155
OPTIMIZATION AND CHARACTERIZATION OF 130 NM CMOS TRANSISTOR DESIGN USING TCAD SIMULATION HANI NOORASHIQIN BINTI ABD. MAJID FACULTY OF SCIENCE UNIVERSITY OF MALAYA KUALA LUMPUR 2007

optimization and characterization of 130 nm cmos transistor design

Embed Size (px)

Citation preview

Page 1: optimization and characterization of 130 nm cmos transistor design

OPTIMIZATION AND CHARACTERIZATION OF 130 NM

CMOS TRANSISTOR DESIGN USING TCAD SIMULATION

HANI NOORASHIQIN BINTI ABD. MAJID

FACULTY OF SCIENCE

UNIVERSITY OF MALAYA

KUALA LUMPUR

2007

Page 2: optimization and characterization of 130 nm cmos transistor design

OPTIMIZATION AND CHARACTERIZATION OF 130 NM

CMOS TRANSISTOR DESIGN USING TCAD

SIMULATION

HANI NOORASHIQIN BINTI ABD. MAJID

DISSERTATION SUBMITTED IN FULFILMENT

OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

FACULTY OF SCIENCE

UNIVERSITY OF MALAYA

KUALA LUMPUR

2007

Page 3: optimization and characterization of 130 nm cmos transistor design

ABSTRACT

Microelectronic device manufacture encompasses the fabrication, testing and simulation,

of structures utilized in a variety of application specific integrated circuits. The

fabrication of transistors, traditionally from elemental silicon wafer is conducted in a

clean room environment utilizing various types of equipment. Process simulation is a

critical element in facilitating the optimization of fabrication stages, confirming test

results and theoretical models and providing physical insight into structural operation.

TSUPREM-4 and MEDICI are two simulators that had been calibrated and used in this

work. As device sizes are scaled down, novel structures are proposed such as halo

(pocket), lightly doped drain (LDD) and source/drain process. In order to characterize

and optimize Silterra 130 nm CMOS transistor design, TCAD simulation is done in this

work and the results is compared with measured data and was verified to be in the

Silterra specification range. Various doses, implant energies and tilt angle is used to

upgrade the transistor performance, reduce the short channel effect and increase the

transistor lifetime. Halo implant, lightly doped drain (LDD) implant and source/drain

(S/D) implant is the three drain engineering structure to be characterized in this work. It

is found that at certain dose, implant energy and tilt angle, the best saturation drain

current, Idsat and threshold voltage, Vt that can match the specification target for NMOS

and PMOS transistor can be achieved. It is also proved in this work that by using

process and device simulator, the correct process can be predicted and the physical

insight of the structure such as current flow lines and doping concentrations can be

further analyzed.

Page 4: optimization and characterization of 130 nm cmos transistor design

ABSTRAK

Pengeluaran peranti mikroelektronik merangkumi pembuatan, menguji dan simulasi,

struktur-struktur yang digunakan dalam pelbagai aplikasi khusus litar-litar bersepadu.

Pembuatan transistor, terdiri daripada wafer silikon asas dikendalikan dalam satu

persekitaran tempat bersih dengan menggunakan pelbagai jenis peralatan. Proses

simulasi adalah satu unsur penting yang membantu dalam mengoptimumkan peringkat-

peringkat pembuatan, mengesahkan hasil percubaan dan model-model teori dan

menyediakan fahaman secara fizikal ke atas operasi struktur. TSUPREM-4 dan

MEDICI adalah dua jenis simulator yang telah ditentukurkan dan digunakan dalam

tugas ini. Setelah saiz-saiz peranti telah diturunkan, struktur-struktur novel telah

dicadangkan seperti proses implant halo ‘halo (pocket)’, ‘lightly doped drain (LDD)’

dan proses ‘source/drain’. Dalam peringkat untuk mencirikan dan mengoptimumkan

rekabentuk transistor CMOS Silterra 130 nm, simulasi TCAD telah dijalankan dalam

projek ini dan hasil-hasilnya telah dibandingkan dengan data yang telah diukur dan

telah disahkan ia berada di dalam julat spesifikasi Silterra. Pelbagai dos, tenaga

mengimplan dan sudut mengimplan digunakan untuk menaiktaraf prestasi transistor,

mengurangkan kesan saluran pendek dan menambah jangka hayat transistor. Implan

halo, ‘lightly doped drain (LDD)’ dan proses ‘source/drain’ adalah tiga jenis struktur

kejuruteraan parit yang akan dikaji dalam kerja ini. Didapati, dengan dos tertentu,

tenaga implan dan sudut condong implan, arus parit tepu, Idsat dan voltan ambang, Vt

yang terbaik dapat menepati sasaran spesifikasi bagi transistor NMOS dan PMOS.

Dalam kerja ini ia dapat dibuktikan bahawa dengan menggunakan proses dan alat

simulator, proses yang menepati spesifikasi dapat diramal dan struktur dalaman seperti

aliran arus dan kepekatan pengedopan seterusnya dapat dianalisa.

Page 5: optimization and characterization of 130 nm cmos transistor design

ACKNOWLEDGEMENT

First of all, I would like to acknowledge my supervisor, Prof Dr. Muhamad Rasat

Muhamad. He has been a great mentor to me with his management skills and

enthusiasm. He has also been an exceptional role model of life. He has done his utmost

to help relieve concerns that his advisees have.

I also would like to thank my co-supervisor, Mr Albert Victor Kordesch, Senior

Manager of Device Modeling Department, Silterra. Throughout my association with

him, he gave me strong motivation in doing my research. I consider him to be a moving

library that has the answers to all my queries. Besides that, his encouragements also

make me feel more comfortable and confidence while carrying out my research in

Silterra.

I am indebted to my second co-supervisor, Mr Binod Kumar Gopalakrishnan, ex-Senior

Engineer of Process Integration Department, Silterra. With all his advice and support I

am able to finish my thesis with a success results.

My special gratitude to Mr Chew Soon Aik, ex-Engineer, Silterra. Being my mentor is

the best thing that had ever happen to me. I would like to thank him a lot for his

guidance and constructive comments in various step of this project. His enthusiasm and

support have made this project a very rewarding, enriching and interesting one. His

useful ideas, advices are a memorable thing. He is also the one who encourage me and

support me to finish this thesis in only one year.

I wish to thank the Silterra (M) Sdn Bhd, pioneer of Malaysia semiconductor industry

for the support of this work through postgraduate internship program. Silterra provides

Page 6: optimization and characterization of 130 nm cmos transistor design

every technical support including financial, equipment, research environment and

technical advice for my research work. I would like to acknowledge the Device

Modeling Department team, (Norliza, Izahan Syemylona, Norhafizah, Wan Rosmaria,

Ayu Ismail, Shahrul Amran, Mohd. Fahmi) and Mrs. SL Lee for her advice on process

and procedure. Mrs. Nor Asmah Zainal Abidin and Mr. Abdullah Lin for the non-

technical support in Silterra. Many thanks to Silterra management for their support on

the managing of the postgraduate internship program.

Last but not least, I express my deepest love and appreciation to my parents and Hailmy

Rizuan for providing me their continuous inspiration and love that keep motivating me

to achieve my goal.

Page 7: optimization and characterization of 130 nm cmos transistor design

TABLE OF CONTENTS

Page

Abstract i

Abstrak ii

Acknowledgement iii

Table of Contents v

List of Tables viii

List of Figures ix

Chapter I INTRODUCTION

1.1 Background 1

1.2 Deep Sub-micron CMOS Process Technology 3

1.3 Overview of Process and Device Simulation 5

1.4 Motivation of this work 6

1.5 Objectives 8

1.6 Thesis Overview 8

Chapter 2 BACKGROUND AND LITERATURE REVIEW

2.1 Introduction 10

2.2 MOSFET Basic Characteristics 10

2.2.1 CMOS 11

2.3 CMOS Electrical Characteristics 13

2.3.1 Threshold Voltage, Vt 13

2.3.2 I-V Characteristics 16

2.4 CMOS Process 18

2.5 MOSFET Ion Implantation 20

2.5.1 Punchthrough Stop (Halo) Implants 21

2.5.2 Source/Drain Extension Implant or Lightly Doped Drain (LDD) 23

2.5.3 Source/Drain Implants 25

2.6 MOSFET Scaling Effects 26

2.6.1 Short Channel Effects 27

2.6.2 On-State Saturation Current 30

Page 8: optimization and characterization of 130 nm cmos transistor design

2.6.3 Punchthrough Effect 31

2.6.4 Hot Carrier Effect 32

2.7 Methods Attempted by Other Researchers 34

2.8 Summary 37

Chapter 3 CALIBRATION AND VALIDATION OF TSUPREM-4

AND MEDICI

3.1 Introduction 38

3.2 TSUPREM-4 38

3.3 TAURUS-MEDICI 39

3.4 Calibration of 130 nm MOSFET 40

3.5 Validation of TSUPREM-4 and MEDICI 49

3.6 Summary 51

Chapter 4 RESULTS

4.1 Investigation of Halo Implant in PMOS and NMOS Device 53

4.1.1 Introduction 53

4.1.2 Simulation Procedure 53

4.1.3 Simulation Results 54

4.1.3.1 PMOS Halo Analysis 55

4.1.3.2 NMOS Halo Analysis 59

4.1.4 Summary 62

4.2 Investigation of Lightly Doped Drain Implant (LDD) in PMOS and NMOS Device

63

4.2.1 Simulation Procedure 63

4.2.2 PMOS Simulation Results 64

4.2.2.1 Boron Di-Fluoride (BF2) dose splits 64

4.2.2.2 Boron Di-Fluoride (BF2) energy splits 72

4.2.3 NMOS Simulation Results 78

4.2.4 Summary 82

4.3 Investigation of Source/Drain Implant (S/D) in PMOS and NMOS

Device 83

4.3.1 Simulation Procedure 83

4.3.2 Simulation Results 84

Page 9: optimization and characterization of 130 nm cmos transistor design

4.3.2.1 PMOS Source/Drain Analysis 85

4.3.2.2 NMOS Source/Drain Analysis 88

4.3.3 Summary 92

Chapter 5 ANALYSIS AND DISCUSSION

5.1 Introduction 93

5.1.1 Analysis and Discussion of Halo Implant in PMOS and NMOS Transistor

93

5.1.1.1 PMOS Device 93

5.1.1.2 NMOS Device 105

5.1.2 Analysis and Discussion of Lightly Doped Drain (LDD) Implant in PMOS and NMOS Device 111

5.1.2.1 PMOS Device 111

5.1.2.2 NMOS Device 115

5.1.3 Analysis and Discussion of Source/Drain (S/D) Implant in PMOS and NMOS Device 118

5.1.3.1 PMOS Device 118

5.1.3.2 NMOS Device 121

5.2 Summary 124

Chapter 6 CONCLUSION

6.1 Conclusion 125

6.2 Limitation and Recommendation 127

6.3 Summary 129

REFERENCES 130

LIST OF PUBLICATIONS 135

APPENDIXES 136

Page 10: optimization and characterization of 130 nm cmos transistor design

LIST OF TABLES

Page Table 3.1 SIMS data obtain for this experiment.

41

Table 3.2 Standard 0.13 µm CMOS process in Silterra.

49

Table 3.3 Saturation current, Idsat data between simulation using MEDICI and real wafer measurement. 51

Table 4.1 Process split for PMOS and NMOS halo implant with various dose. 55

Table 4.2 Dose and energy process split for PMOS LDD implant using Boron Di-fluoride species. 65

Table 4.3 Dose process split for NMOS LDD implant using arsenic species. 78

Table 4.4 Process split for PMOS and NMOS source/drain implant with various implant parameters. 85

Table 5.1 Threshold voltage simulated using MEDICI and measured value for different PMOS halo implant dosage at L=0.13 µm. 97

Table 5.2 Measured and simulation data for Vt with different PMOS halo implant energy at L=0.13 µm. 99

Table 5.3 Measured and simulation data for Vt with different halo implant tilt angle at L=0.13 µm. 100

Table 5.4 Effects of halo implant parameters on threshold voltage, Vt and saturated drain current, Id for PMOS. 105

Table 5.5 Measured and simulation data for Vt with different NMOS halo implant dosage at L=0.13 µm. 107

Table 5.6 Measured and simulation data for Vt with different NMOS halo implant energy at L=0.13 µm. 108

Table 5.7 Measured and simulation data for Vt with different NMOS halo implant tilt angle at L=0.13 µm. 110

Table 5.8 Advantages and disadvantages of BF2 implant in CMOS transistor.

113

Table 5.9 Tilt angle and energy splits for NMOS transistor simulation using TCAD. Fixed parameters: dose at 1.5E+15 atom/cm2, rotation=0. 116

Page 11: optimization and characterization of 130 nm cmos transistor design

LIST OF FIGURES

Page

Figure 1.1

Kilby's Integrated Circuit (flip-flop using two transistors). 3

Figure 1.2 Simple CMOS transistor cross-sectional view. 3

Figure 2.1 (a) Current-controlled and, (b) voltage-controlled amplifiers.

11

Figure 2.2

a) Circuit of CMOS inverter and b) its logic symbol. 12

Figure 2.3

Drain Current (Id) versus gate voltage (Vg) graph. 14

Figure 2.4

Terminology for charges associated with thermally grown SiO2. 15

Figure 2.5

Drain current (Id) versus drain voltage (Vg). 16

Figure 2.6

Transistor in saturation mode (current remains constant or saturates).

18

Figure 2.7

NMOS and PMOS process flow 19

Figure 2.8 Comparisons of ion implantation and diffusion doping profiles 20

Figure 2.9

Halo structure in transistor. 22

Figure 2.10 Source/Drain Extension implants diagram. 24

Figure 2.11

Charge sharing phenomenon. 29

Figure 2.12

Threshold voltage variation with channel length. 30

Figure 2.13

Drift velocity becomes a constant above the critical field. 31

Figure 2.14

Punch-through effect in a short channel transistor. 32

Figure 2.15

Id vs Vd curve due to punch-through effect. 32

Figure 2.16

Impact ionization effect in transistor. 33

Figure 2.17

Hot carrier effect. 33

Figure 3.1

Initial discrepancies between SIMS measurement and profile simulated by TSUPREM4 using default moment parameters. 43

Figure 3.2

Cut line done on PMOS and NMOS device simulated by TSUPREM-4 to obtain doping profile. 45

Figure 3.3 Comparison of Phosphorus Nwell doping concentration profile, simulated and measured by SIMS. 46

Page 12: optimization and characterization of 130 nm cmos transistor design

Figure 3.4 Comparison of PMOS Lightly Doped Drain (PLDD) Boron doping concentration profile, simulated and measured by SIMS. 46

Figure 3.5 Comparison of Boron doping concentration profile by simulated and measured by SIMS for Pwell. 47

Figure 3.6 Comparison of NMOS Source/Drain (NSD) Phosphorus doping concentration profile, simulated and measured by SIMS. 48

Figure 3.7 Comparison of Arsenic doping concentration profile by simulated and measured by SIMS for NMOS Source/Drain (NSD). 48

Figure 3.8 Simulated drain current (Id) versus drain voltage (Vd) using MEDICI for 0.13 µm PMOS device 50

Figure 3.9 Simulated drain current (Id) versus drain voltage (Vd) using MEDICI for 0.13 µm NMOS device. 50

Figure 4.1 Workflow for experimental procedure. 54

Figure 4.2 Arsenic doping profile for three different arsenic doses to optimize the Halo implant step, (a) HaloPD1 (3.3E+13 atom/cm2), (b) HaloPD2 (3.1E+13 atom/cm2) and (c) HaloPD3 (3.5E+13 atom/cm2).

56

Figure 4.3 PMOS Halo dose for 10/0.13 device accuracy plot between measured and simulated. 58

Figure 4.4 Simulated Id versus Vd curve for different PMOS halo dose, (a) HaloPD1 (3.3E+13 atom/cm2), (b) HaloPD2 (3.1E+13 atom/cm2) and (c) HaloPD3 (3.5E+13 atom/cm2).

58

Figure 4.5 Boron concentration for NMOS Halo implants with various doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

60

Figure 4.6 Current flow line (grey lines) in conjunction with electric potential concentration for different boron doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

61

Figure 4.7 Current flow vertical cutline align along the poly gate profile for various halo dose, 1.0E+13, 1.3E+13 atom/cm2 and 1.6E+13 atom/cm2.

61

Figure 4.8 NMOS Halo dose for 10/0.13 µm device accuracy plot between measured and simulated for 3 different boron doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

62

Figure 4.9 Workflow for experimental procedure. 63

Figure 4.10 Doping profile for three different BF2 doses, BF2-PD1 (2.2E+14 atom/cm2), BF2-PD2 (2.5E+14 atom/cm2) and BF2-PD3 (2.8E+14 atom/cm2) compared to boron (2.5E+14 atom/cm2).

66

Page 13: optimization and characterization of 130 nm cmos transistor design

Figure 4.11 Device vertical cut line example; all through the silicon substrate. 67

Figure 4.12 Boron vertical cut line doping profile comparing three different BF2 doses, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) compared to boron (2.5E+14 atom/cm2).

68

Figure 4.13 Current flow profile (grey lines) for PMOS device with different LDD dose, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) compared to compared with without fluorine implant.

69

Figure 4.14 Current flow vertical cutline example done at the drain of the PMOS device. 70

Figure 4.15 Current flow vertical cutline profile comparing various BF2 dose, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) with non-fluorine species dose (boron species).

70

Figure 4.16 Current density vertical cutline profile comparing various BF2 dose, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) with non-fluorine species dose (boron species).

71

Figure 4.17 PMOS LDD dose for 10/0.13 device accuracy plot between measured and simulated. 72

Figure 4.18 Doping profile for three different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared to boron (0.7 KeV).

73

Figure 4.19 Vertical cutline on boron concentration profile for 3 different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared with non-fluorine corporation doping, boron (0.7 KeV).

74

Figure 4.20 Current flow cutline profile and electric potential for three different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared with boron (0.7 KeV).

75

Figure 4.21 Total current profile with the vertical cutline under the gate edge for three different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared with non-fluorine corporation doping, boron (0.7 KeV).

76

Figure 4.22 PMOS total current cutline profile with different LDD energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compare with boron (0.7 KeV).

77

Figure 4.23 PMOS LDD energy for 10/0.13 µm device accuracy plot between measured and simulated data. 77

Page 14: optimization and characterization of 130 nm cmos transistor design

Figure 4.24 Doping profile for two different arsenic doses, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2). 79

Figure 4.25 Vertical cutline under gate for two arsenic concentrations, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

80

Figure 4.26 Current flow line profile for different arsenic dose used, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

81

Figure 4.27 Current flow lateral cutline profile for different NMOS LDD dose, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

81

Figure 4.28 NMOS LDD dose for 10/0.13 µm device accuracy plot between measured and simulated data. 82

Figure 4.29 Workflow of experimental procedure for source/drain process simulation. 84

Figure 4.30 Net doping and depletion region line for three different boron doses for PMOS source/drain process, (a) S/D1 (2.6E+15 atom/cm2), (b) S/D2 (2.7E+15 atom/cm2) and S/D3 (2.8E+15 atom/cm2).

86

Figure 4.31 Id-Vd curve for PMOS source/drain implant process with different dosage used for this process, (a) S/D1 (2.6E+15 atom/cm2), (b) S/D 2 (2.7E+15 atom/cm2) and S/D 3 (2.8E+15 atom/cm2).

87

Figure 4.32 PMOS S/D dose for 10/0.13 µm device accuracy plot between measured and simulated data. 88

Figure 4.33 Current flow profiles for 0.13 µm NMOS device with 3 different source/drain implant energy, (a) S/D_energy1 (45 KeV), (b) S/D_energy2 (50 KeV) and S/D_energy3 (40 KeV

89

Figure 4.34 Current density cutline profiles for 0.13 µm NMOS device with 3 different source/drain implant energy, (a) S/D_energy1 (45 KeV), (b) S/D_energy2 (50 KeV) and S/D_energy3 (40 KeV).

90

Figure 4.35 Drain current, Id versus drain voltage, Vd for NMOS device simulated using MEDICI for 3 different implant energy, (a) S/D_energy1 (45 KeV), (b) S/D_energy2 (50 KeV) and S/D_energy3 (40 KeV).

90

Figure 4.36 PMOS S/D energy for 10/0.13 µm device accuracy plot between measured and simulated data. 92

Figure 5.1 Two different transistors size with halo implant. It is shown here that too small gate length will result in halo overlapping each other and leads to Vt roll-off.

94

Page 15: optimization and characterization of 130 nm cmos transistor design

Figure 5.2 Threshold voltage, Vt versus gate length. This shows the effect of halo to Vt with a different gate length. 95

Figure 5.3 Transistor with different halo depth. Note that for a shallower halo, junction capacitance, Cj will not be affected. For halo deeper than source/drain implant, Cj will increase.

95

Figure 5.4 PMOS halo Vt roll-off curve for different arsenic dose, (a) HaloPD1 (3.3E+13 atom/cm2), (b) HaloPD2 (3.1E+13 atom/cm2) and (c) HaloPD3 (3.5E+13 atom/cm2).

96

Figure 5.5 Measured PMOS threshold voltage (Vt) curve for 10/0.13 µm transistor. 97

Figure 5.6 Measured and simulated PMOS threshold voltage, Vt for different halo dose. The smallest difference is seen at halo dose 3.3E+13 atom/cm2.

98

Figure 5.7 PMOS Vt roll-off for different halo implants energy, 70, 90 and 110 KeV. 99

Figure 5.8 Measured and simulated PMOS threshold voltage, Vt for different halo implants energy, 70, 90 and 110 KeV. 100

Figure 5.9 Measured and simulated PMOS threshold voltage, Vt with different halo implants tilt angle 200, 250 and 300 with rotation=0. 101

Figure 5.10 Transistor with different halo implant tilt angle and rotation=0. Normal tilt is at 300. 102

Figure 5.11 PMOS net doping profile for different implant tilt angle (200, 250 and 300). 103

Figure 5.12 PMOS total doping profile for different implant tilt angle, (200, 250 and 300). 103

Figure 5.13 Transistor showing resists shadowing problem. 104

Figure 5.14 NMOS Vt roll-off for three different halo implant doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

105

Figure 5.15 High dose halo doping on a PMOS transistor, (a) breakdown voltage versus doping concentration and (b) band-to-band tunneling or Zener effect leakage.

106

Figure 5.16 Measured NMOS Vt roll-off for different halo implants doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

107

Figure 5.17 Measured and simulated NMOS threshold voltage, Vt with different halo implants dose, HaloND1 (1.0E+13 atom/cm2), HaloND2 (1.3E+13 atom/cm2) and HaloND3 (1.6E+13 atom/cm2).

108

Figure 5.18 Measured and simulated NMOS threshold voltage, Vt with different halo implant energy, 10, 12 and 15 KeV. 109

Page 16: optimization and characterization of 130 nm cmos transistor design

Figure 5.19 Simulated NMOS Vt roll-off for different halo implants energy, 10, 12 and 15 KeV. 109

Figure 5.20 Measured and simulated NMOS threshold voltage, Vt with different halo implant tilt angle, 150, 250 and 350. 110

Figure 5.21 Threshold voltage, Vt for 1.2 V 10/0.13 µm PMOS LDD for BF2 and boron. 112

Figure 5.22 Threshold voltage, Vt difference for 1.2 V 10/0.13 PMOS for both BF2 and boron. 113

Figure 5.23 Saturation drain current, Idsat for 1.2 V 10/0.13 PMOS for both BF2 and boron. 114

Figure 5.24 Idsat comparison for 1.2 V 10/0.13 PMOS for both BF2 and boron. It can be noted Idsat degrades (~ 4 µA/µm) after changing to BF2 species.

115

Figure 5.25 Simulated NMOS Vt roll-off for different LDD implants tilt angle, 00, 30 and 70. 116

Figure 5.26 Simulated Vt roll-off curves comparing 3 different implant energy used for NMOS LDD process, 2, 3 and 5 KeV. 117

Figure 5.27 Simulated Vt roll-off curve comparing 3 different implant doses used for NMOS S/D process, 2.6E+15 atom/cm2, 2.7E+15 atom/cm2 and 2.8E+15 atom/cm2.

119

Figure 5.28 Simulated PMOS structure showing (a) net doping profile and depletion region and (b) electric potential and current flow lines. 120

Figure 5.29 Simulated Vt roll-off curve comparing 3 different implant energy used for NMOS S/D process, 45, 40 KeV and 50 KeV. 122

Figure 5.30 Simulated NMOS structure showing depletion region for different implant tilt angle. 122

Figure 5.31 Net doping vertical cutline profile for NMOS source/drain implant with different implant energy. 123

Figure 6.1 (a) Denser grid factor, (b) less dense grid factor. 129

Page 17: optimization and characterization of 130 nm cmos transistor design

CHAPTER 1

INTRODUCTION

1.1 Background

Complementary metal-oxide-semiconductor (CMOS) integrated circuits (ICs) have

played a dominant role in the semiconductor world and are likely to retain this position

for the foreseeable future. Nowadays, leading microprocessors, application specific ICs

(ASICs) and dynamic random access memory (DRAMs) larger than 1Mb are most

entirely fabricated using CMOS technology. The key factors contributing to the

popularity of CMOS technology are its low power consumption and ability to

miniaturize.

Sah and Wanlass originally proposed the pairing of complementary n- and p-channel

transistors to form low power ICs in 1963 [1]. The first CMOS IC was fabricated in

1966 and lay dormant for nearly a decade because of several drawbacks that were

difficult to overcome at that time. For many years, CMOS lagged behind the advanced

silicon gate n-channel MOS (NMOS) and bipolar technologies. After the process of

local oxidation of silicon (LOCOS) isolation and ion implantation became available, the

performance and density of CMOS ICs improved dramatically.

From late 1970s, with the dawning of the VLSI era, the power consumption in NMOS

circuits began to exceed tolerable limits. It was apparently impractical to design future

generations of MOS circuits with NMOS technology. Subsequently, CMOS process

technology began to undergo rapid development. IC density and speed increased at a

rapid pace. The electronic industry is now able to fabricate more than 64 billion

transistors on a single chip with the minimum feature size of 90 nm. In the next

Page 18: optimization and characterization of 130 nm cmos transistor design

generations it will be necessary to reduce minimum feature size to 45 nm. Gigabyte

DRAMs and sophisticated ultra large-scale integrated circuit (ULSI) processing using

20nm technology are likely to appear in the near future. Even at the present time,

CMOS transistor designers working in advanced process laboratories are able to

fabricate transistors with channel length smaller than 20nm.

Since the semiconductor devices have evolved tremendously, today transistors are

extremely small and come packed in millions onto tiny silicon chips called IC. In 1958,

Jack Kilby (Nobel Laureate at Texas Instruments) [2] and Robert Noyce (Fairchild)

invented the IC as shown in Figure 1.1. This invention is essential for digital

technologies like computers, mobile phones, CDs, mp3s, DVDs; the list could be made

almost infinite.

“What we didn’t realize then was that the integrated circuit would reduce the cost of

electronic function by a factor of a million to one, nothing had ever done that for

anything before” (Jack Kilby, 1958)

The dramatic advances in integrated circuit IC’s over the past 20 years has propelled the

integrated circuit industry from small-scale integration (SSI) of less than 30 devices per

chip in 1970s, to medium-scale integration (MSI) of 30 to 103 devices per chip, to large-

scale integration (LSI) of 103 to 105 device per chip, to very-large scale (VLSI) of 105 to

107 devices per chip, and now to ultra-large scale (ULSI) of 107 to 109 devices per chip.

The increase in packing density is made possible by reduction in the transistor

dimension. As transistors are being scaled down to the sub micrometer regime, the

fabrication parameters, circuit design constraints and process information become

Page 19: optimization and characterization of 130 nm cmos transistor design

critical to the transistor performance. Therefore, there is a need to consider these factors

before the transistor is fabricated.

Figure 1.1: Kilby's Integrated Circuit (flip-flop using two transistors) [2].

1.2 Deep Sub-micron CMOS Process Technology

A new technology can be considered from three perspectives: performance, reliability

and manufacturability. The continued ability of CMOS to be scaled to 20 nm and below

will require many new innovations in the design of CMOS structures. The cross-

sectional view of the simple CMOS transistor is shown in Figure 1.2.

Figure 1.2: Simple CMOS transistor cross-sectional view.

Page 20: optimization and characterization of 130 nm cmos transistor design

It is likely that many of the innovations will rely heavily on the abilities of high-energy

ion implanters to precisely place dopants just where they are needed. Also, rapid

thermal processing is likely to play a key role in reducing overall thermal budgets. At

the same time, it is probable that the differences between n- and p- channel MOSFETs

will have to be addressed more openly, perhaps leading to different gate material and

drawing schemes for each type. But whichever transistor structure and process

technology is ultimately deemed the best, the choice of production will hinge as much

on manufacturability concerns as it does on reliability and performance.

While the dimension of MOSFET is scaling down steadily, there are two goals for

MOSFET scaling. The first goal is increased transistor saturation drain current per unit

width, Idsat, needed to increase the speed in charging and discharging parasitic

capacitance. This strengthens the major use of short channels and thin gate oxide.

The second goal of scaling is reduced size, needed to increase density. This requires

short channel lengths, and increased current per unit channel width so that less width is

needed to provide the necessary drive current. Reduced size also leads to reduced

transistor and interconnect capacitance for speed and power consumption improvement.

In the case of deep sub-micron transistors, some secondary effects, neglected for long

channel transistor become critical issues. They impose some constraints on the

performance and reliability of scaled transistors. Several new processes and transistor

structures were developed to overcome them. Thus, as the process becomes increasingly

complex for continuously scaled transistors, manufacturability must be taken into

consideration in volume production to guarantee high yield.

Page 21: optimization and characterization of 130 nm cmos transistor design

Two important effects must be avoided for scaled transistors. The first is the short-

channel effect. It causes threshold voltage (Vt) lowering, also known as drain induced

barrier lowering (DIBL), which leads to leakage while transistor is turned off. A thin

gate is required so that the gate control is strong and the leakage is suppressed. Bulk

punch-through, which causes sub-surface drain leakage current is another phenomenon

of the short-channel effect. Shallow source/drain junction depth and high bulk (well)

concentration are required to avoid punch-through. Additional punch-through implant

steps can be introduced to improved punch-through susceptibility independently.

The hot-carrier effect is the other undesired effect for scaled transistors. When the

physical dimensions of the MOS transistor are scaled down and the operation voltage

does not scale proportionally, high electric fields occur in the channel. Hot carriers

created in the high field region result in increased substrate current (Isub) and transistor

degradation, which includes threshold voltage shift, transconductance degradation and

capacitance degradation. The principal mechanism for NMOS field effect transistor

(NMOSFET) degradation appears to be interface state generation in the gate oxide.

1.3 Overview of Process and Device Simulation.

Simulation is the activity of carrying out experiments with the aid of a computer, using

mathematical models formulated to describe the phenomenon being studied. Process

and device simulators can be used to predict the effect of a process change on the circuit

or a process step or device electrical characteristics. Visualizing the details of a process

step and the physical phenomena involved in the transistor operation gives immediate

insight into a miniaturized transistor that is not always accessible by experimental

diagnostics.

Page 22: optimization and characterization of 130 nm cmos transistor design

Process simulations with calibrated models have proven to be valuable tools for new

process development. Process simulation can significantly reduce the cost and cycle

time of developing a new technology. It also allows the analysis of physical effects that

are difficult to be measured, such as lateral impurity profiles, electrical field

distribution, the location of depletion regions and the locations of impact ionization.

Process simulators are capable of simulating many of the individual process steps with a

high degree of accuracy.

On the other hand, many of the physical processes involved in IC fabrication are still

not completely understood. As a result, computer simulations must still be regarded as

useful guides, but not yet perfect representatives of actual process sequences. The

outputs from process simulators are subsequently used as the inputs to device simulators.

These outputs of device simulators can be internal device phenomena such as potential,

charge, and current distributions, as well as device thermal behavior (i.e., threshold

voltage and current-voltage characteristics, etc.).

1.4 Motivation of this work

Basic CMOS processes usually use seven to nine ion implants per wafer, current

leading edge CMOS processes used thirteen to fifteen implants, and some specialized

CMOS circuits use up to twenty implantation steps. The scaling of CMOS required to

move silicon transistors to deep submicron dimensions faces several major problems.

The anticipated performance improvements (e.g., higher speed, lower power usage, etc.)

of scaled transistors are often partially offset by transistor performance degradations

resulting from the scaling. These problems include:

Page 23: optimization and characterization of 130 nm cmos transistor design

• Ultra-shallow junction formation in PMOS transistors.

• Sub-threshold leakage and punch through in transistors.

• Reliability problems caused by “hot electron” effects resulting from higher

peak electric fields in the transistor channel near the drain.

• Performance degradation from the increasingly large contribution of parasitic

resistances and capacitances in the scaled transistor structure.

• Packing density limitations (“layout” constraints) in scaled CMOS circuits

required for avoiding latch-up, leakage between circuit elements, and other

undesirable interactions between different parts of the circuit structure [3].

Ion implantation plays an important role in solving all of these problems. Hence in this

project, particular attention has been given into the investigation of the ion implantation

process as many challenges in submicron CMOS are due to this process.

This work investigates the effect of different dose, energy and tilt angle in Halo, Lightly

Doped Drain (LDD) and Source/Drain (S/D) implant step for PMOS and NMOS

transistors in order to characterize the 0.13 µm CMOS process at Silterra (M) Sdn. Bhd.

The process parameters for these implantation steps will be characterized. This work is

done using a calibrated TCAD simulator, consisting of a process simulator, TSUPREM-

4 and a device simulator, MEDICI. The electrical result (saturation drain current, Idsat

and threshold voltage, Vt) from the simulation will be analyzed and compared with the

measured data obtained from real wafer fabricated at Silterra. From here, optimized

implant parameters for Halo, LDD and S/D in 0.13 µm CMOS process will be

suggested for further use at Silterra.

Page 24: optimization and characterization of 130 nm cmos transistor design

1.5 Objectives

The objectives of this work are as follows:

1) To calibrate and validate the TCAD deck, (TSUPREM-4 and MEDICI) for the

CMOS 0.13 µm Process Technology.

2) To simulate and characterize 0.13 µm CMOS for Halo, lightly doped drain and

source/drain implant.

3) To match the simulated parameters with Silterra (M) Sdn Bhd 0.13 µm CMOS

transistors specification by adjusting the dose, energy, tilt and rotation in the implant

steps.

4) To suggest a practical approach for the design and optimization of sub

micrometer MOSFETs.

1.6 Thesis Overview

This thesis is organized into six chapters.

Chapter 1 provides an introduction to the CMOS transistors, motivation of this work,

and an outline of the thesis.

Chapter 2 focuses on the problems encountered during the process of downscaling

CMOS. A review of other works in process and transistor optimization for CMOS is

also discussed in this chapter.

Page 25: optimization and characterization of 130 nm cmos transistor design

Chapter 3 describes the calibration and validation of TSUPREM-4 and MEDICI. A

characterization technique called Secondary Ion Mass Spectrometry (SIMS) is used for

calibration in this chapter.

In chapter 4, the experimental procedure and results is presented. All halo, LDD and

S/D structure results for NMOS and PMOS is discussed here. In this chapter, the

simulation and experimental results are given and compared. Here, interpretation is

made based on the observation of the structural and electrical analysis.

Chapter 5 focuses on the analysis and discussion. Here, thorough and further analysis is

discussed and from this analysis, the suggested process parameter for all halo, LDD and

S/D structure is determined and it is further clarified with the measured silicon data

from Silterra.

Chapter 6 summarizes the work done in this project. Conclusions of this work and

recommendations for future improvement of Silterra’s process parameters are also

described in this chapter.

Page 26: optimization and characterization of 130 nm cmos transistor design

CHAPTER 2

BACKGROUND AND LITERATURE REVIEW

2.1 Introduction

This chapter covers the literature review of earlier work done on ion implantation with

special focus on drain engineering in submicron CMOS processes. This chapter is split

into four main sections. First section covers the basic MOSFET characteristics. Second

section goes over the 0.13 µm CMOS fabrication process flow. The third reviews the

chronology of ion implantation and the last covers the non-ideal effects due to

MOSFET scaling.

2.2 MOSFET Basic Characteristics

The MOSFET is a four terminal active transistor that is usually used in analog and

digital electronic circuits. It has nearly the same functions as the Bipolar Junction

Transistor (BJT). Kahng and Atalla [4,5] invented the first field effect transistor (FET)

from thermally annealed metal oxide silicon (MOS) in 1960.

There are two types of MOSFET, n-channel and p-channel MOSFET. The n-channel or

NMOS depends on electrons as the majority carrier while p-channel or PMOS depends

on holes for conduction.

The primary conceptual difference between FET and BJT is the fact that the BJT

transistor is a current-controlled transistor as depicted in Figure 2.1(a), while the FET is

a voltage-controlled transistor as shown in Figure 2.1(b). The major advantage of

CMOS is that it uses much less power compared to BJT.

Page 27: optimization and characterization of 130 nm cmos transistor design

Figure 2.1: (a) Current-controlled and, (b) voltage-controlled amplifiers

2.2.1 CMOS

Beginning in the 1970’s, electronic watches and hand-held calculators developed very

quickly. Light-emitting diodes (LEDs) were use for the displays. However, LEDs

consume a lot of power and consequently limit the battery lifetime, so the IC industry

was eager to find a replacement for electronic watch and calculator applications.

Liquid-crystal display (LCD), which consumes much less power than LED, quickly

replaced LED for those applications after its introduction in early 1980 [3].

The need to reduce power consumption of the circuits for calculators and electronic

watches was one of the major driving forces for complementary MOS (CMOS)-based

IC chip development. CMOS is used in logic and memory chips, and dominates the IC

market [3]. Figure 2.2 shows a CMOS inverter circuit. We can see that it has two

transistors, one NMOS and one PMOS. When the input is high voltage, or logic 1, the

NMOS is turn on and the PMOS is turn off. Therefore the output voltage is the ground

voltage, VSS’ and Vout is low voltage, or logic 0. Conversely, if the input is low voltage,

or logic 0, the NMOS will be switch off and the PMOS switched on. The output voltage

will be the high voltage, Vdd, and Vout will be high voltage, or logic 1. Because it inverts

Page 28: optimization and characterization of 130 nm cmos transistor design

the input signal, it is called an inverter. This design is one of the basic logic gates used

in logic circuits.

Figure 2.2: a) Circuit of CMOS inverter and b) its logic symbol [3]

From Figure 2.2, we can see that when the NMOS is on, the PMOS will be off, and vice

versa. This is why the circuit is call complementary MOS, or CMOS. The circuit is

always open between high-voltage biased Vdd and grounded VSS’. Ideally, there is no

current flow between Vdd and VSS, so CMOS has very low inherent power consumption.

The main power consumption of the CMOS inverter comes from leakage or switching,

which has very high frequency. Other advantages of CMOS over NMOS are that it has

higher noise immunity, lower chip temperature, wider operation temperature range, and

less clocking complexity.

Combining both CMOS and bipolar technology, the BiCMOS IC developed rapidly in

the 1990s. The CMOS circuit is use for the logic part, and the bipolar for input/output to

increase transistor speed.

Page 29: optimization and characterization of 130 nm cmos transistor design

2.3 CMOS Electrical Characteristics

During the last three decades, transistor lengths have been reduce from 20 µm to much

less than a micron, which has resulted in high fields in the transistor. Here we will

discuss some main electrical characteristics of the PMOS transistor.

2.3.1 Threshold Voltage, Vt

One of the most important physical parameters of a MOSFET is its threshold voltage Vt,

defined as the gate voltage at which the transistor starts to turn on. Present day MOS

process invariability use ion implantation into the channel region; a step often called the

threshold adjusts implant, which alters the doping profile near the surface of silicon

substrate.

Threshold adjust implantation is a low-energy, low dosage implantation process.

Threshold implantation determines at what voltage a transistor can be turned on or off,

which is called threshold voltage, or Vt. For example, some old electronic transistors

require 12 V DC power supply; the majority of electronic circuits need 5 V or 3.3 V,

and the most advanced IC chips operate at 1.8 V or 1.2V. These operating voltages must

be higher than two times the threshold voltage to make sure these transistors can be

turned on or off, however they can’t be so high that it will break down the gate oxide

and destroy the transistors. By changing dose and energy of the threshold implant, a

desired threshold voltage is achieved [3].

To simplify, threshold voltage, Vt is the minimum gate voltage needed to create a

channel between source and drain. It can be defined as the minimum voltage for strong

inversion to occur. During operation, we supply a voltage between source and gate to let

the inversion occur so that current IDS can flow from source to drain. Vt value is

Page 30: optimization and characterization of 130 nm cmos transistor design

controlled during the fabrication process as mentioned above and is typically VDD/4.

Figure 2.3 shows the threshold voltage point in the drain current, ID versus gate voltage,

VG graph. For PMOS, the value is negative to attract positive charges (holes) in the

channel [6]. In ideal form, Vt is related to physical parameters as follows:

( )B

ox

BAs

tC

qNV ψ

ψε2

22+≈ (2.1)

where sε is the dielectric permittivity of silicon, q is electric charge, AN is impurity

(acceptor) concentration in p-type silicon and Bψ is surface potential at inversion.

However, the real condition is different because the interface charges and the metal or

polysilicon work function that must be taken into consideration.

Figure 2.3: Drain Current (Id) versus gate voltage (Vg) graph.

In reality threshold voltage must be calculated from the flat band voltage due to work

function difference and interface charges,

tFBt VVeffectiveV +≈ (2.2)

Page 31: optimization and characterization of 130 nm cmos transistor design

where FBV is the flat band voltage. FBV can be defined as the function of msφ , where

msφ is the polysilicon work function, Q , for difference defect charges, and oxide

capacitance, oxC .

+++=

ox

m

ox

ot

ox

f

ox

it

msFBC

Q

C

Q

C

Q

C

QV φ (2.3)

As shown in Figure 2.4, the defect charges are classified with respect to their location

and action as follows:

Figure 2.4: Terminology for charges associated with thermally grown SiO2.

1) Interface-state charges Qit, which are located so close to the Si-SiO2

interface and have energy states so close to the EF variation range of the

semiconductor, that is, mainly within the silicon forbidden bandgap as to be

able to exchange charges with the semiconductor in a short time.

2) Fixed charges Qf, which are located at or very near the interface but cannot

exchange charges unlike Qit.

Page 32: optimization and characterization of 130 nm cmos transistor design

3) Dielectric-trapped charges Qot, the trap sites of which are distributed

inside the bulk dielectric and capture (or emit) the charges brought into the

dielectric film by hot-carrier injection, and so on.

4) Mobile ionic charges Qm, such as sodium ions, which are mobile within the

oxide under bias-temperature aging conditions. [40]

From equation (2.1) and (2.2), in order to control tV , the AN and msφ values must be

controlled. We use the implantation of phosphorus ions to control AN and we replaced

aluminum with polysilicon as the gate. This is because the work function for polysilicon

to oxide is smaller than for aluminum to oxide.

2.3.2 I-V Characteristics

Current-Voltage (I-V) characteristics for the PMOS transistor can be divided into two

regions, the linear region and the saturation region. Figure 2.5 shows the relation

between current and voltage for PMOS by dividing it into two different regions.

Figure 2.5: Drain current (Id) versus drain voltage (Vg)

Page 33: optimization and characterization of 130 nm cmos transistor design

The linear region is the region where drain current increase linearly with the drain

voltage for every gate voltage, Vg value. The saturation region in other hand refers to

the region where drain current is almost constant, regardless of the changes of drain

voltage.

(a) Linear Region:

The relation between current that flows from drain to source, Ids and voltage between

drain and source, Vds for linear region ( Vds << ( Vgs-Vt )) is given as:

( )dstgsoxNds VVVC

L

WI −= µ (2.4)

where Nµ is the electron activation, W is the channel width, L is the channel length,

gsV is the voltage between gate and source, oxC is the oxide capacitance density per unit

area which is defined as:

ox

oxox

tC

εε0= (2.5)

where oxε is the oxide permittivity and oxt is the gate oxide thickness.

(b) Saturation region:

Saturation occurs when Vds > Vgs – Vt, the situation where drain-source voltage, Vds is

greater than the gate-source voltage, Vgs, as shown in Figure 2.6. Current will flow

constantly and will not be dependent on the increasing of Vds.

Page 34: optimization and characterization of 130 nm cmos transistor design

Figure 2.6: Transistor in saturation mode (current remains constant or saturates) [9]

The drain current in this region is given by:

( )2

2 tgs

ox

oxN

dsat VVLt

WI −=

εµ (2.6)

2.4 CMOS Process

The basic CMOS process steps are wafer preparation, well formation, isolation

formation, transistor making, interconnection, and passivation, as shown in Figure 2.7.

The wafer preparation includes epitaxial silicon growth, wafer clean, and alignment

mark etch. Well formation defines the substrate type for NMOS and PMOS transistors.

Depending on the technology generation, there are different techniques for well

formation: single well or self-aligned twin well. Shallow Trench Isolation (STI) has

replaced the Local Oxidation of Silicon (LOCOS) process in the manufacturing of deep

submicron transistors [10].

Page 35: optimization and characterization of 130 nm cmos transistor design

Transistor making involves gate oxide growth, polysilicon deposition,

photolithography, polysilicon etch, ion implantation, and thermal annealing. These are

the most crucial process steps in the IC processing sequence. Interconnection processes

use the combination of deposition, photolithography, and etch processes to define metal

wires and connect millions of transistors built on the silicon surface. Finally the

passivation dielectric deposition, photolithography, and etch process seals the IC chip

from the outside world, leaving only the bonding pad open for testing and wire bonding

[3].

Figure 2.7: NMOS and PMOS process flow

Page 36: optimization and characterization of 130 nm cmos transistor design

2.5 MOSFET Ion Implantation

One of the most important properties of semiconductor materials is that adding dopants

can control their conductivity. Semiconductor materials such as silicon, germanium, and

gallium arsenide are doped with either n-type or p-type dopant in IC fabrication. There

are two common methods to dope semiconductors: furnace diffusion from solid sources

and ion implantation. Before 1970, diffusion was use in IC fabrication. Currently

doping is mainly done by ion implantation.

Figure 2.8: Comparisons of doped region for (a) diffusion process and (b) ion implantation process

Ion implantation is processes by which dopant ions are forcefully add into the

semiconductor in the form of energetic ion beam injection. The ion implantation

process provides much more precise control of doping than the diffusion process [3].

For example, the dopant concentration and junction depth cannot be independently

controlled in the diffusion process because both are related to the diffusion temperature

and time; ion implantation can independently control both dopant concentration and

Page 37: optimization and characterization of 130 nm cmos transistor design

junction depth [Figure 2.8]. Dopant concentration can be control by the combination of

ion beam current and implantation time and junction depth can be controlled by the ion

energy. The ion implantation process can dope with a wide range of dopant

concentration, from 1011 to 1017 atoms/cm2.

In this project we are going to discuss the drain engineering implantation which covers

halo implant, source/drain extension implant or LDD, and source/drain implant. By

changing dose, energy and rotation of this implants can change the profile and the

electrical characteristics of the 0.13 µm CMOS.

2.5.1 Punchthrough Stop (Halo) Implants

Halo implants are use in submicron scaled transistors to prevent expansion of the drain

depletion region into the lightly doped transistor channel when the transistor is biased

for operation [3]. This behavior can short the channel (punchthrough) and increase

subthreshold leakage currents from the source to the drain [11,12]. It is also known as

the “drain-induced barrier lowering” (DIBL). In submicron transistors with short

channel lengths, punchthrough stopper implants are very necessary. The punchthrough

condition (channel is shorted) with undesirable leakage current will lead to circuit

failure. It occurs when channel length is reduced for scaling. The high electric fields

occur at the drain end of the channel. In the case of an n-channel MOSFET, the

electrons moving from source to drain are accelerated by this high electric field and by

energetic collision can create a free electron and hole pair. This can cause avalanche

breakdown [13].

Punchthrough stop implants (halo implant) place the dopant just below the active

channel, adjacent to the source and drain, as shown in Figure 2.9, in order to carefully

Page 38: optimization and characterization of 130 nm cmos transistor design

modify the well doping and prevent expansion of the drain depletion region into the

lightly doped channel when the transistor is biased for operation. A boron or indium

implant is employed with NMOS transistors; phosphorus, arsenic or antimony is

employed in PMOS transistors. The active channel region of scaled CMOS transistors is

an area of intense process characterization and development, and advanced ion

implantation applications characterized by complex lateral and vertical doping profiles

in the channel are required to allow implementation of submicron transistors in

mainstream production. Punchthrough stop implants are the best example for this.

Figure 2.9: Halo structure in transistor.

As channel lengths shrink, all MOS transistors become increasingly susceptible

to punchthrough effects. The effect is most severe for buried channel transistors, but

surface channel transistors are affected as well. There are two main methods of

preventing punchthrough. The first is to aggressively scale S/D extension and S/D

junction depth to be as shallow as possible [10,12]. This puts extra burdens on the

implantation and annealing equipment used to manufacture these junctions, especially

for p-MOSFETs. The other method is to increase the well doping just adjacent to and

beneath the S/D extension regions, to minimize the depletion width spread in these

areas. Precise placement of dopant is required to minimize the increases in junction

capacitance and decreases in channel mobility that result from additional well doping

[15-21].

Page 39: optimization and characterization of 130 nm cmos transistor design

Recent punchthrough stop implants are also known as “halo” implants, or “LATIPS”

implants (Large Angle Tilt Implanted Punchthrough Stopper). As this name implies,

these implants are usually performed using higher tilt angles (10-30o) [22,23]. As gate

dimensions shrink, less undercutting of the gate is required to produce the punchthrough

stop effect. Consequently, tilt angle requirements for this application are steadily

decreasing below the 0.25µm node, and are expected to reach the “low tilt” threshold of

10o by the 0.1µm node [24].

2.5.2 Source/Drain Extension Implants or Lightly Doped Drain (LDD)

Lower and lower energy implantation is being used to form the deep “source” and

“drain” regions and their “extensions”, which interface with the channel region in

today’s MOS transistors. The more lightly doped (1019 cm-3 range) extensions provide a

gradual dopant concentration gradient between the S/D and the channel region, which

reduces the maximum electric field. The extension region is also called the lightly

doped drain.

A lightly doped drain (LDD) implant is used when defining the source and drain regions

of the MOS transistors. This type or region is referred to as a source/drain extension as

shown in Figure 2.10. It is needed to achieve dimensional reductions for the scaling of

submicron transistors. The implant places the LDD dopant just to the edge of the

channel region under the gate to provide a gradual dopant concentration to the

source/drain regions. The LDD creates complex lateral and vertical doping profiles in

the interface region at the channel edge. The LDD implants for NMOS and PMOS

transistors occur in two separate masking and implant operations. The gate of each

MOSFET is implanted as the shallow junctions of the source and drain are also formed

[3].

Page 40: optimization and characterization of 130 nm cmos transistor design

Figure 2.10: Source/Drain Extension implants diagram.

The LDD structure is formed by a medium-to-low-dose implant aligned with the gate

structure (this is the n- or p- implant), followed by a high dose (called the n+ or p+)

source/drain implant. The source/drain implant is aligned to the gate by an oxide

sidewall spacer formed between the two implants. If an LDD is not formed, then high

electric fields are present between the junction and channel regions during normal

transistor operation. Electrons exposed to a high electric field during their movement

from source to drain (for an n-channel) are accelerated by this high field. Energetic

electron collisions create a free electron and hole pair (called hot carriers or hot

electrons) [25]. Hot electrons acquire energy from the electric field and cause electrical

performance problems, such as becoming trapped in the gate oxide layer and interfering

with the gate threshold voltage control of the transistor.

The LDD creates a gradual lateral dopant concentration gradient between the high

concentrations in the S/D regions (1020 to 1021 atoms/cm3) and the low concentrations in

the channel region (1016 to 1017 atoms/cm3) under the gate [26]. The reduced doping of

Page 41: optimization and characterization of 130 nm cmos transistor design

LDD decreases the electric field between the junction and channel regions. This

technique separates the maximum current path in the channel from the maximum

electric field location in the junction to avoid creating hot carriers.

In a dual-doped polysilicon process both NMOS and PMOS transistors rely on surface

channel operation to reduce the short-channel effect (SCE) and improve subthreshold

characteristics. The important challenge for this process is the Boron penetration from

the polysilicon gate into the channel area of PMOS transistor. Boron penetration causes

a shift in the threshold voltage (Vt) of the transistor, leads to subthreshold swing

degradation, reduces process control, and degrades gate oxide reliability.

2.5.3 Source/Drain Implants

Source/Drain (S/D) implants form highly doped regions (1020 to 1021 ions/cm3) that

interface with the lightly doped active channel and well regions (1016 to 1017 ions/cm3)

of an MOS transistor. S/D regions are doped to have the opposite conductivity type as

the well that surrounds them. Arsenic implants are often used to form n-type

source/drain for n-channel (NMOS) transistors, and boron or BF2 implants are usually

used to form p-type source/drain for p-channel (PMOS) transistors [10].

These highly doped S/D regions are contacted through cobalt silicide to the rest of the

circuit. In the absence of a bias on the gate and drain, the S/D regions form two back-to-

back semiconductor junction diodes. When a sufficiently large bias is applied to the

gate electrode, a surface inversion layer is formed creating a conducting “channel”

between the source and the drain (for normally-off, enhancement type transistor [12]).

Current can be made to flow in the channel by the application of a bias on the drain and

varying the bias on the gate can modulate the conductance.

Page 42: optimization and characterization of 130 nm cmos transistor design

Scaling of the S/D and LDD implants, to keep up with the transistor scaling

requirements, has posed the biggest challenge in recent years. Present transistor scaling

rules demands that the vertical and lateral dimensions be reduce by a factor of two,

approximately every 5 to 6 years [42]. Accordingly, both the deep S/D and the LDD

junction depths need to be reduce by the same factor. The challenge in scaling the S/D

implant is, primarily, meeting the junction depth requirement while maximizing the

active dopant concentration. The later is necessary to reduce the contact resistance of

the silicide to the doped S/D. For the deep S/D, unlike the LDD, meeting the junction

depth requirement also means not making it too shallow, otherwise the silicidation

process can consume the entire junction. It is also important to minimize the electrically

active residual damage in the S/D regions, which increases the reverse leakage current

of the S/D junctions and thus the leakage current of the transistor in the off state (Ioff).

Reducing the implant energy leads to a reduction in the junction depth. By using

implantation with its excellent control of doping placement, the lateral diffusion of the

source/drain dopant ions into the channel region is minimized.

2.6 MOSFET Scaling Effects

Scaling has major effects on silicon transistors and the wiring interconnects that are

used in silicon technology. In order to increase the semiconductor chips performance

and to match the MOSFET dimension, internal connections have to be scale down along

with the silicon transistor. Scaling doesn’t only change the transistor structure; it also

brings new problems to the transistor performance.

Page 43: optimization and characterization of 130 nm cmos transistor design

2.6.1 Short Channel Effect

A MOS transistor is termed short channel if the channel length is below a minimum

length given by Brew’s rule [27]

( )[ ]3

12

min 4.0 sdoxj WWtxL +≈ (2.7)

where jx = junction depth in µm

oxt = oxide thickness in Å

dW = drain depletion width in µm

sW = source depletion width in µm

For a short channel transistor, the threshold voltage decreases as the channel length

decreases. This is due to charge sharing between the gate region and both the source and

drain depletion region as shown in Figure 2.11. As the channel length decreases, this

effect becomes more pronounced. The equation for the threshold voltage for an NMOS

transistor (assume zero substrate bias, 0=bsV ) is as shown below

ox

ox

OX

B

FMStC

Q

C

QV −++= 0

0 2φφ (2.8)

where

SMMS φφφ −= (2.9)

=

i

A

Fn

N

q

kTlnφ (2.10)

( )FASioB NqQ φεε 220 = (2.11)

Page 44: optimization and characterization of 130 nm cmos transistor design

ox

oxo

oxt

Cεε

= (2.12)

OTMITFox QQQQQ +++= (2.13)

:msφ work function difference between the gate and the channel

:2 Fφ total band bending at surface inversion

:BOQ depletion region charge density at surface inversion

:FQ fixed oxide charge

:ITQ interface charge

:MQ mobile ion charge

:OTQ oxide trapped charge

:/ qkT thermal voltage

:AN substrate doping density

:in intrinsic carrier concentration

:oxC oxide capacitance

:oε permittivity of free space

:siε dielectric constant of silicon

:oxε dielectric constant of silicon oxide

:oxt gate oxide thickness

:k boltzmann constant

:q electronic charge

Page 45: optimization and characterization of 130 nm cmos transistor design

Figure 2.11: Charge sharing phenomenon [27].

However, for short channel transistor, the threshold voltage is given by:

ox

ox

OX

B

FMStC

Q

C

QV −++= 0

0 2φφ (2.14)

where

( )FASi

DS

B NqL

LLQ φεε 22

21' 00

∆+∆−= (2.15)

SL∆ = lateral extension of depletion region of the source

DL∆ = lateral extension of depletion region of the drain

From the above equation, 0BQ ’ becomes smaller than 0BQ as the channel length is

decreased. As result, 0tV decreases for short channel transistor as shown in Figure 2.12.

The reduction of the threshold voltage represents the amount of charge differential

between the rectangular and trapezoidal depletion region.

Page 46: optimization and characterization of 130 nm cmos transistor design

Figure 2.12: Threshold voltage variations with channel length.

2.6.2 On-State Saturation Current

For long channel transistor, the equation for the on-state saturation current is as follows:

( )2

2 togsoxnon VVL

wcI −= µ (2.16)

where w = channel width

L = channel length

nµ = electron mobility

In this equation, it is assumed that the drift velocity, Ev nd µ= as shown in Figure 2.13,

where E is the average electric field inside the channel, for cEE << (critical electric

field). For a short channel transistor, the equation for the on-state saturation current

becomes:

Page 47: optimization and characterization of 130 nm cmos transistor design

( )dtogsoxon vVVwcI −= (2.17)

In this equation, due to high lateral electric field E because of reduction in channel

length, the drift velocity dv becomes saturated ( )Ev nd µ≠ .

Figure 2.13: Drift velocity becomes a constant above the critical field [27].

2.6.3 Punchthrough Effect

In small geometry MOS transistors, a large drain bias voltage will result in the

extension of the drain depletion region towards the source. Once the two depletion

regions meet each other as shown in Figure 2.14, the gate voltage loses control over the

drain current. This can eventually lead to punch-through. If punch-through occurs, there

is a sudden increase in drain current as shown in Figure 2.15.

Page 48: optimization and characterization of 130 nm cmos transistor design

Figure 2.14: Punch-through effect in a short channel transistor.

Figure 2.15: Id vs Vd curve due to punch-through effect [27]. 2.6.4 Hot Carrier Effect

When the lateral electric field becomes high particularly near the reverse biased drain-

to-substrate junction due to dimension reduction, carriers are accelerated into the

depletion region. These carriers possess energy higher than the thermal energy, thus

they are known as hot carriers. When they collide with the ions present in the depletion

region, electron-hole pairs are generated by impact ionization. This is shown in Figure

2.16.

Hot carrier effect causes current to flow in the substrate, Isub which can subsequently

lead to latch-up. It also results in the formation of interface states which reduce the

Page 49: optimization and characterization of 130 nm cmos transistor design

electron mobility and hence decreases the saturation drain current as well as degrade the

transconductance. The threshold voltage will also increase due to electron trapping in

the oxide. This is show in Figure 2.17 [27].

Figure 2.16: Impact ionization effect in transistor.

Figure 2.17: Hot carrier effect

Page 50: optimization and characterization of 130 nm cmos transistor design

2.7 Methods Attempted by Other Researchers

As gates get smaller, the channel length below the gate structure (the silicon region

between the source and drain) also decreases. This reduction, although is good for

manufacturer, it causes many problems for CMOS transistors. Deep sub-micrometer

MOS transistors often need special structures to optimize their performance.

Researchers have been trying to find alternative ways to reduce many downscaling

problems and to optimize CMOS transistors. The halo structure, or halo implant, is

usually adopted to reduce off-state leakage current and enhance on-state drive current.

Jiong-Guang Su [25], mentioned that transistor with higher halo implant tilt angle

reduced body effect and increase source resistance as compared to those with low tilt

angle. They investigate the influence of tilt angle of halo implant on optimizing PMOS

performance based on process and device simulation. If a proper dose is adopted, large

tilt angle halo implant should be adapted to reduced parasitic capacitance and enhanced

transistor performance.

In the work reported by Wen-Kuan Yeh [28], it was proposed that for high performance

sub-0.1µm channel length MOSFETs, we must use optimized indium halo structures

and arsenic implantations. They found that:

1) Compared to B-halo NMOS, In-halo NMOS form localized high-dose structures

that improve transistor punch-through margins and are free of apparent Reverse

Short Channel Effect (RSCE) due to indium deactivation. Thus In-halo

structures reduced Vths while increasing transistor resistance to Short Channel

Effect (SCE).

Page 51: optimization and characterization of 130 nm cmos transistor design

2) In-halo NMOS junction leakage and capacitance were affected by indium

dosage levels and off-state leakage increased due to and increase in the n+ junction

breakdown field.

3) Larger junction fields occurred around deep source/drain junction regions with

larger junction areas due to locally heavy doses of implanted indium and junction

leakage and capacitance were found to be sensitive to In-halo implantation energies.

4) Lower temperature composite liner-oxide/SiN spacer technology is proposed for

PMOS to suppress transistor performance degradation due to high-temperature

processing.

Work from C.Chen [29] also suggest that to improve PMOS Vt roll-off, high energy As

P-halo implant through the polysilicon gate should be used, and this was confirmed by

Secondary Ion Mass Spectroscopy (SIMS) and simulation. They also optimized the

drain engineering with proper BF2 P-extension energy and As P-halo energy, angle and

dose.

Besides using halo structures, researchers also did many studies on source/drain

junctions. Fu-Cheng Wang [26] reported that compared to the conventional single BF2

implant fabrication technique, the double-implanted source/drain junction technique

allows further downscaling of BF2 implant energy without increasing the junction

leakage. This technique also offers the advantages of lower junction capacitance, less

boron penetration, thinner gate oxide, and wider process window.

Page 52: optimization and characterization of 130 nm cmos transistor design

K.K. Bourdelle [30] reported the effect of fluorine from BF2 source/drain extension

implants on performance of PMOS transistors with thin gate oxides. Boron penetration,

which causes a degradation of many transistor parameters, is further enhanced when

BF2 is used to dope the gate electrode. In the paper, a comprehensive study of the

fluorine enhanced boron diffusion in gate oxide is presented. Boron penetration causes a

shift in the threshold voltage (Vt) of the transistor, leads to subthreshold swing

degradation, reduces process control, and degrades gate oxide reliability. A range of

methods was used to suppress fluorine enhanced boron penetration is as follows:

1) nitrogen implantation into poly-Si

2) stacked amorphous/poly gate structure

3) sacrificial gate stack process

These techniques focus on preventing fluorine migration from the implanted gate to the

gate oxide. In order to prevent fluorine from the source/drain (S/D) areas from diffusing

into the gate oxide-channel region, this paper suggests to use W-polycide process, in

which gate and S/D doping is done separately. In this paper they also show that

enhanced boron penetration due to fluorine from the BF2 extension implants degrades

PMOS transistor performance for gate oxide thicknesses less than 27 Å and gate lengths

less than 0.5 µm.

Many researchers have proposed methods not only to optimize the process parameters

but also to control the electrical characteristics, mainly the threshold voltage and

saturation current. Hajime Kurata and Toshihiro Sugii [31], suggested a new method to

control the threshold voltage (Vth) in sub-0.2 µm MOSFET’s. They changed the

concentration of the channel impurity according to the gate length by tilted ion

implantation from two directions after the polysilicon gate formation. Their results

Page 53: optimization and characterization of 130 nm cmos transistor design

show that the short channel effect is effectively reduced by this method without

increasing the S/D junction capacitance.

From the literature review work, the process characterization and device optimization

are developing rapidly for CMOS technology today. It is important for us to

characterize our technology in Silterra as there are many possibilities that problems will

occur when a transistor is scaled down. Based on the literature, equipment and software

availability in Silterra, TCAD simulator was chosen to simulate the process and

transistor for Silterra 0.13 µm CMOS and to compare all the simulated data with the

measured data obtained from the real wafers that were fabricated at Silterra.

2.8 Summary

In this chapter, the basic properties and operation of a transistor have been discussed.

The importance of ion implantation especially Halo, LDD and source/drain implant in

the CMOS process has also been addressed. The importance of characterizing the

electrical characteristics for CMOS has also been discussed. In the following chapter,

the experimental procedure for calibrating and validating the TSUPREM-4 and

MEDICI input files would be described.

Page 54: optimization and characterization of 130 nm cmos transistor design

CHAPTER 3

CALIBRATION AND VALIDATION OF TSUPREM-4 AND

MEDICI

3.1 Introduction

Commercial TCAD tools, TSUPREM-4 [32] and MEDICI [33] are used for 0.13 µm

process and device simulation. Since they are not calibrated to any specific technology,

use of default simulation parameters can lead to significant differences between the

simulated and measured I-V characteristics. Therefore, certain levels of calibration

work are necessary for judicious selection of different models and modification of some

important default parameters. This tuning of model parameters to match measured data

is one of the objectives of the thesis. Simulated results from both TSUPREM-4 and

MEDICI are compared to measured data to evaluate agreement.

Based on optimized models and parameters, simulation work is focused on 0.13 µm

technology mainly for drain engineering consisting halo, lightly doped drain and

source/drain structures. These critical parameters are optimized for performance and

reliability improvement.

3.2 TSUPREM-4 [34]

Taurus TSUPREM-4 is an advanced 1D and 2D process simulator for developing

semiconductor process technologies and optimizing their performance. With a

comprehensive set of advanced process models, Taurus TSUPREM-4 simulates the

process steps used for fabricating semiconductor devices, reducing the need for costly

experiments using silicon. In addition, Taurus TSUPREM-4 has extensive stress

modeling capabilities, allowing optimization of stress to increase transistor performance.

Page 55: optimization and characterization of 130 nm cmos transistor design

Benefits:

• Develop leading edge CMOS, bipolar, and power-device manufacturing

processes cost-effectively.

• Predict 1D and 2D device structure characteristics by accurately simulating ion

implantation, diffusion, oxidation, silicidation, epitaxy, etching, and deposition

processing, reducing experimental runs and technology development time.

• Analyze stress history in all layers as a result of thermal oxidation, silicidation,

thermal mismatch, etching, deposition, and stress relaxation.

• Study impurity diffusion, including oxidation-enhanced diffusion (OED),

transient-enhanced diffusion (TED), interstitial clustering, dopant activation, and

dose loss.

3.3 TAURUS MEDICI [35]

Taurus Medici is a 2D device simulator that models the electrical, thermal, and optical

characteristics of semiconductor devices. A wide variety of devices including

MOSFETs, BJTs, HBTs, power devices, IGBTs, HEMTs, CCDs, and photodetectors

can be modeled. Taurus Medici can be used to design and optimize devices to meet

performance goals, thereby reducing the need for costly experiments. With the

continued scaling of CMOS devices, device design and optimization become more

difficult. Previously unimportant phenomena, such as direct tunneling, can now

dominate device performance and design. The vast array of advanced transport and

quantum models available in Taurus Medici allow users to perform accurate simulations

of deeply scaled devices.

Page 56: optimization and characterization of 130 nm cmos transistor design

Benefits:

• Analyze electrical, thermal, and optical characteristics of devices through simulation

without having to manufacture the actual device.

• Determine static and transient terminal currents and voltages under all operating

conditions of interest.

• Understand internal device operations through potential, electric field, carrier,

current density, and recombination and generation rate distributions.

• Investigate breakdown and failure mechanisms, such as leakage paths and hot-

carrier effects.

3.4 Calibration of 130 nm MOSFET

This section will discuss the calibration done before we start our simulation using

TCAD tools. Calibration of TCAD simulator is done by tuning the implant condition

based on the Secondary Ion Mass Spectrometry (SIMS) implant data taken on finished

wafers.

Silterra SIMS vendor, Charles & Evans, performed the SIMS profiling. Samples were

identified and selected based on the sheet resistance results. Details such as dopant

species, implant doses, implant energy, the expected dopant depth and full process of

the samples were provided in order to get a more accurate result. Samples were also cut

into small pieces at the centre of the wafer before being delivered to the vendor in order

to save the cost. The profiling takes 2 weeks to complete for normal price; extra charges

will be needed if the duration of the profiling is shorter than 2 weeks. Results are

available in both hard copy and soft copy (in the form of excel sheet).

The TCAD simulator is calibrated by comparing the SIMS data with the simulated

result. If the matching does not fit well, the process parameter such as annealing and

Page 57: optimization and characterization of 130 nm cmos transistor design

deposition time was adjusted to make the calibration curve fit well. The implant

condition was not changed, as it will change the process that was used in fabricating the

wafer that was sent for SIMS. In this chapter, we will calibrate one-dimensional as-

implanted profiles. With the help of the optimization function in TSUPREM-4 version

W-2004.12, some key implant moment parameters for Gaussian or Pearson distribution

are optimized to fit the measured data. Implant profiles from secondary ion mass

spectroscopy (SIMS) measurement are selected as the optimization target. The

optimized distributions parameters are subsequently incorporated into the simulations.

The SIMS data that was obtained in order to tune and optimize our TCAD NMOS and

PMOS simulator deck is shown is Table 3.1.

Table 3.1: SIMS data obtain for this experiment.

No. Structure Name Implant Profile

1 Nwell_STI Phosphorus

2 Pwell_STI Boron

3 HP_NSD_S Arsenic, Phosphorus

4 HP_PSD_S Boron

5 HP_Ngate_S Arsenic, Boron, Indium, Phosphorus

6 HP_Pgate_S Arsenic, Phosphorus

7 HP_NLDD Arsenic, Boron

8 HP_PLDD Boron

9 DNWell_STI Phosphorus

During the calibration, some matching problem has occurred and troubleshooting has to

be done in order to get an accurate result with percentage error less than 10%. In this

PMOS and NMOS transistor calibration, the accuracy of as-implanted doping profiles,

Page 58: optimization and characterization of 130 nm cmos transistor design

especially in the channel region, is crucial to the final simulation results. The implant

moment tables for the analytic model in TSUPREM-4 give very accurate results after

implantation into bare, <100> silicon, but are normally not appropriate under other

implant conditions. In this calibration process, the analytic implant models, including

Gaussian, Pearson and dual-Pearson functions, are found to be suitable in predicting

implant profiles. The parameters associated with these models need to be optimized for

best fit to SIMS profile. This analysis is called standard regression analysis. However,

there are some limitations on standard regression analysis method. It is based on some

assumptions, which are not always valid in analysis.

Actually, the importance and the accuracy of the information about the value of y

(concentration) for each x (depth) may differ substantially in SIMS profile. In this study,

it is desirable to give greater emphasis to the data at the near-surface part (from silicon

surface to the depth of 0.1 µm) because the dopants there have more influence on basic

transistor characteristics.

We first select a phosphorus profile implanted at energy of 380 KeV and a dose of

6E+12 atom/cm2 for calibration. Figure 3.1 shows the discrepancy between SIMS

profile and simulated profile before calibration. The default implant table of phosphorus

defined in the implant data file is used for Pearson function. The MOMENT parameters

are fitted to the results of Monte Carlo calculations and hence, no channeled tail can be

simulated and fitted. It is shown that the SIMS profile has a clear channel tail. Thus,

dual-Pearson function should be used to fit it. Appendix B lists the TSUPREM-4

program in which parameters of dual-Pearson functions are determined for the

phosphorus profile. A SIMS data file (concentration versus depth) is used as a target for

the optimization.

Page 59: optimization and characterization of 130 nm cmos transistor design

1E+15

1E+16

1E+17

1E+18

-0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Depth (micron)

Ph

osp

ho

rus c

on

cen

trati

on

(ato

ms/c

c)

Nwell Simulation

Nwell SIMS

Figure 3.1: Initial discrepancies between SIMS measurement and profile simulated by TSUPREM4 using default moment parameters.

During the optimization loop, successive estimates of the Pearson parameters are

introduced into the simulation using the MOMENT statement. The MOMENT statement

sets distribution moments for specific materials. The used of this statement overrides

the reading of implantation data file that would normally occur. The EXTRACT

statement extracts the chemical phosphorus concentration at the locations specified in a

SIMS data file. At each step in the optimization procedure the parameters to be

optimized are varied to match the extracted phosphorus concentration with the target

concentrations given in a specified SIMS data file.

Now the calibration of multiple implanted profiles will be discussed. For a quarter-

micron transistor, doping profile in N-channel region consists of P-well implant, N-

channel anti-punchthrough (NPT) implant and N-channel threshold voltage adjust

(NVT) implant. In designing a new transistor for a specified Vt requirement, we

generally fabricate some wafers simultaneously with different implant conditions.

Among these wafers, the implant energies are fixed. The doses of NVT and NPT

Page 60: optimization and characterization of 130 nm cmos transistor design

implant are varied to ensure that transistors in one or more wafers can hit the target. For

such purpose, accurate prediction of Vt by process and device simulation is desirable.

This cannot be accomplished without calibrations of implant moment parameters related.

In this case, wafers are prepared for each implant and send out for SIMS measurement.

The basic optimization measure is similar to that of previous section and detail steps

will not be described here. The comparison of simulated and SIMS data is discussed

further in section below.

1) PMOS Transistor

Figures below are the calibration graphs comparing the SIMS profile and the simulated

profile obtained from TSUPREM-4. The TCAD deck for PMOS and NMOS were

calibrated focusing more at the implant step, because at this step is the critical part for

optimization and characterization analysis of 130 nm CMOS for this experiment.

We start with PMOS deck, which have three main implant steps to be calibrated, Nwell,

PLDD and PSD. Figure 3.2 shows the cut line that was done on the PMOS transistor

simulated by TSUPREM-4.

Page 61: optimization and characterization of 130 nm cmos transistor design

Figure 3.2: Cut line done on PMOS and NMOS transistor simulated by TSUPREM-4 to obtain doping profile.

As shown in Figure 3.3 to Figure 3.4, the TCAD deck for PMOS transistor is calibrated

as the SIMS profile data fits well with the TSUPREM-4 profile. Figure 3.3 shows the

Nwell matching plot. The maximum percentage difference between simulation and

SIMS is 27.3%.

Cut line

x (µm)

y (µm)

Page 62: optimization and characterization of 130 nm cmos transistor design

1.00E+15

1.00E+16

1.00E+17

1.00E+18

-0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Depth (microns)

Ph

osp

ho

rus c

on

cen

trati

on

(ato

ms/c

c)

Nwell Simulation

Nwell SIMS

Figure 3.3: Comparison of Phosphorus Nwell doping concentration profile,

simulated and measured by SIMS.

0.00E+00

1.00E+19

2.00E+19

3.00E+19

4.00E+19

5.00E+19

6.00E+19

0 0.005 0.01 0.015 0.02

Depth (um)

Bo

ron

co

nc

en

tra

tio

n (

ato

ms

/cc

)

PLDD Simulation

PLDD SIMS

Figure 3.4: Comparison of PMOS Lightly Doped Drain (PLDD) Boron doping

concentration profile, simulated and measured by SIMS. Figure 3.4 shows the comparison for PMOS Lightly Doped Drain (PLDD) implant

which is much shallower than the Nwell implant. The maximum error is 1.36%.

Page 63: optimization and characterization of 130 nm cmos transistor design

2) NMOS Transistor

Figure shown below is for NMOS transistor calibration. It includes the Pwell implant,

NSD and NLDD. Figure 3.5 below shows the comparison for Pwell in NMOS. The

maximum percentage difference for this plot is 5.02%.

0.00E+00

1.00E+17

2.00E+17

3.00E+17

4.00E+17

5.00E+17

6.00E+17

-4.00E-01 -2.00E-01 0.00E+00 2.00E-01 4.00E-01 6.00E-01 8.00E-01 1.00E+00

Depth (um)

Bo

ron

ce

nc

en

tra

tio

n (

ato

ms

/cc

)

Pwell SIMS

Pwell Simulation

Figure 3.5: Comparison of Boron doping concentration profile by simulated and measured by SIMS for Pwell.

For Figure 3.6, shows the comparison between SIMS and simulated of phosphorus

concentration for NMOS Source/Drain (NSD). It is shown that the simulated NSD

implant for phosphorus fits almost 1% over depth of 0.06 µm till 0.1-µm. the maximum

percentage difference for this plot is 3.15%.

Page 64: optimization and characterization of 130 nm cmos transistor design

0.00E+00

2.00E+19

4.00E+19

6.00E+19

8.00E+19

1.00E+20

1.20E+20

1.40E+20

1.60E+20

1.80E+20

2.00E+20

0.00E+0

0

1.00E-

02

2.00E-

02

3.00E-

02

4.00E-

02

5.00E-

02

6.00E-

02

7.00E-

02

8.00E-

02

9.00E-

02

1.00E-

01

Depth (um)

Ph

osp

ho

rus c

on

cen

trati

on

(ato

ms/c

c)

NSD SIMS

NSD Simulation

Figure 3.6: Comparison of NMOS Source/Drain (NSD) Phosphorus doping concentration profile, simulated and measured by SIMS.

Figure 3.7 below shows the comparison for Arsenic concentration. The maximum

percentage error for this plot is 11.62%.

0.00E+00

2.00E+20

4.00E+20

6.00E+20

8.00E+20

1.00E+21

1.20E+21

1.40E+21

1.60E+21

1.80E+21

0 0.02 0.04 0.06 0.08 0.1

Depth (um)

Ars

en

ic c

on

ce

ntr

ati

on

(a

tom

s/c

c)

NSD Simulation

NSD SIMS

Figure 3.7: Comparison of Arsenic doping concentration profile by simulated and measured by SIMS for NMOS Source/Drain (NSD).

Page 65: optimization and characterization of 130 nm cmos transistor design

3.5 Validation of TSUPREM-4 and MEDICI

After calibrating the TSUPREM-4 and MEDICI input files, we have to validate the data

that we simulate to make sure the data is comparable to the data measured on real

wafers fabricated at Silterra. We used the device simulator MEDICI to simulate our

transistor to get Id-Vd curves. Table 3.2 shows the standard process for Silterra 0.13 µm

CMOS process. This same process is used both in the real wafer fabrication and also in

our process simulator, TSUPREM-4.

Table 3.2: Standard 0.13 µm CMOS process in Silterra.

Implant steps Standard Recipe

Pwell-Thin-Imp-HE B11, 1.2E+13(Total),260 KeV,Ti-5,Tw-5,Q+ B11, 7E+12(Total), 180 KeV, Ti-5,Tw-5,Q

VTN-Thin-Imp1 B11, 5.5E+12, 40 KeV, Twist 22, Tilt 7 VTN-Thin-Imp2 In115, 4E+12, 100 KeV, Twist 22, Tilt 7

VTN-Thin-Imp3 BF2, 1.1E+12, 40 KeV, Twist 22, Tilt 7

NPT-Thn-G-Imp B11, 1.3E+13, 15 KeV, Twist 0, Tilt 25, Quad

NPT-Thn-G-Imp2 In115, 2.0E+13, 120 KeV, Twist 22, Tilt 30, Quad

NLDD-Thn-Imp As75, 1.5E+15, 3.5 KeV, Twist 0, Tilt 7, Quad

NSD-Imp1 As75, 4.72E+15, 40 KeV, Twist 0, Tilt 7, Quad NSD-Imp2 P31, 6E+14, 10 KeV, Twist 0, Tilt 7, Quad NSD-Imp3 P31, 9.32E+14, 35 KeV, Twist 0, Tilt 7, Quad PSD-Imp1 B11, 2.7E+15, 2.5 KeV, Twist 0, Tilt 7, Quad PSD-Imp2 B11, 5E+13, 11KeV, Twist 0, Tilt 7, Quad

Figure 3.8 below shows the Id-Vd curve for a PMOS transistor that was simulated using

MEDICI device simulator. The Idsat value was taken at Vd = -1.2 V. The Idsat for this

transistor is 230 µA/µm. Figure 3.9 shows the Id-Vd curve for the NMOS transistor that

was simulated in the device simulator MEDICI. The Idsat value for the NMOS transistor

is 531 µA/µm at Vd = 1.2 V.

Page 66: optimization and characterization of 130 nm cmos transistor design

-2.50E-03

-2.00E-03

-1.50E-03

-1.00E-03

-5.00E-04

0.00E+00

-1.4-1.2-1-0.8-0.6-0.4-0.20

Drain voltage, Vd(V)

Dra

in c

urr

en

t, I

d(A

)PMOS simulated drain

current

Figure 3.8: Simulated drain current (Id) versus drain voltage (Vd) using MEDICI for 0.13 µm PMOS transistor.

0.00E+00

1.00E-03

2.00E-03

3.00E-03

4.00E-03

5.00E-03

6.00E-03

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Drain voltage Vd, (V)

Dra

in c

urr

en

t Id

, (A

)

NMOS simulated drain

current

Figure 3.9: Simulated drain current (Id) versus drain voltage (Vd) using MEDICI for

0.13 µm NMOS transistor.

Page 67: optimization and characterization of 130 nm cmos transistor design

Table 3.3: Saturation current, Idsat data between simulation using MEDICI and real wafer measurements.

Transistor Idsat (MEDICI), µA/µm Idsat (electrical

measurement), µA/µm

PMOS 230 240

NMOS 531 542

From the comparison table shown above, we can conclude that MEDICI simulator is

reliable in its use for the experiments that will be explain later in this work. The

percentage error between simulator MEDICI and real wafer measurement for PMOS is

4.17%. The percentage error for NMOS transistor between the simulated data and the

real wafer measurement is 2.03%.

3.6 Summary

In this chapter, we first make a detailed description of the SIMS measurement and, the

analytic model adopted by TSUPREM-4. Some statistical methods, which are useful for

dopant profile calibration, are introduced. A systematic as-implanted profile calibration

methodology is described.

Generally, we used dual-Pearson functions to model an angle-implanted profile. The

calibration of dual-Pearson functions for a specific energy and impurity is divided into

three phases. First, the moment parameters of the primary Pearson function are

optimized to fit the amorphous part of the SIMS profile. Then, the moment parameters

of the secondary Pearson function are optimized to fit the channeled tail of the SIMS

profile. Finally, some key parameters (or all parameters) in the dual-Pearson functions

are varied in a small range and fine-tuned to fit the whole SIMS profile based on the

previous optimization results.

Page 68: optimization and characterization of 130 nm cmos transistor design

These profiles were matched with the doping profile obtained from TSUPREM-4

simulator and it is proved that the simulator is calibrated and reliable to be used further

in this work.

The use of the device simulator MEDICI has also been described and validated in order

to gain confidence in its use and its application to the measurements that will be

described later in Chapter 4.

Page 69: optimization and characterization of 130 nm cmos transistor design

CHAPTER 4

RESULTS

4.1 Investigation of Halo Implant in PMOS and NMOS Device

4.1.1 Introduction

In this chapter the simulation procedure used to study the Halo implant in the PMOS

and NMOS device is described. The simulated results were analyzed using TCAD

process simulation, TSUPREM-4 and device simulation, MEDICI. The simulation

result is then compared to measure data obtained from the real wafer fabricated in

Silterra. The optimized processes for Halo implant were analyzed and were validated to

be used further in Silterra 0.13 µm CMOS process.

4.1.2 Simulation Procedure

Figure 4.1 below shows the process flow for PMOS and NMOS in this work. First, the

current process recipe for 0.13 µm PMOS and NMOS is obtained from Silterra

specifications. Three splits were done for each device in order to see any discrepancies

between the splits. The simulation was run using the same process recipe running in the

real wafer fabrication in Silterra. Process simulator TSUPREM-4 was used to do the

profile analysis. In this analysis, the dopant concentration was clearly seen, current flow

and many other changing parameters that cannot be seen in real device. Then, the

experiment was continued using device simulator MEDICI to see the electrical

characteristics, which are the threshold voltage, Vt and drain current, Id. From here, the

halo dose, energy or tilt angle was adjusted in order to match with the measured data

from the real wafer.

Page 70: optimization and characterization of 130 nm cmos transistor design

Figure 4.1: Workflow for experimental procedure

Obtain Silterra’s 0.13 µm Halo process splits for NMOS and PMOS

Simulate all PMOS halo process splits using TSUPREM-4

Analyzed Halo doping profiles

from TSUPREM-4

Simulate using device simulator,

MEDICI

Compared and match the measured data with simulation

data

Calculate the best dose to meet Silterra’s specification

target

Repeat the step for NMOS

4.1.3 Simulation Results

In this work, effects of dose, energy and tilt angle were analyzed. Table 4.1 below

shows the split done for different dose for halo implant. Arsenic species was used for

PMOS device and boron was used for NMOS device.

Page 71: optimization and characterization of 130 nm cmos transistor design

Table 4.1: Process split for PMOS and NMOS halo implant with various dose.

Halo Implant Samples

(Splits) Dose

(atom/cm2)

Energy

(KeV)

Tilt Angle

(0)

HaloPD1 Arsenic, 3.3E+13 90 30

HaloPD2 Arsenic, 3.1E+13

90 30 PM

OS

HaloPD3 Arsenic, 3.5E+13

90 30

HaloND1 Boron, 1.0E+13 15 25

HaloND2 Boron, 1.3E+13 15 25

NM

OS

HaloND3 Boron, 1.6E+13 15 25

4.1.3.1 PMOS Halo Analysis

After simulating PMOS device with different dose of Arsenic, the doping profile for

PMOS device was plotted using TV2D in TCAD software. This will allow doping

concentration to be plotted and use to detect any discrepancies when the dose is tuned to

find the optimized and correct dose for PMOS Halo.

Based on Figure 4.2 (a), (b) and (c), it can be summarized that by using HaloPD1 dose

is the best-optimized dose for halo implant in this work. Too high dose concentration

can give a distortion to the channel and will short the channel eventually. So it is

suggested here that arsenic with 3.3E+13 atom/cm2 is used for PMOS halo implant.

Page 72: optimization and characterization of 130 nm cmos transistor design

Junction

Arsenic

Arsenic

(a)

(b)

Page 73: optimization and characterization of 130 nm cmos transistor design

Figure 4.2: Arsenic doping profile for three different arsenic doses to optimize the

Halo implant step, (a) HaloPD1 (3.3E+13 atom/cm2), (b) HaloPD2

(3.1E+13 atom/cm2) and (c) HaloPD3 (3.5E+13 atom/cm2).

In sequence to support the result of using the HaloPD1 dose, the PMOS device was

simulated using MEDICI to further investigate the best dose to be used as an optimized

dose in this work. Figure 4.3 below is Idsat versus PMOS halo dose comparing between

measurement and simulation. From this graph, Idsat is taken at Vd = 1.2 V. It is shown

that Idsat for HaloPD1 (3.3E+13 atom/cm2) dose is near to the value measured from real

wafer. According to the experimental result, conclusion can be made that HaloPD1 is

the recommended dose in PMOS implant with the percentage difference from the real

wafer measurement is less than 2%. This dose is suggested, as it is the best dose that

can match and meet Silterra electrical specification target. Figure 4.4 shows the

simulated I-V curve for Lg=0.13 µm transistor at different halo implant doses, 3.3E+13,

3.1E+13 and 3.5E+13. It is found that decreasing halo implant dosage will increase Id,

which implies Vt will roll-off.

Arsenic (Nearly overlap)

(c)

x (µm)

y (µm)

Page 74: optimization and characterization of 130 nm cmos transistor design

-220

-215

-210

-205

-200

-195

-190

-185

-180

3E+13 3.1E+13 3.2E+13 3.3E+13 3.4E+13 3.5E+13 3.6E+13

PMOS Halo Dose (atom/cm2)

Ids

at

(uA

/um

)

Measured

Simulated

Figure 4.3: PMOS Halo dose for 10/0.13 device accuracy plot between measured

and simulated.

-2.5E-03

-2.0E-03

-1.5E-03

-1.0E-03

-5.0E-04

0.0E+00

-1.4-1.2-1-0.8-0.6-0.4-0.20

Drain voltage, Vd (V)

Dra

in c

urr

en

t, I

d (A

)

As 3.1E+13 atom/cm2

As 3.3E+13 atom/cm2

As 3.5E+13 atom/cm2

Figure 4.4: Simulated Id versus Vd curve for different PMOS halo dose, (a) HaloPD1

(3.3E+13 atom/cm2), (b) HaloPD2 (3.1E+13 atom/cm2) and (c) HaloPD3

(3.5E+13 atom/cm2).

Page 75: optimization and characterization of 130 nm cmos transistor design

4.1.3.2 NMOS Halo Analysis

NMOS is one of the devices that have a stable process developing them. For this work,

analysis for NMOS is done only on different dose of boron. The dose for this implant

has to be optimized in order to upgrade the device performance. Note that from

literature, researchers always work on controlling the dose for NMOS halo implant, as

this dose will affect the device electrical characteristics.

For this work, three different boron doses were used for NMOS halo implant. Boron

dose varies from 1.1E+13 atom/cm2 to 1.6E+13 atom/cm2. After simulating the device

with process simulator TSUPREM-4, device profile analysis will be discussed. Figure

4.5 below shows boron concentration for various doses to optimized NMOS halo

process. It can be observed from this figure that with a medium dose used (1.3E+13

atom/cm2), halo structure can be formed perfectly without overlapping source and drain.

Figure 4.6 below shows the current flow line with existence of electric potential.

Current flow line (grey colour) is different for every dose simulated. To get clearer

picture of which dose have a significant effect to the current flowing in the NMOS

device, a cutline is done under each gate of the device simulated. Figure 4.7 shows the

vertical cutline profile of the current density. It is obvious here that different dose might

affect the current density at the drain. The process of characterizing the correct dose that

match the Silterra specification range using the TCAD simulator have been done. The

measured data and simulated data will now be compared as in Figure 4.8.

Page 76: optimization and characterization of 130 nm cmos transistor design

Figure 4.5: Boron concentration for NMOS Halo implants with various doses, (a)

HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and

(c) HaloND3 (1.6E+13 atom/cm2).

In order to further find the right dose to be used for NMOS device that can match the

electrical specification, simulated data for Idsat that was extracted from MEDICI device

simulator was compared with the measured data obtained from real wafer measurement.

Figure 4.8 shows the accuracy plot for NMOS halo implant. The simulated data show a

good agreement with the measured data at HaloND2, which used 1.3E+13 atom/cm2

boron dose. The percentage error for this dose is only 0.2 %. This shows that this is the

right selected dose that shall be used in NMOS halo process in order to meet the

specification target.

(b) Boron,

1.3E+13atom/cm2

(c) Boron,

1.6E+13atom/cm2

Overlapping

Overlapping

Halo

x (µm)

y (µm)

(a) Boron,

1.0E+13atom/cm2

Page 77: optimization and characterization of 130 nm cmos transistor design

Figure 4.6: Current flow line (grey lines) in conjunction with electric potential concentration for different boron doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

Figure 4.7: Current flow vertical cutline align along the poly gate profile for various halo dose, 1.0E+13, 1.3E+13 atom/cm2 and 1.6E+13 atom/cm2.

x (µm)

y (µm)

Boron,

1.0E+13

Boron,

1.3E+13

Boron,

1.6E+13

Current flow line

Current flow line

Current flow line

Vertical cutline

Vertical cutline

Vertical cutline

Page 78: optimization and characterization of 130 nm cmos transistor design

470

480

490

500

510

520

530

540

550

560

9E+12 1.1E+13 1.3E+13 1.5E+13 1.7E+13

NMOS Halo Dose (atom/cm2)

Idsat

(uA

/um

)

Measured

Simulated

Figure 4.8: NMOS Halo dose for 10/0.13 µm device accuracy plot between measured and simulated for 3 different boron doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

4.1.4 Summary

It has been demonstrated in this chapter how simulations can predict the right dose for

implanting PMOS and NMOS halo. Characterization of the doping profile after the

simulation has been done. From this doping profile, the internal part of the transistor is

clearly analyzed and how the dopant diffuses to form the halo structure is shown. From

here the suggested dose is taken from the dose that form the perfect halo structure. It is

important to determine the right structure as it has an important role for controlling the

Idsat variations. After analyzing the simulation profile, electrical characterization is done

and was compared with the real data measured from the real wafer. The simulation Idsat

value agrees with measured for arsenic 3.3E+13 atom/cm2 in PMOS and boron 1.3E+13

atom/cm2 in NMOS. These doses have been selected as it can give the Idsat value that

can meet and match the Silterra specification target.

Page 79: optimization and characterization of 130 nm cmos transistor design

4.2 Investigation of Lightly Doped Drain (LDD) Implant in PMOS

and NMOS Device

4.2.1 Simulation Procedure

More implant parameter will be tuned in Lightly Doped Drain (LDD) for PMOS and

NMOS. Besides using different dose and energy, a new species will be introduced to

this process by using BF2 species in order to find out the optimized process for this

implant step. More variation of parameters needs to tune up because this is the critical

implant step for the CMOS process. Figure 4.9 below shows the process flow for this

implant step. First, the standard process for 0.13 PMOS and NMOS with splits at the

Lightly Doped Drain (LDD) step is analyzed. All split ran in the TSUPREM-4 and

MEDICI simulator is also in real wafer fabrication. Then the profile for this device is

plotted and analyzed.

Figure 4.9: Workflow for experimental procedure

Simulate using device simulator, MEDICI

Compared and match the measured data with simulation data

Calculate the best dose to meet Silterra’s specification target

Repeat the step for NMOS

Obtain Silterra’s 0.13 µm Lightly Doped Drain (LDD) process splits for NMOS and PMOS

Simulate all the PMOS LDD process splits using TSUPREM-4

Analyzed PMOS LDD doping profiles from TSUPREM-4

Page 80: optimization and characterization of 130 nm cmos transistor design

4.2.2 PMOS Simulation Results

In this section, the work done for PMOS device in optimizing the Lightly Doped Drain

(LDD) implant step is discussed. It will cover the process split using Boron Di-Fluoride

(BF2) species and new BF2 energy process split. The new process using BF2 species,

which has many advantages for PMOS device, will be analyzed and discussed further in

this section.

4.2.2.1 Boron Di-Fluoride (BF2) dose splits

From literature it is known that since the device dimension is scaling down, the use of

p+ polysilicon gate to achieve surface channel PMOS becomes indispensable. The p+

gates surface channel PMOS devices offer the advantages of lower threshold voltage

operation, superior short channel effect, and subthreshold leakage characteristics. On

the other hand, since ultra shallow junction is required, BF2 is often used as the implant

species to achieve self-aligned p+ poly gate MOSFET with shallow junctions. However,

fluorine has been found to enhance the boron diffusion from the polysilicon gate

through thin oxide into silicon substrate. Boron penetration results in the instability of

threshold voltage, the decreasing of low-field mobility, and the increasing of

subthreshold slope and electron trapping rate. Hence, boron-penetration through thin

gate oxide and channel is detrimental to the success of surface channel p-MOSFET. To

prevent the penetration of boron in BF2-implanted p-poly gate MOS device, a co-

implant method is used. The appropriate phosphorus dose or arsenic species is used to

retard this penetration. In Silterra process, phosphorus doses has been set high and co-

implanted with BF2, so that it can effectively reduce the boron-penetration without

degrading the poly-depletion.

Page 81: optimization and characterization of 130 nm cmos transistor design

In this section, the work will focus on finding the correct dose and energy for BF2 that

will have a similar manner as boron species used before this new process is

implemented. Table 4.2 shows the implant split table for this work.

Table 4.2: Dose and energy process split for PMOS LDD implant using Boron Di-

fluoride species.

Lightly Doped Drain (LDD) Implant

Samples

(Splits) Dose

Energy

(KeV)

Tilt Angle

(0)

BF2-PD1 BF2,

2.8E+14 3.5 7

BF2-PD2 BF2,

2.2E+14 3.5 7 (a) Dose Splits

BF2-PD3 BF2,

2.5E+14 3.5 7

BF2-PE1 BF2, 2.8E+14

3.0 7

BF2-PE2 BF2,

2.8E+14 3.5 7

(b) Energy Splits

BF2-PE3 BF2, 2.8E+14

4.0 7

Note that BF2 species is compared with boron species, which was used in the earlier

process before BF2 is implemented for PMOS LDD. Boron dose and energy for this

comparison is 2.5E+14 atom/cm2 and 0.7 KeV respectively.

Page 82: optimization and characterization of 130 nm cmos transistor design

Figure 4.10: Doping profile for three different BF2 doses, BF2-PD1 (2.2E+14 atom/cm2), BF2-PD2 (2.5E+14 atom/cm2) and BF2-PD3 (2.8E+14 atom/cm2) compared to boron (2.5E+14 atom/cm2).

Figure 4.10 above shows doping profile for three different BF2 doses. This result is then

compared to the current process for PMOS LDD, which is using boron. It can be

observed from this figure that the device implanted with BF2 diffuse laterally compared

to device that is implanted using boron only. Even though BF2 is implanted with much

more higher implant energy, the boron only diffuses laterally and not into the junction.

It proves that fluorine incorporated in LDD implant is suitable for shallow junction

devices. This is supported in literature [37], which states that fluorine implantation with

B+ for PMOS transistor combined with conventional spike annealing, produce reduced

junction depths.

x (µm)

y (µm)

Page 83: optimization and characterization of 130 nm cmos transistor design

From Figure 4.11 below shows the cut line done on the PMOS device to study the

doping concentration for boron, which plays a major role in this PMOS LDD

implantation. The cut line is done vertically all through the silicon substrate as shown

below.

Figure 4.11: Device vertical cut line example; all through the silicon substrate.

Figure 4.12 below shows the result of the vertical cut line done on PMOS. It can be

seen clearly the benefit of introducing fluorine into the implantation step, it will

decrease the boron diffusion into the silicon substrate. The blue line as indicated in this

figure is without fluorine incorporation, which clearly shows boron diffuses

considerably and produces a junction deeper than 250 nm. The addition of co-implanted

fluorine significantly reduces the boron diffusion and creates a more box-like profile.

The junction depth is only 200 nm.

Cutline 2

x (µm)

y (µm)

Page 84: optimization and characterization of 130 nm cmos transistor design

Figure 4.12: Boron vertical cut line doping profile comparing three different BF2 doses, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) compared to boron (2.5E+14 atom/cm2).

The results from TSUPREM-4 simulators will be the input to the device simulator

MEDICI. Section below will analyze various internal device phenomena such as

potential, charge and current distributions and also device thermal behavior; (threshold

voltage and current-voltage characteristics). Figure 4.13 shows the current flow profile

for PMOS device with different LDD BF2 dose doping and compared to without

fluorine implant. There is a significant change in current flow for high dose and low

dose implant. The right energy must be identified in order to match the characteristics

with the current process that use non-BF2 implant. Hypothesis can be made that the

current flow pattern shows similar manner between BF2 dose BF2-PD1 (2.8E+14

atom/cm2) and the non-fluorine incorporated, boron with 2.5E+14 atom/cm2 dose.

Page 85: optimization and characterization of 130 nm cmos transistor design

Figure 4.13: Current flow profile (grey lines) for PMOS device with different LDD dose, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) compared with without fluorine implant.

To make the results more reliable, cutline profile of the current flow for PMOS device

was done as in Figure 4.14. From the cutline profile in Figure 4.15, it shows how the

non-fluorine implant (blue line) agrees and overlapping with BF-PD1 dose respectively.

This proves that by using BF2-PD1 dose (2.8E+14 atom/cm2), it can give a similar

electrical characteristics manner as the previous process using non-fluorine incorporated

implant. It also gives more advantages to the process development as the fluorine

incorporated implant is proved to have a good advantage over shallow junction devices.

Figure 4.16 shows another cutline profile done at the same place under the gate edge.

This profile shows current density is high for non-corporate fluorine dopant.

x (µm)

y (µm)

Page 86: optimization and characterization of 130 nm cmos transistor design

Figure 4.14: Current flow vertical cutline example done at the drain of the PMOS device.

Figure 4.15: Current flow vertical cutline profile comparing various BF2 dose, (a) BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) with non-fluorine species dose (boron species).

Current

flow

Cut line

x (µm)

y (µm)

Page 87: optimization and characterization of 130 nm cmos transistor design

Figure 4.16: Current density vertical cutline profile comparing various BF2 dose, (a)

BF2-PD1 (2.2E+14 atom/cm2), (b) BF2-PD2 (2.5E+14 atom/cm2) and (c) BF2-PD3 (2.8E+14 atom/cm2) with non-fluorine species dose (boron species).

In order to prove the accuracy of the simulated electrical data with the measured data,

the Idsat accuracy plot is shown in Figure 4.17 below. The critical parameter such as Idsat

has to be constant, as it will affect the performance of the device. It is proven in the

figure below that the suitable dose to match the previous process that use non-fluorine

incorporated implant is at the 2.8E+14 atom/cm2 BF2 dose. The use of BF2 in this step

has proved to have the same effect as the previous process, which used boron in PMOS

LDD step. The usage of BF2 also will not affect any electrical characteristics for this

device. As stated before, the co-implanted BF2 with heavy dose of phosphorus will

reduce the boron-penetration without degrading the poly-depletion effect.

All these results imply that the proper BF2 dose to be used in order to replace the

previous process is the BF2-PD1 (2.8E+14 atom/cm2) dose.

Page 88: optimization and characterization of 130 nm cmos transistor design

-265

-260

-255

-250

-245

-240

2.1E+14 2.2E+14 2.3E+14 2.4E+14 2.5E+14 2.6E+14 2.7E+14 2.8E+14 2.9E+14

PMOS LDD Dose (atom/cm2)

Ids

at

(uA

/um

)

Measured

Simulated

Figure 4.17: PMOS LDD dose for 10/0.13 device accuracy plot between measured and simulated.

4.2.2.2 Boron Di-Fluoride (BF2) energy splits

Since suggested dose to be used for BF2 has been determined, the implant energy for

BF2 has to be analyzed in order to find the right energy. As stated in Table 4.2, 3

different energies will be used in this section, 3.0 KeV, 3.5 KeV and 4.0 KeV.

TSUPREM-4 was run for process simulation analysis. In this section, the comparison

between 3 different energies was made with the current implant process energy that

used boron species with 0.7 KeV implant energy.

Figure 4.18 shows the three different BF2 implant energy compared with boron species

that used 0.7 KeV implant energy. It can be seen clearly that BF2 species used higher

energy, as it is much heavier species than boron. From the analysis plot, it shows that

even BF2 is implanted using much more higher implant energy; the concentration stays

more at the upper part of the transistor source and drain. This profile proves that BF2,

Page 89: optimization and characterization of 130 nm cmos transistor design

although it is much more heavier and implanted with a high energy than boron, but with

the fluorine incorporation, it will make the boron concentration diffuse laterally. This is

a very important aspect for a shallow junction device that was used today.

In order to strengthen the claim that fluorine incorporation gives great advantages to

device performance, a cutline is done all through the drain to justify this. Figure 4.19

shows PMOS device boron concentration cutline. From this cutline, it proves BF2 gives

great advantage for a shallow device process, as it will stop boron from diffusing further

into the drain. This boron penetration or diffusion will make the electrical

characteristics of the device change, such as shifting Vt and Idsat.

Figure 4.18: Doping profile for three different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared to boron (0.7 KeV).

x (µm)

y (µm)

Page 90: optimization and characterization of 130 nm cmos transistor design

Figure 4.19: Vertical cutline on boron concentration profile for 3 different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared with non-fluorine corporation doping, boron (0.7 KeV).

After characterizing process simulation profile, device simulator, MEDICI will continue

this analysis. Current flow lines will be check in order to match with boron species with

0.7 KeV implant energy. Figure 4.20 shows the current flow and electric potential

profile for PMOS device. It can be seen clearly that the predicted energy to be used in

order to replace boron with BF2 for PMOS LDD implant is by using BF2 with 3.5 KeV

implant energy. The current flow line from BF2 with 3.5 KeV implant energy matches

well with boron profile. This implant energy gives the value that can meet the Silterra

specification target.

Page 91: optimization and characterization of 130 nm cmos transistor design

Figure 4.20: Current flow cutline profile and electric potential for three different BF2

energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared with boron (0.7 KeV).

To support this prediction, the total current cutline is done on this device. One cutline is

done just under the gate for each device as shown in Figure 4.21. The reason to do this

cutline is to see which of the split match with the current from 0.7 KeV boron implant

energy. From Figure 4.22, it can be predicted that the best energy that matches well

with the current process using 0.7 KeV boron implant energy is by using 3.5 KeV BF2

implant energy. The 3.5 KeV energy curve is the nearest to the current process. So this

strengthens the prediction to suggest that in order to replace the current PMOS LDD

process with BF2, 3.5 KeV implant energy for BF2 should be used.

The comparison between simulation and measured data is presented in Figure 4.23.

From this plot, the simulated data shows a good agreement with measured data.

Hypothesis can be made that the predicted energy (3.5 KeV) matches well with the

x (µm)

y (µm)

Page 92: optimization and characterization of 130 nm cmos transistor design

measured data. This energy is suggested to be used as it matches well with the

measurement data and it gives a value that can meet the targeted value in the Silterra

specification.

Figure 4.21: Total current profile with the vertical cutline under the gate edge for

three different BF2 energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compared with non-fluorine corporation doping, boron (0.7 KeV).

x (µm)

y (µm)

Page 93: optimization and characterization of 130 nm cmos transistor design

Figure 4.22: PMOS total current cutline profile with different LDD energy, (a) BF2-PE1 (3.0 KeV), (b) BF2-PE2 (3.5 KeV) and (c) BF2-PE3 (4.0 KeV) compare with boron (0.7 KeV).

-285

-280

-275

-270

-265

-260

-255

-250

2.9 3.1 3.3 3.5 3.7 3.9 4.1

PMOS LDD Energy (KeV)

Ids

at

(uA

/um

)

Measured

Simulated

Figure 4.23: PMOS LDD energy for 10/0.13 µm device accuracy plot between measured and simulated data.

Page 94: optimization and characterization of 130 nm cmos transistor design

4.2.3 NMOS Simulation Results

As the gate length of MOS device is scaled down, issues related to short channel effects

become increasingly important. Some new LDD are designed for better control of short

channel effects and hot carrier injection. Shallow junction is required to control short

channel effects. A simple method is to replace phosphorus with arsenic in LDD implant

for shallow junction. Actually, the reduction of short channel roll-off due to arsenic

LDD cannot compensate the roll-off caused by channel length scaling. On the other

hand, sharper grading, which arises at the edge of arsenic LDD junction, can degrade

hot carrier reliability. Hence profile optimization becomes essential to achieve

performance improvement.

NMOS LDD performance is evaluated by comparing different dose for this implant.

Since NMOS device is widely known as a stable device compared to PMOS,

characterization in order to optimize this process is done just by varying the dose for

LDD implant. As shown is Table 4.3, 2 different arsenic dose is used in order to tune

this process to match with the measured data without disturbing 2D simulation results

and electrical characteristics.

Table 4.3: Dose process split for NMOS LDD implant using arsenic species.

Lightly Doped Drain (LDD) Implant

Samples

(Splits) Dose

(atom/cm2)

Energy

(KeV)

Tilt Angle

(0)

As-NLDD1 Arsenic, 1.5E+15

3.5 7

(a) Dose Splits

As-NLDD2 Arsenic, 1.3E+15

3.5 7

Page 95: optimization and characterization of 130 nm cmos transistor design

From Figure 4.24 below, lightly doped drain structure is clearly visible in the

TSUPREM-4 2D simulation profile. LDD junction will form just at the edge of the

device gate, and under the spacer is the source/drain junction implant. As stated in

literature, this is the cross section of the drain end of an LDD device, showing the

sidewall spacer, the gate overlap length (Lg) and the length of the lightly doped drain

region (Ln-). LDD structure can reduce the maximum electric field along the channel

significantly and can also change the position of the peak lateral electric field. The

lifetime of LDD MOSFET will increased accordingly.

Figure 4.24: Doping profile for two different arsenic doses, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

From the figure above, vertical cutline is done just under the gate edge of the NMOS

device. From this cutline, arsenic concentration can be analyzed between these two

different doses. From Figure 4.25, it is clearly shown that there is no significant effect

between this two dose used for this analysis. The higher dose shows higher

concentration and the concentration curve nearly overlap or act similarly at the channel.

This shows that arsenic species doesn’t change much the doping concentration profile.

The correct dose to be used for this implant will be verified in the next analysis.

(a) As-NLDD1 (b) As-NLDD2

Gate Gate

LDD structure x (µm)

y (µm)

Page 96: optimization and characterization of 130 nm cmos transistor design

Figure 4.25: Vertical cutline under gate for two arsenic concentrations, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

Since from the Figure 4.25, no suitable dose has been proposed, the MEDICI profile is

analyzed. Figure 4.26 shows the current flow line for two different doses used.

Hypothesis can be made that using different dose can change the pattern of the current

flow lines. Arsenic with a higher dose have a better current flow lines just at the

junction, but using the lower dose arsenic, the flow lines seems to be distorted. Vertical

cutline is done under the gate of the device and plotted as in Figure 4.27. It can be seen

from the plot that can be noted that by using higher dose, it can change the current

density drastically and from this, it can be conclude that the higher dose has a better

performance as it produce a better linear current flow line compared with lower dose.

Page 97: optimization and characterization of 130 nm cmos transistor design

Figure 4.26: Current flow line profile for different arsenic dose used, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

Figure 4.27: Current flow lateral cutline profile for different NMOS LDD dose, (a) As-NLDD1 (1.5E+15 atom/cm2) and (b) As-NLDD2 (1.3E+15 atom/cm2).

Gate Gate

Source Drain Source Drain

x (µm)

y (µm)

(a) (b)

Page 98: optimization and characterization of 130 nm cmos transistor design

In order to support the simulated data, the comparison of simulated and measured data

for Idsat comparing both two doses was plotted in Figure 4.28. It can be observed that by

using As-NLDD1 with 1.5E+15 atom/cm2 arsenic doses; the Idsat value is near to the

measured data. It is suggested that best dose to be used in NMOS LDD flow is by using

arsenic with 1.5E+15 atom/cm2 concentrations, as it is also the right dose in order to

meet Silterra specification range.

490

495

500

505

510

515

520

525

530

535

1.25E+15 1.3E+15 1.35E+15 1.4E+15 1.45E+15 1.5E+15 1.55E+15

NMOS LDD dose (atom/cm2)

Ids

at

(uA

/um

)

Measured

Simulated

Figure 4.28: NMOS LDD dose for 10/0.13 µm device accuracy plot between

measured and simulated data.

4.2.4 Summary

The usage of BF2 in PMOS LDD implant step is discussed in this chapter with all the

advantages of using this dopant species is described. The best BF2 energy and dose that

matches well with Silterra specification is recommended for this process. The best

energy is at 3.5 KeV with the dose 2.8E+14 atom/cm2. As for NMOS LDD step,

different arsenic dose is discussed and described. It is recommended that the best dose

to be used is 1.5E+15 atom/cm2 with 3.5 KeV implant energy.

Page 99: optimization and characterization of 130 nm cmos transistor design

4.3 Investigation of Source/Drain (S/D) Implant in PMOS and

NMOS Device

4.3.1 Simulation Procedure As stated in the previous chapter, this chapter will discuss source/drain implant in

CMOS device. The source/drain implantation is a high current, low energy

implantation process. It is probably the last implantation process in the IC chip

fabrication processes. One of the significant differences from the LDD implantation is

that the dosage of the source/drain implantation is much higher, and it is performed

after the sidewall spacer formation. The sidewall spacer separates the heavily doped

source/drain slightly from the channeling area right underneath polysilicon gate and

helps to suppress the hot electron effect. As stated in the previous section, the same

procedure will be done on source/drain implantation process. The same splits will be

ran in the TSUPREM-4 process simulator and also in the real wafer fabricated in

Silterra. The workflow for the procedure is as in Figure 4.29.

Page 100: optimization and characterization of 130 nm cmos transistor design

Figure 4.29: Workflow of experimental procedure for source/drain process simulation.

Simulate all the source/drain process splits using

TSUPREM-4

Analyzed source/drain doping profiles from TSUPREM-4

Analyzed and find the optimized source/drain

process

Simulate using device simulator,

MEDICI

Compared and match the measured data with simulation

data

Repeat the step for NMOS

Obtain Silterra’s 0.13 µm source/drain process splits

for NMOS and PMOS

4.3.2 Simulation Results

For source/drain analysis, different boron dose will be tuned to fine the best dose that

should be used for PMOS source/drain implant. In other hand, for NMOS device, the

characterization is only done on tuning up the energy, as it is too difficult to find the

correct energy for arsenic implant. The parameter values that are used in this

characterization is summed up in Table 4.4.

Page 101: optimization and characterization of 130 nm cmos transistor design

Table 4.4: Process split for PMOS and NMOS source/drain implant with various implant parameters.

Source/drain Implant Samples

(Splits) Dose

(atom/cm2)

Energy

(KeV)

Tilt Angle

(0)

S/D1 Boron, 2.6E+15

2.5 7

S/D2 Boron, 2.7E+15 2.5 7

PMO

S

S/D3 Boron, 2.8E+15

2.5 7

S/D_energy1 Arsenic, 2.7E+15

40 7

S/D_energy2 Arsenic, 2.7E+15

45 7

NM

OS

S/D_energy3 Arsenic, 2.7E+15 50 7

4.3.2.1 PMOS Source/Drain Analysis

PMOS device is simulated using TSUPREM-4. The adjust parameter is as shown in

Table 4.4. Boron dose is varied from 2.6E+15 atom/cm2 to 2.7E+15 atom/cm2 and

2.8E+15 atom/cm2. The same energy and tilt angle is used for the splits to eliminate any

effect on these parameters. Energy used in PMOS simulation is 2.5 KeV and implant tilt

angle is at 70. Boron dose is tune up because the electrical characteristics will be

affected if the wrong dose is used. Results obtained from TV2D in TSUPREM-4 are

presented in Figure 4.28 and will be discussed further in analysis section in chapter 5.

Figure 4.30 shows the net doping and depletion region that was plotted using TV2D.

The distribution of doping is quite similar between three doses used but the depletion

region seems to change. At boron dose 2.6E+15 atom/cm2, the depletion region is wider

and this will affect the current flowing at the junction of PMOS device. This will lead to

device degradation, as the I-V characteristics for PMOS device changes. For boron

dosage 2.8E+15 atom/cm2, the doping concentration is concentrated at the channel of

Page 102: optimization and characterization of 130 nm cmos transistor design

the device as shown in the figure below. The suggested dose to be used in this analysis

is at 2.7E+15 atom/cm2 as the doping profile and depletion is in the right condition. I-V

characteristics will be discussed after the following figure.

Figure 4.30: Net doping and depletion region line for three different boron doses for PMOS source/drain process, (a) S/D1 (2.6E+15 atom/cm2), (b) S/D 2 (2.7E+15 atom/cm2) and S/D 3 (2.8E+15 atom/cm2).

After simulating the device, the electrical characterization of the device is analyzed.

Figure 4.31 shows the Id-Vd curve for PMOS device with different dosage used. In this

plot, Idsat value is taken at Vd=1.2 V. Idsat value for different dose for PMOS device is in

the range from 268 µA/µm to 283 µA/µm. This value is slightly higher compared to

Silterra PMOS Idsat typical value. The best dose to be used in this process is predicted

from the closest value of Idsat, which is 254 µA/µm at Vd=1.2 V. This value is for boron

at dose 2.7E+15 atom/cm2. This result also agrees with the simulation result above that

predicts to use boron at this dose. Too high dose or too low will affect the Idsat and will

(a) 2.6E+15

atom/cm2

(b) 2.7E+15

atom/cm2

(c) 2.8E+15

atom/cm2

x (µm)

y (µm)

Page 103: optimization and characterization of 130 nm cmos transistor design

effect the device performance. It will degrade PMOS device and short channel effect

will occur.

-4.00E-03

-3.50E-03

-3.00E-03

-2.50E-03

-2.00E-03

-1.50E-03

-1.00E-03

-5.00E-04

0.00E+00

-2.5-2-1.5-1-0.50

Drain Voltage, Vd (V)

Dra

in C

urr

en

t, I

d (

uA

/um

)

Boron 2.6E+15 atom/cm2

Boron 2.7E+15 atom/cm2

Boron 2.8E+15 atom/cm2

Figure 4.31: Id-Vd curve for PMOS source/drain implant process with different dosage used for this process, (a) S/D1 (2.6E+15 atom/cm2), (b) S/D 2 (2.7E+15 atom/cm2) and S/D 3 (2.8E+15 atom/cm2).

These results will be clarified by comparing simulation data with the real silicon data

measured in Silterra. Figure 4.32 shows the comparison plot. At boron dose 2.7E+15

atom/cm2, the percentage difference for simulation and measured data is only 0.08%.

This is the lowest percentage error compared with the other doses, 2.6E+15 atom/cm2

and 2.8E+15 atom/cm2, which have 0.13% and 0.18% respectively.

Page 104: optimization and characterization of 130 nm cmos transistor design

-213

-212.5

-212

-211.5

-211

-210.5

-210

-209.5

-209

2.55E+15 2.60E+15 2.65E+15 2.70E+15 2.75E+15 2.80E+15 2.85E+15

PMOS Source/Drain dose (atom/cm2)

Ids

at

(uA

/um

)

Measured

Simulated

1

Figure 4.32: PMOS S/D dose for 10/0.13 µm device accuracy plot between measured and simulated data.

From the percentage error values, suggestion can be made to use boron for PMOS

source/drain implant with the dose 2.7E+15 atom/cm2 as it is also the dose that can meet

the Silterra specification range. It can be noticed that the Idsat value for 2.7 E+15 is low

compared to the other doses. In this case, it can be concluded that there is no significant

effect can be seen within a 1% percentage error.

4.3.2.2 NMOS S/D Analysis

For NMOS source/drain implantation process, different arsenic implant energy is used.

Three splits are done for this step, which are 40, 45 and 50 KeV. In this splits, the

arsenic dose and tilt angle used is at 4.72E+15 atom/cm2 and 70 respectively. This is the

normal condition for NMOS source/drain implant step. Tilt angle can gives significant

difference to the value of Idsat and threshold voltage, which will be discussed further in

this work.

Page 105: optimization and characterization of 130 nm cmos transistor design

The same procedure to run this experiment for source/drain splits is implemented. The

process simulation in TSUPREM-4 is run using the same process recipe as in the

process running in Silterra fab. For electrical characterization, MEDICI is still used in

order to plot the I-V characteristics. These results will then be clarified with the

measurement silicon data.

Figure 4.33: Current flow profiles for 0.13 µm NMOS device with 3 different source/drain implant energy, (a) S/D_energy1 (45 KeV), (b) S/D_energy2 (50 KeV) and S/D_energy3 (40 KeV).

Figure 4.33 above shows the current flow profile for 0.13 µm NMOS device with

different implant energy applied. The similar trend of current flow is shown here. There

is a slightly small difference between those three energies used. For arsenic with

implant energy 50 KeV, the simulation shows that there is a huge current flowing at the

device drain. But for 40 and 50 KeV, the current flow line looks smaller. This is also

proved by the current density cutline profile that is done along the poly gate as in Figure

4.34. It can be noted here that only small differences can be detected when different

energy is varied for source/drain implant.

Page 106: optimization and characterization of 130 nm cmos transistor design

Figure 4.34: Current density cutline profiles for 0.13 µm NMOS device with 3 different source/drain implant energy, (a) S/D_energy1 (45 KeV), (b) S/D_energy2 (50 KeV) and S/D_energy3 (40 KeV).

The prediction of which implant energy that should be used for NMOS source/drain

process will further be analyzed by looking at the I-V characteristics that was simulated

using MEDICI.

0.00E+00

1.00E-03

2.00E-03

3.00E-03

4.00E-03

5.00E-03

6.00E-03

7.00E-03

8.00E-03

0 0.5 1 1.5 2

Drain Voltage, Vd (V)

Dra

in C

urr

en

t, I

d (

uA

/um

)

40 KeV

45 KeV

50 KeV

Figure 4.35: Drain current, Id versus drain voltage, Vd for NMOS device simulated using MEDICI for 3 different implant energy, (a) S/D_energy1 (45 KeV), (b) S/D_energy2 (50 KeV) and S/D_energy3 (40 KeV).

Page 107: optimization and characterization of 130 nm cmos transistor design

Figure 4.35 above shows the simulated drain current versus drain voltage (Id vs Vd)

plots. In this plot, Idsat value can be determined and will be compared with measurement

data. Idsat value is taken at Vd=1.2 V. In this experiment, the value of Idsat varied from

570 µA/µm to 623 µA/µm. This value for Idsat is still in the Silterra NMOS Idsat range,

which is (455 µA/µm – 635 µA/µm). Idsat value for 50 KeV is slightly out of the range,

which it is assume that this device that was simulated is catastrophic. This is because,

Idsat will play a major role in determining the device performance, and if the Idsat is too

high, it will results in device degradation. Conclusion can be made that the best

source/drain implant energy that can be used for NMOS is between 40 and 45 KeV. The

Idsat value for this two energy is 507 µA/µm and 603 µA/µm respectively.

To further clarify this problem, the comparison of Idsat between simulation and

measurement is presented in Figure 4.36. In this figure, the correct energy for

source/drain process can be suggested. The best energy shall be determined by the

percentage error between simulation and measurement. Simulation data must agree with

measurement data in order to implement this suggestion in the real process. From this

figure, the percentage error for 40 and 45 KeV is 0.32% and 0.15% respectively. From

this error, suggestion to used 45 KeV implant energy for source/drain can be verified.

Page 108: optimization and characterization of 130 nm cmos transistor design

556

558

560

562

564

566

568

570

572

574

39 41 43 45 47 49 51

NMOS S/D Energy (KeV)

Ids

at

(uA

/um

)

Measured

Simulated

Figure 4.36: NMOS S/D energy for 10/0.13 µm device accuracy plot between measured and simulated data.

4.3.3 Summary

In this section, the source/drain implant process has been discussed thoroughly. For

PMOS device, implant dose is varied to tune and find the optimum dose for this step.

The suggested dose to be used for PMOS source/drain step is 2.7E+15 atom/cm2. For

NMOS device, a split is done on various implant energy. From the results shown above,

it is suggested to used implant energy at 450. The best S/D implant dose and energy that

suggested here, is the one that meets Silterra specification target.

Page 109: optimization and characterization of 130 nm cmos transistor design

CHAPTER 5

ANALYSIS AND DISCUSSION

5.1 Introduction

This chapter explains the complex results obtained in Chapter 4 in terms of basic

principles. The first section will cover halo implant for PMOS and NMOS transistors.

Discussion of the lightly doped drain (LDD) for PMOS and NMOS transistor will be in

the second section and the last section will analyze the source/drain implant for PMOS

and NMOS.

5.1.1 Analysis and Discussion of Halo Implant Experiment results in

PMOS and NMOS Transistors

A process called “pocket implant” or “halo implant” was introduced in 1985 [50]. The

halo implant is a quad implant that creates a high doping concentration in the channel

near the source/drain junctions. This process improves the short channel performance of

deep submicron MOSFETs. It is widely used to reduce Vt roll-off and punch-through

[40]. But if the dose is too high, it will cause undesirable reverse short channel effect

(RSCE) [41]. Hence, careful tuning of the halo implant parameters is always necessary

in order to obtain the desired electrical performance. PMOS and NMOS halo implant

effects will be explained in this section including simulated and measured TSUPREM-4

profiles and electrical characterization.

5.1.1.1 PMOS Transistor

The purpose of this section is to explain how halo implant parameters influence the

performance of a 130 nm physical gate length PMOSFET. The dose, tilt angle and

implant energy will effect the Vt roll-off and drain saturation current, Id. As in literature,

Page 110: optimization and characterization of 130 nm cmos transistor design

halo implant is employed in the advanced “conventional” MOSFETs in order to

improve the transistor performance. This halo dopant species is introduced only in a

selected local region; namely at the edge of the drain-extension region.

Figure 5.1: Two different transistors size with halo implant. It is shown here that when the gate length becomes small, the two halo implants will overlap and therefore Vt will increase. This is called the “reverse” short channel effect, RSCE. Normally (without halo) Vt should decrease when the gate length becomes smaller.

As shown in Figure 5.1, halo can affect Vt and Vt roll off when the gate length is small.

Halo will overlap each other and increase Vt. In other hand, halo will have very little

effect on Vt in a large gate length transistor. Figure 5.2 shows the relation of Vt with

gate length. As the gate length is reduced, the two halo implants will start to overlap on

each other and Vt will roll up. Eventually, Vt will roll off as the punchthrough effect

comes into play, where LDD implants start to overlap on each other.

Page 111: optimization and characterization of 130 nm cmos transistor design

Figure 5.2: Threshold voltage, Vt versus gate length. This shows the effect of halo to Vt with a different gate length.

Next, the effect of halo on junction capacitance will be briefly discussed. Figure 5.3

shows junction capacitance, Cj, value that changes due to different depth of halo

implant. Halo will increase Cj value a lot if halo is deeper than source/drain implant. For

halo that is shallower, Cj value is not affected.

Figure 5.3: Transistor with different halo depth. Note that for a shallower halo, junction capacitance, Cj will not be affected. For halo deeper than source/drain implant, Cj will increase.

Table 4.1 in the previous chapter gives the arsenic halo implant dose splits for PMOS

transistors. In this section, further analysis has been made to analyze and find the effect

of this dose to the transistor performance. It can be observed from Figure 5.4, which is

Vt

Gate Length, L

Begin punchthrough

Left and right halo overlap (double dose)

Cj Cj

Source

Gate

Drain

P-sub

Deep halo Shallow halo

Page 112: optimization and characterization of 130 nm cmos transistor design

the simulated Vt roll-off curve for different doses used in PMOS halo, that too much

halo implant dose will produce excessive RSCE, which is undesirable for circuit design.

-0.5

-0.45

-0.4

-0.35

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0 0.1 0.2 0.3 0.4 0.5 0.6

Drawn Gate Length (um)

Thre

shold

Voltage (V

t)

As 3.3E+13 atom/cm2

As 3.1E+13 atom/cm2

As 3.5E+13 atom/cm2

Figure 5.4: PMOS halo Vt roll-off curve for different arsenic doses, (a) HaloPD1 (3.3E+13 atom/cm2), (b) HaloPD2 (3.1E+13 atom/cm2) and (c) HaloPD3 (3.5E+13 atom/cm2). The bumps in the curve happen because each point is simulated with different gate length, so it is a grid or numerical error.

Figure 5.5 below shows threshold voltage (Vt) that was measured using a tester (Agilent

4072 and Keithley S680) and prober (Tokyo Electron Limited, TEL P-8 and Electro

Glass, EG). In this plot, with lower halo dose, Vt is roll-off speedily as the channel

lengths become shorter. Here, the Vt is defined as the gate voltage, Vg, at which the

drain current, Id, reaches 0.1 µA per µm of gate width. Although the Vt roll-off can be

improved at higher halo implant dose, nevertheless such a high Vt will not meet the

performance requirements. With halo implant dose 3.3E+13 atom/cm2, it can effectively

improve Vt roll-off and hold the threshold voltage at about 0.38 V, which is desirable

for 0.13 µm transistors.

RSCE

Page 113: optimization and characterization of 130 nm cmos transistor design

Figure 5.5: Measured PMOS threshold voltage (Vt) curve for 10/0.13 µm transistor. Measured data and Simulation Program with Integrated Circuit Emphasis (SPICE) model are shown.

Table 5.1 shows the threshold voltage (Vt) value for measured data and simulated data

for the same wafer that is run with the same recipe.

Table 5.1: Threshold voltage simulated using MEDICI and measured value for different PMOS halo implant dose at L=0.13 µm.

Doses 3.1E+13

atom/cm2

3.3E+13

atom/cm2

3.5E+13

atom/cm2

Measured threshold voltage, Vt (V)

0.376 0.387 0.409

Simulated threshold voltage, Vt (V)

0.301 0.393 0.433

Page 114: optimization and characterization of 130 nm cmos transistor design

0.2

0.25

0.3

0.35

0.4

0.45

0.5

3E+13 3.1E+13 3.2E+13 3.3E+13 3.4E+13 3.5E+13 3.6E+13

PMOS halo dose (atom/cm2)

Th

resh

old

vo

ltag

e (

Vt)

Measured

Simulated

Figure 5.6: Measured and simulated PMOS threshold voltage, Vt for three different halo doses. The smallest difference is seen at halo dose 3.3E+13 atom/cm2.

It can be seen in Figure 5.6, the measured value for arsenic with 3.3E+13 atom/cm2

implant dose are in good agreement with the simulated value using MEDICI.

Percentage error between these two values is 1.5%. This value is suggested as it is the

right dose that can meet the specification target.

There are 3 independent input parameters; dose, tilt angle/rotation and energy. Above

studies are based on a single variable, implant dose. The other 2 variables implant tilt

angles/rotation and implant energy have also been studied. It is assumed that the

variables are independent (no interactions) in this work, so only single-variable

experiments are implemented. Figure 5.7 shows Vt roll-off curve for different halo

implant energy. This split only ran on 2 energy splits, 90 and 110 KeV, in the real wafer

fabrication process. For simulation, 70 KeV is introduced to see the effect of the halo

using the lowest implant energy. This curve has a similar trend to the dose splits. In this

experiment, the best energy to meet the Vt target as in Silterra’s specification is at 90

KeV. This is because by using 90 KeV energy, threshold voltage value obtained is in

Page 115: optimization and characterization of 130 nm cmos transistor design

Silterra specification range. This is the desired threshold voltage for this device as

higher Vth value may affect the device performance. The summary for energy tuning is

shown is Table 5.2 and the comparison graph between measured and simulated is

plotted in Figure 5.8. The evaluations are continued by tuning the tilt angle to see how it

can affect the Vt for the transistor.

-0.35

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0 0.1 0.2 0.3 0.4 0.5 0.6

Drawn Gate Length (um)

Th

res

ho

ld v

olt

ag

e (

Vth

)

70KeV

90KeV

110KeV

Figure 5.7: PMOS Vt roll-off for different halo implants energy, 70, 90 and 110 KeV.

Table 5.2: Measured and simulated data for Vt with different PMOS halo implant

energy at L=0.13 µm.

Implant Energy 70 KeV 90 KeV 110 KeV

Measured threshold voltage, Vt (V)

Not Available 0.254 0.300

Simulated threshold voltage, Vt (V)

0.230 0.248 0.296

It can be seen in this plot that the simulation data is far off from the measured value.

Measured Vt value for 70 KeV is not available in Silterra split table obtained. In spite of

Page 116: optimization and characterization of 130 nm cmos transistor design

that, the Vt trend can be seen in the plot. As the implant energy is increased, Vt value is

also increased.

0.2

0.22

0.24

0.26

0.28

0.3

0.32

0.34

60 70 80 90 100 110 120

PMOS halo implant energy (KeV)

Th

res

ho

ld v

olt

ag

e (

Vt)

Measured

Simulated

Figure 5.8: Measured and simulated PMOS threshold voltage, Vt for different halo implants energy, 70, 90 and 110 KeV.

As for different tilt angle splits, Vt data that obtained from measurement is compared

with simulation data. Table 5.3 shows the Vt measurement for halo implant with

different tilt angle. By using different tilt angle, it gives no significant change of the

threshold voltage. As for simulated data, the same trend is shown. There is only a small

change in threshold voltage for each tilt angle used as shown in Figure 5.9.

Table 5.3: Measured and simulation data for Vt with different halo implant tilt angle at L=0.13 µm.

Tilt Angle (0) 20 25 30

Measured threshold voltage, Vt (V)

0.288 0.285 0.281

Simulated threshold voltage, Vt (V)

0.261 0.259 0.251

Page 117: optimization and characterization of 130 nm cmos transistor design

0.2

0.22

0.24

0.26

0.28

0.3

0.32

0.34

18 20 22 24 26 28 30 32

PMOS halo tilt angle (0)

Th

resh

old

vo

ltag

e (

Vt)

Measured

Simulated

Figure 5.9: Measured and simulated PMOS threshold voltage, Vt with different halo implants tilt angle 200, 250 and 300 with rotation=0.

Figure 5.9 above shows the difference of threshold voltage, Vt between measured and

simulated. The difference between the measured value and simulated value is in the

range of 9%. The trend for this work can be seen, as the halo implant tilt angle is

increased, Vt value will decrease. This is because as the tilt angle decreased, the halo

will be formed too far away from the channel. This will not affect the Vt value. In other

hand, if too high tilt angle is used, halo will form too close to the channel, thus effecting

Vt value. This can be illustrated clearly in Figure 5.10.

Figure 5.10 shows different implant tilt angle that can affect transistor threshold

voltage, Vt. It is shown here that too high tilt angle for example, 800 will make the halo

formed too close to the channel. For too low tilt angle, the halo will form too far away

from the channel. When the halo is form too far from the channel, it will not give any

effect to the transistor.

Page 118: optimization and characterization of 130 nm cmos transistor design

Figure 5.10: Transistor with different halo implant tilt angle and rotation=0. Normal tilt is at 300.Too high tilt angle (800) resulting halo forming too close to the channel. Too low tilt angle (50), halo is too far from the channel.

The structure is then characterized by using TV2D in TSUPREM-4. The results are

shown in Figure 5.11 and 5.12. There are significant changes that can be seen at the

PMOS channel when low tilt angle (200), medium tilt angle (250) and high tilt angle

(300) are used. The doping concentration spreads all over the channel as this can effect

the current flowing, as the main purpose of halo implant is to stop the leakage current

from drain to source when the transistor is off.

From this figure, it is shown that at a higher tilt angle value, resist shadowing problem

occurs. The halo implant can’t form perfectly at the channel. The halo overlaps each

other in the channel. This can be explained from Figure 5.13, which explains more

about the resist shadowing problem.

Page 119: optimization and characterization of 130 nm cmos transistor design

Figure 5.11: PMOS net doping profile for different implant tilt angle (200, 250 and 300).

Figure 5.12: PMOS halo doping profile for different implant tilt angle, (200, 250 and 300). The halo formed too closed to the channel when the implant tilt is too high and halo is too far from the channel when the implant tilt angle is too low.

x (µm)

y (µm)

x (µm)

y (µm)

Page 120: optimization and characterization of 130 nm cmos transistor design

Figure 5.13: Transistor showing resists shadowing problem. Noted that if the tilt angle is larger than 270, there will be no halo in the transistor. Photoresist thickness is 0.60 µm and space from gate is 0.30 µm.

Figure 5.13 above shows the resist shadowing effect that happens when ions is

implanted on a transistor. Before implant, photoresist is used to block any implant to be

implanted in the wrong place. High-energy implant requires thick photoresist to mask

the penetration of high-energy ions. In most cases, the ion will be implanted into a

wafer at a tilt angle to minimize channeling. It is common practice to implant with 70

tilt angle, with 20 degrees rotation to avoid dopant channeling. However, halo implant

must be done with a larger angle so shadowing from photoresist will not be a potential

problem. As in the figure above, the shadowing creates shadow region at the channel. If

tilt angle is larger than ~270, there will be no halo in the transistor. Table 5.4 below

summarizes halo effect on Vt and Id for PMOS.

Page 121: optimization and characterization of 130 nm cmos transistor design

Table 5.4: Effects of halo implant parameters on threshold voltage, Vt and saturated drain current, Id for PMOS.

Implant Parameters Threshold voltage, Vt Saturated Drain

Current, Id

Higher dose Higher Lower

Higher implant energy Higher Lower

Larger tilt angle Higher Lower

5.1.1.2 NMOS Transistor

Extensive studies of varying implant dose, tilt angle and energy have also been taken

for NMOS transistors. The experiments were single variable, without combinations. As

stated in Table 4.1 in chapter 4, three different Boron doses are used for characterizing

NMOS transistors. Figure 5.14 shows the simulated Vt roll-off curve when different

dose is used to tune NMOS halo implant process.

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Drawn Channel Length (um)

Th

resh

old

Vo

ltag

e (

V)

Boron 1.0E+13

Boron 1.3E+13

Boron 1.6E+13

Figure 5.14: NMOS Vt roll-off for three different halo implant doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

NMOS halo implant dose shows results similar to PMOS halo implant. Higher dose will

lead to higher threshold voltage; Vt. Too high dose can make the halo catastrophic and

will lead to RSCE and junction leakage. Too much halo dose will not only make the

Page 122: optimization and characterization of 130 nm cmos transistor design

junction leak, but will produce low breakdown voltage or band-to-band tunneling also

called Zener effect leakage. This can be illustrated in Figure 5.15 below.

(a) (b) Figure 5.15: High dose halo doping on a PMOS transistor, (a) breakdown voltage

versus doping concentration [51] and (b) band-to-band tunneling or Zener effect leakage [51].

In this case, by using boron with 1.3E+13 will give the Vt that is closest to the target of

Silterra specifications. Figure 5.16 shows the measured silicon data. In this plot,

1.3E+13 atom/cm2 boron will have the Vt value closest to the Silterra target. The

comparison between these two data is shown in Table 5.5.

Page 123: optimization and characterization of 130 nm cmos transistor design

0.2

0.22

0.24

0.26

0.28

0.3

0.32

0.34

0.36

0.38

0.1 1 10

Length (um)

NM

OS

Th

res

ho

ld V

olt

ag

e (

Vth

@

VD

=0

.1V

)

Vtlin_HP(1.0E+13)

Vtlin_HP(1.3E+13)

Vtlin_HP(1.6E+13)

Figure 5.16: Measured NMOS Vt roll-off for different halo implants doses, (a) HaloND1 (1.0E+13 atom/cm2), (b) HaloND2 (1.3E+13 atom/cm2) and (c) HaloND3 (1.6E+13 atom/cm2).

As in Table 5.5, it is shown that the measured silicon data for 1.3E+13 atom/cm2 boron

is comparable with the simulated data. The percentage difference is only 1.76%

compared to other dose that is used in this experiment.

Table 5.5: Measured and simulation data for Vt with different NMOS halo implant dose at L=0.13 µm.

Doses 1.0E+13 atom/cm2 1.3E+13 atom/cm

2 1.6E+13 atom/cm

2

Measured threshold voltage, Vt (V) 0.320 0.341 0.364

Simulated threshold voltage, Vt (V)

0.310 0.335 0.347

Page 124: optimization and characterization of 130 nm cmos transistor design

0.25

0.27

0.29

0.31

0.33

0.35

0.37

0.39

0.41

9.0E+12 1.0E+13 1.1E+13 1.2E+13 1.3E+13 1.4E+13 1.5E+13 1.6E+13 1.7E+13

NMOS halo dose (atom/cm2)

Th

resh

old

vo

ltag

e, (V

t)

Measured

Simulated

Figure 5.17: Measured and simulated NMOS threshold voltage, Vt with different halo implants dose, HaloND1 (1.0E+13 atom/cm2), HaloND2 (1.3E+13 atom/cm2) and HaloND3 (1.6E+13 atom/cm2).

In order to fine-tune the halo step, halo implant tilt angle splits and halo implant energy

splits are also tuned accordingly on NMOS transistors in single-variable experiments.

The results for NMOS transistor are similar to PMOS transistor. Figure 5.18 shows the

simulated Vt roll-off for different NMOS halo implant energies, while Table 5.6 shows

measured and simulated threshold voltage, Vt values for different implant tilt angles and

implant energies.

Table 5.6: Measured and simulation data for Vt with different NMOS halo implant energy at L=0.13 µm.

Implant Energy 10 KeV 12 KeV 15 KeV

Measured threshold voltage, Vt (V) 0.311 0.338 0.357

Simulated threshold voltage, Vt (V)

0.313 0.369 0.374

Page 125: optimization and characterization of 130 nm cmos transistor design

0.3

0.31

0.32

0.33

0.34

0.35

0.36

0.37

0.38

0.39

9 10 11 12 13 14 15 16

NMOS halo implant energy (KeV)

Th

resh

old

vo

ltag

e (

Vt)

Measured

Simulated

Figure 5.18: Measured and simulated NMOS threshold voltage, Vt with different halo

implant energy, 10, 12 and 15 KeV.

As shown in figure above, the average percentage difference between simulated and

measured value is 4.86%. The best implant energy to match with the targeted

specification is at 10 KeV with the smallest percentage difference (0.64%) . Figure 5.19

below shows the simulated threshold voltage that is obtained from MEDICI simulator.

0.2

0.22

0.24

0.26

0.28

0.3

0.32

0.34

0.36

0.38

0.4

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Drawn Channel Length (um)

Th

resh

old

vo

ltag

e (

V)

10 KeV

12 KeV

15 KeV

Figure 5.19: Simulated NMOS Vt roll-off for different halo implants energy, 10, 12 and 15 KeV.

Table 5.7 shows the measurement and simulation data for NMOS halo with different

implant tilt angles. The average percentage difference between simulation and

Page 126: optimization and characterization of 130 nm cmos transistor design

measurement value is slightly bigger with the difference of 7.11%. The data is tabulated

in Figure 5.20. The expected value that can match with the specification is at 250, which

have 5% percentage difference.

Table 5.7: Measured and simulation data for Vt with different NMOS halo implant

tilt angle at L=0.13 µm.

Tilt Angle (0) 15 25 35

Measured threshold voltage, Vt (V)

0.288 0.320 0.351

Simulated threshold voltage, Vt (V)

0.311 0.339 0.377

0.25

0.27

0.29

0.31

0.33

0.35

0.37

0.39

0.41

0.43

12 17 22 27 32 37

NMOS halo tilt angle (0)

Th

resh

old

vo

ltag

e (

Vt)

Measured

Simulated

Figure 5.20: Measured and simulated NMOS Vt roll-off for different halo implants tilt angle, 150, 250 and 350. This shows that larger tilt angles increase the threshold voltage.

In this section, simulated results of how the halo implants doses, tilt angles and energy

influences the characteristics of sub-micron MOSFET is reported. The conclusion is

that properly tuned halo implant can effectively suppress the short channel effect (SCE)

in MOSFETs with 130 nm physical gate length and maintain the desired level of

threshold voltage. Measured data have been taken to compare with this simulation data

and the results are summed up in tables at the end for each section.

Page 127: optimization and characterization of 130 nm cmos transistor design

5.1.2 Analysis and Discussion of Lightly Doped Drain (LDD) Implant

in PMOS and NMOS Transistor

Lightly doped drain (LDD) is another commonly used structure that can reduce hot

electron problems in sub-micron transistors by lowering the peak longitudinal electric

field. We can form the LDD junction by using low energy, low current ion

implantation. It is a shallow junction with a very low dopant concentration, extended

just underneath the gate. In this section, PMOS and NMOS LDD structures will be

analyzed further from the results obtained in chapter 3. Simulated data will be compared

with measured silicon data for verification.

5.1.2.1 PMOS Transistor

As stated in Figure 4.3 in chapter 4, dose split and energy split has been done for PLDD

process in order to meet the Id and Vt specification targets. Before using boron di-

fluoride (BF2) as indicated in Figure 4.3, previous process used boron for this step. In

this section, PLDD species changes from boron to BF2 will be further analyzed. It is

reported that, to achieve higher throughput and better negative bias temperature

instability (NBTI) lifetime, BF2 species usage is promising [52].

Page 128: optimization and characterization of 130 nm cmos transistor design

Figure 5.21: Threshold voltage, Vt for 1.2 V 10/0.13 µm PMOS LDD for boron di-fluoride (BF2) and boron (B11). This shows that B11 and BF2 can produce very similar Vt.

The comparison of threshold voltage between boron (B11) and boron di-fluoride (BF2)

is shown in Figure 5.21. There is almost no difference between using boron in the

former process and by using BF2 in the current process. Threshold voltage for 1.2 V

PMOS 10/0.13 µm for both BF2 and boron are quite similar. The analysis will go

further in investigating the difference of using BF2 and boron. Figure 5.22 shows the

zoom-in for threshold voltage value between BF2 and boron species. It can be seen

clearly that ~5 mV lower threshold voltage is observed for BF2 implant. From this,

conclusion can be made that different species (BF2 and boron) did not give much effect

on threshold voltage. BF2 can be used to replace boron species in PMOS LDD process

as it can give many advantages as well as disadvantages to the transistor performance.

Silterra is changing the process technology from boron to BF2 is PMOS LDD step in

order to reduce Negative Bias Temperature Instability (NBTI) effect and BF2 is useful

to form ultra-shallow junction in small-scale device. The boron penetration effect that

happen when BF2 is applied can be stop by co-implanting BF2 with higher dose of

Page 129: optimization and characterization of 130 nm cmos transistor design

phosphorus. This is done to retard the boron penetration thus eliminating the device

degradation effect.

Table 5.8: Advantages and disadvantages of BF2 implant in CMOS transistor.

Advantages Disadvantages

1 Improves NBTI performance [52]. Formation of silicide is worse on BF2 implanted surface [52].

2 Reduces oxide leakage current in polyoxide [53].

Junction leakage is generally higher for BF2 implanted surface [52].

3 Used for ultra-shallow junction formation [54].

Enhances boron diffusion into SiO2/Si [30].

Figure 5.22: Threshold voltage, Vt difference for 1.2 V 10/0.13 PMOS for both boron

di-fluoride (BF2) and boron (B11). The difference is about 5 mV.

To continue the analysis, Idsat is measured from the wafer and was plotted box plot as in

Figure 5.23 for 10/0.13 µm PMOS. It is shown in this plot that Id values between this

two species have the same characteristics. Analysis goes further and found that a

marginal of (~ 4 µA/µm) degradation of Idsat observed on changing thin PLDD species

from boron to BF2 as shown in Figure 5.24. This is also assumed to be a minor impact

on the saturation drain current and it will give no big effect on the transistor. Conclusion

Page 130: optimization and characterization of 130 nm cmos transistor design

can be made that the used of BF2 will be implemented in current PLDD process and as

many wafers have been evaluated, the results obtained are excellent.

Figure 5.23: Saturation drain current, Idsat for 1.2 V 10/0.13 PMOS for both boron di-fluoride (BF2) and boron (B11). The average difference between BF2 and B11 is only 4 µA.

After determination of BF2 usage as the standard species in PLDD process, simulation

of different dose and energy of BF2 is done as in Section 4.2.3 in chapter 4. Results and

analysis has been done and the best dose and energy that meets the specification target

is already determined as in previous chapter.

Page 131: optimization and characterization of 130 nm cmos transistor design

Figure 5.24: Idsat comparison for 1.2 V 10/0.13 PMOS for both BF2 and boron. It can

be noted Idsat degrades (~ 4 µA/µm) after changing to BF2 species. 5.1.2.2 NMOS Transistor

Since in the measured silicon data, only two splits were done on NLDD, simulation is

done on different implant energy and implant tilt angle in order to see any effect that

can change the transistor performance by tuning these parameters. In chapter 4, it had

been shown that the arsenic splits for NMOS LDD implant can change the current flow

line and furthermore, it will affect the drain current, Id, for the transistor. The best dose

to meet the spec target is 1.5E+15 atom/cm2.

In this analysis, tilt angle and energy will be tuned by using the TCAD simulator and

the threshold voltage obtained for these splits shall be analyzed in order to find the right

tilt angle and implant energy that should be used in this process. The splits that have

been done are shown in Table 5.9.

Page 132: optimization and characterization of 130 nm cmos transistor design

Table 5.9: Tilt angle and energy splits for NMOS transistor simulation using TCAD. Fixed parameters: dose at 1.5E+15 atom/cm2, rotation = 0 degrees.

Samples

(Splits)

Tilt Angle (0) 0 3 7

Energy (KeV) 2 3 4

Figure 5.25 below shows the Vt roll-off curve for different tilt angles used in the NMOS

LDD process. From this graph, the best tilt angle can be predicted according to the

threshold voltage value obtained. The higher the tilt angle causes Vt to shift up and can

cause degradation to the transistor performance. The best tilt angle that is matched with

Silterra specification is by using 70 tilt angle. This result is further verified by the

existing works [30], which stated that larger tilt angle of LDD implant should be

adopted to get reduced parasitic capacitance and enhanced transistor performance, but

with a proper dose adopted.

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Drawn Channel Length (um)

Th

resh

old

vo

ltag

e (

V)

0 degree

3 degree

7 degree

Figure 5.25: Simulated NMOS Vt roll-off for different LDD implants tilt angle, 00, 30 and 70. Higher angle causes higher Vt.

Page 133: optimization and characterization of 130 nm cmos transistor design

0.15

0.2

0.25

0.3

0.35

0.4

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Drawn Gate Length (um)

Th

resh

old

vo

ltag

e (

V)

2 KeV

3 KeV

4 KeV

Figure 5.26: Simulated Vt roll-off curves comparing 3 different implant energy used for NMOS LDD process, 2, 3 and 5 KeV. Higher energy causes higher Vt.

From Figure 5.26 above, 3 different implant energies are used to tune the LDD process.

In this work, 2, 3 and 4 KeV implant energies are used. From the Vt roll-off curve, it is

clearly shown that the Vt value shifts tremendously when 4 KeV energy is used to

implant arsenic to create an LDD channel. Too high energy can degrade the transistor

performance [29]. At higher arsenic LDD implantation energies, deep source/drain

junction regions were locally implanted with arsenic, resulting in larger junction depth.

In this study, the best LDD NMOS implant energy should be used is in between 3 to 4

KeV. Too low energy resulting in Vt value did not match with the Silterra specification.

In this case, my conclusion is to use 3 KeV implant energy as it can give Vt value 0.28

V for L=0.13 µm, which is in the Silterra’s specification range of Vt for NMOS

(Min=0.27 V, Typ=0.33 V, Max=0.40 V).

Page 134: optimization and characterization of 130 nm cmos transistor design

5.1.3 Analysis and Discussion of Source/Drain (S/D) Implant in PMOS

and NMOS Transistor

Source/drain implant is commonly known also as highly doped drain (HDD) implant. It

is implanted after the sidewall spacers step. Sidewall spacers will be used alongside the

poly gates to prevent the higher source/drain (S/D) implant from penetrating too close

to the channel where S/D punch through could occur. In order to complement the

retrograde well implant techniques, medium-dose implants are made to penetrate the

silicon slightly beyond LDD junction depth, but not as deep as the original twin-well

implants. Retrograde well is an approach to well formation in CMOS structures; highest

concentration of dopant (implanted) in the well is located at a certain distance from the

surface. The result is a denser circuit and less susceptible to latch-up. The objective of a

retrograde well is to get a minimum doping in the channel, so that better mobility

caused by less collisions can be achieved. It also can stop the punchthrough effect. In

the S/D implant step, the spacer oxide from the previous step will protect the channel

from the dopant atoms during the implant process.

5.1.3.1 PMOS Transistor

S/D implant is a medium energy implant step that penetrates the silicon deeper than the

LDD junction depth. Spacer oxide prevents the boron dopant that is used in PMOS

transistor from encroaching into the channel. For PMOS source/drain analysis, as

shown in Table 4.4 in chapter 4, boron dose is split into 3 experiments: 2.6E+15

atom/cm2, 2.7E+15 atom/cm2 and 2.8E+15 atom/cm2. Different dose of boron can give

a slight impact on Idsat (as shown in chapter 4) and in threshold voltage, Vt that is be

discussed in this chapter. The main effect is on the sheet resistance of the source and

drain.

Page 135: optimization and characterization of 130 nm cmos transistor design

In analyzing source/drain implant, all the other implants (Vt adjust implant, halo and

LDD) are included. Simulation results in Figure 5.27 below show with the help of Vt

implant and the correct dose of halo implant, indicates that the source/drain implant

dose can be correctly determined.

From this result, it is suggested to use boron at 2.7E+15 atom/cm2 dose. Although the Vt

value looks significantly higher compared to the other two doses, but at this dose only

the Vt value is comparable to Silterra’s specification. The existence of halo/pocket

implant can act as a stopper that effectively resists the punchthrough from drain to

source.

-0.35

-0.3

-0.25

-0.2

-0.15

-0.1

0 0.1 0.2 0.3 0.4 0.5 0.6

Drawn Gate Length (um)

Thre

shold

voltage (V

t)

Boron 2.6E+15 atom/cm2

Boron 2.7E+15 atom/cm2

Boron 2.8E+15 atom/cm2

Figure 5.27: Simulated Vt roll-off curve comparing 3 different implant doses used for NMOS S/D process, 2.6E+15 atom/cm2, 2.7E+15 atom/cm2 and 2.8E+15 atom/cm2.

The sidewall of LDD region is not surrounded by n-type dopant, thus less p-dopant is

compensated. Source/drain series resistance does not increase much and the transistor

will suffer less current degradation. On the other hand, the depth of p-extension region

is typically less than half of the source/drain junction depth and contributes less to short

channel because the depletion region in the vertical direction has no impact on short

Page 136: optimization and characterization of 130 nm cmos transistor design

channel effect as shown in Figure 5.28. At the overlap region where PMOS S/D and n-

halo implant meet, the concentration of the n-region is 2 to 3 orders of magnitude lower

than that of p+ region. Then at S/D, compensation by n-halo is not significant and the

increase in parasitic S/D junction capacitance is negligible, which would otherwise be a

serious problem.

(a) (b)

1) Boron 2.7E+15 atom/cm2

2) Boron 2.6E+15 atom/cm2

x (µm)

y (µm)

x (µm)

y (µm)

Gate Gate

Gate Gate

Depletion region

Depletion region

Page 137: optimization and characterization of 130 nm cmos transistor design

3) Boron 2.8E+15 atom/cm2

Figure 5.28: Simulated PMOS structure showing (a) net doping profile and depletion region and (b) electric potential and current flow lines for three different S/D implant doses.

5.1.3.2 NMOS Transistor

The steps of creating the NMOS source/drain region are similar to PMOS S/D

formation. After the S/D implant, the implanted wafer is annealed in a rapid thermal

process (RTP) tool. This step is important to activate the dopants while preventing

structures from spreading by diffusion. In this experiment for NMOS transistors, 3 splits

were done using different implant energy values (40, 45 and 50 KeV) as shown in Table

4.4 in chapter 4. In chapter 4, only a small significant change at the transistor profile is

detected. In this analysis chapter, Vt roll-off characteristic will be characterized in order

to see if any short channel effect occurs in the NMOS transistor.

In Figure 5.29 below, it is shown that low implant energy (40 KeV), combined with a

proper halo implant and Vt adjust in the earlier process step gives excellent Vt roll-off

characteristics. This Vt value is also within our Silterra specification.

x (µm)

y (µm)

Gate Gate

Depletion region

Page 138: optimization and characterization of 130 nm cmos transistor design

0.2

0.22

0.24

0.26

0.28

0.3

0.32

0.34

0.36

0.38

0 0.1 0.2 0.3 0.4 0.5 0.6Drawn Gate Length (um)

Th

res

ho

ld v

olt

ag

e (

V)

VTN1(40)

VTN2(45)

VTN3(50)

Figure 5.29: Simulated NMOS Vt Roll-off curves for three different source/drain implant energies, (a) 45 KeV (b) 40 KeV (c) 50 KeV. Higher energy gives higher Vt.

Figure 5.30: Simulated NMOS structure showing depletion region for three different source/drain implant energies, 40, 45 and 50 KeV.

When the transistor is off (i.e., zero gate bias), the channel region under the gate is

depleted to a certain depth. Low NMOS implant energy is used. The junction depth of

the n-region under the gate is less than half of the depth of depletion region. Thus, the

Page 139: optimization and characterization of 130 nm cmos transistor design

charge in n-region under the gate will contribute less to the charge sharing effect with

the channel, which is the origin of short-channel effect. Under such condition, charges

in NMOS source/drain play a dominant role on short-channel effect. Vertical cutline is

done on this transistor in order to see the dopant distributions when low and higher

energy is used. Cutline profile is shown in Figure 5.31. From this figure, it shows that

higher energy will produce deeper junction and will create larger depletion region for

the transistor. From this explanation, the perfect implant energy that can meet Silterra

specification is suggested, that is 40 KeV.

Figure 5.31: Net doping vertical cutline profile for NMOS source/drain implant with different implant energy. When higher energy is used, dopant will diffuse further to the drain and creates deeper junction.

5.2 Summary

In this chapter, all halo, LDD and S/D implant results that were presented in chapter 4 is

extensively analyzed and explained. The best parameter that should be used is

determined by the impact that it gives to the NMOS and PMOS transistor. Electrical

characteristics especially Vt roll-off is discussed further in order to find the limits of the

Page 140: optimization and characterization of 130 nm cmos transistor design

implant parameters used. All the suggested work is applied to Silterra 0.13 µm CMOS

current process and it is proved to improve the transistor performance.

Page 141: optimization and characterization of 130 nm cmos transistor design

CHAPTER 6

CONCLUSION AND RECOMMENDATION FOR FUTURE WORK

This chapter is divided into 3 parts. The first part describes the conclusion of the

research followed by the second part, which deals with the recommendation for future

work. The last part is the summary of this work.

6.1 Conclusion

Process and device simulation are extensively used in this study. The TCAD simulator

that was used in this work must be calibrated before quantitatively accurate results can

be obtained. In this work, the drain engineering work in CMOS 130 nm technology is

discussed and analyzed. It covers all PMOS and NMOS halo implant, lightly doped

drain (LDD) and lastly source/drain (S/D) implant. All simulated data is compared with

the measured silicon data on the wafer that is fabricated in Silterra.

From the simulation results for PMOS halo implant that used arsenic with different

doses, it is suggested that in order to get a better Idsat value and match with the measured

data, arsenic with dose 3.3E+13 atom/cm2 is used. On the other hand, for NMOS

device, it is suggested that the boron dose to be used in halo implant is at 1.3E+13

atom/cm2. This dose also agrees with the measured data as shown in Figure 4.8, with

the percentage error ~0.2%.

As for the lightly doped drain implant step, the same analysis is done for PMOS and

NMOS device. There are two sets of results used to compare the usage of the previous

Page 142: optimization and characterization of 130 nm cmos transistor design

implant dopant, boron (B11) with boron di-fluoride (BF2). BF2 is introduced in this step

since it has many advantages to the quarter-micron technology, such as it can reduce the

NBTI and increase the device lifetime [26, 36-37]. All the consequences and differences

of changing these two dopants are described in chapter 4 and 5. The result from the

simulation and also supported with measured data suggest that, BF2 can be used in

PMOS LDD implant step to replace boron that was implemented in the previous

process. The variations of Id and Vt from these two processes are in the same manner.

The suggested BF2 dose that can meet the specification target is 2.8E+14 atom/cm2,

with fixed energy at 3.5 KeV. Implant energy that should be used in this similar step is

also analyzed and compared with boron. The suggested energy to be used is at 3.5 KeV.

By using high energy as suggested by other work, NBTI effect and short channel effect

that always happen in short channel devices can be reduced. For NMOS LDD step, only

two arsenic doses is used and analyzed. Since NMOS transistor is a stable device in

Silterra, the tuning process for this device is only to determine the suitable arsenic dose

in NMOS LDD step. The proposed arsenic dose is 1.5E+15 atom/cm2, with fixed

energy at 3.5 KeV.

Source/drain implant (S/D) is a heavier implant compared to LDD. It has less impact

compared to LDD as it is located further away from the device channel and contributes

less effect to the short channel device. But tuning up this process step is still important,

as it is one of the main implant which creates source and drain junction in the PMOS

and NMOS device. For PMOS device, various boron doses are used to tune up this

process step. Medium dose is suggested for this step as it can give better Idsat value and

also this value is comparable to Silterra specification. The suggested boron dose is at

2.7E+15 atom/cm2. Different energy is used to characterize NMOS S/D process step. 45

Page 143: optimization and characterization of 130 nm cmos transistor design

KeV implant energy is suggested to be used for this implant, as the Idsat value produced

is comparable to Silterra specification.

In this work, 0.13 µm TCAD deck is calibrated using SIMS data for both PMOS and

NMOS device. The TSUPREM-4 and MEDICI is then validated to be use further in this

work. Halo, lightly doped drain (LDD) and source/drain (S/D) implant is simulated

using the same process recipe splits that runs in the real wafer fabricated in Silterra.

Electrical characteristics, Idsat and Vt are then characterized in order to see the effect that

occur by changing the implant parameters such as dose, energy and implant tilt angle to

the PMOS and NMOS devices. The implant parameter is then tuned and optimized in

order to match with Silterra specification target. An optimized implant parameters for

each of the implant step; halo, lightly doped drain (LDD) and source/drain (S/D) are

then suggested to the process engineer for future used in Silterra. The objectives for this

work is achieved by having a calibrated 0.13 µm TCAD deck and the optimized implant

parameters to be used further in sub micrometer CMOS. It is also understood that by

using TSUPREM-4 and MEDICI as the process and device simulator in this work, the

behavior of the finished structure under multitude of process recipe and biasing

conditions can be analyzed.

6.2 Limitation and Recommendation The result of this work is limited by a few factors. As mention earlier, the accuracy of

the simulated results is reduced since some models used in the simulation are

uncalibrated. Thus, further calibration of simulator is necessary to ensure reliable

simulation results. A simple yet useful methodology can be adopted to calibrate the 2D

profiles. Since SIMS profiles are expensive and measured device characteristics are

relatively easier to obtain, it is possible to calibrate the impurity profiles by matching

short channel and reverse short channel effects.

Page 144: optimization and characterization of 130 nm cmos transistor design

Another source of error that need to be deal with for future work in TCAD calibration is

to accurately measure poly line width and oxide thickness to the necessary resolution.

Many important MOSFET electrical characteristics are first-order related to these

parameters (e.g., a 1% change in either Lgate or Tox will result in approximately a 1%

change in Idsat), thus these parameters need to be characterized thoroughly to get

accurate results.

Grid factor is also a limitation. In this work the experiment is initially used a grid factor,

which is not dense enough. This cause simulation results to be inaccurate, especially at

short channels. Consequently, device performance parameters for 0.13 µm and 0.18 µm

could not be obtained and for other channel lengths, values were illogical. Thus, a new

experiment, with a better grid factor can be simulated and corresponding result will be

more meaningful. A comparison of the two grid factors is shown in Figure 6.1.

The next limitation of this work is that the analysis is only based on a few variations of

implant dose, energy and tilt. However, the advantage of using TCAD simulator is that

a new experiment can be easily duplicated and simulated for other values of implant

dose, energy and tilt.

Page 145: optimization and characterization of 130 nm cmos transistor design

Figure 6.1: (a) Denser grid factor, (b) less dense grid factor.

6.3 Summary

In summary, TSUPREM-4 and MEDICI for 0.13 µm Silterra’s Process Technology in

TCAD deck has been calibrated and can be further used by the process engineer to

analyzed and verify Silterra’s process variations. Three engineering structure; halo

implant, lightly doped drain (LDD) implant and source/drain (S/D) implant has been

simulated and characterized. By implementing certain value of implant dose, implant

energy and implant tilt angle gives the best saturation drain current, Idsat and threshold

voltage, Vth value that also meet the Silterra’s specification range. By using process and

device simulator such as TCAD, it is proved that the correct process parameter can be

determined by using TCAD. The inner part of a transistor can also be seen and analyzed

to debug the transistor performance and speed; hence the device performance can be

upgraded as the channel length shrink in the future.

Page 146: optimization and characterization of 130 nm cmos transistor design

REFERENCES

1 Wanlass, F.M., and C. T. Sah, Nanowatt logic using field-effect metal-oxide semiconductor triodes, (1963). International Solid State Circuit Conf, Philadelphia, Pa.; Digest of Technical Papers, Feb 1963, pp 32-33.

2 Mary Bellis, (2006). Inventors of the Modern Computer. http://inventors.about.com/library/weekly/aa080498.htm

3 Xiao, H. (2001). Introduction to Semiconductor Manufacturing Technology. New Jersey: Prentice Hall.

4 M.M. Atalla, “ Semiconductor Devices Having Dielectric Coatings“, U.S. Patent 3,206,607, filed Mar.8, 1960, issued Sept.14, 1965.

5 D. Kahng, “Electric Field Controlled Semiconductor Device”, U.S. Patent 3,102,230, filed May 31, 1960, issued Aug. 27, 1963.

6 N. Kamal, (2005), Fabrikasi & Pencirian Elektrik Peranti PMOS 0.24 µm, BSc.

Thesis, Malaysia: Universiti Kebangsaan Malaysia.

7 S. A. Chew, (2003), Optimization of Rapid Thermal Processing in 0.13 micron

CMOS Poly Silicon Gate, MSc. Thesis, Malaysia: Universiti Teknologi Petronas. 8 A.F. Abdul Rahim (2002), Simulation, Fabrication and Electrical

Characterization of Silicon Bipolar Transistor and Design of SiGe

Heterojunction Bipolar Transistor, MSc. Thesis, Malaysia: Universiti Sains Malaysia.

9 Peggy Gui, (2004) VLSI Design and Lab Fall.

http://engr.smu.edu/~pgui/class_notes_pdf/EE7356-05%20IC%20manufactureing.pdf

10 Quirk, A., Serda, J. (2001). Semiconductor Manufacturing Technology, New Jersey: Prentice Hall.

11 Klaasen, et.al. (1984) “Scaling of Compensated MOSFETs Towards Submicron Dimensions”, IEDM Tech. Digest, p.613.

Page 147: optimization and characterization of 130 nm cmos transistor design

12 .M. Sze. (1981). Physics of Semiconductor Devices, 2nd edn. New York: John Wiley and Sons.

13 R. Simonton and F. Sinclair, (1992). “Ion Implantation Applications”, In J.

Ziegler (ed), Handbook of Ion Implantation Technology, Amsterdam: Elsevier Publ. (p.282).

14 Muhlhoff, et.al., (1987) “Submicron P-MOSFETs Under Static and Swap Stress”, Symp. VLSI Tech., Digest of Tech. Papers, p.57.

15 K. Kaga, et. al., (1998). “Effects of Lightly Doped Drain Structure With Optimum Dose on P-channel MOSFETs”, IEEE Elec. Dev. Lett., Vol.35 #12,

p.2384.

16 S. Odanaka, et.al., (1985). “A New Half-Micron P-channel MOSFET With Efficient Punchthrough Stops”, Symp. VLSI Tech., Digest of Tech. Papers, p.62.

17 M. Brassington, et.al., (1988). “An Advanced Submicron BiCMOS Technology for VLSI Applications”, Symp. VLSI Tech., Digest of Tech. Papers, p.62.

18 S. Ogura, et.al., (1982). “A Half-micron MOSFET Using Double Implanted LDD”, Symp. VLSI Tech., Digest of Tech. Papers, p.42.

19 M. Miyake, et.al., (1990). “Sub-quarter-micrometer Gate Length p-channel MOSFETs With Shallow Boron Counter-doped Layer Fabricated using Channel Preamorphization”, IEEE Trans. Elec. Dev., Vol.37 #9, p.2007.

20 K. Tsuda, et.al., (1987). “0.5 Micron CMOS Technology for 5.6 Nanosecond High Speed 16x16 Bit Multiplier”, Symp. VLSI Tech., Digest of Tech. Papers,

p.109.

21 K. Cham, et.al., (1984). “Device Design for the Submicrometer P-channel FET with N+ Polysilicon Gate”, IEEE Elec. Dev. Lett., Vol. EDL-31, p.964.

22 T. Hori, et.al., (1988). “A New P-channel MOSFET With Large Tilt Implanted Punchthrough Stopper (LATIPS)”, IEEE Elec. Dev. Lett., Vol.9 #12, p.641.

23 T. Hori, (1994). “A 0.1 mm CMOS Technology With Tilt Implanted Punchthrough Stopper (TIPS)”, IEDM Tech. Digest, p.75.

Page 148: optimization and characterization of 130 nm cmos transistor design

24 J.F. Ziegler (ed) (2004), Ion Implantation Science and Technology, Ion

Implantation Technology. Amsterdam: Elsevier Publ. (p.282).

25 Jiong-Guang Su, Shyh-Chyi Wong, Chi-Tsung Huang, Chang-Ching Cheng, Chih-Chiang Wang, Shiang Huang-Lu, Bing-Yui Tsui, (1997). “Tilt Angle Effect on Optimizing HALO PMOS Performance”, Simulation of

Semiconductor Processes and Devices (SISPAD), p.33 – 36.

26 Fu-Cheng Wang, Constantin Bulucea, (2000). “BF2 and Boron Double-

Implanted Source/Drain Junctions for Sub 0.25 µm CMOS Technology”, IEEE Electron Device Letters, Vol.21, No.10, October 2000.

27 K.Y. Kam, Y.M. Son (2000), TCAD Synthesis Approach to Deep Submicron

MOSFET Design and Optimization, Singapore: Nanyang Technological University.

28 Wen-Kuan Yeh, Jih-Wen Chou, (2001). “Optimum Halo Structure for Sub-0.1 µm CMOSFETs”, IEEE Transaction on Electron Devices, Vol.48, No.10,

October 2001.

29 C. Chen, C.Y. Chang, (2000). “Optimization of Short Channel Effect by Arsenic

P-Halo implant Through Polysilicon Gate for 0.12um P-MOSFET”, Electron

Devices Meeting. Proc. 2000 p.44 – 47.

30 K.K. Bourdelle, H.-J.L. Gossmann, S. Chaudry, A. Agarwal, (2001).”The Effect of Fluorine from BF2 Source/Drain Extension Implants on Performance of PMOS Transistors with Thin Gate Oxides”, IEEE Electron Device Letters,

Vol.22, No.6, June 2001.

31 Hajime Kurata, Toshihiro Sugii, (1998). “Self-Aligned Control of Threshold

Voltages in Sub-0.2-µm MOSFET’s”, IEEE Transactions on Electron Devices,

Vol. 45, NO. 10.

32 Synopsys, “TAURUS TSUPREM-4 User’s Guide 2006”

33 Synopsys, “TAURUS MEDICI User’s Guide 2006,”

34 Synopsys Products and Solutions (2007) http://www.synopsys.com/products/tcad/taurus_tsuprem4_ds.html

Page 149: optimization and characterization of 130 nm cmos transistor design

35 Synopsys Products and Solutions (2007) http://www.synopsys.com/products/tcad/taurus_medici_ds.html

36 Hani Noorashiqin Abd. Majid, Albert Victor Kordesch, Muhamad Rasat

Muhamad, (2006) “Process Optimization of p+LDD in 130 nm Process Technology using TCAD Simulation”, IEEE International Conference on

Semiconductor Electronics (ICSE), Kuala Lumpur, Malaysia, 29 Nov – 1 Dec,

2006.

37 S. Ogura, P. J. Tsang, W. W. Walker, P. L. Critchlow, and J. F. Shepard, (1980).

“Design and Characteristics of the Lightly Doped Drain-Source (LDD) Insulated Gate Field-Effect Transistor,” IEEE Trans. Electron Devices, Vol.27, p. 1359.

38 S. B. Felch, H. Graoui, S. Severi, T. Hoffmann, P. Eyben, B. Pawlak, 2006. Meeting Future SDE Requirements Using Co-implantation and RTA. Solid State

Magazine October: 45-49. 39 Hu, C. (2000). MOS Device Research, Final Report of Micro Project 99-052,

Department of Electrical Engineering and Computer Sciences University of California, CA 94720.

40 Chiah, S.B, Zhou, X. et al, (2001). Semi-empirical Approach to Modeling

Reverse Short-Channel Effect in Submicron MOSFET’s, Modeling and Simulation Microsystem, pp 486-489.

41 International Technology Roadmap for Semiconductors (Semiconductor

Industry Association, San Jose, 1999). Available at http://public.itrs.net/.

42 Lilienfeld, J.E. US patent 174175 (1925).

43 Kilby, J.S. US patent 3138743 (1964).

44 Kamins T, (1998). “Polycrstalline Silicon for Integrated Circuits and Displays”, Kluwer Academic Publishers, 1998.

45 Fiory A.T. and Bourdelle K. K., "Activation of Implanted Poly Gates by Short Cycle Time Annealing", Mat. Res. Soc. Symp, Vol. 610, 2000, pp. B3.3.1-B3.3.6.

Page 150: optimization and characterization of 130 nm cmos transistor design

46 Yu Bin, Wan Yun, Wang Haihong, Xiang Qi, Riccobene Concetta, Talwar Somit, Lin Min-Ren, "70nm MOSFET with Ultra-Shallow, Abrupt, and Super-Doped S/D Extension Implemented by Laser Thermal Process (LTP)", IEDM

1999, 1999, pp 20.4.1-20.4.4.

47 M.E. Law, (2002). Process Modeling for Future Technologies, IBM J.Res &

Dev., Vol.46, No. 213.

48 A. Narain (1993). MOSFET Models for VLSI Circuit Simulation, Theory and

Practice, Springer-Verlag Wien, New York.

49 H.C. Chooi (2002), Modeling of Nanoscale MOSFETs, Stanford University. 50 S. Wolf, Silicon Processing for the VLSI Era: Vol. 3 – The Submicron MOSFET,

Lattice Press, Sunset Beach CA, 1995.

51 A.S.Kyuregyan, S.N.Yurkov, Sov. Phys.Semicond., Vol. 23, no.10, pp.1126-1132 (1989).

51 Indirect Electron Tunneling Poster, University of Minnesota, Twin City, (1998). mxp.physics.umn.edu/s98/projects/menz/poster.htm

52 A. Scarpa, D. Ward, J. Dubois, M. Leo, G. Steven, C. Richard, Y. S. Kwang, C. Antonio, K. Ramun, B. Mike, (2006) ”Negative-Bias Temperature Instability Cure by Process Optimization”, IEEE Trans. Electron Devices, Vol.53, NO. 6.

53 N. C. Horng, L. L. Chung, F. L Tung, (1994) “Improvement of Polysilicon

Oxide Characteristics by Fluorine Incorporation”, IEEE Trans. Electron Devices,

Vol.15, NO. 5.

54 S. B. Felch, B. S. Lee, (1996) “Fluorine Effects in BF2

+ Implants at Various Energy”, XI International Conference on Ion Implantation Technology.

Page 151: optimization and characterization of 130 nm cmos transistor design

LIST OF PUBLICATIONS

1 Hani Noorashiqin Abd. Majid, Albert Victor Kordesch, Muhamad Rasat

Muhamad, (2006) “Process Optimization of p+LDD in 130 nm Process

Technology using TCAD Simulation”, IEEE International Conference on

Semiconductor Electronics (ICSE), Kuala Lumpur, Malaysia, 29 Nov – 1 Dec,

2006.

2 Hani Noorashiqin Abd. Majid, Albert Victor Kordesch, Muhamad Rasat

Muhamad, (2006) “Optimization of Lightly Doped Drain (LDD) process in

130nm PMOS using TCAD Simulation”, 2nd

Mathematical and Physical

Science Graduate Conference (MPSGC), National University of Singapore,

Singapore, 12 Dec – 14 Dec, 2006.

3 Hani Noorashiqin Abd. Majid, Albert Victor Kordesch, Muhamad Rasat

Muhamad, (2007) “Optimum Halo Structure for 0.13 µm CMOS Technology

using TCAD”, UNITEN Student’s Conference on Research and Development

(UNITEN SCOReD), College of Graduate Studies, UNITEN, Malaysia, 14 – 15

May, 2007.

4 Hani Noorashiqin Abd. Majid, Albert Victor Kordesch, Muhamad Rasat

Muhamad, (2007) “Optimization of Halo Structure for 0.13 µm PMOSFET

using TCAD Simulation”, International Conference on Advancement of

Material and Nanotechnology (ICAMN), The City Bayview Hotel, Langkawi,

Malaysia, 29 May – 1 June, 2007.

Page 152: optimization and characterization of 130 nm cmos transistor design

APPENDIX A

Periodic Table of the Elements

Group**

Period 1 IA 1A

18

VIIIA 8A

1 1 H

1.008

2

IIA 2A

13

IIIA 3A

14

IVA 4A

15

VA 5A

16

VIA 6A

17

VIIA 7A

2 He 4.003

2 3

Li 6.941

4 Be 9.012

5 B

10.81

6 C

12.01

7 N

14.01

8 O

16.00

9 F

19.00

10 Ne 20.18

8 9 10 3

11 Na 22.99

12 Mg 24.31

3

IIIB 3B

4

IVB 4B

5

VB 5B

6

VIB 6B

7

VIIB 7B

------- VIII -------

------- 8 -------

11

IB 1B

12

IIB 2B

13 Al 26.98

14 Si

28.09

15 P

30.97

16 S

32.07

17 Cl

35.45

18 Ar 39.95

4 19 K

39.10

20 Ca 40.08

21 Sc 44.96

22 Ti

47.88

23 V

50.94

24 Cr 52.00

25 Mn 54.94

26 Fe 55.85

27 Co 58.47

28 Ni 58.69

29 Cu 63.55

30 Zn 65.39

31 Ga 69.72

32 Ge 72.59

33 As 74.92

34 Se 78.96

35 Br

79.90

36 Kr 83.80

5 37

Rb 85.47

38 Sr

87.62

39 Y

88.91

40 Zr

91.22

41 Nb 92.91

42 Mo 95.94

43 Tc (98)

44 Ru 101.1

45 Rh 102.9

46 Pd 106.4

47 Ag 107.9

48 Cd 112.4

49 In

114.8

50 Sn 118.7

51 Sb 121.8

52 Te 127.6

53 I

126.9

54 Xe 131.3

6 55 Cs 132.9

56 Ba 137.3

57 La* 138.9

72 Hf 178.5

73 Ta 180.9

74 W

183.9

75 Re 186.2

76 Os 190.2

77 Ir

190.2

78 Pt

195.1

79 Au 197.0

80 Hg 200.5

81 Tl

204.4

82 Pb 207.2

83 Bi

209.0

84 Po (210)

85 At (210)

86 Rn (222)

7 87 Fr

(223)

88 Ra (226)

89 Ac~ (227)

104 Rf (257)

105 Db (260)

106 Sg (263)

107 Bh (262)

108 Hs (265)

109 Mt (266)

110 ---

()

111 ---

()

112 ---

()

114 ---

()

116 ---

()

118 ---

()

Lanthanide Series*

58 Ce 140.1

59 Pr

140.9

60 Nd 144.2

61 Pm (147)

62 Sm 150.4

63 Eu 152.0

64 Gd 157.3

65 Tb 158.9

66 Dy 162.5

67 Ho 164.9

68 Er

167.3

69 Tm 168.9

70 Yb 173.0

71 Lu 175.0

Actinide Series~ 90

Th 232.0

91 Pa (231)

92 U

(238)

93 Np (237)

94 Pu (242)

95 Am (243)

96 Cm (247)

97 Bk (247)

98 Cf (249)

99 Es (254)

100 Fm (253)

101 Md (256)

102 No (254)

103 Lr (257)

Page 153: optimization and characterization of 130 nm cmos transistor design

APPENDIX B $driver ts4 $module NTub_PTub:0 $step load init in.f=STI:0_0 tif method unrefine=0 material mat=poly grasz.0=0.02 gamma.0=5.6e-5 impurity mat=poly imp=phosphorus q.sites=8e15 gseg.ini=25 impurity mat=poly imp=gb_phosphorus cm.sec dix.0=6e3 impurity mat=poly imp=boron q.sites=2e16 gseg.ini=25 impurity mat=poly imp=gb_boron cm.sec dix.0=420 $step Variables assign name=device c.val="n" assign name=gox c.val="thin" assign name=region c.val="2d" assign name=Lgate n.val=0.13 assign name=Lpolywin n.val=0.10 assign name=Lwin n.val=0.38 assign name=Lwintox n.val=0.1 assign name=Liso n.val=0.28 assign name=includeSTI n.val=0 $step Method Method pd.full act.full itrap source bas.inp $step NTUB1-L05 deposit photo thick=2.00 if ("@device"=="p") etch photo all if.end $step NTUB1-IMP $moment fitted to UT-MARLOWE assign name=NW1ds n.val=6e12 assign name=NW1en n.val=600 assign name=NW2ds n.val=6e12 assign name=NW2en n.val=380 $ Nwell 1st Implant implant phos dose=@NW1ds/4 ener=@NW1en tilt=4 rota=0 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print implant phos dose=@NW1ds/4 ener=@NW1en tilt=4 rota=90 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print implant phos dose=@NW1ds/4 ener=@NW1en tilt=4 rota=180 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print implant phos dose=@NW1ds/4 ener=@NW1en tilt=4 rota=270 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print $ Implant 2 implant phos dose=@NW2ds/4 ener=@NW2en tilt=4 rota=22 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print implant phos dose=@NW2ds/4 ener=@NW2en tilt=4 rota=112 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print implant phos dose=@NW2ds/4 ener=@NW2en tilt=4 rota=202 damage +

Page 154: optimization and characterization of 130 nm cmos transistor design

taur.mod=DUALPEARSON taur.exe=tauruss4 print implant phos dose=@NW2ds/4 ener=@NW2en tilt=4 rota=292 damage + taur.mod=DUALPEARSON taur.exe=tauruss4 print $step PTHRU2-IMP implant arse dose=5e12 ener=150 tilt=7 rota=22 damage taur.mod=DUALPEARSON + taur.exe=tauruss4 print $step PVT1-IMP implant arsenic dose=5e12 ener=150 tilt=7 rota=22 damage taur.mod=DUALPEARSON + taur.exe=tauruss4 print $step NTUB1-CSTR etch photo all $step PTUB1-L15 deposit photo thick=1.7125 if ("@device"=="n") etch photo all if.end $step PTUB1-IMP assign name=PW1ds n.val=8e12 assign name=PW1en n.val=260 assign name=PW2ds n.val=5e12 assign name=PW2en n.val=120 $ High Energy Pwell Implant $ PWELL IMPLANT IMPLANT Boron DOSE=@PW1ds/4 ENERGY=@PW1en taur.mod=DUALPEARSON tilt=4 + rotation=0 damage taur.exe=tauruss4 print IMPLANT Boron DOSE=@PW1ds/4 ENERGY=@PW1en taur.mod=DUALPEARSON tilt=4 + rotation=90 damage taur.exe=tauruss4 print IMPLANT Boron DOSE=@PW1ds/4 ENERGY=@PW1en taur.mod=DUALPEARSON tilt=4 + rotation=180 damage taur.exe=tauruss4 print IMPLANT Boron DOSE=@PW1ds/4 ENERGY=@PW1en taur.mod=DUALPEARSON tilt=4 + rotation=270 damage taur.exe=tauruss4 print $ 2ND IMPLANT IMPLANT Boron DOSE=@PW2ds/4 ENERGY=@PW2en taur.mod=DUALPEARSON tilt=4 + rotation=0 damage taur.exe=tauruss4 print IMPLANT Boron DOSE=@PW2ds/4 ENERGY=@PW2en taur.mod=DUALPEARSON tilt=4 + rotation=90 damage taur.exe=tauruss4 print IMPLANT Boron DOSE=@PW2ds/4 ENERGY=@PW2en taur.mod=DUALPEARSON tilt=4 + rotation=180 damage taur.exe=tauruss4 print IMPLANT Boron DOSE=@PW2ds/4 ENERGY=@PW2en taur.mod=DUALPEARSON tilt=4 + rotation=270 damage taur.exe=tauruss4 print $step NTHRU-IMP implant boron dose=5e12 ener=60 tilt=7 rota=22 damage taur.mod=DUALPEARSON +

Page 155: optimization and characterization of 130 nm cmos transistor design

taur.exe=tauruss4 print $step NVT1-IMP implant boron dose=5e12 ener=20 tilt=7 rota=22 damage taur.mod=DUALPEARSON + taur.exe=tauruss4 print implant imp=indium dose=1e13 ener=120 tilt=7 rota=22 damage impl.t=tr.indium + print $step PTUB1-CSTR etch photo all $step Doping select z=phos print x.v=0 layers $step save save out.f=NTub_PTub:0_0 tif