively [1]. The basic steps of this integration scheme are
illustrated in Fig. 5.1. One metal gate material (Metal-A) is deposited on the whole
wafer firstly, followed by selective etching from one side (NMOS or PMOS). Then
the other metal gate (Metal-B) is deposited, and finally it is the gate patterning and
source/drain (S/D) formation. Sometimes a poly-Si capping layer is deposited on top
of Metal-B to reduce the thicknesses of Metal-A and Metal-B layers; or a
planarization process can be inserted after step (c) to reduce the height difference
between NMOS and PMOS for the depth-of-focus (DOF) consideration in gate
lithography. This approach is a gate-first integration scheme and is generally used in
integrating those direct metal gate candidates with good thermal stability, like TiN-
TaSiN, TaSiN-Ru, etc [2]-[3].
Fig. 5.1 Process flow of dual metal gate integration by direct etching method. PR
denotes photoresist, HM denotes hard-mask, and HK denotes high-κ dielectric.
133
The most challenging step in this direct-etching integration scheme is the
selective etching of Metal-A from the gate dielectric and the subsequential process to
strip the photoresist (PR) or hard-mask (HM) used to pattern the Metal-A (Fig. 5.1
(b)). Dry etching of Metal-A from gate dielectric will be easy to cause a loss of gate
dielectric, leading to different EOT for NMOS and PMOS, as observed in [1]. A wet
etching process will be more gentle and safe for the gate dielectric, but still there may
be some reliability concerns because the gate dielectric is exposed to a series of wet
chemicals during processing. Moreover, appropriate chemical solution need to be
identified for the specific metal material so that high selectivity can be achieved
among Metal-A, dielectric and hard-mask (HM) [4]. In addition, a proper plasma
etching process also needs to be developed to enable good selectivity to be achieved
on NMOS and PMOS regions simultaneously, given that the different metal materials
with different thicknesses are formed in NMOS and PMOS regions (referring to step
(d)).
In order to avoid possible etching damage to the gate dielectric, a dual high-κ
(HK) dual metal gate integration process was proposed [5]. The basic idea of this
process is to remove the high-κ dielectric together with Metal-A during the metal
etching process, and then deposit a new high-κ dielectric layer and Metal-B material
as the new gate stack, as illustrated in Fig. 5.2. Note that the Metal-B and HK-B
layers must be selectively removed from the top of Metal-A to make the Metal-A
stack conductive, as depicted in step (d)-(e) of Fig. 5.2. By using this process, the
gate dielectric in either NMOS or PMOS region will not be exposed during the metal
etching process, thus less reliability problem would be expected. An additional
advantage of this integration scheme is that the metal gate/high-κ stacks for n-
MOSFET and p-MOSFET can be engineered independently to obtain optimized
134
transistor characteristics. However, one more lithography step and CMP step is
required in this scheme and thus the process complexity increases.
HM-A HM-A
HK-A selective HK-B
etching
Metal-A Metal-A Metal-A Metal-B
(a) (b) (c)
Poly-Si
HM-B HM-A HM-B Capping
Metal-A Metal-B Metal-A Metal-B Metal-A Metal-B
(d) (e) (f)
Fig. 5.2 Process flow of the dual metal gate/dual high-κ integration scheme. (a)
Metal-A/HK-A deposition; (b) Metal-A/HK-A selective etching from one side of
CMOS; (c) Metal-B/HK-B deposition; (d) hard-mask-B deposition and patterning; (e)
Metal-B/HK-B selective removal; (f) hard-mask removal, thick poly-Si top-up, and
gate patterning.
In the above process flows, the first-deposited metal gate (Metal-A) are all
etched away in order to put another metal gate to get dual work function. But actually
this may not be necessary. Another stratagem to achieve dual work function for
CMOS is to “modulate” the work function of the first-deposited metal, instead of
“replace” it by others. The metal inter-diffusion method is such an integration scheme
with no step to expose the gate dielectric during the whole process, and hence less
process complexity [6]. As shown in Fig. 5.3, a metal gate (Metal-A) is first
deposited on top of the gate dielectric on the whole wafer. Then the work function of
Metal-A can be modulated by putting the 2nd metal (Metal-B) on top of Metal-A and
135
performing a subsequential annealing process to form a new metal alloy (Alloy-AB).
In some extreme cases, Metal-B can even segregate at the metal-dielectric interface
and push the Metal-A atoms away from the interface.
Fig. 5.3 Process flow of the dual metal gate integration via metal inter-diffusion.
The metal inter-diffusion approach has been applied to many metal
combinations, such as Ru-Ta alloy [7], Ta-Pt alloy [8], Hf-Mo alloy [9], and even
Mo-Si pair [10] where Mo and MoSix is formed for PMOS and NMOS, respectively.
However, the concern of this integration scheme is the thermal stability of the N-type
metal gate candidate. Though a P-type metal gate candidate with good thermal
stability can be utilized as the bottom metal (Metal-A), the stability of the Alloy-AB
is still a concern for the applications of this approach in a gate-first CMOS process.
Therefore it may be difficult to apply such an integration scheme into a gate-first
CMOS process. However, it is at least quite useful in a gate-last process.
Another example to “modulate” the work function with no need to etch away
the first-deposited layer is the FUSI gate process, where the conventional poly-Si gate
is used as the bottom protective layer instead of the Metal-A in the inter-diffusion
process discussed above. The use of poly-Si as the bottom layer makes the FUSI
process most compatible with the conventional CMOS process flow. As illustrated in
Fig. 5.4, a poly-Si gated CMOSFET is first fabricated using a conventional gate-first
136
CMOS process. After S/D silicidation, the devices are wrapped by the post-metal-
dielectric (PMD). A planarization step is then performed using oxide CMP until the
HM on top of poly-Si is exposed. After HM stripping and pre-silicidation cleaning,
metal film(s) is deposited on top of poly-Si followed proper rapid thermal annealing
(RTA) steps to form the desired silicide phases. Some optional process steps, such as
an ion implantation (I/I) process (to introduce impurities) or a poly-Si etch-back step
(to adjust the poly-Si thickness for desired silicide phases) [11], can be introduced
before the metal deposition, as shown in Fig. 5.4 (c). From the integration point of
view, the FUSI process is less challenging thanks to its compatibility with the
conventional CMOS process. However, the work function tunability on high-κ
dielectric still needs to be improved, as well as the precise process control of the
impurities and phases in MOSFETs with sub-20-nm Lg.
I/I (optional)
HM
Poly-Si
(a) (b) (c)
Single Metal (e.g. Ni) Multi-Metals (e.g. Pt / Ni)

after silicidation
(d1) (d2) (e)
Fig. 5.4 Process flow of the FUSI process. (a) CMOS fabrication conventionally; (b)
oxide re-flow and planarization by CMP; (c) hard-mask stripping followed by ion-
implantation or poly-Si etch-back; (d1) deposition of the same metal, e.g. Ni, for both
NMOS and PMOS; (d2) deposition of different metals for NMOS and PMOS,
respectively (in parallel with step (d1)); (e) silicidation and unreacted metal stripping.
137
Finally, the gate-last replacement gate process has also been widely studied for
long time. The process flow of the replacement gate process is described in Fig. 5.5.
First, CMOS transistors with dummy SiO2 gate dielectric and poly-Si gate electrodes
are fabricated using a conventional process, followed by the PMD deposition and
planarization. After that, the dummy poly-Si and SiO2 dielectric are selectively
removed to create a groove. New high-κ and metal gate materials are then filled into
the groove to replace the original SiO2 and poly-Si gate, followed by CMP to pattern
the gate electrode. Dual metal gates can be integrated one by one by utilizing the
above “replacement gate” process for two times. Alternatively, one can open the
grooves of NMOS and PMOS together and adopt a metal inter-diffusion process to
obtain dual work function metal gates.
Dummy poly-Si
& oxide removal
HM
Poly-Si
(a) (b) (c)
HK-A
Metal-A Metal-A Metal-B
Metal-A
(d) (e) (f)
Fig. 5.5 Process flow of the replacement gate process. (a) CMOS fabrication with
poly-Si as dummy gate; (b) oxide re-flow and planarization; (c) dummy poly-Si &
SiO2 removal; (d) filling the groove with new high-κ and metal gate; (e) metal CMP
to pattern the metal gate; (f) dual metal gate CMOS formation by repeat steps (c)-(e).
138
From the process integration perspective, the replacement gate process can
bypass the metal selective etching step required in the direct-etching approach, and
allow the metal/high-κ gate stacks for NMOS and PMOS to be optimized individually.
From the material selection point of view, many metal gate candidates, including both
the direct metal gates and the binary inter-diffusion metal alloys, can all be used in a
replacement gate process, thanks to the low thermal budget used in this process.
Therefore, the replacement gate process could be a valuable solution for the metal
gate integration. Unfortunately, the concerns of process complexity and cost may
prevent this process being utilized for mass-production.
In summary, integration of dual metal gate in CMOS process is a very
challenging task and many material and process issues need to be considered
systematically. Innovations in both the novel materials and process integration
schemes will be highly valued. In order to address or improve some of the issues and
limitations in the existing integration schemes mentioned above, we propose two
novel dual metal gate integration methods in this chapter. The first one is a gate-first
dual metal gate CMOS integration process with the aim to solve the etching damage
issue associated with the direct-etching method. In this process, an ultra-thin TaN
buffer layer is utilized to protect the gate dielectrics during the metal etching process
and a subsequential high-temperature intermixing technique is used to tune the work
function of TaN for dual-WF CMOS. The other one is a novel replacement gate
process utilizing HfN as the dummy gate to enable the large work function tuning
range and excellent EOT scalability to be achieved simultaneously. It is believed that
these attempts will be of practical values for the development of the dual metal gate
CMOS technology.
139
5.2 A Gate-First Dual Metal Gate Integration Scheme by High-
Temperature Metal Intermixing Technique
5.2.1 Motivation
Several approaches have been studied to integrate dual work function metal
gate (MG) electrodes into the CMOS platform using a gate-first process flow, as
discussed in the Section 5.1. The conventional direct-etching method is to selectively
remove the first-deposited metal from the gate dielectric, followed by the deposition
of the second metal material [1]-[3]. However, this may introduce a potential
reliability concern as the gate dielectric is exposed to plasma or several chemical
solutions during the metal etching process. To avoid the exposure of the gate
dielectric during processing, the metal inter-diffusion process was proposed, as
depicted in Fig. 5.3 [6]-[9]. But the concern is that the thermal stability of these metal
alloys is always limited by the properties of the low WF metal counterpart and not
adequate for a gate-first CMOS process. Park et al. has proposed a novel method in
which an ultra-thin AlN buffer layer is used to protect the gate dielectric during metal
etching, and is later consumed by Ta and Hf in an alloying process [12]. However,
obtaining a thermal stability which is adequate for the source/drain (S/D) dopant
activation step in a gate-first CMOS process is still challenging.
In order to develop a dual metal gate integration process which can not only
avoid the gate dielectric being exposed during the process but also fulfill the thermal
stability requirement for the gate-first CMOS process, the thermal/chemical stabilities
of the buffer metal layer which protects the gate dielectric need to be carefully
optimized. Using the well-known thermally stable materials, such as TaN or TiN, as
the buffer layer could be helpful to address this problem.
140
In this work, we report a dual metal gate integration scheme using ultra-thin
TaN as the buffer layer and a subsequently high-temperature metal intermixing (InM)
process to modulate the WF of TaN. TaN/Tb/TaN and TaN/Ti/HfN [13] stacks are
employed for NMOS and PMOS, respectively. Thanks to the good diffusion barrier
properties of TaN, the temperature for the metal InM process can be elevated to
1000oC, comfortable for the gate-first CMOS process. Several factors affecting the
metal intermixing process will also be discussed. In principle, this technique can be
further generalized to other MNx/metal stacks for performance optimization.
5.2.2 Proposed Integration Flow and Device Fabrication
buffer MNx ~2 nm P-gate Mat’l

Wet Etch & Capping
dielectric (e.g. TaNx)
N-well N-well
(1) P-well (2) P-well
N-gate Mat’l Top Capping

(e.g. Tb) (e.g. TaN)
N-gate P-gate
Intermixing
N-well N+ N+ P+ P+
(3) P-well (4) P-well N-well
Fig. 5.6 Dual metal gate integration process flow by high-temperature metal
intermixing technique: (1) TaN buffer layer deposition; (2) P-type metal gate stack
(e.g TaN/Ti/HfN) formation followed by selective etching; (3) N-type metal gate
stack formation (e.g TaN/Tb/TaN) and capping layer deposition; (4) gate etching, S/D
implantation, and dopant activation annealing (also for intermixing).
Fig. 5.6 shows the proposed dual MG integration process by high-temperature
metal intermixing technique. In this prototype demonstration, we used TaN/Tb/TaN
and TaN/Ti/HfN structures for NMOS and PMOS, respectively. Firstly, an ultra-thin
141
TaN buffer layer with thickness of ~2 nm was deposited on the gate dielectric using a
precisely controlled PVD facility, Anelva C-7100 GT PVD system, followed by the
deposition of Ti and HfN to form laminate gate stack [13]. Secondly, HfN and Ti
were selectively removed by diluted HF (DHF) solution (1:200) from NMOS region.
Terbium with thickness of ~2.5 nm was then deposited on TaN buffer layer to form
TaN/Tb stack in NMOS region, followed by the deposition of an in-situ TaN capping
layer to finalize the metal gate stack. The TaN capping can also be replaced by other
metals or a TaN/Poly-Si stack. After gate patterning, metal intermixing was carried
out during the S/D activation annealing process at 950oC or 1000oC for 1~5 sec.
To study the feasibility of this high temperature metal intermixing process in
tuning the WF of TaN, MOS capacitors with TaN/Tb/TaN and TaN/Ir/TaN gate
electrode stacks on thermally grown SiO2 were fabricated. TaN with thickness of
2~2.5 nm was first deposited on SiO2 by dc sputtering of Ta target in Ar/N2 ambient,
where the nitrogen N2 flow rate was varied during sputtering. Tb or Ir of ~2.5 nm was
then in-situ deposited on top of TaN, followed by a thick TaN capping layer. After
gate patterning, the capacitors were subjected to RTA up to 1000oC for metal
intermixing study. On the other hand, in order to investigate the feasibility of this
InM process in adjusting Vth of the MOSFET with high-κ dielectric, HfTaON/HfO2
(30Å/15Å) dielectric stack (HfTaON on top) was deposited directly on p-Si by PVD
process for MOSFET fabrication [14], followed by post-deposition-annealing (PDA)
in N2/O2 ambient (5% O2) to improve the film quality. TaN, TaN/Tb/TaN, and co-
sputtered Ta0.9Tb0.1Ny were then deposited as the gate electrodes of n-MOSFETs for
comparison. Intermixing of TaN/Tb/TaN stack was carried out at 1000oC for 1s
together with the S/D activation annealing. NiSi was performed in S/D regions to
reduce the sheet resistance. All the samples received a forming gas annealing (FGA)
142
at 420 oC for 30 min prior to fab-out. Cross-sectional transmission electron
microscope (XTEM) and energy dispersive X-ray (EDX) were utilized to study the
intermixing process. Atomic force microscopy (AFM) was used to characterize the
film roughness. WF was extracted from EOT versus flat band voltage (VFB) plots.
5.2.3 Feasibility Study of the High-Temperature Intermixing Method
Fig. 5.7 presents the XTEM images of TaN/Tb stack capped by thick TaN
layer before and after high-temperature intermixing annealing. As shown in the
XTEM images, smooth and clear interface between TaN and SiO2 can be found even
after 1000 oC RTA, indicating good barrier properties of TaN. Unlike HfN/Ti/TaN
structure studied in [13], the XTEM image of the TaN/Tb/TaN stack does not show
significant crystallization effect after 1000 oC RTA treatment. This could be due to
the large atom size difference between Tb and Ta [15].
Fig. 5.7 XTEM images of the TaN/Tb/TaN stack on SiO2 (a) as-deposited and (b)
after 1000 oC RTA in N2 ambient for 5 sec.
143
Intermixing of Tb with TaN buffer layer after 1000 oC RTA can be evidenced
by STEM/EDX depth profiles, as shown in Fig. 5.8. The Si-Ta profiles represent
counts from Si (K) line and/or Ta (M) line due to the overlapping of the two signals at
~1.7 keV in EDX spectrum. The EDX count of oxygen is low because of the lower
X-ray generation rate of oxygen compared with that of Ta or Tb. It is observed that
the centers of Tb and Ta profiles separates for ~2.5 nm for the as-deposited condition
(Fig. 5.8 (c)), corresponding with the XTEM result in Fig. 5.7 (a). After annealing,
the profile of Tb becomes broader and tends to mix with that of Ta (N), as shown in
Fig. 5.8 (d). Note that a small amount of Tb at the metal-dielectric interface would be
sufficient to modify the WF of TaN towards the conduction band of Si as
demonstrated in Chapter 4.
Counts (a.u.)
Counts (a.u.)
Fig. 5.8 (a), (b) STEM and (c), (d) EDX depth profiles of the TaN/Tb/TaN stack on
SiO2 as-deposited and after 1000 oC RTA in N2 ambient for 5 sec. In (a) & (b), the
dark layer in the sandwich structure denotes Tb element.
144
Fig. 5.9 summarizes the WF tunability of the metal intermixing approach. WF
of pure Tb and Ir elements are ~3.1 eV and ~5.27 eV, respectively. WF of ~4.25 eV
and ~4.75 eV were achieved in the TaN/Tb and TaN/Ir stacks, respectively, indicating
that the WF of TaN can be adjusted by the proposed high-temperature InM technique.
There are several factors which may affect the intermixing process as well as the final
WF values of the metal stacks. The first one is the thickness ratio between TaN and
the pure metal layers. If TaN is too thick or Tb (Ir) is not thick enough, metal on top
of TaN will be prevented from reaching the metal/dielectric interface due to the
diffusion barrier properties of TaN. For example, no significant VFB shift was
observed in a TaN/Tb stack with 40 Å TaN under 25 Å Tb even after 1000 oC RTA
for 5 sec. Secondly, the properties of the TaN bottom layer can also affect the
intermixing process. By reducing the N2 flow rate from 5 sccm to 4 sccm during
sputtering deposition of the TaNx buffer layer, it was observed that WF of TaNx with
less nitrogen (fN2 = 4 sccm) can be modified more easily with a lower thermal budgets
(Fig. 5.9), suggesting that N concentration determines the immunity of TaNx to
intermixing. Thirdly, the choice of the pure metal used to tune the WF is also of
importance. Although Ir can be used to tune the WF of TaN, the magnitude of WF
tuning is smaller than that of Tb; this will be discussed later. Moreover, TaN/Ir
devices show worse uniformity and larger EOT increase compared with TaN/Tb
capacitors. Breakdown was also observed in TaN/Ir gated capacitors after 1000oC
RTA for 10 sec, presumably due to the high diffusivity of the Ir element [16].
Therefore the P-type metal candidate may need to be further developed. On the other
hand, TaN/Tb stack with optimized metal thickness shows stable gate leakage even
after 1000oC 10 sec anneal as depicted in Fig. 5.10, implying good thermal stability
suitable for gate-first CMOS process.
145
5.0
Effective Work Function ΦM (eV)

TaN Control
4.8 TaN / Ir
4.6
4.4
4.2 TaN / Tb
TaNx (less N%) / Tb
4.0
FGA 900 950 1000
o
Anneal Temperature ( C)
Fig. 5.9 Work function versus annealing temperature for different TaN/Metal stacks.
Thickness of TaN or TaNx (less N%) bottom layers are about 2.0~2.5 nm, and that of
Tb or Ir are about 2.5 nm. The N2 gas flow rate during deposition of thin TaN layer is
5 sccm, while that for TaNx (less N%) is 4 sccm. All samples are capped with thick
TaN film of ~100 nm.
TaN/Tb InM.
Current Density (A/cm )
-1
2
10 ~2.5 nm/~2.5 nm
-3
SiO2 ~3.4 ± 0.1 nm
10
FGA
-5 o
10 900 C 20 Sec.
o
950 C 10 Sec.
-7 o
10 1000 C 5 Sec.
o
1000 C 10 Sec.
-9
10
-6 -4 -2 0
VG-VFB (V)
Fig. 5.10 Gate leakage characteristics of TaN/Tb stack after different RTA treatments.
The corresponding WF of the sample is denoted by open circle in Fig. 5.9.
146
A self-consistent charge transfer calculation [17] suggests that the WF of a
binary metal AB alloy is determined by not only the WF of each species, but also the
effective density of states at the Fermi level, as approximately expressed below [8][9]:
(Φ m, A − Φ m, B )( ρ A − ρ B )
Φ m ( x) = xΦ m, A + (1 − x)Φ m, B + x(1 − x) (5-1)
x ⋅ ρ A + (1 − x) ρ B
where x is the fraction of metal A in AB alloy, ρA and ρB are the effective densities of
states at the Fermi level for metal A and B, which are proportional to their respective
electronic specific heat coefficients γ. In the dilute-A limit, where a small
concentration x of metal A is incorporated into metal B to tune the WF, Eq. (5-1) can
be simplified as follow:
Φ m ( x ) ≅ Φ m , B + x (Φ m , A − Φ m , B ) ρ A / ρ B (5-2)
It is thus preferable to have ρA/ρB > 1 (hence γA > γB) for the introduction of a
small amount of A to modify the WF of metal B significantly. In other words, the
element with the larger effective density of states at the Fermi Level would have a
higher weighting in determining the final WF of AB alloy, and vice versa. It is
known that γ of Ta, Ir and Tb are 5.9, 3.1 and 9.05 [18], respectively. Therefore, the
different behaviors of Tb and Ir in modifying WF of Ta(N) can be plausibly explained
by the theory above, although the metal system in our study is not strictly a binary
alloy and the situation could be more complicated than described. Moreover, larger
WF tunability might be achievable by replacing Ir with Pt (γPt=6.8) or using TiN
(γTi=3.35) to replace TaN, as demonstrated in [19].
5.2.4 Compatibility with High-κ Dielectrics
Although the InM approach is effective in modulating work function on SiO2,
147
it is still not clear whether this method is compatible with high-κ dielectric or not.
Therefore we further investigated the effectiveness of this InM method in adjusting
the Vth of MOSFET with high-κ HfTaON dielectric. HfTaON has been reported as a
promising high-κ candidate with high mobility, good thermal and electrical stability
[14]. In this work, HfTaON/HfO2 (30Å/15Å) stack was used as the gate dielectric to
reduce the N concentration at the Si surface. TaN, TaN/Tb/TaN, and co-sputtered
Ta0.9Tb0.1Ny were used as the gate electrodes for comparison. The InM annealing and
S/D activation were performed at 1000oC for 1s at the same time. Fig. 5.11 compares
the C-V measurements of the n-MOSFETs with different metal gates. It is found that
the VFB of TaN/Tb/TaN-gated MOSFET shifts negatively compared with that of the
TaN-gated devices, showing the effectiveness of the InM method in tuning WF on
high-κ dielectric. However, the EOT of the Ta0.9Tb0.1Ny and TaN/Tb/TaN gated
nFET is smaller than that of TaN gated devices, possibly due to the scavenging effect
of Tb element [20].
TaTbN co-sput. EOT: 1.53 nm

2
18.0
TaN/Tb/TaN InM.
15.0 TaN
12.0
9.0 EOT:
2.25 nm
6.0
3.0
HfTaON/HfO2 EOT: 1.85 nm
0.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Gate Voltage (V)
Fig. 5.11 C-V characteristics of n-MOSFETs with TaN, TaN/Tb/TaN and co-
sputtered Ta0.9Tb0.1Ny metal gates on HfTaON after RTA at 1000 oC for 1 sec.
148
Fig. 5.12 compares the IDS-VGS characteristics of n-FET with three kinds of
metal gates: TaN, TaN/Tb/TaN by InM, and co-sputtered Ta0.9Tb0.1Ny with uniform
Tb composition. Good subthreshold swing of ~75 mV/dec has been achieved. It is
noted that a significant shift of Vth can be obtained by introducing Tb into TaN gate,
either by the InM approach or by the co-sputtering process. Statistic Vth distributions
of the three kinds of devices are compared in Fig. 5.13. The results show that the Vth
can be adjusted successfully with a magnitude of ~300mV by the InM method, while
larger Vth difference of ~400mV can be achieved using the pre-doped Ta0.9Tb0.1Ny
metal gate. This implies that only a small amount of Tb diffuses through TaN and
reaches the MG/high-κ interface during the InM annealing. Further adjustment of Vth
could be realized by optimizing the thickness ratio of TaN and Tb layers.
-4
10 HfTaON/HfO2 VD=1 V
Drain Current IDS (A/μm)
-5 W/L=320 μm/5 μm
10
-6
10
-7
VD=0.05 V
10
10
-8 SS: ~ 75 mV/dec.
10
-9
TaN
-10 TaN/Tb/TaN InM.
10 TaTbN co-sput.
-11
10
-0.4 0.0 0.4 0.8 1.2 1.6 2.0
Gate Voltage VGS (V)
Fig. 5.12 IDS-VGS characteristics of n-MOSFETs with TaN, TaN/Tb/TaN (InM), and
co-sputtered Ta0.9Tb0.1Ny gates on HfTaON high-κ dielectric.
The electron mobility in the three kinds of devices has also been evaluated and
shown in Fig. 5.14. It was observed that the electron mobility in TaN/Tb/TaN gated
n-MOSFET is a little bit lower than that in TaN and Ta0.9Tb0.1Ny gated n-MOSFETs.
149
This could be correlated with the higher Dit level in TaN/Tb/TaN gated devices,
measured by charge pumping method. The reason is still not very clear, but a
possible reason could be the stress effect associated with the InM process. Therefore
the annealing condition needs to be further optimized to minimize this problem.
1 TaTbN) TaN
co-sput.
0
Ln(-Ln(1-F))
-1 ~ 300 mV
~ 400 mV
-2
TaN/Tb/TaN
InM.
-3
-200 0 200 400 600 800
Vth_Lin (mV)
Fig. 5.13 Vth distribution of n-MOSFETs with TaN, TaN/Tb/TaN (InM) and co-
sputtered Ta0.9Tb0.1Ny gates on HfTaON/HfO2 dielectric. (W/L=320 μm / 5 μm)
Electron Mobility (cm / V⋅s)
400 HfTaON/HfO Universal

2
2
300
200
TaN
100 TaN/Tb/TaN InM.
TaTbN co-sput.
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Effective Field (MV/cm)
Fig. 5.14 Effective electron mobility in n-MOSFETs with TaN, TaN/Tb/TaN (InM)
and co-sputtered Ta0.9Tb0.1Ny metal gates on HfTaON dielectric, measured by split C-
V measurement. (W/L = 200 μm / 20 μm)
150
5.2.5 Dual Work Function Metal Gate Integration using InM
Dual metal gate integration using the proposed InM method is also
successfully demonstrated. TaN/Tb/TaN and TaN/Ti/HfN [13] laminate structures
were integrated on a single wafer for NMOS and PMOS, respectively, as a prototype
demonstration of the integration scheme depicted in Fig. 5.6. In this process, it is
important to note that the capping layer on top of the P-type metal gate should be
thick enough to prevent the N-type material in affecting the WF of the P-type metal
gate stack (Fig. 5.6). In the other words, the HfN layer in the TaN/Ti/HfN stack
should be thick enough to prevent Tb from diffusing down. It is observed in Fig.
5.15 (a) that HfN of 15 Å is not thick enough to block Tb from reaching the metal-
dielectric interface, as evidenced by the negatively VFB shift after 1000 oC RTA. To
eliminate this problem, we increases the HfN thickness to about 100 Å, and found
that the Tb layer no longer affect the EWF of the TaN/Ti/HfN stack in PMOS region,
as shown in Fig. 5.15 (b). Therefore HfN with thickness of 100 Å was used in our
process, though this may not be an optimal value.
Now we start to discuss the integration process. After gate dielectric growth,
TaN/Ti/HfN (~2nm/1nm/10nm) stack was first deposited on SiO2. HfN and Ti were
then selectively removed by DHF (1:200) solution in NMOS region, where the ultra-
thin TaN of ~2nm was used to protect the underlying gate dielectric in metal etching
process. In order to check whether the DHF process can cause any change to the
TaN buffer layer, root mean square (rms) roughness of the ultra-thin TaN film has
been examined by AFM measurement. As illustrated in Fig. 5.16, the rms roughness
of the TaN film shows no significant difference after the DHF process, indicating
TaN is robust enough to the DHF solution used in our process.
151
TaN/Ti/HfN gate stack

with and w/o Tb layer on top
6 w/o Tb
with Tb (2.5 nm)
4

2
HfN ~15 Å
2
o
after 1000 C RTA a)
0
10
w/o Tb
8 with Tb (2.5 nm)
6
HfN ~100 Å
4
2 o b)
after 1000 C RTA
0
-2 -1 0 1
Gate Voltage (V)
Fig. 5.15 C-V characteristics of TaN/Ti/HfN metal stack with and without Tb on top,
where the HfN thickness is (a) ~15 Å and (b) ~100 Å.
Fig. 5.16 AFM images of TaN (~ 2 nm) deposited on the bare-Si wafer before and
after wet etching in DHF (1:200) for 30 sec.
After removal of HfN and Ti layer, ultra-thin Tb and the TaN capping was
then deposited to form TaN/Tb/TaN in NMOS region, followed by gate patterning.
Fig. 5.17 illustrates the final structures of the N-type and P-type metal gate stacks
152
integrated on a same wafer. C-V and I-V characteristics were measured in the as-
deposited condition and no obvious difference observed between the capacitors with
TaN/Tb/TaN (N-type) and TaN/Ti/HfN (P-type) metal gate stacks, as shown in Fig.
5.18. This indicates clearly that the wet etching process by DHF does not cause any
change to the thin TaN buffer layer as well as the underlying dielectric.
Fig. 5.17 XTEM images of as-deposited dual metal gate stacks on a single wafer:
TaN/Tb/TaN (left) for NMOS and TaN/Ti/HfN (right) for PMOS on SiO2.
10
2
0
As-deposited 10
Gate leakage (A/cm )
8
SiO2 ~ 3.3 nm
-2
10
6
-4
4 TaN/Tb/TaN 10
TaN/Ti/HfN/Tb/TaN
2 -6
10
0 a) b)
2
-8
10
-2 -1 0 1 -4 -3 -2 -1 0
Gate Voltage (V)
Fig. 5.18 (a) C-V and (b) I-V characteristics of TaN/Tb/TaN (N-type) and
TaN/Ti/HfN (P-type) metal gate stacks on SiO2 in as-deposited condition.
153
Finally the devices were subjected to a RTA annealing at 1000 oC for 1 sec for
metal intermixing. This annealing also simulates the S/D dopant activation process
used in a gate-first CMOS fabrication. Fig. 5.19 shows the C-V measurements of the
TaN/Tb/TaN (NMOS) and TaN/Ti/HfN (PMOS) stacks before and after InM
annealing. A VFB difference for above 0.5 V can be achieved through InM, suggesting
that the work function of TaN buffer layer is modulated by intermixing with the
metals on top. The work function values of the two metal gate stacks are determined
to be 4.15 eV and 4.72 eV, respectively, from the VFB-EOT plots shown in Fig. 5.20.
Note that the work function of 4.15 eV is ideal for NMOS. But the WF value of 4.72
eV is still not optimal for PMOS and the investigation on other laminate combinations
[13] could be one of the possible solutions.
12
2
As-deposited
10 TaN/Tb/TaN
TaN/Ti/HfN
8
6 TaN/Ti/HfN
o
RTA@1000 C
4
2
TaN/Tb/TaN InM.
0 RTA@1000oC
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Gate Voltage (V)
Fig. 5.19 C-V characteristics of TaN/Tb/TaN (NMOS) and TaN/Ti/HfN (PMOS)

stacks on SiO2 before and after 1000 oC RTA for 1 sec.
154
0.0 o
after 1000 C spike anneal
-0.2 TaN/Ti/HfN
~ 4.72 eV
VFB (V) -0.4
-0.6
-0.8
~ 4.15 eV TaN/Tb/TaN InM.
-1.0
0 1 2 3 4 5 6 7 8 9
EOT (nm)
Fig. 5.20 Work Function extraction for TaN/Tb/TaN (NMOS) and TaN/Ti/HfN
(PMOS) metal stacks on SiO2 after metal intermixing process.
5.2.6 Summary
In summary, we demonstrate a CMOS compatible dual work function metal
gate integration process by using a high-temperature metal intermixing technique. In
this process, an ultra-thin TaN buffer layer is used to protect the gate dielectric during
metal etching process, and a subsequential high-temperature metal intermixing
process is performed to modify the work function of TaN for the requirement of
CMOS. Work function of 4.15 and 4.72 eV can be achieved by intermixing of
TaN/Tb/TaN (NMOS) and TaN/Ti/HfN (PMOS) metal stacks, respectively, after a
high-temperature annealing process which can be performed together with the S/D
activation process used in conventional CMOS fabrication. Successful Vth adjustment
for ~300 mV compared with TaN has also been demonstrated on HfTaON dielectric
using this metal intermixing technique.
155
5.3 A Gate-Last Dual Metal Gate Integration Process
Employing a Novel HfN Replacement Gate
5.3.1 Motivation
As we discussed in Chapter 2, dual metal gate electrodes with band-edge work
functions within ± 0.2 eV from the Ec or Ev of Si are required for the threshold voltage
considerations in plenary bulk-Si CMOS technology. Immense challenges have been
faced in identifying the potential metal gate candidates with band-edge work
functions and integrating them into CMOS process. To date, it is still quite difficult
to integrate dual metal gates with work function difference (ΔWF) larger than 0.8 eV
on high-κ dielectrics using a gate-first process due to the thermal stability and Fermi-
pinning considerations. On the other hand, the gate-last integration process shows
great advantages in achieving wide-range work function modulation, thanks to its low
thermal budget process flow. From this point of view, the gate-last integration
process still stands the chance to be a potential integration scheme for the dual metal
gate CMOS and is hence worthy to be studied [21]-[24].
The process flow for a conventional gate-last integration process is illustrated
in Fig. 5.5. In this process, the metal gate electrodes and high-κ dielectric are used to
replace the poly-Si dummy gate and underlying sacrificial SiO2, respectively, after the
formation of the S/D regions. Nevertheless, a high-temperature annealing of the high-
κ dielectric is desirable for better carrier mobility, less fixed charge and less C-V
hysteresis [24]-[26]. A high-temperature post-deposition-annealing (PDA) of high-κ
dielectric without metal gate capping, however, will cause an unwanted EOT increase
which is a serious concern for aggressive CMOS scaling [24]. As of the starting of
this work, there had been no solution to achieve high-quality high-κ dielectric with
sub-1-nm EOT and wide-range WF difference above 0.8 eV simultaneously.
156
In this work, a novel replacement gate process employing HfN as the dummy
gate electrode is demonstrated to integrate dual metal gate electrodes on HfO2
dielectric with sub-1-nm EOT. The HfN/HfO2 gate stack shows excellent thermal
stability during the S/D dopant activation annealing, allowing sub-1 nm EOT and
good dielectric characteristics to be achieved [27]. In addition, the high etching
selectivity of HfN with respect to HfO2 enables the high-quality HfO2 dielectric to be
left after the removal of dummy HfN electrode, with no degradation to the leakage or
TDDB reliability characteristics. Replacement of the dummy HfN gate by Ta and Ni
has also been demonstrated, with ΔWF of ~0.8 eV achieved. These results make the
HfN dummy gate process more attractive than the conventional poly-Si dummy gate
process.
5.3.2 Proposed Integration Flow and Device Fabrication
Fig. 5.21 Proposed replacement gate process using HfN as dummy gate: (a) CMOS
fabrication using TaN/HfN/HfO2 as the gate stacks; (b) high selective etching of TaN
and HfN by wet chemicals; (c) new metal gate deposition and CMP planarization; (d)
dual metal gate integration by repeating steps (b)-(c).
157
Fig. 5.21 illustrates a simplified process flow of the proposed HfN
replacement gate process. The feasibility of this process was demonstrated using
MOS capacitors. First, the HfN/HfO2 gate stack was fabricated [27]. Metal-organic
chemical vapor deposition (MOCVD) HfO2 with thickness of ~ 25 Å was deposited
on p-Si (100) substrate (NA=6×1015 cm-3) followed by PDA at 700 °C in N2 ambient.
A TaN/HfN stack (1000 Å/500 Å) was then deposited on HfO2 by dc magnetron
sputtering. After gate patterning, all the devices were subjected to RTA in N2 at
1000°C for 20 sec, which is adequate for S/D dopant activation. The TaN/HfN
dummy stack was then removed by standard cleaning-1 (SC-1) (NH4OH+H2O2+H2O)
solution at 80oC and DHF (1:100) solution at room temperature, respectively. It
should be noted that the SC-1 solution does not attack the HfN metal. Finally, Ta and
Ni were deposited on HfO2 gate dielectric as the new gate electrodes. For some
devices, HfN was re-deposited onto HfO2 to compare with those fresh devices to
check the impact of the DHF etching process to the dielectric properties. Backside Al
metallization and FGA were also performed to all the devices.
5.3.3 Results and Discussions
HfN has been demonstrated as a thermally robust metal gate electrode on
HfO2 [27], making it feasible to scale down the EOT of the HfN/HfO2 gate stack to
less than 1 nm after 1000 °C RTA annealing. Good transistor electrical characteristics
have also been achieved [28]. The work function of HfN, however, is close to the
mid-gap level of silicon. By replacing the HfN with Ta or Ni using the proposed
replacement gate process, dual metal gates with large work function difference can be
realized while maintaining the good properties of the HfO2 gate dielectric.
158
To study the feasibility of the proposed HfN dummy gate process, we first
investigate the selectivity of HfN with respect to HfO2 in DHF solution. Fig. 5.22
compares the etch rates of HfN and HfO2 in DHF (1:100) solution. The etch rate of
HfN is around 12 nm/min while that of HfO2 is almost negligible if 1000 °C PDA
treatment has been performed. This demonstrates the very high etching selectivity of
the HfN over HfO2. Fig. 5.23 examines the surface morphology of the HfO2 films
under three different conditions: as-deposited HfO2 film, HfO2 after 1000 °C RTA
anneal, and HfO2 in HfN/HfO2 stack with HfN removed by DHF solution after
1000 °C RTA anneal. The rms variation induced by the DHF etching process is only
about 0.05 nm, indicating that the HfN removal process has negligible physical
impact to the HfO2 film.
56 o
HfN/HfO2 stack (1000 C RTA)
54
HfN etching:
Thickness (nm)
~12 nm/min
7
6
o
HfO2: 1000 C RTA
5
0 5 10 15 20
Etch Time in DHF (minute)
Fig. 5.22 Etching properties of the HfN/HfO2 gate stack (open triangle symbol) and
the HfO2 (solid symbol) film after 1000 oC RTA process in diluted HF solution
(1:100). The etch rate of HfN is determined by surface profiler, and the remaining
HfO2 thickness is measured by ellipsometer.
159
Fig. 5.23 AFM images of HfO2 with different process history: as-deposited HfO2
film, HfO2 after 1000 °C RTA anneal, and HfO2 in HfN/HfO2 stack with HfN
removed by DHF solution after 1000 °C RTA anneal.
To further examine the potential influence of the wet etching process to the
electrical properties of HfO2, HfN metal gate is re-deposited onto HfO2 dielectric after
removing the HfN dummy gate. Fig. 5.24 compares the C-V and I-V characteristics of
the “control” HfN/HfO2 devices (1000 oC RTA, fresh device without HfN removal
process) and the “re-deposited” HfN/HfO2 devices (re-deposit HfN after the removal
of HfN dummy gate by DHF solution). The C-V characteristics were measured at
different frequencies (100 KHz and 1 MHz). No significant frequency dispersion
observed from both of the samples, indicating the series resistance from the gate and
substrate is not an issue to the C-V measurement. The EOT of the two gate stacks can
thus be extracted and a low EOT value of ~0.83 nm is obtained by fitting the C-V
measurement with simulated curves which takes quantum mechanical effect into
account. It is found that the EOT and the gate leakage of the two devices are almost
identical, as shown in Fig. 5.24. These results suggest that the ultra-thin HfO2 film is
not etched or damaged. The small VFB difference between the “control” and “re-
deposited” HfN/HfO2 devices could be due to the different thermal history employed
160
to the HfN metal gate in these two devices, or a possible plasma damage effect during
the re-deposition of HfN.
30
-1
25 10
Jg (A/cm )
Capacitance (fF/μm )
2
2
20 -5
10 Control
Re-Dep
15 -9
10
"Control" HfN/HfO2 -2.0 -1.5 -1.0 -0.5 0.0
Gate Voltage (V)
10 100 KHz
1 MHz Simulation:
EOT=0.83 nm
5 "Re-Dep" HfN/HfO
2
100 KHz
0 1 MHz
-1.5 -1.0 -0.5 0.0 0.5
Gate Voltage (V)
Fig. 5.24 C-V and I-V (inset) characteristics of the “control” HfN/HfO2 devices and
“re-deposited” HfN/HfO2 devices with HfO2 EOT~0.83 nm. The C-V curves were
measured at 100 kHz and 1 MHz on devices with an area of 50×50 μm2.
The above observations allow the HfN to be used as a dummy gate material on
HfO2 using the proposed replacement gate process. Replacement of HfN dummy gate
by Ta and Ni are then demonstrated for the applications of NMOS and PMOS,
respectively. Fig. 5.25 shows the high frequency C-V measurements for the Ta/HfO2
and Ni/HfO2 devices using the HfN replacement gate process. The C-V
measurements of all the devices fit well with the simulation curves, indicating
negligible changes of the HfO2/Si interface quality during the HfN removal process.
The C-V hysteresis for all the devices is less than 20 mV thanks to the good dielectric
quality of HfO2. The work function shifts attributed to Ta and Ni with respect to HfN
161
are -0.35 eV and +0.45 eV, respectively. The work function difference of 0.8 eV
between the two gate electrodes could be adequate for good device performance for
bulk-Si CMOS transistors. More importantly, the working devices with “re-
deposited” Ni on ultra-thin HfO2 (EOT ~0.9nm) are successfully demonstrated using
this novel replacement gate process (inset of Fig. 5.25), indicating good scalability of
this process for next generation MOSFETs. However, the “re-deposited” Ta-gated
device did not survive on such a thin HfO2, which is presumably associated with the
higher tunneling current through HfO2 due to the relatively lower WF of Ta than that
of Ni, and/or some possible plasma damage effects. Process optimization will be
needed in future study.

Capacitance (fF/ μm )
20 25 EOT~0.9 nm
2
20 Ni /HfO2
15
Capacitance (fF/μm )
ΔVFB=+0.45 V
2
10
15 5
0
HfN/HfO2
-1.5 -1.0 -0.5 0.0 0.5 1.0
Gate Voltage (V)
10
ΔVFB=0.8 V
5 Ta /HfO2
Ni /HfO2
0 HfN /HfO2(control)
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Gate Voltage (V)
Fig. 5.25 High frequency C-V curves of the HfN/HfO2 “control” devices and “re-
deposited” Ta/HfO2, Ni/HfO2 devices. The inset compares the C-V curves measured
from the “re-deposited” Ni/HfO2 devices with ultra-thin HfO2 (EOT~0.9 nm) and that
of a “control” HfN/HfO2 device. All the C-V curves were measured at 100 kHz on
devices with an area of 50×50 μm2.
162
(A/cm )
Poly-Si/SiO2
2
3
10
1 Bench Mark
1
10
FB
Jg @ -1V + V
-1
0 10
Ln(-Ln(1-F))
-3 (Control) HfN/HfO2
10 (Re-Dep) HfN Ta Ni
0.5 1.0 1.5 2.0
-1 EOT (nm)
HfN/HfO2 (Control)
-2 Re-Dep HfN/HfO2
Re-Dep Ta/HfO2
CCS Stress
Re-Dep Ni/HfO2
Gate Injection
-3 1 2 3 4
10 10 10 10
2
Charge to Breakdown (C/cm )
Fig. 5.26 Comparison of TDDB and gate leakage (inset) characteristics between the
“control” HfN/HfO2 devices and “re-deposited” HfN/HfO2, Ta/HfO2, Ni/HfO2
devices. For the TDDB study, CCS with a current density of ~8 A/cm2 was
performed on devices with an area of 100×100 μm2 at room temperature.
Finally, we compare the gate leakage and TDDB characteristics of the “re-
deposited” HfN/HfO2, Ta/HfO2 and Ni/HfO2 devices with that of the “control”
HfN/HfO2 stack, as depicted in Fig. 5.26. It is observed that the gate leakage of the
Ta/HfO2 (or Ni/HfO2) devices is slightly higher (or lower) than that of the HfN/HfO2
“control” devices (inset of Fig. 5.26). This can be attributed to the lower (or higher)
work function of Ta (or Ni) with respect to HfN. The TDDB characteristics are also
investigated by applying a constant current stress (CCS) to the samples with different
gate electrodes. The EOT of these devices is about 1.35 nm. No significant
degradation in TDDB characteristics was observed by using this HfN replacement
gate process, as shown in Fig. 5.26. More reliability study will be required to qualify
this process systematically.
163
It should be noted that other metal gate candidates, like Ta-Ru or Hf-Mo
alloys, can also be integrated using this HfN replacement gate process as potential
solutions for the dual metal gate CMOS technology towards sub-1-nm EOT regime.
5.3.4 Summary
In summary, we proposed a novel replacement gate process employing HfN as
the dummy gate material for the integration of dual metal gates on HfO2 gate
dielectric with sub-1-nm EOT. The excellent thermal stability of the HfN/HfO2 gate
stack and the high etch selectivity between HfN and HfO2 allows an ultra-thin, high-
quality, and damage-free HfO2 gate dielectric to be achievable. Replacing the HfN
dummy gate by Ta and Ni metal gates using the proposed process has been
successfully demonstrated, with large work function difference for about 0.8 eV
achieved.
5.4 Conclusion
In this Chapter, we demonstrated two integration schemes for dual metal gate
CMOS integration. The first one is a novel gate-first integration process using a high-
temperature metal intermixing technique. In this process, a TaN buffer layer is used
to avoid the gate dielectric being exposed during the metal etching process. This
addresses the etching damage concerns associated with the conventional direct-
etching integration scheme. The work function of the TaN buffer layer can be
modulated by a following high-temperature metal intermixing process, which is
compatible with the conventional gate-first CMOS process flow. By using this
integration scheme, dual work function of 4.15 and 4.72 eV have been achieved in
164
TaN/Tb/TaN (NMOS) and TaN/Ti/HfN (PMOS) metal stacks, respectively.
Successful Vth adjustment on HfTaON dielectric has also been demonstrated.
The second integration scheme proposed in this thesis is a gate-last
replacement gate process employing HfN as a novel dummy gate electrode, which
enables the high-quality HfO2 gate dielectric with sub-1 nm EOT and a wide work
function tuning range to be achieved simultaneously for bulk-Si CMOS applications.
High quality HfN/HfO2 stack with HfO2 EOT less than 1 nm can be achieved, due to
the good thermal stability of the HfN/HfO2 stack, and large work function difference
for about 0.8 eV can be realized by using Ta and Ni to replace the HfN dummy gate
electrode. The EOT, gate leakage, and TDDB characteristics of the ultra-thin HfO2
dielectric are observed not to be affected by the HfN dummy gate removal process.
These proposed novel integration schemes may provide some useful
discussions to address some of the major issues associated with the conventional
integration schemes, and are believed to make a contribution to the development of
the dual metal gate integration processes for the future CMOS technology.
165
References
1. Q. Lu, Y.-C. Yeo, P. Ranade, H. Takeuchi, T.-J. King, C. Hu, S. C. Song, H. F.

Luan, and D.-L. Kwong, “Dual-metal gate technology for deep-sub-micron
CMOS transistors,” in VLSI Tech. Dig., pp. 72-73, 2000.
2. S. B. Samavedam, L. B. La, J. Smith, S. Dakshina-Murthy, E. Luckowski, J.

Schaeffer, M. Zavala, R. Martin, V. Dhandapani, D. Triyoso, H. H. Tseng, P. J.
Tobin, D. C. Gilmer, C. Hobbs, W. J. Taylor, J. M. Grant, R. I. Hegde, J. Mogab,
C. Thomas, P. Abramowitz, M. Moosa, J. Conner, J. Jiang, V. Arunachalam, M.
Sadd, B.-Y. Nguyen, and B. White, “Dual-metal gate CMOS with HfO2 gate
dielectric,” in IEDM Tech. Dig., pp. 433-436, 2002.

electrodes on HfO2 gate dielectric,” in VLSI Tech. Dig., pp. 50-51, 2005.
4. M. M. Hussain, N. Moumen, J. Barnett, J. Saulters, D. Baker, M. Akbar and Z.

Zhang, “Metal wet etch process development for dual metal gate CMOS using
standard industry tools,” Proc. ECS 207th meeting, 629, Quebec, Canada, 2005.
5. S. C. Song, Z. B. Zhang, M. M. Hussain, C. Huffman, J. Barnett, S. H. Bae, H. J.

Li, P. Majhi, C. S. Park, B. S. Ju, H. K. Park, C. Y. Kang, R. Choi, P. Zeitzoff, H.
H. Tseng, B. H. Lee, and R. Jammy, “Highly manufacturable 45nm LSTP
CMOSFETs using novel dual high-k and dual metal gate CMOS integration,” in
VLSI Tech. Dig., pp. 16-17, 2006.
22, pp. 444-446, Sep. 2001.

CMOS,” in IEDM Tech. Dig., pp. 359-362, 2002.
8. B.-Y. Tsui, and C.-F. Huang, “Wide range work function modulation of binary
alloys for MOSFET application,” IEEE Electron Device Lett., vol. 24, pp. 153-
155, Mar. 2003.
9. T.-L. Li, C.-H. Hu, W.-L. Ho, H. C.-H. Wang, and C.-Y. Chang, “Continuous and
precise work function adjustment for integratable dual metal gate CMOS
technology using Hf–Mo binary alloys,” IEEE Trans. Electron Devices, vol. 52,
pp. 1172-1179, 2005.
10. T.-L. Li, W.-L. Ho, H.-B. Chen, H. C.-H. Wang, C.-Y. Chang, and C. Hu, “Novel
dual-metal gate technology using Mo-MoSix combination,” IEEE Trans. Electron
Devices, vol. 53, pp. 1420-1426, 2006.
166
11. A. Veloso, T. Hoffmann, A. Lauwers, S. Brus, J.-F. de Marneffe, S. Locorotondo,

C. Vrancken, T. Kauerauf, A. Shickova, B. Sijmus, H. Tigelaar, M. A. Pawlak,
H.Y. Yu, C. Demeurisse, S. Kubicek, C. Kerner, T. Chiarella, O. Richard, H.
Bender, M. Niwa, P. Absil, M. Jurczak, S. Biesemans and J. A. Kittl, “Dual work
function phase controlled Ni-FUSI CMOS (NiSi NMOS, Ni2Si or Ni31Si12 PMOS):
manufacturability, reliability & process window improvement by sacrificial SiGe
cap,” in VLSI Tech. Dig., pp. 96-97, 2006.
12. C. S. Park, B. J. Cho, A. Y. Du, N. Balasubramanian, and D.-L. Kwong, “A novel

approach for integration of dual metal gate process using ultra thin aluminum
nitride buffer layer,” in VLSI Tech. Dig., pp. 149-150, 2003.
13. W. P. Bai, S. H. Bae, H. C. Wen, S. Mathew, L. K. Bera, N. Balasubramanian, N.

Yamada, M. F. Li, and D.-L. Kwong, “Three-layer laminated metal gate
electrodes with tunable work functions for CMOS applications,” IEEE Electron
Device Lett., vol. 26, pp. 231-233, Apr. 2005.
14. X. Yu, C. Zhu, M. Yu, M. F. Li, A. Chin, C. H. Tung, D. Gui, and D.-L. Kwong,
“Advanced MOSFETs using HfTaON/SiO2 gate dielectric and TaN metal gate
with excellent performances for low standby power application,” in IEDM Tech.
Dig., 2005, pp. 31-34.
15. H. J. Lee, T. Cagin, W. L. Johnson, W. A. Goddard III, “Criteria for formation of

metallic glasses: the role of atomic size ratio,” J. Chem. Phys., vol 119, pp. 9858-
9870, Nov. 2003.
16. J. K. Schaeffer, S.B. Samavedam, D.C. Gilmer, V. Dhandapani, P.J. Tobin, J.

Mogab, B.-Y. Nguyen, B.E. White, Jr., S. Dakshina-Murthy, R.S. Rai, Z.-X. Jiang,
R. Martin, M.V. Raymond, M. Zavala, L.B. La, J.A. Smith, R. Garcia, D. Roan,
M. Kottke, and R.B. Gregory, “Physical and electrical properties of metal gate
electrodes on HfO2 gate dielectrics,” J. Vac. Sci. Technol. B, vol. 21, pp. 11-17,
Jan./Feb. 2003.
17. C. D. Gelatt, Jr. and H. Ehrenreich, “Charge transfer alloys: AgAu,” Phys. Rev. B,
vol. 10, pp. 398-415, Jul. 1974.
18. O. V. Lounasmaa and P. R. Roach, “Specific heat of terbium metal between 0.37
and 4.2 oK,” Phys. Rev., vol. 128, pp. 622-626, Oct. 1962.

Dig., pp. 303-306, 2004.
20. H. Kim, P. C. Mclntyre, C. O. Chui, K. C. Saraswat, and S, Stemmer,

“Engineering chemically abrupt high-k metal oxide/silicon interfaces using an
oxygen-gettering metal overlayer,” J. Appl. Phys., vol. 96, pp. 3467-3472, 2004.
21. A. Chatterjee, R.A. Chapman, K. Joyner, M. Otobe, S. Hattangady, M. Bevan,

G.A. Brown, H. Yang, Q. He, D. Rogers, S.J. Fang, R. Kraft, A.L.P. Rotondaro,
167
M. Terry, K. Brennan, S.-W. Aur, J.C. Hu, H.-L. Tsai, P. Jones, G. Wilk, M. Aoki,
M. Rodder, and I.-C. Chen, “CMOS metal replacement gate transistors using
tantalum pentoxide gate insulator,” in IEDM Tech. Dig., pp. 777-780, 1998.
22. A. Yagishita, T. Saito, K. Nakajima, S. Inumiya, Y. Akasaka, Y. Ozawa, K. Hieda,

Y. Tsunashima, K. Suguro, T. Arikado, and K. Okumura, “High performance
damascene metal gate MOSFET’s for 0.1 μm regime,” IEEE Trans. Electron
Devices, vol. 47, pp. 1028-1034, May 2000.
23. J. Pan, C. Woo, C.-Y. Yang, U. Bhandary, S. Guggilla, N. Krishna, H. Chung, A.

Hui, B. Yu, Q. Xiang, and M.-R. Lin, “Replacement metal-gate NMOSFETs with
ALD TaN/EP-Cu, PVD Ta, and PVD TaN electrode,” IEEE Electron Device Lett.,
vol. 24, pp. 304-305, May 2003.
24. B. Guillaumot, X. Garros, F. Lime, K. Oshima, B. Tavel, J. Chroboczek, P.

Masson, R. Truche, A. M. Papon, F. Martin, J. F. Damlencourt, S. Maitrejean, M.
Rivoire, C. Leroux, S. Cristoloveanu, G. Ghibaudo, J. L. Autran, T. Skotnicki, and
S. Deleonibus, “75nm damascene metal gate and high-k integration for advanced
CMOS devices,” in IEDM Tech. Dig., pp. 355-358, 2002.
25. Y. Kim, G. Gebara, M. Freiler, J. Barnett, D. Riley, J. Chen, K. Torres, J. Lim, B.

Foran, F. Shaapur, A. Agarwal, P. Lysaght, G. A. Brown, C. Young, S. Borthakur,
H.-J. Li, B. Nguyen, P. Zeitzoff, G. Bersuker, D. Derro, R. Bergmann, R. W.
Murto, A. Hou, H. R. Huff, E. Shero, C. Pomarede, M. Givens, M. Mazanec, and
C. Werkhoven, “Conventional n-channel MOSFET devices using single layer
HfO2 and ZrO2 as high-k gate dielectrics with polysilicon gate electrode,” in
IEDM Tech. Dig., pp. 455-458, 2001.
26. G. D. Wilk, M. L. Green, M.-Y. Ho, B. W. Busch, T. W. Sorsch, F. P. Klemens, B.

Brijs, R. B. van Dover, A. Kornblit, T. Gustafsson, E. Garfunkel, S. Hillenius, D.
Monroe, P. Kalavade, and J. M. Hergenrother, “Improved film growth and
flatband voltage control of ALD HfO2 and Hf-Al-O with n+ poly-Si gates using
chemical oxides and optimized post-annealing,” in VLSI Tech. Dig., pp. 88-89,
2002.

Dig., pp. 99-103, 2003.
28. J. F. Kang, H. Y. Yu, C. Ren, M.-F. Li, D. S. H. Chan, X. Y. Liu, and D.-L.
Kwong, “Ultrathin HfO2 (EOT<0.75 nm) gate stack with TaN/HfN electrodes
fabricated using a high-temperature process,” Electrochemical and Solid-State
Lett., vol. 8, G311-G313, 2005.
168
Ch 6: Conclusion
Chapter 6
Conclusion
6.1 Summary
This work has sought to address some of the most pressing issues in advanced
metal gate technology. As the time of writing this dissertation, the 65 nm CMOS
technology is being introduced into production, and the 45 nm CMOS technology is
under development at most leading semiconductor manufacturers. Concurrently, the
whole semiconductor community is initiating the research and development efforts for
the 32 nm technology node and beyond, where it is likely that some of the materials
and process integration challenges discussed herein must be addressed. The task for
finding a replacement material for poly-silicon gate electrode is by no means trivial.
Numerous problems related to the material selection and the process integration of
metal gates (deposition, lithography, etching, etc.) need to be addressed before a
suitable metal gate solution can be transferred to manufacturing.
6.1.1 Understanding the Metal-Dielectric Interface
The metal-dielectric interface is important since it directly affects the effective
work function of metal gates. Although the metal-induced gate states and the related
interface dipole theory has been pretty successful in describing the dependence of
169
Ch 6: Conclusion
metal effective work function on underlying dielectric, the variation of metal gate
work function during the thermal annealing process used in CMOS fabrication still
need to be understood and addressed with scientific and technological importance.
In Chapter 3, the dependence of metal gate work function on the annealing
temperature was investigated on different dielectrics. The extrinsic states created at
the metal-dielectric interface was identified to be the major factor responsible for the
instability of metal gate effective work function during the high-temperature
annealing process. The creation of extrinsic states appears to be thermodynamically
driven and becomes more pronounced when the annealing temperature is higher,
making the Fermi-level pinning effect to be more serious after RTA at higher
temperature. The chemical bond configurations at the metal-dielectric interface could
be correlated with the creation of extrinsic states. In general, the Hf-Si bond tends to
create extrinsic states upon annealing while Hf-Hf or Si-Si bonds would be less
pronounced. A model considering the impact of extrinsic states has also been
proposed to qualitatively explain the dependence of metal effective work function on
annealing process. This understanding on the metal-dielectric interface could be
useful for work function tuning and interface engineering of the metal gate electrodes
in future MOSFETs.
6.1.2 Lanthanide-Incorporated Metal Nitride Gate Electrodes
Identifying the proper metal gate candidates for CMOS application is a key
challenge for the implementation of metal gates in 32 nm technology node and
beyond. Many issues need to be considered systematically in developing the
appropriate metal gate materials. One of the most challenging issues for the
development of a N-type metal gate candidate is the difficulty to achieve good
170
Ch 6: Conclusion
thermal stability required for gate-first process and a low work function suitable for
NMOS applications simultaneously.
Chapter 4 discussed a possible solution of the N-type metal gate materials for
the applications in a gate-first CMOS process. For the first time, we demonstrated
that lanthanide elements can be very useful in modulating the work function of metal
gates, and this provides a new way for metal gate work function engineering. In this
work, lanthanide elements with very low work function are combined with metal
nitride materials to get a best trade-off between thermal stability and low work
function. The physical and electrical properties of lanthanide-incorporated metal
nitrides are systematically studied. By varying the lanthanide concentration in
lanthanide-MNx, the work function of lanthanide-MNx can be tuned continuously and
a work function value of 4.2~4.3 eV can be obtained even after a 1000 oC RTA
treatment, promising for NMOS devices using a gate-first CMOS process flow. Our
study also indicates that the nitrogen concentration is an important parameter in
influencing the properties of lanthanide-MNx, especially its thermal stability. Hence
the nitrogen content needs to be carefully engineered in process. Moreover, we also
demonstrated the compatibility of this method with high-κ dielectrics. It is believed
that these results will be of considerable practical value for the development of metal
gate technology.
6.1.3 Process Integration of Dual Metal Gates
Another effort of this thesis is a discussion on the dual metal gate integration
issues for CMOS applications. Introducing novel materials into the well-established
CMOS platform is always accompanied by innovations in process integration
technology. However, to date, there has been no integration scheme which addresses
171
Ch 6: Conclusion
all the concerns in dual metal gate integration. The goal of this thesis has been to
identify the issues and limitations in the existing dual metal gate integration schemes
and propose novel integration schemes to address these potential problems.
Two dual metal gate integration schemes have been studied in Chapter 5. The
first one is aimed at addressing the etching damage issues associated with the
conventional direct-etching method used for dual metal gate integration. A novel
gate-first dual metal gate CMOS integration process by using a high-temperature
metal intermixing process was demonstrated for the first time. In this process, a TaN
buffer layer is used to protect the gate dielectric and avoid the exposure of gate
dielectric during the selective metal etching process. The work function of the TaN
buffer layer can be modulated by a subsequent high-temperature metal intermixing
technique, which is compatible with the conventional gate-first CMOS process flow.
By using this integration scheme, dual work function of 4.15 and 4.72 eV has been
achieved in TaN/Tb/TaN (NMOS) and TaN/Ti/HfN (PMOS) metal stacks,
respectively. Successful Vth adjustment on high-κ HfTaON dielectric has also been
demonstrated using this method.
In the second part of Chapter 5, the feasibility of a gate-last replacement gate
process employing HfN as a novel dummy gate electrode has also been investigated.
The advantage of this process is that the high-quality HfO2 gate dielectric with EOT
scalability down to sub-1 nm regime can be integrated with dual metal gates with a
wide range work function difference for above 0.8 eV, which is very attractive for
future bulk-Si CMOS devices. Due to the good thermal stability of HfN/HfO2 stack,
high-quality HfO2 dielectric with EOT less than 1 nm can be achieved using a gate-
first process. The dummy HfN gate can then be selectively removed from HfO2 so
that other metal gate candidates with suitable WF for bulk-Si CMOS can be integrated
172
Ch 6: Conclusion
on the high-quality HfO2 dielectric. As a prototype demonstration, Ta and Ni are used
to replace HfN and large work function difference for about 0.8 eV has been
demonstrated. The EOT, gate leakage, and TDDB characteristics of the ultra-thin
HfO2 dielectric are also not affected by the HfN dummy gate removal process,
suggesting the potential of this process for bulk-Si CMOS applications.
3.2 Suggestions for Future Work
The work in this thesis has been very exploratory. More detailed investigation
and more rigorous characterization will be necessary to further optimize the processs
described in this thesis. Suggestions for future work will be directly or indirectly
related to the concerns described earlier in this thesis.
In Chapter 3, a metal-dielectric interface model taking the role of extrinsic
states into account has been proposed to explain the work function thermal instability.
However, in order to precisely engineer the effective work function of metal gates in
process, the source of the extrinsic defects need to be further identified and the
properties of these extrinsic states, such as their energy levels, need to be studied by
physical analysis or atomistic simulation. On the other hand, approaches to
minimizing the impact of the extrinsic defects and improving the thermal stability of
metal gate work functions during process would be an important question for
discussion in future work.
For the development of lanthanide-MNx metal gates, the etching and cleaning
issues need to be well optimized in order to examine the feasibility of these materials
in short-channel transistor fabrication. Moreover, the co-optimization of these metal
173
Ch 6: Conclusion
gates with state-of-the-art high-κ candidates, like HfSiON or HfSiO, also needs to be
emphasized in future work. The mobility and charge-trapping characteristics in
MOSFETs with lanthanide-MNx metal gates should be investigated. In addition, the
possible interface interaction between lanthanide elements and the underlying
dielectrics needs to be further characterized and its impact on device reliability should
be evaluated.
In addition, developing a suitable P-type metal gate material with work
function higher than 5.0 eV would be an urgent task for the dual metal gate CMOS
technology. It has been reported that the effective work function of many P-type
metal gate candidates, such as Pt, Ru, and Re, on high-κ HfO2 will be affected by the
oxygen vacancy induced dipoles, rendering the resulting threshold voltage too high
for p-MOSFET [1]. This problem could possibly be addressed by engineering the
high-κ dielectric to reduce the oxygen vacancies near the metal-dielectric interface [2].
Therefore, the co-optimization of the metal/high-κ stack should be emphasized in
future research to get optimal device performance.
Moreover, for the proposed dual metal gate integration scheme using high-
temperature metal intermixing technique, some interesting phenomenon are still
worthy to be studied in future work. The dependence of the work function
modulation range on a number of variables, including the buffer layer thickness, the
buffer layer composition, the selection of metal materials, and the annealing
conditions, should be systematically investigated in more detail. A model to depict
the work function tunability of this intermixing process should be developed to guide
174
Ch 6: Conclusion
the selection of metal materials. The performance of the P-type metal stacks also
needs to be further optimized.
Finally, the applications of metal gate electrodes in multi-gate or other non-
planar CMOS devices should also be studied. FUSI gate process could be a potential
solution for these kinds of applications. Process integration of FUSI metal gates in
these 3-D structures could be challenging due to the difficulties in controlling the
phase/dopant precisely in 3-D structures, and hence it needs to be carefully developed,
particularly for devices in 32-nm technology node and beyond where the gate length
will be less than 15 nm.
175
Ch 6: Conclusion
References

Dig., pp. 230-231, 2005.
2. X. P. Wang, C. Shen, M.-F. Li, H. Y.Yu, Y. Sun, Y. P. Feng, A. Lim, W.S.

Hwang, A. Chin, Y. C. Yeo, G. Q. Lo and D. L. Kwong, “Dual metal gates with
band-edge work functions on novel HfLaO High-k gate dielectric,” in VLSI Tech.
Dig., pp. 12-13, 2006.
176
APPENDIX
Publications Related to This Thesis
Referred Journal Publications

Lett., vol. 25, pp. 123-125, Mar. 2004.
2. C. Ren, H. Y. Yu, J. F. Kang, X. P. Wang, H. H. H. Ma, Yee-Chia Yeo, D. S. H.

Chan, M.-F. Li, and D.-L. Kwong, “A dual-metal gate integration process for
CMOS with sub-1 nm EOT HfO2 by using HfN replacement gate,” IEEE
Electron Device Lett., vol. 25, pp. 580-582, Aug. 2004.
3. C. Ren, H. Y. Yu, X. P. Wang, H. H. H. Ma, D. S. H. Chan, M.-F. Li, Yee-Chia

Yeo, C. H. Tung, N. Balasubramanian, A. C. H. Huan, J. S. Pan, and D.-L.
Kwong, “Thermally robust TaTbxN metal gate electrode for n-MOSFETs
applications,” IEEE Electron Device Lett., vol. 26, pp. 75-77, Feb. 2005.
4. C. Ren, D. S. H. Chan, X. P. Wang, Faizhal B. B., M.-F. Li, and Y.-C. Yeo, A. D.
Trigg, A. Agarwal, N. Balasubramanian, J. S. Pan, P. C. Lim, A. C. H. Huan, and
D.-L. Kwong, “Physical and electrical properties of lanthanide-incorporated
tantalum nitride for n-channel metal-oxide-semiconductor field-effect-
transistors,” Appl. Phys. Lett., vol. 87, 073506, Aug. 2005.
5. C. Ren, Faizhal B. B., D. S. H. Chan, M.-F. Li, Y.-C. Yeo, A. D. Trigg, N.

Balasubramanian, and D.-L. Kwong, “Work function tuning of metal nitride
electrodes for advanced CMOS devices,” Thin Solid Films, vol. 504, pp. 174-
177, May 2006.
6. C. Ren, D. S. H. Chan, M.-F. Li, W. Y. Loh, S. Balakumar, C. H. Tung, N.

Balasubramanian, and D.-L. Kwong, “Work function tuning and material
characteristics of lanthanide -incorporated metal nitride gate electrodes for NMOS
device applications,” IEEE Trans. Electron Devices, vol. 53, pp. 1877-1884,
Aug. 2006.
7. C. Ren, D. S. H. Chan, W. Y. Loh, S. Balakumar, A. Y. Du, C. H. Tung, G. Q.

Lo, R. Kumar, N. Balasubramanian, and D.-L. Kwong, “Work function tuning of
TaN by high-temperature metal intermixing technique for gate-first CMOS
process,” to be published in IEEE Electron Device Lett.

May 2004.
177
Referred Conference Publications
1. C. Ren, J. F. Kang, H. Yang, D. D. Han, X. Y. Liu, R. Q. Han, D. S. H. Chan,

and D.-L. Kwong, “Carrier transport and anomalous C-V characteristics in PVD
AL2O3 gate dielectrics,” in Int. Conf. on Materials for Advanced Technologies
(ICMAT-2003), Dec. 2003, Singapore.
2. C. Ren, D. S. H. Chan, Fazhal B. B., H. M.-F. Li , Y.-C. Yeo, A. D. Trigg, A.

Agarwal, N. Balasubramanian, J. S. Pan, P. C. Lim, and D.-L. Kwong,
“Lanthanide-incorporated metal nitrides with tunable work function and good
thermal stability for NMOS devices,” in Symposium on VLSI Technology
(VLSI-2005), pp. 42-43, Jun. 2005, Kyoto, Japan.
3. C. Ren, Faizhal B. B., D. S. H. Chan, M.-F. Li, Y.-C. Yeo, A. D. Trigg, N.

Balasubramanian, and D.-L. Kwong, “Work function tuning of metal nitride
electrodes for advanced CMOS devices,” in Int. Conf. on Materials for
Advanced Technologies (ICMAT-2005), July. 2005, Singapore.
4. C. Ren, D. S. H. Chan, W. Y. Loh, J. W. Peng, S. Balakumar, Y. Jiang, C. H.

Tung, A. Y. Du, G. Q. Lo, R. Kumar, N. Balasubramanian, and D.-L. Kwong,
“CMOS compatible dual metal gate integration with successful Vth adjustment on
high-k HfTaON by high-temperature metal intermixing,” accepted by 36th
European Solid-State Device Research Conference (ESSDERC-2006), Sep.
2006, Montreux, Switzerland.
5. H.Y. Yu, C. Ren, J. F. Kang, Y.-C. Yeo, D. S. H. Chan, M.-F. Li, and D.-L.
Kwong, “Thermal stability of metal gate work functions,” in Int. Conference on
Solid State Devices & Materials (SSDM-2004), pp. 712-713, Sep. 2004, Tokyo,
Japan.
6. J. F. Kang, C. Ren, H. Y. Yu, X. P. Wang, M-F. Li, D. S. H. Chan, Yee-Chia

Yeo, Y. Y. Wang, and D.-L. Kwong, “A novel dual-metal gate integration process
for sub-1 nm EOT HfO2 CMOS devices,” in Int. Conference on Solid State
Devices & Materials (SSDM-2004), pp. 198-199, Sep. 2004, Tokyo, Japan.
178

Work Function and Process Integration

Uploaded by

Copyright:

Available Formats

Work Function and Process Integration

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Work Function and Process Integration

Uploaded by

Copyright:

Available Formats

WORK FUNCTION AND PROCESS INTEGRATION

ISSUES OF METAL GATE MATERIALS

NATIONAL UNIVERSITY OF SINGAPORE

ISSUES OF METAL GATE MATERIALS

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

First of all, I would like to express my sincere thanks to my advisors, Prof.

guidance, encouragement, knowledge, freedom and all kinds of support during my

technology, which has helped me to avoid many detours in my research work. I do

all the opportunities provided in developing my potential and personality, especially

Chan and Prof. Kwong always.

suggestions and inspirational discussions which had been indispensable for my

research work. My special acknowledgement goes to Dr. Yu Hongyu, who is my

A significant part of my research was performed in Institute of

Microelectronics (IME), Singapore. Many of my thanks also go to the managers and

like to appreciate Dr. Balasubramanian Narayanan, Dr. Lo Guo-Qiang, Dr. Rakesh

the course of my doctoral research.

their love and enduring supports.

Table of Contents ........................................................................................................ iii

Summary .................................................................................................................... viii

List of Tables ................................................................................................................. x

List of Figures ............................................................................................................... xi

List of Symbols ........................................................................................................ xviii

List of Abbreviations ................................................................................................... xx

Chapter 1. Introduction ...................................................................................... 1

1.1. Overview ............................................................................................................ 1

1.2. MOSFET Scaling: Challenges and Opportunities ............................................. 2

1.2.1. Leakages in Deep-Submicrometer MOSFET ............................................. 4

1.2.2. Vertical Scaling of MOSFET Gate Stack ................................................... 6

1.2.3. Innovations in Device Structures .............................................................. 10

1.2.4. Mobility Enhancement for Performance Gain .......................................... 11

1.3. Summary .......................................................................................................... 13

Chapter 2. Developments in Metal Gate Materials for CMOS

2.1. Limitations of Poly-Si Electrode ..................................................................... 22

2.1.1. Poly-Si Depletion Effect ........................................................................... 23

2.1.3. Gate Electrode Resistivity......................................................................... 25

2.1.4. Compatibility with High-κ Dielectrics...................................................... 27

2.2. Post Polysilicon Era: Metal Gate Technology ................................................. 30

2.2.1. Historical Perspective of Metal Gate Electrodes ...................................... 30

2.2.2. Considerations for Metal Gate Candidates ............................................... 31

2.2.2.1. Work Function Requirement.............................................................. 31

2.2.2.2. Thermal Stability Considerations ...................................................... 33

2.2.2.3. Process Integration Issues .................................................................. 34

2.2.2.4. Co-optimization of Metal Gate/High-κ Gate Stack ........................... 36

2.2.3. Research Status of Metal Gate Technology .............................................. 37

2.2.3.1. Direct Metal Gates ............................................................................. 38

2.2.3.2. Binary Metal Alloys ........................................................................... 40

2.2.3.3. Fully-Silicided (FUSI) Metal Gates ................................................... 40

2.3. Challenges in Metal Gate Technology ............................................................. 43

2.3.1. Understanding of the Metal-Dielectric Interface ...................................... 43

2.3.2. Developing Appropriate Metal Gate Materials......................................... 44

2.3.3. Dual Metal Gate Integration Issues........................................................... 44

2.4. Research Scope and Major Achievements in this thesis .................................. 45

Chapter 3. The Metal-Dielectric Interface and Its Impact on the

Effective Work Function of Metal Gates...................................................... 56

3.1. Introduction ...................................................................................................... 56

3.2. Theoretical Backgrounds ................................................................................. 58

3.2.2. Definition of Effective Work Function ..................................................... 58