Advanced physical modelling of step graded Gunn Diode for high power TeraHertz Sources

Advanced physical modelling of step graded Gunn

Diode for high power TeraHertz Sources

A thesis submitted to The University of Manchester for the degree of

Doctor of Philosophy

in the Faculty of Engineering and Physical Sciences

2011

FAISAL AMIR

SCHOOL OF ELECTRICAL AND ELECTRONIC ENGINEERING

TABLE OF CONTENTS

TABLE OF CONTENTS .......................................................................................... 2

LIST OF FIGURES .................................................................................................. 6

LIST OF TABLES .................................................................................................. 12

LIST OF ABBREVIATIONS................................................................................... 13

ABSTRACT ........................................................................................................... 15

DECLARATION ..................................................................................................... 17

ACKNOWLEDGEMENTS ..................................................................................... 18

DEDICATION ......................................................................................................... 19

Chapter 1 Introduction ........................................................................................ 20

1.1

Project Overview ................................................................................. 20

1.2

Project Motivation ................................................................................ 20

1.3

Research Papers ................................................................................. 21

1.4

Prizes / Awards .................................................................................... 23

1.5

Thesis Organization ............................................................................. 24

Chapter 2 Terahertz Generation and Applications ............................................ 26

2.1

Introduction .......................................................................................... 26

2.2

The Terahertz (THz) Spectrum ............................................................ 27

2.3

THz Generation ................................................................................... 29

2.4

Solid State two Terminal Active Devices for Terahertz Generation ..... 29

2.4.1

Transferred Electron Devices (Gun Diodes) ............................................................29

2.4.2

Tunnelling Devices ...................................................................................................32

2.4.3

Transit – Time Diodes ..............................................................................................32

2.5

Monolithic Microwave Integrated Circuit (MMIC) ................................. 33

2.6

Applications of THz Sources ................................................................ 34

2.6.1

THz Space Applications ...........................................................................................34

2.6.2

THz Security Applications ........................................................................................35

2.6.3

THz Imaging and Spectroscopy Systems ................................................................35

2.6.4

Communications Systems ........................................................................................36

2.6.5

MM-Wave Automotive RADAR Systems .................................................................36

2.6.6

MM-Wave RADAR Industrial Applications ...............................................................36

2.7

Conclusion ........................................................................................... 37

Table of Contents

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency ........ 38

3.1

Introduction .......................................................................................... 38

3.2

Gunn Diode Oscillator Design ............................................................. 38

3.2.1

Gunn Diode Operation and Hot-Electron Injection ...................................................41

3.2.2

Gunn Diode Electromagnetic Modelling ...................................................................43

3.2.3

Accounting for Interaction between the Diode and Oscillator Circuit .......................45

3.2.4

Proposed Modelling Techniques ..............................................................................47

3.3

Frequency Multipliers ........................................................................... 49

3.3.1

Harmonic Balance Simulation Tool ..........................................................................51

3.3.2

Linear EM Structure Simulations ..............................................................................53

3.3.3

Semiconductor Component – Schottky Diode Varactor ...........................................53

3.4

Schottky Diode SILVACO

Modelling ................................................ 56

3.5

Conclusion ........................................................................................... 60

Chapter 4 Gunn Diode Theory ............................................................................ 61

4.1

Introduction .......................................................................................... 61

4.2

Gunn Diode as Transferred Electron Effect (TED) Device ................... 61

4.2.1

Electric Field Effects on Electron Drift Velocity ........................................................61

4.2.2

High Field Transport in n-GaAs ................................................................................64

4.2.3

Negative Differential Resistance (NDR) ...................................................................65

4.2.4

Conditions for Negative Differential Mobility ............................................................67

4.3

Instability and Domain Formation ........................................................ 67

4.3.1

Domain Dynamics ....................................................................................................69

4.3.2

Device Oscillating Frequency ...................................................................................72

4.3.3

The Doping – Length (N

L) Product .........................................................................72

4.3.4

Stable Operating Point of Domain ............................................................................73

4.4

Conventional Gunn Diode .................................................................... 75

4.4.1

Temperature Effects on Conventional Gunn Diode .................................................76

4.4.2

Limitations of the Conventional Gunn Diode ............................................................77

4.5

Gunn Diode Oscillation Modes ............................................................ 78

4.5.1

Transit Time Mode ...................................................................................................78

4.5.2

Delayed Domain mode .............................................................................................80

4.5.3

The Quenched Domain Mode ..................................................................................80

4.5.4

Low Space-charge Accumulation (LSA) mode ........................................................81

4.5.5

Operating Modes Summary .....................................................................................82

4.6

Conclusion ........................................................................................... 83

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron

Injection ................................................................................................................ 84

5.1

Introduction .......................................................................................... 84

5.2

AlGaAs/GaAs Graded Gap Heterostructure ........................................ 85

5.3

Hot Electron Injection........................................................................... 86

5.3.1

The Graded Gap Hot Electron Injector Concept ......................................................86

5.3.2

The AlGaAs/GaAs Hot Electron Injector ..................................................................87

Table of Contents

5.4

I-V Characteristics of graded gap injector GaAs Gunn Diode .............. 89

5.4.1

High Frequency Investigations of GaAs Gunn Diodes .............................................90

5.4.2

Hot Electron Injector Barrier Height..........................................................................90

5.4.3

Doping Spike ............................................................................................................91

5.5

Drift Velocity Computation and Operation Mode Classification ............ 93

5.6

Conclusion ........................................................................................... 94

Chapter 6 Gunn Diode Model developments ..................................................... 96

6.1

The SILVACO

TCAD Suite ............................................................... 96

6.2

Development of the Simulation Model ................................................. 99

6.2.1

High Speed Heterostructure Graded Gap Injector Gunn Diode ...............................99

6.2.2

Gunn Diode 2D Model ..............................................................................................99

6.2.3

2D Model with Heat Sink ........................................................................................103

6.2.4

3D Rectangular Model Development .....................................................................104

6.2.5

3D Cylindrical Model Development ........................................................................105

6.3

Initial Device Solution ........................................................................ 106

6.3.1

DC Simulation Results ...........................................................................................107

6.3.2

Transient solutions .................................................................................................107

6.3.3

Transient Response – free running oscillations .....................................................108

6.4

Free-running frequency of oscillation ................................................. 110

6.5

Conclusion ......................................................................................... 112

Chapter 7 Physical Models................................................................................ 114

7.1

Physical Models Used for Device Simulation ..................................... 114

7.2

Mobility Models Used ......................................................................... 115

7.2.1

Low field Mobility Models .......................................................................................116

7.2.2

AlGaAs Default Low field Mobility Model ...............................................................117

7.2.3

GaAs Concentration Dependent Low Field Mobility Model ....................................118

7.2.4

GaAs Analytic Low Field Mobility Model ................................................................118

7.2.5

Parallel Electric Field Dependent Mobility Model ...................................................119

7.3

Carrier Generation – Recombination Models ..................................... 121

7.3.1

Shockley-Read-Hall (SRH) Recombination ...........................................................121

7.3.2

SRH Concentration-Dependent Lifetime model .....................................................122

7.4

Carrier Statistics and Transport ......................................................... 123

7.4.1

Drift Diffusion Model ...............................................................................................124

7.4.2

The Energy Balance and Hydrodynamic Transport Models ..................................124

7.5

GIGA

– Self Heating Simulator ....................................................... 125

7.5.1

The Lattice Heat Flow Equation .............................................................................126

7.5.2

Heat Capacity .........................................................................................................127

7.5.3

Thermal Conductivity ..............................................................................................128

7.5.4

Heat Generation .....................................................................................................129

7.5.5

Electron Energy Relaxation Time ...........................................................................130

7.5.6

Thermal Boundary Conditions ................................................................................131

7.6

C-Interpreter Functions ...................................................................... 132

7.7

Conclusion ......................................................................................... 133

Table of Contents

Chapter 8 Gunn diode Results: DC Analysis ................................................... 134

8.1

Introduction ........................................................................................ 134

8.2

Advanced Step Graded Gunn Diode Modelled–Measured Results ... 134

8.2.1

Doping Spike Carrier Concentration.......................................................................135

8.2.2

DC I-V Characteristics of 77 GHz second Harmonic GaAs Gunn .........................136

8.3

Doping Spike Effects ......................................................................... 136

8.4

Measured 77 GHz Devices Data Comparison with Modelled Devices139

8.4.1

Doping Spike CC 1×10

-3

Modelled Device Results (VMBE 1928A) ..............139

8.4.2

Doping Spike CC 5×10

-3

Modelled Device Results (VMBE 1900) .................141

8.4.3

Doping Spike CC 7.5×10

-3

Modelled Device Results (VMBE 1909) ..............142

8.5

Higher Frequency Measured–Modelled DC I-V Characteristics ........ 144

8.5.1

1.65 µm Transit Region Device DC Results (VMBE 1901 – 77 GHz Device) .......145

8.5.2

1.1 µm Transit Region Device DC Results (VMBE 1950 - 125GHz Device) .........146

8.5.3

0.9 µm Transit Region Device DC Results (VMBE 1897- 125 GHz Device) .........147

8.5.4

0.7 µm transit region device DC results (VMBE 1898 - 100GHz device) ..............148

8.5.5

0.4 µm Transit Region Device DC Results (XMBE 189 – 200 GHz Device) .........149

8.6

Conclusion ......................................................................................... 151

Chapter 9 Gunn diode Results: Time-domain analysis .................................. 153

9.1

Time-Domain (Transient) Solutions ................................................... 153

9.2

Time-Domain Analysis of various Epitaxial Structures....................... 153

9.2.1

1.65 µm Transit Region Device Results (VMBE1901) ...........................................154

9.2.2

1.1 µm Transit Region Device Results (VMBE1950) .............................................155

9.2.3

0.9 µm Transit Region Device Results (VMBE1897) .............................................156

9.2.4

0.7 µm Transit Region Device Results (VMBE1898) .............................................156

9.2.5

0.4 µm – 0.6 µm Transit Region Device Results (XMBE189) ................................158

9.3

Time-Domain Response with a Resonant Cavity .............................. 159

9.3.1

Dipole Domain Formation .......................................................................................159

9.3.2

Modelled Device with Resonant Cavity Time-Domain Results ..............................161

9.4

Conclusion ......................................................................................... 161

Chapter 10 Conclusion and future work .......................................................... 163

10.1

Conclusions ....................................................................................... 163

10.2

Directions for Future Research .......................................................... 165

10.2.1

Lumped Element Model for SILVACO

............................................................166

10.2.2

Taking into account Gunn Diode Domain Growth and Propagation ..................166

Appendix – A ...................................................................................................... 168

Appendix – B ...................................................................................................... 174

Appendix – C ...................................................................................................... 177

References.......................................................................................................... 179

Words count, including footnotes and endnotes: 36,822

LIST OF FIGURES

Fig. 2.1

Attenuation due to atmospheric absorption at microwave and mm-

wave frequencies. At mm-wave range, attenuation not only increases, but

becomes more dependent upon absorbing characteristics of H

O & O

.. ... 27

Fig. 2.2

TeraHertz (THz) Gap ..................................................................... 28

Fig. 2.3

Compilation of published state-of-the-art results between 30 and 400

GHz for GaAs and InP Gunn diodes under CW operation. Legend format:

‘mode of operation (‘1’ denotes fundamental, ‘2’ second-harmonic, etc.),

package type, heatsink technology’. Solid lines outline the highest powers

and frequencies achieved experimentally to-date from each material in

fundamental and second-harmonic. ............................................................ 31

Fig. 2.4

Typical Advanced Gunn Diode Structure ........................................ 31

Fig. 3.1

Simplified schematic of the mechanical outline of a second harmonic

resonant disk Gunn diode oscillator. ........................................................... 39

Fig. 3.2

Equivalent electrical circuit model of a second harmonic resonant

disk Gunn diode oscillator. .......................................................................... 39

Fig. 3.3

(a) Ansoft HFSS

model of a second harmonic resonant disk

millimetre-wave oscillator with tuning pin and circular waveguide sliding

backshort. (b) Simulated field electric field plot for the oscillator at second

harmonic frequencies. ................................................................................ 44

Fig. 3.4

Ansoft HFSS

simulations of a reduced-height waveguide model

developed at e2v (a) second harmonic resonant disk millimetre-wave

oscillator (b) simulated field electric field plot. ............................................. 45

Fig. 3.5

Gunn diode packages developed at e2v (a) Standard Alumni

Packages (b) Quartz Package.(c) Power Combining .................................. 45

Fig. 3.6

Frequency multiplier design methodology developed at e2v showing

both hydrodynamic and liner electromagnetic structure modelling. ............ 50

List of Figures

Fig. 3.7

A harmonic balance simulation tool schematics created using

Microwave Office ® by AWR, specifically for analysis and prediction of GaAs

Schottky diode performance given device’s basic electrical characteristics.52

Fig. 3.8

Result of a harmonic balance parametric sweep to study output

power with variation in input embedding impedance .................................. 52

Fig. 3.9

(a) Varactor Diode Chip and the waveguide circuit model created

using HFSS

at e2v Technologies Plc (b) HFSS

diode chip model

embedded in waveguide diode mount. ....................................................... 53

Fig. 3.10

2D GaAs diode structure topped by a heavily doped region of In

As graded from x=0 at the GaAs interface to x=0.53 at the upper surface. 54

Fig. 3.11

(a) C-V plot of 4 µm Schottky device. Junction capacitance of

fabricated single anode device (bias - 1/C

inset) (b) image of a 4 µm

Schottky device with 2 µm bridge to the anode........................................... 55

Fig. 3.12

Six anode anti-series device (a) design (b) fabricated device. ......... 56

Fig. 3.13

GaAs Schottky diode SILVACO

2D model. Equal area rule was

maintained same as the manufactured device (a) 2D device structure for a 4

µm diameter anode Schottky diode (b) Conduction band diagram for GaAs

Schottky contact (c). Conduction band diagram at zero bias ...................... 57

Fig. 3.14

Forward bias I-V curve for a 4 µm diameter anode Schottky diode.

Series resistance

was calculated from curve’s slope. ............................ 57

Fig. 3.15

C-V plot for a 4 µm diameter anode Schottky diode ......................... 58

Fig. 3.16

Schottky diode C-V plots for 14 µm, 13 µm, 12 µm anode dia.. ....... 59

Fig. 4.1

Carrier Drift Velocity verses Electric Field graph ............................. 63

Fig. 4.2

Electron occupations for (a) GaAs, (b) InP and (c) GaN ................ 63

Fig. 4.3

Electron occupations under various Electric Field levels - n-GaAs 65

Fig. 4.4

Negative Differential Resistance region .......................................... 66

Fig. 4.5

Stable dipole domain formation with the space-charge growth ....... 69

Fig. 4.6

Dynamic characteristics curve and electron drift velocity – electric

field curve plotted in SILVACO using MOCASIM for GaAs 1.1x10

-16

doped.

Equal area relationship between

and

is shown. ............................... 70

List of Figures

Fig. 4.7

Zero Diffusion Domain profiles ........................................................ 70

Fig. 4.8

The domain potential

and minimum electric field

................. 74

Fig. 4.9

Conventional Gunn Diode structure ................................................ 75

Fig. 4.10

GaAs temperature dependant electron drift velocity curves plotted in

SILVACO using MOCASIM for GaAs 1.1x10

-16

doped. .............................. 77

Fig. 4.11

Transit Time Mode - I-V and Time Evolution .................................... 79

Fig. 4.12

Delayed Domain Mode- I-V and Time Evolution .............................. 79

Fig. 4.13

Quenched Domain Mode- I-V and Time Evolution ........................... 81

Fig. 4.14

Gunn Diode Operating Modes Summary ......................................... 82

Fig. 5.1

Potential profile of Hot Electron Injector .......................................... 87

Fig. 5.2

(a) Structure of a GaAs Gunn Diode with step graded hot electron

injector (b) Conduction band profile of a Step Graded AlGaAs

heterostructure Gunn Diode. ....................................................................... 88

Fig. 5.3

Typical advanced step graded injector Gunn Diode structure ......... 89

Fig. 5.4

Gunn Diodes I-V characteristics ..................................................... 90

Fig. 5.5

Conductance and Susceptance versus Frequency plot .................. 91

Fig. 5.6

Low voltage I-V characteristics ....................................................... 91

Fig. 5.7

Electron Concentration – without doping spike ............................... 92

Fig. 5.8

Electron Concentration – with doping spike .................................... 92

Fig. 5.9

Diode with Injector ........................................................................... 93

Fig. 5.10

Electrons occupations in the L- valley .............................................. 94

Fig. 6.1

ATLAS

Input-Output hierarchy ..................................................... 97

Fig. 6.2

ATLAS

command groups ............................................................. 98

Fig. 6.3

Gunn Diode model development process flow ................................ 98

Fig. 6.4

77 GHz Gunn advanced step graded injector Gunn Diode str. ....... 99

Fig. 6.5

Graded Gap Heterostructure Gun Diode ....................................... 100

Fig. 6.6

TonyPlot

of modelled device structure showing impurity doping

profile (a) Device planar view (b) Device cross section view .................... 100

List of Figures

Fig. 6.7

Device structure with mesh density defined (a) using ‘constraint’ (b)

using ‘refine’, plotted in TonyPlot

........................................................... 101

Fig. 6.8

Conduction band comparative data showing two models with course

and dense meshing in the step-graded launcher ...................................... 102

Fig. 6.9

Modelled device structure with heat sinks ..................................... 103

Fig. 6.10

3D rectangular model epitaxial structure ........................................ 104

Fig. 6.11

3D rectangular modelled device structure with heat sinks ............. 105

Fig. 6.12

3D cylindrical modelled device structure with heat sinks ................ 105

Fig. 6.13

Device model structure cross section view plotted in TonyPlot

.. 106

Fig. 6.14

Model forward-reverse asymmetrical I-V characteristics ................ 107

Fig. 6.15

Device model transient voltage ramp in ATLAS

......................... 108

Fig. 6.16

Transient response - critical current at 2V ...................................... 109

Fig. 6.17

Transient response - damping frequency at 2V ............................. 109

Fig. 6.18

Transient response - stable oscillations at 4V ................................ 110

Fig. 6.19

Monte Carlo simulation results for GaAs Gunn diodes with and

without hot electron injection. It is noted that the power curve is taken from

measurements of a typical device with hot electron injection, and that an

oscillator circuit was not included in the simulation. .................................. 111

Fig. 8.1

The effects of the doping spike (cc 1×10

-3

) on electron

concentration in the transit region (—; spike present, ----; spike absent) .. 135

Fig. 8.2

Modelled and measured forward - reverse I-V curves................... 136

Fig. 8.3 Electron concentration with changing doping spike cc ................. 137

Fig. 8.4

Effects of variation in doping spike cc on simulated I-V characteristics

(nos next to forward bias curves represent calculated asymmetry values) 137

Fig. 8.5

Simulated - measured asymmetry values versus doping spike cc 138

Fig. 8.6

VMBE1928A Modelled and measured forward-reverse I-V curves 140

Fig. 8.7

VMBE 1928A Temperature forward - reverse I-V curves .............. 140

Fig. 8.8

VMBE 1900 Modelled and measured forward - reverse I-V curves141

Fig. 8.9

VMBE 1900 Temperature forward - reverse I-V curves ................ 141

List of Figures

Fig. 8.10

VMBE 1909 Modelled and measured forward - reverse I-V curves 142

Fig. 8.11

VMBE 1909 Temperature forward - reverse I-V curves ................. 142

Fig. 8.12

Simulated-measured asymmetry values versus doping spike cc.143

Fig. 8.13

VMBE 1901 Forward and reverse I-V curves for a 70-80 GHz 2

harmonic device at 300K .......................................................................... 145

Fig. 8.14

VMBE 1901 Temperature forward and reverse I-V curves ............. 145

Fig. 8.15

VMBE 1950 Forward and reverse I-V curves for a 62.5GHz

fundamental, 125GHz 2

harmonic device at 300K ................................. 146

Fig. 8.16

VMBE 1950 Temperature forward and reverse I-V curves ............. 146

Fig. 8.17

VMBE 1897 Forward and reverse I-V curves for a 62.5 GHz

fundamental, 125GHz 2

harmonic device at 300K ................................. 148

Fig. 8.18

VMBE 1897 Temperature forward and reverse I-V curves ............. 148

Fig. 8.19

VMBE 1898 Forward and reverse I-V curves for a 100 GHz

fundamental device at 300K ..................................................................... 149

Fig. 8.20

VMBE 1898 Temperature forward and reverse I-V curves ............. 149

Fig. 8.21

XMBE 189 Modelled forward and reverse I-V curves. Effects of

variation in doping spike cc on simulated I-V characteristics (numbers next

to forward bias curves represent calculated asymmetry values). Transit

region cc was 5.4 x10

-3

. .................................................................... 150

Fig. 8.22

XMBE 189 Modelled forward and reverse I-V curves for a 200 GHz

fundamental device at 300K ..................................................................... 151

Fig. 8.23

XMBE 189 Temperature forward and reverse I-V curves ............... 151

Fig. 9.1

The simulated fundamental frequency time-domain response of a

device fabricated from VMBE1901 (1.65 µm transit length) under a 2.2V

external bias ............................................................................................. 154

Fig. 9.2

The simulated fundamental frequency time-domain response of a

device fabricated from VMBE1901 (1.65 µm transit length) under a 3.2 V

external bias. ............................................................................................ 155

List of Figures

Fig. 9.3

The simulated fundamental frequency time-domain response of a

device fabricated from VMBE1950 (1.1 µm transit length) at an external bias

of 2.5V, temperature 300K. ....................................................................... 155

Fig. 9.4

The simulated fundamental frequency time-domain response of a

device fabricated from VMBE1897 (0.9 µm transit length) at an external bias

of 2.2V, temperature 300K. ....................................................................... 156

Fig. 9.5

The simulated fundamental frequency time-domain response of a

device fabricated from VMBE1898 (0.7 µm transit length) at an external bias

of 1.7V, temperature 300K. ....................................................................... 157

Fig. 9.6

The simulated fundamental frequency time-domain response of a

device fabricated from VMBE1898 (0.7 µm transit length) at an external bias

of 2.75 V, temperature 300K. .................................................................... 157

Fig. 9.7

The simulated fundamental frequency time-domain response of a

device fabricated from XMBE189 (0.4 µm transit length) at an external bias

of 3.5V, temperature 300K. ....................................................................... 158

Fig. 9.8

Spatial dipole domain profiles in the transit region, for a 0.7 µm

transit length device. These are plotted at an applied bias of 3V at time,

starting at 0.156ps, in one oscillation period T=4.68 ps. The domain first

grows to a maximum size and then drifts to the anode before it collapses

there and a new domain nucleates. .......................................................... 160

Fig. 9.9

Time-domain stable oscillation at 3 V with C

diode

= 0.158 pF. The

lumped-element LC parallel circuit values are L= 3.16 pH and C= 0.403 pF.

The main loss R=0.9 Ohms is in series with the L. The current overshoot

peaks are in agreement with [9] which are due to electron experiencing

energy and momentum inertial effects at high frequencies. ...................... 161

LIST OF TABLES

Table 2.1

Terahertz terminology ................................................................... 26

Table 3.1

Schottky diode models with 14µm, 13µm and 12µm anode

diameters showing an additional area and calculated cut off frequencies. . 59

Table 7.1

Physical models summary .......................................................... 115

Table 7.2

Mobility values for Al

1-x

As regions ....................................... 117

Table 7.3

Caughey Thomas analytic low field mobility model parameters . 119

Table 7.4

GIGA

– self heating simulator parameters .............................. 126

Table 7.5

Heat Capacity parameters .......................................................... 127

Table 7.6

Thermal Conductivity parameters ............................................... 128

Table 7.7

Electron energy relaxation time parameters ............................... 131

Table 7.8

C-Interpreter Functions used in model development .................. 132

Table 8.1

Measured devices with slightly different epitaxial structures –

doping spike carrier concentration variation.............................................. 139

Table 8.2

Measured devices with slightly different epitaxial structures – transit

region length and its carrier concentration variation ................................. 144

Table 9.1

Measured devices with slightly different epitaxial structures – transit

region length and its carrier concentration variation ................................. 153

LIST OF ABBREVIATIONS

2D Two Dimensional

ACC Adaptive Cruise Control

ACCS Adaptive Cruise Control Systems

ADS

Advanced Design System by Agilent

Technologies

AlAs Aluminium arsenide

AlGaAs Aluminium gallium arsenide

AlSb Aluminium Antimonide

As Arsenic

AWR Advancing the Wireless Revolution

BARITT Barrier Injection Transit Time

CB Conduction band

cc Carrier Concentration

CdTe Cadmium Telluride

CMS Collision Mitigation Systems

CO2 Carbon dioxide

CVD Chemical Vapour Deposition

CW Continuous Wave

DC Direct Current

EM ElectroMagnetic

fF Femto Farad

FMCW Frequency Modulated Continuous Wave

Ga Gallium

GaAs Gallium arsenide

GaAsP Gallium Arsenide Phosphide

GaInAs Gallium Indium arsenide

GaN Gallium Nitride

GHz Gega Hertz

HFSS High Frequency Structure Simulator

IMPATT Impact Ionization Transit Time

InAs Indium arsenide

List of Abbreviations

Inc Incorporation

InP Indium Posphide

I-V Current-Voltage

K Kelvin

L Satellite

LSA Low Space-charge Accumulation

MBE Molecular Beam Epitaxy

micron micrometre

MITATT Mixed Tunnelling-Avalanche Transit-Time

mm milli metre

MMIC Monolithic Microwave Integrated Circuit

mW milli watt

NDR Negative Differential Resistance

Plc public limited company

RF Radio Frequency

RIE Reactive Ion Etching

RTD Resonant Tunnelling Diode

SI Semi-Insulating

SRH Shockley-Read-Hall

STFC Science and Technology Facilities Council

TADAR Tactical Area Defence Alerting Radar

TCAD Technology Computer-Aided Design

TD Tunnel diode

TDTS Time-domain THz Spectroscopy

TED Transferred Electron Devices

TEM Transverse ElectroMagnetic

THz Terahertz

Ti Titanium

TM Trademark

TUNNETT Tunnel Injection Transit Time

UMS Universal Marking Systems Ltd

VWF Virtual Wafer Fab

ZnSe Zinc Selenide

Γ Gamma

ABSTRACT

The mm-wave frequency range is being increasingly researched to close the gap

between 100 to 1000 GHz, the least explored region of the electromagnetic

spectrum, often termed as the ‘THz Gap’. The ever increasing demand for

compact, portable and reliable THz (Terahertz) devices and the huge market

potential for THz system have led to an enormous amount of research and

development in the area for a number of years. The Gunn Diode is expected to

play a significant role in the development of low cost solid state oscillators which

will form an essential part of these THz systems.

Gunn and mixer diodes will “power” future THz systems. The THz frequencies

generation methodology is based on a two-stage module. The initial frequency

source is provided by a high frequency Gunn diode and is the main focus of this

work. The output from this diode is then coupled into a multiplier module. The

multiplier provides higher frequencies by the generation of harmonics of the input

signal by means of a non-linear element, such as Schottky diode Varactor. A

realistic Schottky diode model developed in SILVACO

is presented in this

work.

This thesis describes the work done to develop predictive models for Gunn Diode

devices using SILVACO

. These physically-based simulations provide the

opportunity to increase understanding of the effects of changes to the device’s

physical structure, theoretical concepts and its general operation. Thorough

understanding of device physics was achieved to develop a reliable Gunn diode

model. The model development included device physical structure building,

material properties specification, physical models definition and using

appropriate biasing conditions.

The initial goal of the work was to develop a 2D model for a Gunn diode

commercially manufactured by e2v Technologies Plc. for use in second harmonic

mode 77GHz Intelligent Adaptive Cruise Control (ACC) systems for automobiles.

Abstract

This particular device was chosen as its operation is well understood and a

wealth of data is available for validation of the developed physical model. The

comparisons of modelled device results with measured results of a manufactured

device are discussed in detail. Both the modelled and measured devices yielded

similar I-V characteristics and so validated the choice of the physical models

selected for the simulations. During the course of this research 2D, 3D

rectangular, 3D cylindrical and cylindrical modelled device structures were

developed and compared to measured results.

The injector doping spike concentration was varied to study its influence on the

electric field in the transit region, and was compared with published and

measured data.

Simulated DC characteristics were also compared with measured results for

higher frequency devices. The devices mostly correspond to material previously

grown for experimental studies in the development of D-band GaAs Gunn

devices. Ambient temperature variations were also included in both simulated

and measured data.

Transient solutions were used to obtain a time dependent response such as

determining the device oscillating frequency under biased condition. These

solutions provided modelled device time-domain responses. The time-domain

simulations of higher frequency devices which were developed used modelling

measured approach are discussed. The studied devices include 77GHz (2

harmonic), 125 GHz (2

harmonic) and 100 GHz fundamental devices.

During the course of this research, twelve research papers were disseminated.

The results obtained have proved that the modelling techniques used, have

provided predictive models for novel Transferred Electron Devices (TEDs)

operating above 100GHz.

DECLARATION

No portion of the work referred to in the thesis has been submitted in support of

an application for another degree or qualification of this or any other university or

other institute of learning.

I. The author of this thesis (including any appendices and/or schedules to

this thesis) owns certain copyright or related rights in it (the “Copyright”)

and he has given The University of Manchester certain rights to use such

II. Copies of this thesis, either in full or in extracts and whether in hard or

electronic copy, may be made only in accordance with the Copyright,

Designs and Patents Act 1988 (as amended) and regulations issued

under it or, where appropriate, in accordance with licensing agreements

which the University has from time to time. This page must form part of

any such copies made.

III. The ownership of certain Copyright, patents, designs, trade marks and

other intellectual property (the “Intellectual Property”) and any

reproductions of copyright works in the thesis, for example graphs and

tables (“Reproductions”), which may be described in this thesis, may not

be owned by the author and may be owned by third parties. Such

Intellectual Property and Reproductions cannot and must not be made

available for use without the prior written permission of the owner(s) of the

relevant Intellectual Property and/or Reproductions.

IV. Further information on the conditions under which disclosure, publication

and commercialisation of this thesis, the Copyright and any Intellectual

Property and/or Reproductions described in it may take place is available

in the University IP Policy, The University Library’s regulations and in The

University’s policy on presentation of Theses.

(see http://www.campus.manchester.ac.uk/medialibrary/policies/intellectual-property.pdf),

(see http://www.manchester.ac.uk/library/aboutus/regulations).

ACKNOWLEDGEMENTS

I would like to thank my supervisor Prof M Missous, for his guidance and

encouragement in the project, which motivated me to learn and develop the skills

for modelling performed using the SILVACO

software. These physically based

engineering model simulations provided me an opportunity to increase

understanding of changes to the device physical structure, theoretical concepts

and its general operation. Thanks also to Dr. Novak Farrington who remained

involved with the research work and provided the best possible help and support.

His valuable time spent on different aspects of project discussions, provision of

experimentally measured data and suggestions for the write-up resulted in high

quality research.

I would like to thank the M&N group PhD students performing semiconductor

device modelling for their encouragement, motivation and assistance when

required during the device model development phase.

The National University of Sciences and Technology (NUST, Pakistan) is greatly

acknowledged for supporting my PhD studies under ‘faculty development

programme’. Thanks to my supervisor who arranged finances from the School of

Electrical and Electronic Engineering (E&EE) and M&N Group, which allowed me

to remain focused and achieve research goals. Thanks are also due to the IEEE

Electron Devices Society for the award of a PhD fellowship in August 2009.

Thanks are also due to Mike Carr (e2v Technologies Plc, UK) for providing the

experimentally measured data for the Gunn diodes.

DEDICATION

This thesis is dedicated to my family.

In particular to my

caring parents and beloved wife Fatima.

gorgeous daughter Farheen Wardah and

adorable son Muhammad Daud, both born

during the course of this research work

Chapter 1 Introduction

1.1 Project Overview

The aim of this project was to develop a physical model for an advanced GaAs

hot electron injector Gunn diode to be used as a high power terahertz source.

The physically-based model has been developed in SILVACO

, Inc.-TCAD

(Technology Computer-Aided Design) using the ATLAS

Virtual Wafer Fab

(VWF

) device simulation software. The model I-V simulation response was

then compared to the fabricated devices measured data to validate the various

models used during simulations and their associated material parameters. The

injector performance was evaluated and the effects of doping spike carrier

concentration were studied. Time domain transient simulations were performed

to determine modelled devices operating frequency and to investigate transit

length scaling effects on device performance.

A Schottky diode 2D model was also developed in ATLAS

to study the

underlying device physics and extract parameters to be used in harmonic

balance simulations created using Microwave Office ® by AWR (Advancing the

Wireless Revolution) Corporation.

1.2 Project Motivation

The research work was undertaken as part of the STFC (Science and

Technology Facilities Council) funded project, ‘High Power Semiconductor

Terahertz Frequency Sources for Imaging Applications’. The research was aimed

at accessing the THz region through the use of high frequency Gunn diodes in

conjunction with frequency multipliers. The project was a joint collaboration

between The University of Manchester (UoM) and e2v Technologies (UK) Plc.

The main objective of this project was to design, fabricate and test the

components (Gunn diodes and multipliers) covering the range 100 to 600 GHz.

This was planned to be achieved by the development and delivery of high power

graded gap Gunn diodes with measurable output of at least 20 to 40 mW at 200

Chapter 1 Introduction

GHz and novel multipliers sources with overall efficiency of at least 5% for 600

GHz operation.

e2v Technologies (UK) Plc have been working on step-graded gap diodes for

automotive applications at 77 GHz for over a decade whilst developing

technologies to increase the operational frequency of these GaAs devices to well

over 100 GHz [13].

This research work presented here is focused on predictive modelling of Gunn

diodes using the SILVACO

software package. The developed model was

ruggedly tested against actual data at 77GHz to demonstrate its worth to forming

a bench mark for future predictive device modelling and research. It was then

used as the basis for predicting the response and performance characteristic of

higher frequency Gunn diodes prior to their fabrication. The model proved to be

an extremely useful tool for the optimisation of the required epitaxial structures.

1.3 Research Papers

The following papers have been presented during the course of this research

work.

1. N. Farrington, F. Amir, J. Sly and M. Missous, ‘SILVACO modelling of a

Gunn diode with step-graded hot-electron injector, and pulsed DC testing of on-

wafer quasi-planar mm-wave Gunn diode structures’, 16th European Workshop

on Heterostructure Tech., HETECH’07, 2-5 Sept 07, Fréjus, France, pp. Tu2-7.

Research undertaken as part of MSc Dissertation project ‘Terahertz (THz)

Sources for Space and Security Applications’.

http://www.crhea.cnrs.fr/hetech07/index.htm

2. F. Amir, N. Farrington, J. Sly and M. Missous, ‘Step-graded Hot Electron

Injector Gunn Diode Modelling in SILVACO’, Workshop on Theory, Modelling

and Computational Methods for Semiconductor Materials And Nanostructures,

31 Jan – 1 Feb 08, The University of Manchester, UK.

http://www.eee.manchester.ac.uk/research/groups/mandn/docs/abstractsv3.pdf

Chapter 1 Introduction

3. F. Amir, N. Farrington, J. Sly and M. Missous, 'Physical Modelling of a

GaAs Gunn Diode with a Step-graded AlGaAs Hot Electron Injector', UK

semiconductors 2008, 2-3 Jul 08, University of Sheffield, Sheffield, UK.

http://www.uksemiconductors.com/

4. F. Amir, N. Farrington, T. Tauqeer, M. Missous, ‘Physical Modelling of a

Step-Graded AlGaAs/GaAs Gunn Diode and Investigation of Hot Electron

Injector Performance’, Advanced Semiconductor Devices and Microsystems,

2008. ASDAM 2008. International Conference pp.51-54, 12-16 Oct 08.

http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4743356

5. T. Tauqeer, J. Sexton, F. Amir, M. Missous, ‘Two-Dimensional Physical

and Numerical Modelling of InP-based Heterojunction Bipolar Transistors,’

Advanced Semiconductor Devices and Microsystems, 2008. ASDAM 2008.

International Conference, pp.271-274, 12-16 Oct 08.

http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4743335

6. F. Amir, N. Farrington, T. Tauqeer and M. Missous, ‘A Novel Physical

Model Developed of an Advanced Cylindrical Step Graded Heterostructure mm-

wave Gunn Diode’, 17th European Workshop on Heterostructure Technology,

HETECH’07, 2-5 Nov 08, Venice, Italy.

http://www.hetech2008.org/

7. F. Amir, C. Mitchell, N. Farrington and M. Missous, ‘Advanced Step-

graded Gunn Diode for mm-wave Imaging Applications’, 5th ESA Workshop on

Millimetre Wave Technology and Applications and 31st ESA Antenna

Workshop, 18-20 May 09, ESTEC, Noordwijk, The Netherlands, pp. 201-205.

http://www.congrex.nl/09c05/

8. T. Tauqeer, J. Sexton, M. Mohiuddin, R. Knight, F. Amir and M.

Missous, 'Physical modelling of base-dopant out diffusion in Single

Heterojunction Bipolar Transistors', UK semiconductors 2009, 1-2 Jul 09,

University of Sheffield, Sheffield, UK.

http://www.uksemiconductors.com/

Chapter 1 Introduction

9. F. Amir, C. Mitchell, N. Farrington and M. Missous, ‘Advanced Gunn

Diode as High Power Terahertz Source for a Millimetre Wave High Power

Multiplier’, SPIE Europe Security + Defence 2009, 31 Aug to 03 Sept 09,

Berliner Congress Centre, Berlin, Germany, Proc. SPIE, vol. 7485, pp 748-50I.

http://spie.org/x648.html?product_id=830296

10. F. Amir, C. Mitchell, N. Farrington, T. Tauqeer and M. Missous,

‘Development of Advanced Gunn Diodes and Schottky Multipliers for High

Power THz sources’, 2nd UK/Europe-China Workshop on Millimetre Waves and

Terahertz Technologies, 19-21 Oct 09, Rutherford Appleton Laboratory (RAL),

Oxford, UK pp. 80.

http://www.sstd.rl.ac.uk/mmt/ukchinathz2009.php

11. F. Amir, N. Farrington, C. Mitchell and M. Missous, ‘Time-domain

analysis of sub-micron transit region GaAs Gunn diodes for use in Terahertz

frequency multiplication chains’, SPIE Europe Security + Defence 2010, SD108

‘Millimetre Wave and Terahertz Sensors and Technology’ (Session I - Millimetre

and THz Devices), 20-23 Sept 10, Centre de Congrès Pierre Baudis, Toulouse,

France. Proc. SPIE 7837, 783702 (2010).

http://spie.org/x648.html?product_id=864872

12. F. Amir, C. Mitchell and M. Missous, ‘Development of Advanced Gunn

Diodes and Schottky Multipliers for High Power THz sources’, Advanced

Semiconductor Devices and Microsystems (ASDAM), 2010. 8

International

Conference, Smolenice, Slovakia, 25-27 Oct 10, pp. 29-32.

http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5667005

1.4 Prizes / Awards

a. MSc Dissertation project ‘Terahertz (THz) Sources for Space and Security

Applications’, research work adjudged best in the class and presented in 16th

European Workshop on Heterostructure Technology (HeTech 07) held in Nice,

France, September 2007.

Chapter 1 Introduction

b. Awarded 2009 IEEE EDS PhD Student Fellowship representing entire

region 8 (Europe, Africa and Middle-East), for the demonstration of significant

ability to perform independent research in the field of electron devices and a

proven record of academic excellence.

http://eds.ieee.org/eds-phd-student-fellowship-program.html

c. The fellowship award was announced in January 2010 issue of EDS

newsletter. A brief research progress report was published in July 2010 issue.

http://eds.ieee.org/eds-newsletters.html

d. Awarded 1

Prize during Post Graduate Research (PGR) Poster

Conference 2009 at The University of Manchester, School of E&EE.

http://www.eee.manchester.ac.uk/research/pgr_conference/pgrconf09/

e. SPIE Europe Security + Defence 2009 Symposium’s paper ‘Advanced

step-graded Gunn diode for millimetre applications’ was published on the SPIE

Newsroom page http://spie.org/x36521.xml?highlight=x2404&ArticleID=x36521.

f. In February 2010, nominated for the 2010 University of Manchester

Distinguished Achievement Awards in category ‘Postgraduate Research Student

of the year’, representing the School of E&EE, The University of Manchester.

1.5 Thesis Organization

The organization of the remainder of this thesis is as follows:

Chapter 2 briefly covers Terahertz technology and its applications with an

emphasis on Gunn diodes. The multiplier and harmonic generation methodology

is covered in Chapter 3. Chapter 3 presents an oscillator cavity developed in

HFSS

at e2v. The harmonic balance simulation tool created in Microwave

Office ® by AWR is discussed and both semiconductor components of the

multiplier are also presented. Chapter 4 provides a detailed account of

conventional Gunn diode theory with an explanation of the Negative Differential

Resistance (NDR) effects, which depends on the bulk material properties.

Chapter 1 Introduction

Various device operating modes are presented with emphasis on the high

frequency oscillations before summarizing the limitations of conventional Gunn

diode. Chapter 5 will discus the concept and theory behind the hot electron

injector. It would provides a literature survey of the step graded hot electron

injector Gunn diode that has been developed to overcome the limitations of the

conventional Gunn diode. Chapter 6 will provide a step-by-step process used to

develop a model in SILVACO

. Chapter 7 would outline device physical models

and discusses them in detail along with the material parameters used. Gunn

diode DC I-V characteristics that are compared to the measured results will be

presented in chapter 8. The doping spike carrier concentration optimization is

also presented. Chapter 9 will discuss Gunn diode Time-domain analyses. Also

presented would be the results of a novel 100 GHz fundamental device that has

been simulated in an oscillator cavity. Chapter 10 will provide the conclusion and

summary of the Gunn diode modelling work. It also gives a brief account of

proposed future research work.

The appendices of this thesis contain SILVACO

codes.

Chapter 2 Terahertz Generation and Applications

2.1 Introduction

The Terahertz (THz) spectrum is being increasingly researched to close the

frequency gap between microwaves and infrared. The THz region, typically

referred to as the frequencies from 100 GHz to 10 THz, is the least explored

region of the electromagnetic spectrum, often termed as the ‘THz Gap’. Table 1

list the common terminologies used to describe the frequency bands within this

gap.

Frequency Wavelength Used term

30 to 300 GHz 10 mm to 1 mm Millimetre

300 GHz to 3 THz 1 mm to100 µm Sub millimetre [14]

100 GHz to 10 THz 3 mm to 30 µm Terahertz [15]

3 to 30 THz 100 µm to 10 µm Far infrared

Table 2.1 Terahertz terminology

THz waves can penetrate through dielectric materials such as fabrics, plastics

and cardboard, which makes it an ideal choice to replace X-ray imaging and

ultrasound imaging for security applications such as detecting concealed objects.

Its non-ionizing properties and lower photon energy levels, milli-electron volt,

makes it safe for people during security checks, inspection of various biological

samples, etc. At THz frequencies the living tissues are semi-transparent and

have ‘terahertz fingerprints’, permitting them to be imaged, identified, and

analyzed. THz spectral imaging technology not only differentiates objects

morphology, but it also identifies their composition. Additionally, it provides

higher spatial resolution, and therefore an ideal choice for non-destructive

testing.

THz radiation has higher frequency and bandwidth that provides short-distance

high-capacity wireless communications. An increasingly widespread application

is automotive FMCW (Frequency Modulated Continuous Wave) radar in the low

Chapter 2 Terahertz Generation and Applications

atmospheric attenuation window at 77GHz (Fig. 2.1). The unique characteristics

of THz radiation have important applications in the field of astrophysics, plasma

physics, materials science, information science and engineering etc.

The demand for high-frequency, high output power device technology has

increased enormously in the past decade due to an array of emerging

applications, which has resulted in focused and intense research in the field. This

has led to an interest in the development of Physically-based models to aid in the

development of next generation devices.

The Gunn Diode has long been considered the heart of mm-wave power

generation. However, in recent years Monolithic Microwave Integrated Circuit

(MMIC) solutions for power generation at mm-wave frequencies have become

commercially viable but suffer from high cost and low output power levels as

discussed later.

2.2 The Terahertz (THz) Spectrum

As shown in Fig. 2.2, THz frequencies correspond to photon energy levels from

approximately 4 to 40 milli-electron volts or to an equivalent black body

temperature between 50 and 500 Kelvin. The frequency range spans 100 GHz to

Fig.

Attenuation due to atmospheric absorption at microwave and

millimetre wave frequencies [7]. At millimetre

wave range, the attenuation not

only increases, but becomes more dependent upon absorbing characteristics

of H

O and O

Chapter 2 Terahertz Generation and Applications

10 THz, which corresponds to wavelengths 3 mm to 30 µm [16]. The potential

applications for THz devices have increased many fold during the past few years

and range from security radar to medical imaging for tumour detection [17]. Other

potential applications include high bandwidth communications, high resolution

radar systems, security imaging systems and space exploration.

The transition from electronics to optics takes place in the infrared region of the

electromagnetic spectrum where the wavelengths are less than 1mm. The lower-

frequency region of the infrared is known as ‘far infrared’ and is generally

considered as extension of the microwave region. Originally, the edge of the

microwave band (300 GHz) was considered the highest viable frequency for

electronics, but as technology has progressed, the limits of electronics have

been pushed further into the infrared. However, the lower efficiency of both

optical and electronic devices in the THz region has been a big impediment in

the development of Terahertz systems. Therefore, research has been focused on

the THz gap (Fig. 2.2) with aims to improve device efficiency and increase its

output power levels [15]. In optics, apart from a few electron lasers, which

reached the kilowatt power range [18], other laser sources are limited to milliwatt

and microwatt power levels [8]. In electronics fabrication of 77 & 125 GHz

systems using Gunn diode & MMIC technology have been developed. Although

these currently operate in the mm-wave frequency range, they are being

researched to extend in to the THz region [19].

Fig. 2.2 TeraHertz (THz) Gap [8]

Chapter 2 Terahertz Generation and Applications

2.3 THz Generation

Terahertz sources can be broadly divided into three categories namely, vacuum

tube, optical and solid state sources. The vacuum tube sources are bulky, large

and need huge power to generate substantial electric and magnetic fields and

current densities. Additionally, their physical scaling is very difficult. The optical

sources operate at very low energy levels ~meV. The low energy levels requires

cryogenic cooling due to the effect of lattice phonons [16]. The electronic solid

state sources are preferred for room temperature operation. They are limited due

to parasitics and transit time effects. Their engineering becomes challenging due

to power rolling off exponentially as the frequency increases.

2.4 Solid State two Terminal Active Devices for Terahertz Generation

Millimetre-wave frequencies can be generated using various two-terminal solid

state active devices. The most commonly used devices include Transferred

electron or Gunn diodes (TEDs) [20], Esaki tunnel diodes (TDs) [21], resonant

tunnelling diodes (RTDs) [22] and transit-time diodes. The transit-time diodes

includes Impact Ionization Transit Time (IMPATT) [23], Barrier Injection Transit

Time (BARITT) [24], Tunnel Injection Transit Time (TUNNETT) [25] and Mixed

Tunnelling-Avalanche Transit-Time (MITATT) [26, 27] diodes. All of these

devices display the property of Negative Differential Resistance (NDR).

The transferred electron effect, tunnelling and transit-time diodes are discussed

next;

2.4.1 Transferred Electron Devices (Gun Diodes)

The Gunn Diode named after J. B. Gunn [20] who discovered the effect named

after him, is an active solid state two terminal device, classed as a Transferred

Electron Device (TED). The most conspicuous feature of the device is the

negative differential resistance, which depends on bulk material properties [28].

The Gun diode in its basic form is a homogenous two terminal device with an

ohmic contact at each end. The device has a sandwich-like structure and

Chapter 2 Terahertz Generation and Applications

comprises of n

-n-n

semiconductor materials and is commonly grown using

Molecular Beam Epitaxy (MBE). GaAs and InP are the most commonly used

material systems although other materials such as CdTe, ZnSe, GaAsP and GaN

may possibly also exhibit the transferred electron effect and are being

researched for use at higher frequencies [29]. State of the art GaAs and InP

Gunn diode output powers, and DC to RF conversion efficiencies are shown in

Fig. 2.3. It can be seen that the output power of Gunn devices falls rapidly with

the increase in frequency due to material limitations such as energy relaxation

time. Generally two or more diodes are employed in conjunction with a frequency

multiplier to achieve frequency generation above the fundamental limit e.g. from

200 GHz to 1 THz, with power levels of 0.1 to 1 mW at 400 GHz [6] being

achieved.

Although GaAs or InP Gunn diode devices are simple in physical structure, they

require extreme care during growth for optimization of physical parameters such

as doping concentration and device length. The ohmic contact resistance needs

to be as small as possible, as this can significantly affect device resistance and

thus has considerable effect on the device operating frequency range. Devices

operating in W-band (60 to 110 GHz) have typically specific contact resistivity

values of less than

6 2

5 10

− −

× Ω

, while for W-band and above (i.e. above 110

GHz) values of specific contact resistivity of

7 2

5 10

− −

× Ω

or less are required

[30].

GaAs Gunn diodes currently being commercially fabricated provide moderate

power levels of typically around 60 mW at 94 GHz for GaAs in second harmonic

mode [31]. Due to low phase noise (-88 dBc / Hz at 100 KHz offset) at mm-wave

frequencies, Gunn diodes are considered extremely suitable for Frequency

Modulated Continuous Wave (FMCW) radar and imaging systems. However, the

biggest problem of the Gunn diode is its low DC-to-RF conversion efficiency and

associated high operational temperatures. Due to the high cost associated with

device fabrication, physically-based predictive modelling can be used to

complement experimentally verified devices. A typical model of an advanced

Gunn Diode developed at Manchester, and used in commercial Adaptive Cruise

Control Systems (ACCS) in BMW and Audi cars, is shown in Fig. 2.4.

Chapter 2 Terahertz Generation and Applications

Doping Concentration except launcher

Contact Layer

Graded AlGaAs Launcher

Undoped

Buffer

Substrate

Contact Layer

Transit Region

Device Depth

Doping Spike

Hot Electron Injector

Fig.

Typical Advanced Gunn Diode Structure (not to scale)

Fig.

Compilation of published state

the

art results between 30

and 400 GHz for GaAs and InP Gunn diode

s under CW operation. Legend

format: ‘mode of operation (‘1’ denotes fundamental, ‘2’ second-

harmonic,

etc.), package type, heatsink technology’. Solid lines outline the highest

powers and frequencies achieved experimentally to-

date from each material in

fundamental and second-harmonic [6]

30 100 400

0.1

100

1000

RF Output Power (mW)

Frequency (GHz)

1, Alumina Ring, IHS

1, Alumina Ring, Diamond

2, Alumina Ring, IHS

3, Package unknown

GaAs, InP

denotes e2v technologies result

with hot-electron injection

Novak E. S. Farrington, e2v technologies, Lincoln, UK, September 2009

1, Open Quartz, IHS

1, Open Quartz, Diamond

2, Quartz Ring, IHS

2, Open Quartz, Diamond

2, None, IHS

2, None, Diamond

3, None, IHS

Chapter 2 Terahertz Generation and Applications

2.4.2 Tunnelling Devices

Both Esaki Tunnel diodes (TDs) [21] and Resonant Tunnelling Diodes (RTDs)

[22] fall under this category. These devices generate oscillations by exhibiting

negative differential resistance in their I-V characteristics. They are active solid

state two terminal devices and work on different tunnelling mechanisms. The

TDs comprises of a heavily doped (degenerate) p

-n

junction where interband

tunnelling takes place from the valence band to the conduction band. In RTDs,

the tunnelling occurs in the conduction bands of a double-barrier heterostructure.

In contrast to other solid state two terminal devices, both TDs and RTDs provide

very low output power. The output power is limited by the small voltage swing

(RF) and large junction capacitance. State of the art RTDs have achieved power

levels of 0.3 µW at 712 GHz using InAs/AlSb RTD [32] and 1 µW at 831 GHz

using GaInAs / AlAs double-barrier RTD [33].

2.4.3 Transit – Time Diodes

The transit-time diodes include devices that use special current injection

mechanism and are categorized accordingly. These devices have carriers

injected into a depletion region which drift through the device active region with

the drift velocity. The drift velocity depends on the applied electric field in the

active region. This creates a phase shift between device terminal voltage and

current, which in turn creates a NDR and generates RF oscillations [34].

Various transit-time diodes carrier generation and injection is discussed as

follows;

• In IMPATT diodes [23] carriers are generated and injected due to avalanche

multiplication through impact ionization occurring in a reverse-biased p-n

junction. These devices are known to be capable of providing high power at

mm-wave frequencies. However, they not only require a high current source

to operate but also due to the avalanche effect they have an inherently high

phase-noise. Thus, their application as Local Oscillators (LOs) in FMCW

systems is not a viable option. However, presently Si IMPATT diodes are

Chapter 2 Terahertz Generation and Applications

being used in passive mm-wave radiometric imaging systems as incoherent

noise sources for illumination [3].

• The BARITT diodes [24] generate microwave oscillations by using thermionic

emission of carriers over a forward biased barrier. The barrier is formed due

to p-n or Schottky junction or heterojunction, which contributes to positive

active resistance by forming an RC circuit. Other BARITT diode limitations

include longer active region, small NDR, low output power and efficiency as

compared to IMPATT diodes [34].

• The TUNNETT diodes [25] have carriers injected by tunnelling. The tunnelling

is band-to-band in case of a p-n junction and through the barrier for Schottky

barriers. The TUNNETT diodes have lower noise then IMPATT diodes but are

limited by low output power due to low tunnelling current. However, they can

work at lower operating voltages and theoretically can achieve 1 THz.

Experimentally, it has been shown that the TUNNETT diodes have achieved

power levels of 7.9 mW at 655 GHz [35].

• MITAT diodes [26, 27] use both tunnelling and impact ionization mechanism

for carrier generation. These devices have smaller carrier generation region

and are designed for high frequency operation [26]. Their limitations include

high noise level, small NDR, series resistance and poor impedance matching

characteristics. It has been shown that the MITATT diodes have achieved

power levels of 3 mW at 150 GHz [36].

2.5 Monolithic Microwave Integrated Circuit (MMIC)

MMIC solutions are generally preferred over hybrid technologies due to their

reliability and increased functionality. This was reflected in the 2004 GaAs device

market segmentation, which showed that 83% of the market was taken by MMIC

technology, with the remaining 16% and 1% accounted for by discrete and digital

ICs respectively [19].

In contrast to Gunn diode based systems, MMIC solutions have a compact

design, can be easily integrated into planar circuits, do not require a cavity and

Chapter 2 Terahertz Generation and Applications

can provide a higher degree of functionality. However, commercial MMIC

solutions have yet not reached the output power levels provided by Gunn Diodes

at mm-wave frequencies [37] and generally cost more. Recently, planar Gunn

diodes work has produced promising results [38-41] but still the output power

levels are low. The output power of state of the art planar Gunn has been

reported as 1.58 µW at 115.5GHz [42] .

Commercially available low phase-noise pHEMT-based MMIC chip-sets

generating a frequency of 77 GHz have been developed for automotive and

general radar applications, where they exhibits a slightly lower phase noise than

Gunn diode based oscillators. United Monolithic Semiconductors (UMS) has

developed a MMIC based FMCW chipset using a low frequency oscillator MMIC

(38.5 GHz) and a x2 frequency multiplier MMIC [19, 31].

2.6 Applications of THz Sources

Applications of THz Sources are discussed, including space, security, Imaging

and spectroscopy systems, communications systems, mm-wave automotive

RADAR systems and mm-wave RADAR industrial applications.

2.6.1 THz Space Applications

Receiver diode based heterodyne receivers have been extensively used by radio

astronomers to investigate the composition and chemistry of interstellar matter,

the energy budget of interstellar clouds and the process of star formation [43].

Other important scientific applications include diagnostics for nuclear fusion

experiments and particle accelerators. Technology advancements continue due

to upcoming challenging projects. For instance a new generation ‘Microwave

Limb Sounder’ is being developed by NASA to monitor additional molecular

species (such as hydroxide ion OH¯ ) at frequencies as high as 2.5 THz, which is

expected to be installed on NASA’s ‘Earth Observing System’. Additionally, a

Millimetre Array (MMA) radio telescope is being developed by the National Radio

Astronomy Observatory. The MMA would cover all atmospheric windows from

below 100 GHz through 1 THz. Although, ultra-low noise cryogenic

Chapter 2 Terahertz Generation and Applications

superconductive junctions would be used for the mixers in MMA, LO powers

would still be provided by the GaAs multiplier diodes [43].

Recently a mm-wave diode system was used to detect Space shuttle insulation

foam defects. The system satisfactorily detected defects by using 12 mW, 0.2

THz Gunn diode oscillator [44].

2.6.2 THz Security Applications

The current geopolitical situation has increased the desire to have sophisticated

and reliable mm-wave imaging systems in sensitive places. In particular both

airport and sea port security markets are expected to grow exponentially [37].

Applications of THz imaging include detection of concealed weapons and

recording the signatures of explosives and chemical & biological agents, which

unlike x-ray imaging, causing no harm to the person or object being scanned.

2.6.3 THz Imaging and Spectroscopy Systems

Security screening of people is generally accomplished using stand off

surveillance or portal screening. As an example the Smiths Detection company

has developed a passive TADAR (Tactical Area Defence Alerting Radar) system

with a detection range of 25 metres, the system uses 77 GHz, 94 GHz and 140

GHz sensors [45]. Portal screening generally uses a passive system optimized

for wavelengths of around 77 GHz to 140 GHz). Clothes are transparent at these

wavelengths and denser objects such as plastics and metals become clearly

visible after image processing. Gunn diode oscillators are preferred for LO

frequencies from 94 GHz upwards in both portal and stand off screening

systems. However, higher frequency sources are desirable due to the higher

spatial resolutions achievable.

The detection of explosives and their related compounds has been accomplished

using Time-domain THz Spectroscopy (TDTS). Previous detection was limited to

spectral regions less than 3 THz. However, it has been reported [46] that due to

the advancement of emitters and sensors the spectra has now been extended

from 0.5 to 6 THz. Thus four new explosives compounds, RDX (1, 3, 5 –

Trinitroperhydro - 1, 3, 5 - Triazine), HMX (1, 3, 5, 7 – Tetranitroperhydro -1, 3, 5,

Chapter 2 Terahertz Generation and Applications

7 - tetrazocine), PETN (Pentaerythritol Tetranitrate), and TNT (2, 4, 6 -

Trinitrotoluene) were successfully studied and evaluated [46].

2.6.4 Communications Systems

The relatively low costs associated with point-to-point line-of-sight ‘last mile’

communication links mean they are preferred over optical fibre connections,

especially in urban built-up areas. Last mile communication links are currently

available from Bridgewave Communications Inc. (30 GHz) and E-Band

communications corp. (94 GHz) [19]. The offered per link cost is approximately

one tenth of the MMIC chip-sets fibre link commercially available (UMS Ltd. and

Northrop Grumman corporation) [19]. Operation of these links at higher

frequencies is desirable due to the higher achievable bandwidths and narrow

beam widths. The later being an important security consideration.

2.6.5 MM-Wave Automotive RADAR Systems

Radar is one of the most widely used mm-wave systems, whose spatial

resolution increases with frequency due to narrowing of the achievable

beamwidth. The main utilization of mm-wave radar systems to date has been in

the automotive industry. A recent study by TRW Automotive ® 2011 showed that

in North America road accidents involving heavy vehicles were reduced by 70

percent due to the installation of collision warning and avoidance systems on

vehicles [47]. Radar systems using both Pulse Doppler and FMCW modes,

operating at 77 GHz have been successfully installed and used in top of the

range automobiles for over a decade. Future deployment of Collision Mitigation

Systems (CMS), which would initiate various safety actions such as seatbelt pre-

tensioning and airbags deployment upon detection of a probable collision has

already begun [47].

2.6.6 MM-Wave RADAR Industrial Applications

Another significant FMCW radar application includes its utilization in a factory for

vibration and displacement measurements. In this regard coherent radar

operating at 37.5 GHz with 50 mW power has been developed and tested, which

would be an alternate to traditional piezoelectric sensor measurements [48].

Chapter 2 Terahertz Generation and Applications

2.7 Conclusion

The THz spectrum is being increasingly researched to close the frequency gap

between microwaves and infrared. Terahertz technology and its applications with

an emphasis on Gunn diodes have been presented. It was shown that THz

waves can be used for in the field of astrophysics, plasma physics, materials

science, information science and engineering etc. Applications of THz Sources

were discussed, including space, security, Imaging and spectroscopy systems,

communications systems, mm-wave automotive RADAR systems and mm-wave

RADAR industrial applications.

THz generation was discussed with a review of solid state two terminal active

devices. The commonly used devices discussed included TEDs, Esaki TDs,

RTDs and transit-time diodes. The transit-time diodes presented include

IMPATT, BARITT, TUNNETT and MITATT diodes. It was shown that the Gunn

Diode has long been considered the heart of mm-wave power generation. The

development of Physically-based models is shown to aid in the development of

next generation devices.

The MMIC solutions for THz generation were discussed. It was shown that MMIC

solutions have a compact design, can be easily integrated into planar circuits, do

not require a cavity and can provide a higher degree of functionality. However,

commercial MMIC solutions have yet not reached the output power levels

provided by Gunn Diodes at mm-wave frequencies.

Chapter 3 Multiplier and Harmonic generation for

Terahertz frequency

3.1 Introduction

Gunn and mixer diodes will make a significant contribution to “power” future THz

systems. The THz frequencies considered (up to 600 GHz) are to be generated

in a two-stage module with the initial frequency source provided by the high

frequency Gunn diodes. The output from these diodes is then coupled into a

multiplier module. The realisation of high power Gunn diodes operating at higher

frequency, enable THz generation with a single stage multiplier module. A single

stage multiplier potentially reduces losses and increases output power, whilst

also providing a simpler, lighter and cheaper device for room temperature

operation. The key technology issues are therefore Gunn diode designs and

packaging to provide adequate power at high frequencies and the efficiency of a

single stage multiplier. The multiplier provides higher frequencies by the

generation of harmonics of the input signal by means of a non-linear element. In

this case the non-linear element is provided by a Schottky diode Varactor with

low RC time constants to increase efficiency.

Design considerations of the microstrip and multiplier block are discussed in the

next section. The semiconductor materials for such a complete device including

modelling, design and testing; focusing in depth on the Gunn diode module and

Schottky diode are also presented.

3.2 Gunn Diode Oscillator Design

The accurate physical modelling of Gunn diodes requires them to be mounted in

an oscillator circuit. The circuit provides DC power, and couples out the

generated RF power. The circuit will have reactive elements and a resonant

frequency associated with it. The circuit can therefore be designed to resonate

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

with the Gunn diode in order to optimise output power and efficiency at a given

frequency. At millimetre-wave frequencies the most common circuit configuration

used to do this is the waveguide-based second-harmonic resonant disk

oscillator. A schematic diagram of the cross-section of a second harmonic

resonant disk oscillator is shown in Fig. 3.1-3.2.

In a resonant disk oscillator the packaged diode will typically be positioned on or

near the waveguide floor and the DC power for the diode is supplied through the

Load

Backshort

Bias

choke

filter

Disk

Waveguide

-R

Backshort

distance

Distance

to load

Fig.

quivalent electrical circuit model

of a second harmonic

resonant disk Gunn diode oscillator.

Rectangular

waveguide

Backshort

Resonant disk

Bias choke filter

Gunn diode

package

Gunn diode

Fig.

Simplified schematic of the mechanical outline

of a second

harmonic resonant disk Gunn diode oscillator.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

bias post and choke assembly (often a single piece of machined metal), which is

insulated from the waveguide block. The operational principle is such that three

distinct electromagnetic modes of propagation exist [12, 49]: a TE01 rectangular

waveguide mode (in the rectangular waveguide section) at second harmonic

frequency, a radial mode (between the resonant disk and the waveguide floor),

and a quasi-coaxial (quasi-TEM) mode (between the resonant disk and the

bottom of the choke). Operation is such that a fundamental frequency resonance

exists in the TEM (Transverse ElectroMagnetic) mode along the quasi-coaxial

line between the choke and the disk. The radial disk couples energy between the

magnetic fields of the radial mode and the TE01 waveguide mode at the second

harmonic frequency.

The equivalent electrical circuit for a second harmonic resonant disk oscillator is

given in Fig. 3.2. Here the Gunn diode itself is represented as a capacitance C

and a negative resistance –R

, while package capacitance and bond wire

inductance are modelled as C

and L

respectively. The capacitance C

associated with the resonant disk while the inductance L

is that of the post.

Generally speaking, the impedances of the bias post, disk and cavity are much

greater than those associated with the packaged diode. The frequency of

oscillation is therefore determined mainly by a combination of the resonant disk

diameter, post width, and the length of the post between the resonant disk and

the bottom of the lowest choke section. It is however noted that as frequency is

increased, the impedance’s related to the packaged Gunn diode becomes

increasingly significant compared to those of the post, disk and cavity: it has

been shown (through measurement and simulation) that a small, controlled

change in bond ribbon profile can easily produce a frequency shift of around 800

MHz in a 77 GHz oscillator. In order to obtain maximum second harmonic power

at the output, the dimensions and position of the radial disk must be optimised

not only to obtain, in conjunction with the post geometry, a particular

fundamental frequency resonance in the quasi-coaxial region, but also to

maximise the coupling between the radial mode and the TE01 waveguide mode

at the second harmonic frequency. This coupling is also dependent on the

backshort position: it is noted that in practical oscillators, the coupling between

the radial mode and waveguide mode is detuned (decoupled) from the optimum

to reduce load pull effects. It is also noted that the fundamental frequency is

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

below the cut-off of the rectangular waveguide and so is confined to the quasi-

TEM region.

The Gunn diode operation and hot-electron injection, Gunn diode

electromagnetic modelling, interaction between the diode and oscillator circuit

and proposed modelling techniques are discussed below.

3.2.1 Gunn Diode Operation and Hot-Electron Injection

A brief description of Gunn diode and Gunn diode oscillator operation is included

here as this is required to adequately explain the non-trivial nature of accurate

oscillator simulation. The details including device physics are discussed in

chapters 4 and 5. The Gunn diode can, at its most basic, be thought of as a DC-

to-RF converter: when an applied bias voltage exceeds a certain threshold,

oscillation will occur, the free-running frequency of which will depend on the

material properties and the geometry of the device itself. This is due to the

transferred electron effect exhibited by certain binary and ternary compound

semiconductors.

As an electron is accelerated by an electric field through a transferred electron

effect supporting material, it accumulates energy and so the probability of it

being scattered (transferred) from the central conduction band valley to the

nearest (in terms of energy and momentum) satellite valley increases. This

probability increases dramatically when the electric field reaches a certain

threshold value related to the distance in energy and momentum, between the

conduction band’s central valley and the nearest satellite valley. The scattering

mechanism is columbic in nature and involves a change in both momentum and

energy and so the transfer process is extremely inefficient with large amounts of

energy being lost to phonon excitation (and consequent elevation of the lattice

temperature).

The transferred electron effect results in oscillation because the effective electron

mass (related to the parabolicity of the conduction band profile) is greater in the

satellite valley than in the central valley. This means electron velocity in the

satellite valley due to an applied electric field is lower than that in the central

valley. The effect of this is a ‘bunching’ of electrons as they travel along the

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

length of the device, with those in the satellite valley moving at a slower rate than

those in the central valley. This leads to a depleted region (of electrons) followed

by a region of accumulation (known collectively as a high-field domain), which

sets up an internal electric field opposing that applied externally. The effect of

this is a continuing reduction in the net electric field across the device which

continues until a certain threshold electric field is crossed (preventing the

nucleation of any further high-field domains). At this point the high-field domain

stops expanding and propagates at a constant velocity through the device until it

reaches the anode contact layer at which point it collapses and the electric field

through the device increases again. Oscillation is observed in the current and

voltage at the device terminals as the domain forms at the cathode, expands,

propagates and collapses at the anode, before another domain nucleates at the

cathode.

The oscillation frequency and RF output power level are therefore fundamentally

governed by a combination of the transit region length and carrier concentration:

these define the high-field domain’s transit time through the diode, and the size

(in the direction of the electric field gradient) of the high-field domain. However,

the maximum useable frequency of a transferred electron device is generally

limited by the inter conduction band energy relaxation time: the time constant

governing the return of an electron from the satellite valley to the central valley.

The fundamental physics behind conventional Gunn device operation are

therefore very much centred on electron energy, lattice temperature, and random

scattering processes. These give rise to the following inherent limitations of

conventional Gunn devices:

• The onset of oscillation (turn-on voltage) in a Gunn device varies greatly with

ambient temperature.

• The oscillation frequency is highly dependent on temperature as it is defined

(in part) by the length of the region in which the domain propagates (transit

region). A small portion (referred to as the ‘dead zone’) of this region is

required to accelerate electrons to the point at which they have sufficient

energy to transfer to the satellite valley. The length of the dead zone, and

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

therefore the effective transit region length and oscillating frequency are also

dependent on temperature.

• The random nature of the scattering mechanisms which facilitate domain

nucleation, lead to small variations in the point at which domain nucleation

occurs. This effectively varies the length of the transit region from cycle-to-

cycle leading to enhanced phase noise at the output.

In order to address these issues the use of a hot electron injection structure

based on a graded gap AlGaAs launcher was proposed and demonstrated in the

1980’s [12, 50, 51]. The reasoning behind this was to raise the electron energy

sufficiently to greatly increase the probability of their direct entry to the

conduction band satellite valley upon delivery to the transit region. The injected

electrons have greater energy than those at equilibrium with the transit region

lattice, and so are referred to as ‘hot’. With the majority of the injected electrons

directly entering the satellite valley, the dead-zone is effectively eliminated along

with the temperature dependency of the oscillation frequency and turn-on

voltage. In addition the overall conversion efficiency is increased (due to the

reduced parasitic positive resistance) and because the random factors of domain

nucleation are reduced, the phase noise is decreased.

3.2.2 Gunn Diode Electromagnetic Modelling

In order to accurately model the behaviour of a waveguide oscillator circuit, a full-

wave, 3D high-frequency electromagnetic solver, such as Ansoft (ANSYS

) High

Frequency Structure Simulator (HFSS

), is required. To this end, numerous

HFSS

models of different second harmonic resonant disk oscillators that

correspond to the oscillators used to obtain measured results have been

developed at e2v Technologies Plc. These are typically 3-port models with ports

at the output waveguide, the top end of the choke (to evaluate filtering

effectiveness at both fundamental and harmonic frequencies) and at the Gunn

diode location. The Gunn diode port is, as in [52-54], set up as a coaxial

transmission line (the effects of which are de-embedded post-process) to

accurately represent the radial mode in the region of the diode and the resonant

disk, and allow the impedance in the region of the diode to be evaluated.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

Fig. 3.3a shows an ANSYS

High Frequency Structure Simulator (HFSS

)

model and electric field plot of a second harmonic resonant disk oscillator

configuration using a circular waveguide contacting back short and frequency

tuning pin. Although not showing the resonance in the quasi-coaxial TEM mode

(the plot illustrates the electric field at second harmonic frequencies, not

fundamental), Fig. 3.3b shows power being coupled between the diode and the

waveguide output (left), and a degree of interaction between the resonant disk

and the waveguide floor. For future work it is proposed to use these models to

establish the diode embedding impedance at both fundamental and second

harmonic frequencies. This will enable equivalent lumped element circuits for the

fundamental and second harmonic resonators to be defined, which can then be

coupled with the SILVACO

model for subsequent time-domain simulations

(chapter 9).

Oscillators were modelled and simulated at e2v to assess the performance of

resonant disk Gunn diode oscillators (Fig. 3.4). It was believed that a reduced-

height waveguide oscillator could be used to increase the output power obtained

from the current D-band oscillators. An initial design was outlined and finalized

for fabrication. Additionally, Quartz packages (Fig. 3.5b) were tested at e2v for

(a) (b)

Fig.

(a)

Ansoft HFSS

model of a second harmonic resonant disk millimetre

wave oscillator with tuning pin and circular waveguide sliding backshort. (b)

Simulated

field electric field plot for the oscillator at second harmonic frequencies.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

use at frequencies above 125 GHz allowing increased power and frequencies

than achieved with standard alumina packages (Fig. 3.5a). Power combining of

two GaAs Gunn oscillators was achieved at 62.5 GHz with 210 mW output (Fig.

3.5c). That was intended for use as a high-power source for a frequency

multiplier.

3.2.3 Accounting for Interaction between the Diode and Oscillator Circuit

Previous modelling efforts have been concentrated on electromagnetic (EM)

modelling of the oscillator using idealized approximations for the diode [12, 55-

57]. Most advanced diode modelling work typically used either a hydrodynamic

(c)

Fig.

Gunn diode packages developed at e2v (a)

tandard

lumni

ackage

DA810xx Bias-tuned Oscillators (b) Quartz Package.(c) Power Combining

(a)

• 125GHz ± 400MHz

• 20mW min. output power

• Phase Noise <-75dBc/Hz at

100kHz offset

• -40°C to +70°C

• 76.5GHz ± 400MHz

• 40mW min. output power

• Phase Noise <-80dBc/Hz at

100kHz offset

• +10°C to

•

62.5GHz ± 200MHz

•

100 mW min. output power

•

Phase Noise <-80dBc/Hz at

100kHz offset

•

+10°C to +70°C

(b)

(a) (b)

Fig.

Ansoft HFSS

simulations of a

reduced

height waveguide

model

developed at e2v (a) second harmonic resonant disk millimetre-wave oscillator (b)

simulated field electric field plot.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

based approach [9, 57, 58] or Monte Carlo simulation techniques [59-61]. A

novel approach is proposed here which will be the first time a semiconductor

device modelling technique has been combined with an accurate 3D

electromagnetic modelling tool to accurately compute the behaviour of a Gunn

diode mounted in an oscillator.

Work has been reported where efforts have either been concentrated on EM

modelling of the oscillator using idealised approximations for the diode, or

concentrated on modelling the diode with approximations used for the oscillator

circuit. Of the work reported on the EM modelling of second harmonic Gunn

diode oscillators (mainly using HFSS

), either the impedance of the Gunn

device is estimated as a static value [52], or a non-linear harmonic balance

simulator has been used to estimate the reactance of the device mounted in an

equivalent circuit [54]. Although the latter [54] attempts to incorporate the effects

of the dynamic reactance of the device in the EM simulations, the device model

used in the harmonic balance simulation was an idealised spice model (i.e. not a

true semiconductor device modelling tool) as the research focussed mainly on

the optimisation and behaviour of wideband tuneable Gunn oscillators rather

than evaluating the performance of the Gunn device itself. However, the latter

[54], represents the most accurate modelling technique yet demonstrated,

predicting oscillation frequency to a few GHz and RF output power to within 20%.

The majority of the work that concentrated on modelling the diode itself, which

typically used either a hydrodynamic based approach [57] or Monte Carlo

simulation techniques [12, 55, 59-62], generally relies on estimated values for

the parameters of an approximate equivalent oscillator circuit. These

investigations centred on analysis of the behaviour and performance of the diode

in isolation rather than in conjunction with an accurately modelled equivalent

circuit: this is partly due to the difficulties in combining Monte Carlo simulation

results with external circuit models due to the solution noise generated by the

stochastic nature of the technique. The aim of the majority of previous work was

to enable a study of domain formation and to compare the performance of

different device variations, and not to enable accurate modelling of the Gunn

diode and associated oscillator as a whole.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

3.2.4 Proposed Modelling Techniques

The oscillator circuit effectively determines the oscillation frequency and so in

turn affects both the dynamics of the high-field domain within the device’s transit

region and therefore the dynamic reactance of the device itself. Here, due to the

cycle of domain nucleation, growth, propagation and collapse during each

oscillation cycle, the device capacitance (itself partly dependent on the oscillator

circuit parameters) varies throughout the oscillation cycle. It is therefore not

accurate to simply extract a small-signal s-parameter model of the Gunn device

from the SILVACO

model and use this in conjunction with the HFSS

model

of the oscillator (i.e. terminate the diode port in the HFSS

simulation with a

SILVACO

generated s-parameter matrix). This would not enable the effects of

the oscillator circuit to be taken into account when computing a device’s time-

domain response.

As SILVACO

has the ability to include the effects of an external circuit on a

simulated device through the definition of a lumped element circuit in which the

device can be embedded, the effects of the external circuit can be considered

during computation of device behaviour. The equivalent frequency-domain

behaviour of a second harmonic oscillator circuit can therefore be simulated

using HFSS

at the fundamental oscillation frequency and the equivalent circuit

parameters at this frequency extracted: this can then be used to define a suitable

circuit model for inclusion in the SILVACO

simulations. From simulation of the

same HFSS

model at second harmonic frequencies, the values of the

equivalent circuit elements at this frequency can be extracted and used in a

lumped element equivalent circuit model to estimate the conversion efficiency to

second harmonic. This will allow estimation of the second harmonic output power

for a device mounted in the oscillator. This will ultimately enable the effects of

both variations in the semiconductor device and the oscillator geometry to be

evaluated simultaneously.

The HFSS

models for second harmonic resonant disk oscillators described in

section 3.2 accurately represent the geometry and materials of the circuit

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

including the package reactance values. The Gunn diode itself is currently

modelled using conductive GaAs for the contact layers and launcher, and bulk

semi-insulating GaAs (representing the depletion region in the high-field domain)

for the transit region. From the resulting small-signal frequency-domain s-

parameter file, all equivalent circuit element values (with the exception of the

diode parameters C

and -R

) can be accurately extracted at both fundamental

and second harmonic frequencies. It is noted here that modelling the whole

transit region as dielectric (representing the depletion region of the high-field

domain) is an approximation: in reality the depletion region does not fill the whole

transit region, only a varying portion of it. The depletion region increases through

the oscillation cycle as the high-field domain grows until, depending on the length

and doping of the transit region, it reaches a constant value that traverses the

length of then transit region, or (for short transit region devices) potentially

continues to grow until it collapses at the anode. This will influence the shape of

the fundamental frequency waveform and therefore its harmonic content

meaning it will also affect the level of second harmonic power that can be

extracted by the resonant disk circuit.

As the frequency of operation increases, not only do the impedances related to

the packaged diode become increasingly significant in relation to those of the

circuit, but in the shorter (sub-micron) transit region lengths the required mean

spread in variation of device reactance over the oscillation cycle will also

increase. This is due to a smaller percentage of the cycle time being devoted to

the steady propagation of the high-field domain through the transit region (i.e.

more of the cycle time will be consumed by domain formation and collapse as

opposed to steady domain propagation through the device). It is currently

thought that neglecting this variation when simulating the EM model will not

significantly affect the behaviour of the EM model produced, so long as an

accurate approximation of the diode structure is made. To achieve this, the

micron-scale transit region will be modelled as a suitably conductive bulk GaAs

material with an embedded insulating layer representing the high-field domain.

The size and position of this layer, along with suitable material properties

(permittivity and loss factor) can be estimated from an initial, free-running

SILVACO

simulation.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

As the dynamic nature of the depletion region is not negligible when simulating

the EM behaviour of the oscillator circuit (leading to degradation in accuracy of

the model as a whole) the growth of the high-field domain will need to be

adequately represented within the EM model. A proposed technique to achieve

this is to approximate the continuous process of domain growth and propagation

in the HFSS

model by treating it as a sequence of discrete stages, each with

an individual HFSS

model in which the length and/or the position of the

dielectric region is altered. The size, position and material properties of the

depletion region would again be estimated from an initial SILVACO

result. A

time-varying series of equivalent circuit parameters could then be extracted, and

using curve fitting, continuous functions approximated for each of the oscillator

circuit spice model parameters. Although initially seeming labour intensive, this

process could be simplified through exploitation of the parametric sweep

functionality of HFSS

. Suitable code would then be written to update the

values of the equivalent spice circuit model used to represent the oscillator circuit

in the SILVACO

time-domain simulations, as the oscillation cycle progresses.

This approach is proposed as future work while developing and designing high

power THz sources using multipliers.

3.3 Frequency Multipliers

Multipliers were designed at e2v and several key doubler components (DC bias

filter, output waveguide taper, and a broadband cross-waveguide coupler) were

modelled for output at 140 GHz. The approach allowed verifying the design at

relatively low frequency, before it could be scaled to higher frequencies.

The fabrication technique used for the multiplier split block waveguide circuit is

dictated by the smallest feature size (typically the reduced height waveguide

required for the broadband performance of the cross-waveguide coupler). Thin-

film production of quartz circuits was routinely carried out at e2v on 127 µm

substrates. It was thought that for prototyping and higher-frequencies, circuits

using thinner substrates could be fabricated using standard semiconductor

processing techniques.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

The design methodology used to design multiplier is shown in Fig 3.6 and

discussed below:

• Based on the operational frequency and power requirements, a suitable

diode or diode array geometry is identified. The diode size and number of

anodes are determined.

• The diodes C-V, I-V and RF characteristics are measured. The SPICE

parameters (series resistance, series current, built-in potential, ideality factor,

and capacitance) are fitted to the measurements.

• Non-linear circuit analysis is performed using harmonic balance simulator

tool such as Microwave Office ® by AWR. The harmonic balance simulator

determines optimum diode embedding impedance, especially for resistive (real)

impedance component. It calculates maximum theoretical conversion efficiency

Non

line

ar circuit

analysis

(Harmonic Balance

using MWO

)

Linear electromagnetic

structure simulations

(FEM

, HFSS

)

Diode Impedance

Generalized n

port

s-matrix

•

Linear circuit simulations

AWR Microwave Office

(MWO®)

• Circuit tuning

•

Input impedance

• Diode mount impedance

• Power balance between

diodes

Fig.

Frequency

multiplier d

esign methodology

developed at e2v showing

both hydrodynamic and liner electromagnetic structure modelling.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

and power. It can be used to study the effects of chip parasitics on maximum

theoretical efficiency.

• HFSS

is used to create a model of the diode chip and the waveguide

circuit in which it is embedded. It enables impedance seen at the individual

anodes to be evaluated.

• Simulation and tuning of entire multiplier circuit.

Millimetre-wave frequency multiplication using Schottky diode technology design

methodology was extensively verified against published designs and data at e2v.

The harmonic balance simulation tool, Linear EM Structure Simulations and

Schottky diode Varactor are discussed;

3.3.1 Harmonic Balance Simulation Tool

A harmonic balance simulation tool was created using Microwave Office ® by

AWR (Fig. 3.7), specifically for analysis and prediction of GaAs Schottky diode

performance given device’s basic electrical characteristics. The extensive

comparison to published work at e2v allowed accurate prediction of maximum

conversion efficiency for a given device, under specified conditions. The tool is

fully parametric and can be used to study the effects of operating conditions etc.

on the final performance (Fig. 3.8), as well as calculating the optimum

embedding impedances in the multiplier circuit. The results of the harmonic

balance simulations can therefore be used along with electromagnetic simulators

to design and tune circuit parameters for optimal circuit performance. Other

circuit components such as input and output waveguide tapers, microstrip-to-

waveguide coupling structures, and hammerhead bias filters have all been

designed at e2v Technologies Plc. at frequencies up to 140 GHz.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

Fig.

Result of a harmonic balance parametric sweep to study output power

with variation in input embedding impedance (with other operating conditions fixed) P

= 21 mW at 100 GHz, F

OUT

= 200 GHz

Fig.

A harmonic balance simulation tool

schematics created

using

Microwave Office ® by AWR

, specifically for analysis and prediction of GaAs Schottky

diode performance given device’s basic electrical characteristics.

IND

ID=L1

L=lf pH

DCVS

ID=V1

V=bias V

RealSweep=30

ImSweep=50

Cmmwave=2.24

Cpp=0

lf=10

bias=3.7

ZinR=8 ZinIm=52

ZoutR=13 ZoutIm=29

ZinNorm=complex(ZinR, ZinIm)/50

ZoutNorm=complex(ZoutR, ZoutIm)/50

GammaN=(ZinNorm-1)/(ZinNorm+1)

GammaOUT=(ZoutNorm-1)/(ZoutNorm+1)

M1=abs(GammaN)

A1=deg(angle(GammaN))

M2=abs(GammaOUT)

A2=deg(angle(GammaOUT))

Xn. . .

SWPVAR

ID=1

VarName="ZoutR"

Values=stepped(1, RealSweep, 1)

UnitType=None

PORT1

P=1

Z=50 Ohm

Pwr=14.5 dBm

1 2

CIRC

ID=U1

R=50 Ohm

LOSS=0 dB

ISOL=1000 dB

PORT

P=2

Z=50 OhmPORT

P=3

Z=50 Ohm

Xn. . .

SWPVAR

ID=2

VarName="ZoutIm"

Values=stepped(1, ImSweep, 1)

UnitType=None

M_PROBE

ID=VP1

3:Bias

HBTUNER2

ID=OutputTuner

Mag1=1

Ang1=180 Deg

Mag2=M2

Ang2=A2 Deg

Mag3=1

Ang3=180 Deg

Fo=100 GHz

Zo=50 Ohm

3:Bias

HBTUNER2

ID=InputTuner

Mag1=M1

Ang1=A1 Deg

Mag2=1

Ang2=180 Deg

Mag3=1

Ang3=180 Deg

Fo=100 GHz

Zo=50 Ohm

CAP

ID=C2

C=Cmmwave fF

CAP

ID=C1

C=Cpp fF

SDIODE

ID=SD1

100

(

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

3.3.2 Linear EM Structure Simulations

Other circuit components such as input and output waveguide tapers, microstrip-

to-waveguide coupling structures, and hammerhead bias filters have all been

designed at e2v Technologies Plc at frequencies up to 140 GHz. Initial designs

outlining the geometry of Schottky diode arrays suitable for use in frequency

doublers and triplers have been produced (Fig. 3.9).

HFSS

was used to create a model of the diode chip and the waveguide circuit

in which it is embedded, which enables impedance seen at the individual anodes

to be evaluated (Fig. 3.9a). Varactor Chip Waveguide Mount Modelling (Fig.

3.9b) shows diode chip model embedded in a waveguide diode mount.

3.3.3 Semiconductor Component – Schottky Diode Varactor

The harmonic oscillations of the input signal created in the multiplier are

produced by a non-linear component, in this case a Schottky diode Varactor. In

order to achieve a low RC time constant, both the capacitance and series

resistance need to be kept as low as possible. In practice this is achieved by

using small area anodes for the lowest possible capacitance, and although this

(a)

(b)

Fig.

(a)

Varactor Diode Chip

and the waveguide circuit

model

created

using

HFSS

at e2v Technologies Plc (b) HFSS

iode chip model embedded in waveguide diode

mount.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

does increase the series resistance, the overall effect is still to reduce the RC

time constant in properly designed structures. The challenge is then to reduce

the series resistance as much as possible for a specific Schottky contact.

The device structure shown in Fig. 3.10 consists of a traditional GaAs diode

structure topped by a heavily doped region of In

1-x

As graded from x=0 at the

GaAs interface to x=0.53 at the upper surface. A subsequent heavily doped layer

was grown to form the low resistance non-alloyed Ti/Au Ohmic contact. The

fabrication process was such that the metallisation layer used to form the Ohmic

contact was also utilised to provide the Schottky contact, after the appropriate

patterned wet recess etch though the InGaAs layers to the GaAs. A

methane/hydrogen RIE (Reactive Ion Etching) step was then used to etch

through the lightly doped GaAs layer to the highly doped conductive region

below (conventionally the Ohmic contact layer with alloyed contacts). Importantly

the contact metals are used as an etch mask. An air bridge (2 µm wide) was then

formed between the anode and a bond pad to complete the device.

Titanium 0.04um + Gold 0.5µm

Substrate GaAs 1x0

-3

8µm

Buffer GaAs undoped 0.05µm

Etch Stop AlAs undoped 0.08µm

Membrane GaAs 1x10

-3

3µm

N+ GaAs 5x10

-3

3µm

Contact (n) GaAs 1.8x10

-3

0.35µm

Graded InGaAs In 53% to 0% 5x10

-3

0.0605µm

Ohmic n++ InGaAs In 53% 5x10

-3

0.033µm

100µm

106µm

2µm

Tn+Au

Schottky

4µm

Cathode

Anode

Fig.

GaAs diode structure topped by a heavily doped region of In

1-x

graded from x=0 at the GaAs interface to x=0.53 at the upper surface.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

Developing such a device, the doping and thickness of the GaAs layer beneath

the anode becomes critically important in reducing the series resistance. Carriers

have to pass through the undepleted region of this layer to the highly doped

GaAs region as in a conventional Schottky, but then back through this lightly

doped region to the InGaAs contact layer. Calculating the exact value of the

depletion region at breakdown is therefore of importance in reducing the series

resistance. Increasing the doping in this region reduces the resistivity of the

material and the thickness of material required (reduced depletion width) further

reducing the series resistance. Unfortunately the breakdown voltage is also

reduced. However, using the anodic metal as a mask in the RIE means that the

maximum breakdown voltage is achieved by increasing uniformity of current

density under the anode. Effective balancing of these parameters provides

optimum device performance. Shown below (Fig. 3.11a) is the C-V plot for a 4

µm device obtained by extracting the S11 parameter from RF testing and fitting

the data to an ADS

simulation at each bias point with constant values of

inductor, resister and co-planar waveguide elements in the simulation.

These small area Schottky diodes were fabricated and improved over several

iterations to give a reduced RC time constant producing a cut-off frequency

approaching 900 GHz (thus operation up to ~300 GHz). The design was

developed to utilise a non-alloyed self aligned contacts structure with the

smallest anode diameter of 4

µm and a thin 2 µm contacting air-bridge (see Fig.

(b)

Fig.

(a)

plot of 4 µm Schottky device

Junction capacitance of fabricated

single anode device (bias - 1/C

inset) (b) image of a 4 µm Schottky device with 2

µm bridge

to the anode.

(a)

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

3.11b). It produced low contact resistance and capacitance with a still relatively

high breakdown voltage. That material was decided to be used to fabricate

smaller anode devices for reduced RC time constant and higher frequency

operation.

At the University of Manchester new process techniques were developed for

undercut air bridges to reduce device parasitics, process complexity and series

resistance. Yield and uniformity was increased in order to repeatedly deliver the

more complicated multiple device anti-series structures required and thick (> 2.5

µm) lift-off processes were developed. Those processes allowed the design of

new devices in collaboration with multiplier design requirements across a wide

range of chip size and device combination totalling over 700 individual device

structures (such as Fig. 3.12). Junction capacitance was extracted from RF

testing indicating 4.8 fF at breakdown (Fig. 3.11). These devices gave a

theoretical cut-off frequency in excess of 1.1 THz.

3.4 Schottky Diode SILVACO

Modelling

A Schottky diode 2D model has been developed in SILVACO

. Its epitaxial

structure is shown in Fig. 3.13. The conduction band diagram of both ohmic and

Schottky contacts are shown in Fig. 3.13b and 3.13c. The models and material

parameters are given in Appendix 2.

(b)

(a)

Fig.

Six

anode

anti

series device single anode (full circuit inset)

(a)

design (b)

fabricated device.

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

The modelled versus measured results matched very well (Fig 3.14 - 3.16). The

Schottky contact breakdown voltage analysis is shown in Fig. 3.14. The model

Fig.

Forward bias I

V curve

for a 4 µm diameter anode Schottky diode

Series resistance

was calculated from curve’s slope.

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

100

110

120

Modelled fwd current

Measured fwd current

Anode Current (uA)

Voltage (V)

Titanium 0.04um + Gold 0.5µm

N+ GaAs 5x10

-3

3µm

Contact (n) GaAs 1.8x10

-3

0.35µm

Graded InGaAs In 53% to 0% 5x10

-3

0.0605µm

Ohmic n++ InGaAs In 53% 5x10

-3

0.033µm

µm

54µm

2µm

Tn+Au

2µm

Schottky

Conduction band energy (eV)

Conduction

band energy (eV)

SILVACO

Anode Schottky

contact Conduction band

diagram at 0V

SILVACO Cathode graded

InGaAs ohmic contact

Conduction band diagram at 0V

Device width µm

Conduction Band

Fig.

GaAs Schottky diode SILVACO

2D model. Equal area rule was

maintained same as the manufactured device (a) 2D device structure for a

4 µm diameter

anode Schottky diode (b) Conduction band diagram for GaAs

Schottky contact (c).

Conduction band diagram at zero bias showing graded InGaAs ohmic contacts

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

gives a series resistance of 6 Ohms, substantially less than the measured value

of 9.5 Ohms. This is attributed to external device resistances and modelled

device rectangular geometry. The model can be used to study underlying device

physics and extract parameters such as ideality factor, zero voltage junction

capacitance, built in voltage and breakdown voltage. These parameters will be

used in harmonic balance simulations created using Microwave Office ® by

AWR, which will allow analysing and prediction of the GaAs Schottky diode

performance at mm-wave frequencies.

The modelled Schottky diode capacitance-voltage (C-V) plot matched very well

with measured data (Fig. 3.15). The junction capacitance indicated 4.6 fF at

breakdown and a C

of ~17 fF giving a theoretical cut-off frequency in excess of

1.1 THz.

Three Schottky models listed in table 3.1 were developed and compared with

measured results (Fig. 3.16). They were designed in collaboration with multiplier

design requirements for ~120 GHz output. An additional area was included in

modelled devices that were in line with fabricated devices. The C-V plots were

compared with measured results and junction capacitance C

was determined.

The cut-off frequencies were calculated as shown in table 3.1

Fig.

plot for a 4 µ

m diameter anode Schottky

diode

-11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0

Modelled Capacitance

Measured Capacitance

Capacitance

(

)

Voltage (V)

Anode diameter 4um

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

Parameters

Schottky

node diameter (d)

14µm 13µm 12µm

Anode area calculated using

154 133 113

Anode additional area (A) added in SILVACO

model, which is due to processing in

49 29 28

Total Area (d+A) in

203 162 141

Capacitance (C

)

obtained from

plot in

286 223 198

Cut off frequency

s j

R C

GHz

with

0.9

= Ω

618

793

893

Table 3.1 Schottky diode models with 14µm, 13µm and 12µm anode diameters

showing an additional area and calculated cut off frequencies.

-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0

100

125

150

175

200

225

250

275

300

Capacitance

(

)

Voltage (V)

Modelled Capacitance

Measured capacitance

Anode 14um dia (d) + area (A) ~ 203 (154+49)

Anode 13um dia (d) + area (A) ~ 162 (133+29)

Anode 12um dia (d) + area (A) ~ 141 (113+28)

Fig.

Schottky diode C

V plots for 14

m, 13

m and 12

m anode diameters

Schottky anode

Chapter 3 Multiplier and Harmonic generation for Terahertz frequency

3.5 Conclusion

The Gunn diode operation and hot-electron injection, Gunn diode

electromagnetic modelling, interaction between the diode and oscillator circuit

and proposed modelling techniques were discussed. Two novel modelling

techniques to merge SILVACO

and HFSS

models and to account for

interaction between the Gunn diode and oscillator circuit were presented. The

techniques include the lumped element model for SILVACO

and taking into

account Gunn diode domain growth and propagation in the HFSS

model

Multipliers designed at e2v and several key doubler components (DC bias filter,

output waveguide taper, and a broadband cross-waveguide coupler) modelled

for output at 140 GHz have been presented in detail. Oscillator cavity developed

in HFSS

and harmonic balance simulation tool created in Microwave Office ®

by AWR at e2v Technologies Plc. were discussed. The harmonic balance

simulation tool, Linear EM Structure Simulations and Schottky diode Varactor

were presented.

Both semiconductor components of the multiplier namely the Gunn diode and the

Schottky diode have been presented. It was shown that the modelled Schottky

diode C-V plot matched very well with measured data. The junction capacitance

indicated 4.6 fF at breakdown and a C

of ~17 fF giving a theoretical cut-off

frequency in excess of 1.1 THz. Additionally, three Schottky models, designed in

collaboration with multiplier design requirements for ~120 GHz output, were also

compared with measured results. An additional area was included in modelled

devices that were in line with fabricated devices. The C-V plots were compared

with measured results and junction capacitance C

was determined. The cut-off

frequencies were shown.

Chapter 4 Gunn Diode Theory

4.1 Introduction

The Gunn Diode, named after J. B. Gunn, who discovered the Gunn Effect, is an

active solid state two terminal device, classed as Transferred Electron Device

(TED). The most conspicuous feature of the device is the negative differential

resistance, which depends on the bulk material properties [28]. In this chapter

after covering the basic Gunn diode theory, the negative differential resistance

phenomenon will be analyzed with relation to material band structure. Finally the

various device operating modes will be discussed with emphasis on the high

frequency oscillations before summarizing the limitations of conventional Gunn

diodes.

4.2 Gunn Diode as Transferred Electron Effect (TED) Device

The Gunn diode can, at its most basic, be thought of as a DC-to-RF converter:

when an applied bias voltage exceeds a certain threshold, voltage oscillation will

occur, the free-running frequency of which will depend on the material properties

and the geometry of the device itself. This is due to the transferred electron effect

exhibited by certain binary and ternary compound semiconductors. The

transferred electron effect is reviewed in detail through Electric Field Effects on

Electron Drift Velocity, High Field Transport in n-GaAs, Negative Differential

Resistance (NDR) and Conditions for Negative Differential Mobility.

4.2.1 Electric Field Effects on Electron Drift Velocity

At low electric-fields the electron drift velocity is linearly proportional to the

applied electric field. However, it departs from the linear relationship with the

increase in applied electric-field, which causes the drift velocity to approach the

thermal velocity value. Before considering the high electric field effects the

Chapter 4 Gunn Diode Theory

following gives a brief account covering the essentials of low field electron drift

velocity and its corresponding high field effects [3].

The low field electron drift velocity is given as:













−=

cm/s ………………….…………..……………..(4.1)

where

is electron drift velocity,

is the applied electric field,

is the mean

free time and

is the effective electron mass. The drift velocity is proportional

to the electric field and the proportionality factor is known as the mobility, which

is given by:

………………………………………………………….. (4.2)

Thus the drift velocity can be written as:

−

………………………….……..……………………..... (4.3)

where the negative sign indicates that the electrons move in the opposite

direction to the applied electric field. The mobility determines the influence of

electric field on the motion of an electron. At low electric-field values the drift

velocity of electrons is linearly proportional to the applied electric field and is

smaller than the thermal carrier velocity (which is

scm /107.7

[1] for n-type

GaAs). As the electron drift velocity approaches the thermal velocity, the linear

relationship due to the corresponding constant mobility breaks down. The drift

velocity increases with the increase in electric field until reaching a saturation

velocity value. The plots of carrier drift velocity versus electric field for Gallium

Arsenide (GaAs), Indium Phosphide (InP) and Gallium Nitride (GaN) are shown

in Fig. 4.1, where it can be seen that the drift velocity after reaching a maximum

value decreases with an increase in applied field.

Chapter 4 Gunn Diode Theory

Note that the threshold field at which NDR occurs for GaN is very high implying

large voltages for operation (about 50 times larger than those of GaAs).

(a) (b)

(c)

GaAs InP GaN

Fig.

Electron

ccupations for (a) GaAs, (b) InP and (c) GaN

[3]

Fig.

Carrier Drift Velocity verses Electric Field graph

[1]

Chapter 4 Gunn Diode Theory

4.2.2 High Field Transport in n-GaAs

The negative differential mobility phenomenon can be explained through

consideration of the band diagrams shown in Fig. 4.2. Fig. 4.2a shows that GaAs

is a direct band gap material with an energy gap, between the top of valence

band and bottom of conduction band, of 1.43 eV at k=0 (zero wave vector). The

satellite conduction band minimum (L) is 0.32 eV higher than the conduction

band minimum (Γ) and is at

2 /

k a

= ±

(where a is the lattice constant). At 300K,

n-type GaAs has its valence band highly occupied with few electrons in the

conduction minimum (central

band) and nil in the satellite L band. However, as

the increase in applied electric field reaches ~3kV/cm, some electrons from the

valley gain sufficient energy and are scattered into the L valley.

As shown by equation 4.2, electron mobility is dependent on effective electron

mass, which is different in the different conduction band valleys and depends on

the local curvature of the band structure. The effective electron mass is given by

[29]:













∂

∗



…………...……………………….………………(4.4)

Where

∗

is the effective mass,

2/h



is planks constant and k is the wave

vector. In the satellite valley the curvature is higher than the central valley, which

results in a higher effective mass compared to the central

valley. For GaAs,

electrons have an effective mass of

4.0 m

(

∗

) in the L valley and

068.0 m

(

∗

)

in the

valley [2].

The density of states of each

and L valley represent a number which is

proportional to energy interval available to the conduction electrons and number

of allowed energy states per unit volume. Its ratio, R, is given by:













∗

R ……………………………..……………………….(4.5)

Chapter 4 Gunn Diode Theory

Where

and

are the equivalent

and L valley whose values for GaAs

are 1 and 4 respectively [34]. Thus the density of states ratio becomes R=94 [34,

63], which is much higher than the corresponding

valley. In the L valley in

addition to the higher electron effective mass, the electrons are also subjected to

stronger scattering processes [64]. These effects collectively result in lower

electron mobility (up to 70 times) in the L valley compared to the

valley.

4.2.3 Negative Differential Resistance (NDR)

In Fig. 4.3a applied electric-field is low (

) thus all the electrons are in the

central valley. As the field is increased (

) some electrons gain the

energy required to transfer to upper valley as shown in Fig. 4.3b. Finally when

the field is large enough (

) effectively all the electrons are transferred to

the upper valley. The effect of this on the drift velocity is shown in Fig. 4.4, which

can be equated as [3]:

≅

when

……….………………….………………(4.6)

≅

when

…………….…………………………….….(4.7)

(a) (b) (c)

Fig. 4.3 Electron occupations under various Electric Field levels for n-GaAs [3]

Chapter 4 Gunn Diode Theory

The electron drift velocity increases linearly until the onset of negative differential

resistance where the electric filed value is given as

. The drift velocity

continues to decrease until the field acquires the maximum value

as shown in

Fig. 4.4.

It can be seen from Fig. 4.4 that

is greater than

, which results in a

decrease of drift velocity between

and

, The NDR exists between the

threshold field (

) and final valley field (

) as shown in Fig. 4.4 [1, 3].

If the electron mobility, effective mass and density for both central (

band) and

satellite (L band) valleys are denoted by

111

,, nmu

and

222

,, nmu

respectively then

the steady state conductivity of n-type GaAs is given as [3]:

uqnnunuq =+= )(

2211

………………………………....…………(4.8)

Where

is the average mobility and is given by:

)(

2211

nunu

………………………………………….…………(4.9)

qnu

Electric Field KV/cm

Carrier Drift Velocity cm/s

Fig. 4.4 Negative Differential Resistance region [3]

Chapter 4 Gunn Diode Theory

Thus the average drift velocity can be written as:

……………………………………….……….……………(4.10)

4.2.4 Conditions for Negative Differential Mobility

For a material to exhibit the transferred electron effect and therefore negative

differential mobility, it must display the following characteristics [65]:

• The material satellite or L band energy minimum is above the normal

conduction band (

) minimum.

• The E verses k curvature for L band is less than the

band minimum,

thereby providing greater effective mass values.

• The density of states in L band has a higher value than the

conduction

band.

• The energy gap between

and L valley must be greater than kT to avoid

occupation of L valley due to thermal occupations at low electric field: This

ensures that the occupation of L band is due to the applied electric-field.

• The fundamental energy gap,

, must be greater than

∆

to avoid impact

ionization of electrons across

before inter-valley transfer occurs.

• Some mechanism must be present to supply the k-value (momentum) change

required for transition to the satellite valley.

Besides GaAs materials such as InP and GaN depict the above characteristics

and thus are the favoured materials choices for the implementation of

Transferred Electron Devices (TED) [1, 3, 66].

4.3 Instability and Domain Formation

Semiconductors such as GaAs and InP, which exhibit the Transferred Electron

Effect are inherently unstable. The instability is due to the formation of a space

Chapter 4 Gunn Diode Theory

charge region that grows exponentially with time due to a small, random change

in carrier density in the semiconductor. This instability has been explained in [2],

and is discussed below:

In a uniformly doped semiconductor device with an electron concentration

, a

noise process or variation of doping level or some crystal defects can cause a

spontaneous fluctuation in the electron density (

). This spontaneous

fluctuation forms a dipole, consisting of an accumulation and a depletion region

with associated electric field gradient. The non-uniformity in the space charge is

related to the electric field by the following Poisson equation:

NNq

∂

=−

εε

)(

………………………………………………(4.11)

where

and

are the vacuum and relative permittivity respectively, and

the doping density. If the mean electric field is less than the threshold field then

electrons subjected to a higher electric field will move faster than those in the

low-field region. The space charge would then fill the depletion region resulting in

the dampening of the fluctuation due to dielectric relaxation.

Now considering the case when the applied electric field is greater than the

threshold electric field

: the electron drift velocity reduces in the region of

higher field, causing the space charge fluctuation to grow along with time

associated local electric field in the region as shown in Fig. 4.5b. This

exponential growth of electric field continues until a stable domain is formed as

shown in Fig. 4.5c. Stability arises because to keep the total voltage drop across

the device the same, the electric field outside the domain remains below the

threshold level. It can be seen in Fig. 4.5c that the electric field within the domain

reaches a peak value

, while outside the domain the electric field (

) is less

than the threshold value.

is therefore less than

removing the drift velocity

difference (which causes the space charge to grow) and a stable domain is

formed. Additionally this means that when the domain is stable another domain

does not grow as

remains below

Chapter 4 Gunn Diode Theory

4.3.1 Domain Dynamics

Once stable, the domain drifts through the device at a constant velocity. The

electric field is less than the threshold level outside the domain. The electric field

is at its peak at the outside edge of the accumulation region, where it is above

the threshold value. It drops across the domain until it becomes less than the

threshold field (at the outer edge of the depletion region). Butcher [58] has done

a detailed analysis of domain formation, which concluded that if the diffusion

parameter D is assumed constant and independent of electric field, then the

electron velocity outside domain,

, is the same as the domain drift velocity,

Secondly

and

are related by the dynamic characteristic. The dynamic

characteristic curve is related to velocity electric field (

−

) curve by equal area

rule shown in Fig. 4.6 [1]. Using these results Hobson explained the current

conduction process within the domain as [2]:

( )

R r

J qNv qD

t x

ε ε ε

∂ ∂

= + −

∂ ∂

…………………….……….……(4.12)

where, the current density J is constant throughout the device, and the three

terms represent, drift, displacement and diffusion components respectively.

Outside the domain both

and

are independent of position thus the current is

(a)

(b)

(c)

Cathode

Anode

Cathode

Anode Cathode

Anode

Stable Domain

Fig.

Stable

ipole

omain

ormation with the growth of space

charge

[2]

Chapter 4 Gunn Diode Theory

carried as conduction current and current density is due to the drift component

( )

qNv

. Within the domain all three components play roles due to the gradient

in both doping density (

) and

. The change in electric field gradient causes a

displacement current to flow. At the peak electric field,

, value no

displacement current flows because

∂

. However, at

an electron

density gradient causes a diffusion current to flow. Thus in the depletion region,

the current is due to the displacement current while in the accumulation region

the current is due to the large conduction (drift) current, which is opposed by the

displacement and diffusion currents [4].

Conduction current

Diffusion current

Equal areas

Electron drift velocity

Electric field (E KV/cm)

Dynamic

characteristic curve

Fig.

Dynamic characteristics curve and electron drift velocity

–

electric

field curve. It is plotted in SILVACO using MOCASIM (Monte Carlo simulator) for

GaAs 1.1x10

-16

doped. Equal area relationship between

and

is shown.

∞

Electric Field

Accumulation Region

(a)

(b)

Cathode Anode Cathode Anode

Electron Density

Fig.

Zero Diffusion Domain p

rofiles

[5]

Chapter 4 Gunn Diode Theory

Finally, if the diffusion current is neglected i.e. diffusion coefficient D is zero, then

the domain travels at the same velocity (

) as the electrons outside the domain

(

). The electron density is zero in the depletion region and maximum (

∞

) at

the accumulation region. Thus a triangular shaped electric field is formed within

the domain, while outside domain the carrier concentration remains unchanged

as shown in Fig. 4.7. The current density outside the domain is given as

qvNJ =

∗

………………………...………………...……………(4.13)

while inside the domain due to fully depleted region the current density is given

0 0

r r D

J v

t x

ε ε

ε ε ε ε

∂ ∂

= = −

∂ ∂

……………………………..………….(4.14)

where

is the domain velocity. Poisson’s equation requires that

−=

∂

εε

………………………………...…………………(4.15)

Thus the current density is given as

qvNJ =

…………………………………………...…………….(4.16)

whereas outside the domain at

∂

and carrier density

NN =

, using the

current continuity equation

vv =

……………………………….………………………….....(4.17)

the current is given as

vqNJ =

…………………………….……………………………(4.18)

Chapter 4 Gunn Diode Theory

4.3.2 Device Oscillating Frequency

Generally in GaAs

scm /101~

for moderately doped material. As the

domain grows the current in an external circuit falls due to the fall in the field

outside the domain. As the domain reaches the anode, the electric field gradient

becomes zero and the domain collapses. The external electric field and current

values increase until reaching the threshold value and allowing another domain

to form at the cathode. The external current therefore oscillates at a frequency,

which is inversely proportional to the transit length

(the distance between the

domain formation point and anode). The oscillating frequency can therefore be

given as:

f =

……..………………………….…….…………………….(4.19)

Maximum operating frequency is determined by the material’s inter-valley energy

relaxation time. The electrons transfer from the

-valley to the L-valley and vice

versa in a finite time. At high frequencies ~60 GHz the energy relaxation time is

~3×10

-12

sec [2], which has a significant effect on energy relaxation time and as

a result the RF cycle does not follow the

−

curve. The NDR is reduced since

the current cannot rise as the electric field falls thus reducing the device

efficiency and power [5].

4.3.3 The Doping – Length (N

L) Product

Domain formation takes a finite time and distance and so to achieve stability

requires a minimum device length

for a given doping density

. To determine

the minimum device length, it is compared with the domain length under applied

electric field. For a fully depleted triangular domain with width

, peak electric

field

and external field

(as shown in Fig. 4.7b), the domain potential

given as:

)(

RpD

εεφ

−=

……..………………….…………………………(4.20)

Chapter 4 Gunn Diode Theory

From Poisson’s equation

Rpr

)(

−

……………………………….…………………(4.21)

And so solving equations 4.20 and 4.21, the domain potential becomes;

Rpr

wqN

εε

εεεε

)(

−

……………………..……………(4.22)

The ratio of peak to valley current is ~2 in GaAs, a threshold field twice as high

as the electric field outside the domain. Mathematically the domain potential is

given as;

……………………………………………..…………….(4.23)

Thus it can be stated that

12 2

2.4 10

r T

N l cm

ε ε ε

−

> = ×

……………………….……………(4.24)

4.3.4 Stable Operating Point of Domain

After analyzing the domain potential, the stable value of the outside field

with

regard to the applied voltage can be determined. If

is the device length,

the

width,

the domain potential and

the external bias voltage, then while

maintaining the electrical boundary condition, the external bias voltage,

, is

given as:

)(

wlV

tRDB

−

………………………………..……………(4.25)

As domain length is very small (

<<<

) it can be neglected and the bias

voltage given as:

tRDB

……………………………………...………………(4.26)

Chapter 4 Gunn Diode Theory

The minimum electric field value outside the domain required to sustain the

domain is shown in Fig. 4.8, where for a constant voltage

, the domain

potential

is plotted against the minimum electric field

. The three load lines

correspond to three different terminal bias voltages (

). The intercepts are for a

given doping density,

, and device length,

. Therefore, the following three

configurations are possible [2, 29]:

•

The average electric field,

lV /

, is greater than the threshold field yielding

the load line 1, which intersects at a point higher than

on the

axis. In this

condition only one stable solution is possible.

•

The average electric field

lV /

is greater than the sustaining voltage

StS

but less that the threshold field (giving load line 2). For

lV /

less

than

, two solutions are possible. The first intersection point of the

curve and load line gives a stable operating point. It can be seen that for a

steep load line i.e. a long device, the stable operating point is approximately

half the threshold value. Therefore, as shown earlier in equation 4.26, the

Stable

Unstable

Load Lines

Domain Potential

Electric Field

Fig. 4.8 The domain potential

and minimum electric field

(solid line) relationship due to the space-charge dynamics [2]

Chapter 4 Gunn Diode Theory

domain will sustain and continue to propagate even if the external voltage

drops below the threshold voltage. The unstable intersection is explained

through consideration of a small noise fluctuation, which causes

increase, thus decreasing

. A decrease in the value of

shifts the state

higher than the

curve. It shows that outside the domain the electrons

are moving faster than those inside the domain. It will cause the accumulation

and depletion regions to grow to keep the Poisson’s equation satisfied,

resulting in a further increase in

. Eventually it will reach steady state and

the stable point on the load line will be achieved.

•

The average electric field

lV /

is equal to the sustaining voltage

StS

which gives line 3. This condition causes the domain to terminate while still

growing, with

lV /

greater than or less than the sustaining voltage. At the

sustaining voltage the load line is tangent to the

curve. Thus

considered very important for the domain to grow and be sustained.

4.4 Conventional Gunn Diode

The Gun diode in its basic form is a homogenous device with ohmic contacts at

each end. The device has a sandwiched structure and comprises of

contact

layers at the ends with a

−

transit region in between, as shown in Fig. 4.9.

Doping Concentration

Dead Zone

Contact Layer

−

Transit Region Length

Device Depth

Contact Layer

Effective Transit Length

Fig.

Conventional Gunn Diode s

tructure

Chapter 4 Gunn Diode Theory

As previously described, the transit region length partially determines the

device’s operating frequency. However, domain formation in conventional Gunn

diodes does not occur in the initial portion of the transit region, where the

electrons are being accelerated to the energy levels required for inter-valley

transfer. This initial portion is termed as the dead zone. Thus domain formation

occurs at a certain distance away from the cathode depending on the applied

bias. This reduces the effective length of the transit region and acts as parasitic

positive resistance, which degrades the overall negative resistance and results in

a reduction in DC-to-RF conversion efficiency. Generally the length of the dead

zone is about 40 percent of the transit length in high frequency devices

depending on the transit length and the applied bias [67]. The applied bias

determines the point of domain nucleation, making the effective transit length

and therefore operating frequency very sensitive to the applied bias.

The temperature effects on conventional Gunn diodes and their limitations are

discussed below;

4.4.1 Temperature Effects on Conventional Gunn Diode

The temperature dependence of the velocity-field characteristic of low doped

GaAs is shown in Fig. 4.10 [2]. The average electron velocity is lower than the

formation energy of optical phonons (~0.035 eV in GaAs) at low lattice

temperature and low electric field [68]. The electrons attain high drift velocity

before acquiring sufficient energy for inter-valley transfer. At lower temperatures

the threshold field and the onset of NDR decreases due to the shift of the domain

nucleation point towards the cathode. This shift increases the device’s effective

transit length and reduces the operating frequency [5]. However, the device

critical current at the onset of NDR increases. Likewise, the device turn on

voltage, required to achieve stable oscillations, also increases with reducing

temperature. Thus a high turn on voltage results in lower device efficiency and

output power. These factors make the conventional diodes operational

characteristics extremely sensitive to changes in ambient temperature.

Chapter 4 Gunn Diode Theory

4.4.2 Limitations of the Conventional Gunn Diode

Conventional GaAs Gunn Diodes can provide around 500 mW power at 20 GHz

with 5.5% efficiency. The efficiency reduces to about 1.5% at 60 GHz and power

to 50 mW. It has also been observed that at 77 GHz with 0.8% efficiency, powers

of up to 50 mW have been achieved. The conventional Gunn Diode limitations

are as follows [5]:

•

The device has a high turn on voltage.

•

The device exhibits poor bias and temperature stability

•

Thermal parasitics value is high at higher frequencies.

•

The device operating current is high thus leading to more power dissipation.

•

At mm frequency the device has lower efficiency and provides lower power.

A number of these limitations are directly related to the presence of the transit

region dead zone [67]. To overcome the limitations imposed by this the graded

gap GaAs/AlGaAs hot electron injector Gunn diode was proposed [69]. The hot

electron injector effect on dead zone reduction has been demonstrated by Monte

Carlo analytical simulations [70] and experimental results [11, 69]. The state of

0 2 4 6 8 10 12 14

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Electron drift velocity (10

cm.s

−1

)

Electric field (KV.cm

−1

)

temperature 77K

temperature 150K

temperature 225K

temperature 300K

temperature 400K

temperature 500K

Fig.

GaAs temperature d

ependant electron drif

t velocity curves. The

curves are plotted in

SILVACO using MOCASIM (Monte Carlo simulator) for GaAs

1.1x10

-16

doped.

Chapter 4 Gunn Diode Theory

art Gunn diode with step graded gap hot electron injector is the topic of

discussion in the next chapter.

4.5 Gunn Diode Oscillation Modes

A Gunn diode generally needs to be placed in a resonant circuit to generate

usable microwave power. In a controlled circuit various modes of operation are

possible. The major factors affecting the mode of operation are:

•

Device transit region length

•

The resonant frequency of the external circuit

•

Applied bias voltage

•

Device doping uniformity and concentration

The various possible modes of operation will now be discussed.

4.5.1 Transit Time Mode

If the external circuit is purely resistive (

→

∞

→

) the device will operate in

transit time mode and the domain transit-time across the transit region

determines the operating frequency as long as the bias voltage is kept greater

than the threshold field. Narrow current spikes are observed when the domain

collapses at the anode and the next is being nucleated near the cathode as

shown in Fig. 4.11. The oscillating frequency is therefore inversely proportional to

the space-charge or domain transit time

and is given as:

drift

f ==

……………………………………..………………(4.27)

where

drift

is the electron drift velocity and

is the transit region length [29]. In

this mode the voltage amplitude is small and current pulses narrow, indicating

Chapter 4 Gunn Diode Theory

the conversion efficiency of this mode is low with a high harmonic content.

Disadvantages of this mode is that the transit time frequency strongly depends

on the temperature and operating voltage rather than the characteristics of the

external circuit, thus poor bias and temperature stability are observed.

time

Fig. 4.12 Delayed Domain Mode- I-V and Time Evolution [4]

time

Fig. 4.11 Transit Time Mode - I-V and Time Evolution [4]

Chapter 4 Gunn Diode Theory

4.5.2 Delayed Domain mode

In delayed domain mode the external bias voltage drops below the threshold

field level during part of RF cycle as shown in Fig. 4.12. This results in the

delayed nucleation of the second domain at the cathode. Care is taken to avoid

reduction of bias voltage below the sustaining voltage. Thus as the field drops

below the threshold as the first domain reaches the anode and collapses, the

new domain at the cathode does not nucleate until the field rises above the

threshold value. This effect causes a reduction in the device operating frequency

which is always below the transit time frequency. In this mode the resonant

frequency of the external circuit determines the operating frequency. In other

words the resonant period of circuit

fT /1=

is greater than

, and so the

device frequency is given by;

drift

=<<

ττ

……………………..……………(4.28)

Generally in delayed domain mode, devices can be operated over a wide range

of frequencies depending on the external circuit. Additionally, the current spikes

are broader with low harmonics thus yielding higher conversion efficiency.

4.5.3 The Quenched Domain Mode

In this mode the domain is quenched during the RF cycle due to a decrease in

terminal voltage below the sustaining voltage. A new domain at the cathode does

not nucleate until the terminal voltage rises above the threshold value as shown

in Fig. 4.13. Operating devices in quenched domain mode yields higher

operating frequencies and efficiencies than in transit time mode. In this mode the

period of the resonant circuit

fT /1=

is less than

, but is greater than the

domain nucleation and extinction time

drift

ττ

<<=

……………………….……….……………(4.29)

Chapter 4 Gunn Diode Theory

Conversion efficiency is lower than in delayed domain mode due to the tendency

of quenched domain devices to generate narrow spike currents. It can be shown

by considering the area under the pulse that the efficiency is about 5% higher

than that of transit-time mode [2].

4.5.4 Low Space-charge Accumulation (LSA) mode

In LSA mode the external circuit configuration again plays a vital role. The effect

has been explained by Copeland [71], in that the circuit causes the electric field

to rise above the threshold field value and drops down instantly such that the

space charge distribution associated with the high field domain does not have

enough time to form. In this mode the operating frequency is much greater than

that set by the device transit length. In other words the frequency is so high that

the domain does not get enough time to form fully, while the field is above the

threshold value. In LSA mode the I-V characteristics follows the velocity electric

field (

−

), which exhibits the negative differential mobility region.

time

Fig. 4.13 Quenched Domain Mode- I-V and Time Evolution [4]

Chapter 4 Gunn Diode Theory

The biggest advantage of this mode is the very high frequencies achievable,

since the device length is not related to the operating frequency. Additionally, the

mode provides high output powers since high terminal voltages can be applied

without causing impact ionization [2] and efficiencies of up to 18.5% have been

demonstrated [72].

4.5.5 Operating Modes Summary

The various operating modes discussed for GaAs are illustrated in Fig. 4.14 in

terms of the stable oscillation, domain oscillation and LSA oscillation regions. In

the stable region the device is used as an amplifier, since no domain formation

occurs for

12 2

N l cm

−

. While for

12 2

N l cm

−

three oscillation modes (Transit

time, delayed domain and quenched) are possible. In GaAs case, at higher

frequencies and

31110

]102,102[/

−

××∈ smfN

, the device operates in LSA mode

[73]. In all the modes the device boundaries vary depending on the applied bias

voltage, circuit loading and environmental effects.

Frequency f x device length [m/s]

Electron Concentration

x Device length (

2−

)

Stable

Fig. 4.14 Gunn Diode Operating Modes Summary [2]

Chapter 4 Gunn Diode Theory

4.6 Conclusion

The transferred electron effect was reviewed in detail through Electric Field

Effects on Electron Drift Velocity, High Field Transport in n-GaAs, Negative

Differential Resistance (NDR) and Conditions for Negative Differential Mobility. A

detailed account of conventional Gunn diode theory has been provided with an

explanation of the Negative Differential Resistance (NDR) effects, which

depends on the bulk material properties.

A Gunn diode generally needs to be placed in a resonant circuit to generate

usable microwave power. It was shown that in a controlled circuit various modes

of operation are possible. The major factors affecting the mode of operation were

shown as device transit region length, the resonant frequency of the external

circuit, applied bias voltage and device doping uniformity and concentration. The

various possible modes of operation were discussed.

The temperature effects on conventional Gunn diodes and their limitations were

presented. The limitations included high turn on voltage, poor bias and

temperature stability, high thermal parasitics value at higher frequencies, high

operating current thus leading to more power dissipation and lower efficiency at

mm-wave frequencies. It was shown that number of these limitations is directly

related to the presence of the transit region dead zone. To overcome the

limitations imposed by this the state of art Gunn diode with step graded gap hot

electron injector is the topic of discussion in the next chapter.

Chapter 5 Advanced GaAs Gunn Diodes with Graded

AlGaAs Hot Electron Injection

5.1 Introduction

A semiconductor structure formed between two dissimilar semiconductors joined

together is known as a heterojunction, which gives rise to a heterostructure

device. The heterostructure formed through epitaxial growth of one

semiconductor on top of the other such as the wide bandgap material (AlGaAs)

and narrow band gap material (GaAs) exhibits special properties. The

heterojunction properties depend on the distribution of the band gap discontinuity

(

E∆

) between the conduction band offset (

∆

) and valence band offset (

∆

The conduction band wave function primarily evolves from the atomic wave

function of cations (Ga and Al) and the valence band wave function evolves from

the atomic wave function of anions [74]. Thus the valence band structure of

GaAs–AlGaAs heterostructure sharing a common anion element (As), is smaller

than the conduction band offset i.e.

1/2/

≅

∆Ε

for direct gap range of

)4.0(

−

xAsGaAl

The maturity of device growth techniques such as Chemical Vapour Deposition

(CVD) and the more precise Molecular Beam Epitaxy (MBE) has led to the

commercial manufacturing of devices which rely on bandgap engineering.

Increase in RF power and improved efficiency of a Gunn diode with the

introduction of a heterojunction based hot electron injector was first reported in

1988 [75]. It was shown that under forward bias electron energy increases by an

amount equal to the AlGaAs-GaAs conduction band offset as the electron

passes through the injector region. Thus by keeping the conduction band offset

at the end of the AlGaAs injector close to the GaAs inter-valley separation,

electrons will be injected directly into the satellite valley. Therefore, the device

dead zone is effectively eliminated.

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

5.2 AlGaAs/GaAs Graded Gap Heterostructure

AlGaAs is capable of forming a lattice matched heterostructure with GaAs (lattice

mismatch is

102.1/

−

×≤∆ aa

) and provides a band gap range of 1.424 to 2.1 eV

through variation of the Al components. The graded gap AlGaAs alloy is formed

by using the chemical formula

AlAsGaAs

−1

)()(

, where

is the mole fraction (or

compound chemical composition). If the semiconductors are not lattice matched

then microscopic defects occur in the material, which significantly degrade the

device performance. The dependence of forbidden-gap energy on the Al fraction

‘x’ is usually expressed in terms of parabolic (linear plus quadratic) dependence.

The magnitude of the parabolic factor is termed as the bowing parameter

. The

bandgap energy of the alloy

AsGaAl

xx −1

is given as:

AsGaAl

ExxAlAsxGaAsxE

)1()(1)(

)(

−

eV……………….(5.1)

where term

Exx )1(

−

describes the quadratic dependence of the gap. However,

for

AsGaAl

−

the bowing parameter is very small thus its band gap is given as:

)(1)(

)(

AlAsxGaAsxE

AsGaAl

−

eV…………………..…….….(5.2)

At room temperature, the bandgap is given by the following empirical formula

[37]:

AsGaAl

25.142.1

)(

−

eV………………………...…………….(5.3)

The distribution of the bandgap discontinuity (

E∆

) between the conduction (

∆

) and valence (

∆

) band has been researched extensively. Experimental

data suggests that the ratio is approximately 60:40 for Al fractions less than 0.45

[4, 76, 77]. The conduction band offset is given as:

][6.0

)(

GaAs

AsGaAl

EEE

−

∆

−

eV………………….……………...(5.4)

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

Thus from equations 5.3 and 5.4 the conduction band offset can be written as:

75.0

∆

eV…………………………………….………………..(5.5)

5.3 Hot Electron Injection

Electrons with kinetic energy greater than kT, where K

is the Boltzmann’s

constant and T is the lattice temperature, are known as hot electrons. The

process of raising the electron energy to the level of the satellite L band and its

introduction into the transit region is known as hot electron injection. By

employing a hot electron injector just before the transit region the device dead

zone can be effectively eliminated. Additionally it results in the following

advantages [4]:

•

The turn on voltage improves (reduces), especially at low temperatures.

•

Increase in device efficiency.

•

Reduction in phase noise.

•

Improved temperature stability characteristics.

The concept of graded gap hot electron injector and AlGaAs-GaAs hot electron

injector are discussed.

5.3.1 The Graded Gap Hot Electron Injector Concept

The potential profile of a linearly graded hot electron launcher can be

approximated to a triangular shape as discussed below [5]:

At zero bias condition Fig. 5.1 (a) shows the right hand side (RHS) and left hand

side (LHS) of the barrier height

with lengths

and

respectively. A forward

bias

results in a potential shift

given as:

φφ

………………………………………………………(5.6)

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

A potential

changes the electric field on RHS given as:

212

…………………….………………..………………(5.7)

The change in

becomes vanishingly small if the length

ll >>

. The longer

region on the LHS gives a rising potential profile and the small

results in an

abrupt interface at the injector transit region interface. Thus the injector structure

with long rising potential region from cathode and abrupt interface gives an ideal

potential profile for a heterostructure Gunn diode.

5.3.2 The AlGaAs/GaAs Hot Electron Injector

The AlGaAs/GaAs hot electron injector consists of two components: a step-

graded AlGaAs barrier (launcher) and a thin

GaAs doping spike. The graded

AlGaAs barrier is used to form a launcher to ensure that the electrons gain

sufficient energy to transfer directly in to the transit region conduction band

satellite valley. The width of the launcher is large enough to avoid tunnelling at

low energy levels while being narrow enough to avoid significant increase in the

device’s serial resistance. The

doping spike is required to modify the

downstream electrical field in the transit region and eliminate the depletion region

RHS

LHS

(b) Forward Bias

(a) Zero Bias

Fig. 5.1 Potential profile of Hot Electron Injector [5]

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

which would be formed in its absence [70]. The hot electron injector structure is

shown in Fig. 5.2.

The linear graded Al

1-x

As launcher was introduced with Al fraction ‘x’

increased from the cathode side to its maximum value at the injector-transit

region. The maximum Al composition used was 30% to form a conduction band

discontinuity of 0.32 eV at the injector-transit interface [69]. Couch further

concluded that the thickness and doping of

GaAs layer played crucial part in

the device performance.

The structure similar to the proposed by Couch was optimized [69] to yield high

RF power and efficiency. The optimized device structure is shown in Fig. 5.3. It

Doping Concentration except launcher

Contact Layer

Graded AlGaAs Launcher

Undoped

Buffer

Substrate

Transit Region

Device Depth

Doping Spike

Hot Electron Injector

Fig.

(a) Structure of a GaAs Gunn Diode with step gra

ded hot electron injector

(b) Conduction band profile of a Step Graded AlGaAs heterostructure Gunn Diode.

0.5µm

GaAs

Contact

~5x10

-3

0.5µm

GaAs

Contact

~5x10

-3

50nm

1-x

Launcher

(nominally

undoped)

5nm

spike

~1x10

-3

GaAs transit region

(a)

(b)

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

has an undoped AlGaAs step graded barrier (from 1.7% to a maximum value of

Al ~32%). The maximum Al composition 0.32 formed the conduction band

discontinuity of 0.38 eV at the injector-transit interface, which improved the

device performance significantly [69]. The structure has two spacers to avoid

dopant diffusion into the graded barrier, which is followed by highly doped delta

(δ) layer and a thick transit region with low doping.

5.4 I-V Characteristics of graded gap injector GaAs Gunn Diode

Typical DC characteristics of Gunn Diodes with and without injector are shown in

Fig. 5.4. The injector gives rise to asymmetrical I-V characteristics due to the

Schottky like behaviour of the injector in reverse bias. The asymmetrical I-V plot

indicates different electron occupations in the L-valley for forward and reverse

currents due to the position of injector in the device. The graded gap injector

starting from cathode is only effective in forward bias: in reverse bias it does not

influence election transfer into the transit region. It may be noted that the injector

effectiveness can be evaluated from the ratio of forward and reverse currents (

) and, to a point, high values are desired for high overall device

efficiency [11]. This will be discussed in more detail later.

Device length in

0.25 um

Gold Cathode

0.5 um GaAs n=5×10

Contact Layer

0.01 um

GaAs undoped Spacer

0.05 um

0.017

0.983

As -----------------

Step Graded Barrier

--------------------

0.32

0.68

0.01 um

GaAs undoped Spacer

0.005um GaAs n=1×10

δ- doping

1.65 um GaAs n=1.1×10

-3

Transit Region

0.5 um GaAs n=5×10

Contact Layer

6 um GaAs n=2×10

Substrate

0.25 um

Gold Anode

0.0

9.625

0.0 0.38

Conduction Band

Fig.

Typical

advanced s

tep

raded

njector Gunn

iode

tructure

Chapter 5

The heterostructure graded gap Gunn diode

investigations, effects of hot electron injector barrier height and carrier

concentration in the doping spike are

5.4.1

High Frequency Investigations of GaAs Gunn Diodes

The Gunn diode with graded gap injector admittance graphs are shown in

.5 for negative (forward

bias

graph is similar to the diode without graded gap injector. However, in

forward bias a

sharp negative

reduction in dead zone due to the graded gap injector. Additionally a second

harmonic signal can also be observed

5.4.2

Hot Electron Injector Barrier Height

The step graded heterostructure barrier with 32% maximum Al concentration

provides a conduction band offset of ~0.38eV

measurements for Al content in the range 28

be seen that,

as expected

behaviour more pronounced in the

n the forward direction the peak saturation current reduces with the increase in

Fig. 5.4

Gunn Diodes

Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

The heterostructure graded gap Gunn diode

[11, 12]

investigations, effects of hot electron injector barrier height and carrier

concentration in the doping spike are

discussed.

High Frequency Investigations of GaAs Gunn Diodes

The Gunn diode with graded gap injector admittance graphs are shown in

.5 for negative (forward

bias) and positive (reverse bias

) voltages. The reverse

graph is similar to the diode without graded gap injector. However, in

sharp negative

peak in the

response at 60 GHz shows the

reduction in dead zone due to the graded gap injector. Additionally a second

harmonic signal can also be observed

at ~110 GHz [11].

Hot Electron Injector Barrier Height

The step graded heterostructure barrier with 32% maximum Al concentration

provides a conduction band offset of ~0.38eV

[78]

measurements for Al content in the range 28

32% is shown in

as expected

increasing Al content makes the Schottky like

behaviour more pronounced in the

reverse direction due to higher barrier height.

n the forward direction the peak saturation current reduces with the increase in

Gunn Diodes

I-V characteristics [11]

Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

[11, 12]

, high frequency

investigations, effects of hot electron injector barrier height and carrier

High Frequency Investigations of GaAs Gunn Diodes

The Gunn diode with graded gap injector admittance graphs are shown in

Fig.

) voltages. The reverse

graph is similar to the diode without graded gap injector. However, in

response at 60 GHz shows the

reduction in dead zone due to the graded gap injector. Additionally a second

The step graded heterostructure barrier with 32% maximum Al concentration

[78]

. The comparative

32% is shown in

Fig. 5.6 [10]. It can

increasing Al content makes the Schottky like

reverse direction due to higher barrier height.

n the forward direction the peak saturation current reduces with the increase in

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

Al content due to higher transfer rate of electrons from the

to the L valley and

higher injector efficiency [10].

5.4.3 Doping Spike

In addition to injector barrier height, the doping level of the spike plays an

important role in controlling the electric field potential behind the graded gap

Fig. 5.6 Low voltage I-V characteristics [11]

Conductance (solid curve)

Susceptance (broken curve)

Fig. 5.5 Conductance and Susceptance versus Frequency plot [11]

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

barrier. The appropriate electric field potential behind the graded gap barrier

retain the hot electron properties [10, 51, 79]. It has been shown [78] that the

absence of an

doping spike results in the opposite electric field gradient to

that required for domain formation and in forward bias a dead zone is formed at

the start of the transit region as shown in Fig. 5.7. The

doping spike

eliminates the depletion region as shown in Fig. 5.8. The thickness of the

doping spike is kept lower than the injected electrons mean free path [51, 78].

Cap

−

Transit Region

Injector with spike

Length

Electron Concentration cm

Fig.

Electron Concentration

–

ith doping spike

[10]

Electron Concentration cm

Cap

−

Transit Region

Depletion zone

Length

Injector without spike

Fig.

Electron Concentration

–

ithout doping spike

[10]

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

5.5 Drift Velocity Computation and Operation Mode Classification

In the Gunn diode, electron drift velocity in the transit region is dependent on

factors such as injector barrier height, electric field and temperature [10, 11]. A

small signal analysis is shown in Fig. 5.9 illustrating measured drift velocity

against applied voltage at negative conductance frequency (

NCM

77 GHz) for an

active region length of 1.6 µm. The Gunn diode with injector achieved an

asymmetric response and in forward bias the drift velocities were smaller due to

the injector’s hot electrons entering the active region, while in reverse bias the

response was similar to a diode without injector. The graph represents operation

in transit time mode for drift velocities of

scmxv

/1023.1

. The experimental

data indicated that the injector with Al fractions of 32% and 34% provided stable

77 GHz oscillations in the Quenched domain mode [10, 11].

Fig. 5.10 shows electron occupation in the L-valley obtained from drift velocity

data and plotted for different electric fields [10]. Due to its dynamic nature

prediction of the electric field in the transit region is complicated and so data was

obtained using SILVACO

simulations [11]. The emphasis was on the effect of

carrier concentration (

) and mobilities (

) on device operating

frequency (

NCM

), which changed due to the change in ratio of

and

Fig. 5.9 Diode with Injector [11]

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

5.6 Conclusion

The Advanced GaAs Gunn Diodes with graded AlGaAs hot electron injector

concept and theory was discussed in detail. It was shown that under forward bias

electron energy increases by an amount equal to the AlGaAs-GaAs conduction

band offset as the electron passes through the injector region. Thus by keeping

the conduction band offset at the end of the AlGaAs injector close to the GaAs

inter-valley separation, electrons will be injected directly into the satellite valley.

Therefore, the device dead zone is effectively eliminated and the device was

shown to have overcome the limitations of the conventional Gunn diode.

Additionally, dead zone elimination was shown to be advantageous due to

improvement in the device turn on voltage, increased device efficiency, reduction

in phase noise and improved temperature stability characteristics. The concept of

graded gap hot electron injector and AlGaAs-GaAs hot electron injector was

presented.

Typical DC characteristics of Gunn Diodes with and without injector were

compared. It was shown that the injector gives rise to asymmetrical I-V

characteristics due to the Schottky like behaviour of the injector in reverse bias.

The heterostructure graded gap Gunn diode’s high frequency investigations;

effects of hot electron injector barrier height and carrier concentration in the

Fig.

Electrons occupations in the L

valley

[11]

Chapter 5 Advanced GaAs Gunn Diodes with Graded AlGaAs Hot Electron Injection

doping spike were presented. Finally drift velocity computation and operation

mode classification were discussed for an advances AlGaAs/GaAs hot electron

injector Gunn diode.

Chapter 6 Gunn Diode Model developments

6.1 The SILVACO

TCAD Suite

SILVACO

is a software package which provides a Virtual Wafer Fabrication

(VWF

) simulation environment in which two or three dimensional device

simulations can be performed (using the ATLAS

simulation engine) [80]. It is

physically-based and allows the electrical, optical, thermal characteristics of a

semiconductor device to be simulated under given bias conditions to obtain DC,

RF and time domain responses. The response of the model is simulated through

definition of data on mesh points (nodes) throughout the volume (or area if 2D) of

the structure and the subsequent solution of differential equations. The

simulations are cost effective and quick in contrast to performing experiments

and besides potentially yielding a predictive model for a given device, can

provide insight into device operation and analysis of the underlying theoretical

concepts. Data which is difficult to obtain from actual experiments can also be

obtained using simulation. However, despite numerous advantages, these

simulations are difficult to perform because they require thorough understanding

of the device physics and knowledge of suitable physical model parameters.

In SILVACO

’s VWF

environment ATLAS

is the core tool for ATLAS

physically-based two or three dimensional device simulation. It determines the

electrical behaviour of a structure created by DevEdit

and provides insight into

the device’s internal physical mechanisms. Fig. 6.1 shows the types of

information which flow in and out of ATLAS

[81]. The two input files are the

device structure and input text file containing structure and execution commands.

During device simulations, ATLAS

creates three output files: The Runtime

output file provides the simulation progress, error or warning messages. The

solution file stores the solution variables data for the given conditions. The log

file stores all terminal voltages and currents from the device analysis.

Chapter 6 Gunn Diode Model developments

Fig. 6.1 ATLAS

Input-Output hierarchy [81]

In the ATLAS

program flow, the order in which statements are defined is

crucial. There are five groups of statements which need to be specified in the

order shown in Fig. 6.2 to ensure correct execution. After specifying the device

structure in DevEdit

, the material and model statements are defined in

DeckBuild

. The material statements are used to define basic material

parameters relating to band structure, mobility, recombination and carrier

statistics etc. The model statement specifies the inclusion of various physical

mechanisms, models, and other parameters such as the global temperature for

the simulation. The method statement is then used to set the numerical methods

to calculate the solution under various bias conditions. Parameters in the method

statement also set the tolerances for data convergence. Finally the results are

extracted and presented using TonyPlot

[81].

ATLAS

uses its sub-engine ‘Blaze’ for heterostructure device simulations.

Blaze is a general purpose 2D device simulator used for III-V materials and is

invoked by default for heterostructures with a position dependent band structure.

During simulations it modifies the charge transport equations to account for the

effects of positionally dependent band structures [81].

Chapter 6 Gunn Diode Model developments

Fig. 6.2 ATLAS

command groups [81]

Fig. 6.3 Gunn Diode model development process flow

Chapter 6 Gunn Diode Model developments

6.2 Development of the Simulation Model

The device structure was first defined in ‘DevEdit

’, which provides a simple

Graphical User Interface (GUI) development environment [82]. A device structure

file is then constructed and saved by ‘DevEdit

’. This is then opened in

‘DeckBuild

’ and simulated using ‘ATLAS

’ as illustrated in Fig. 6.3. Finally the

results are plotted using ‘TonyPlot

’.

The 77GHz second harmonic graded gap injector Gunn diode epitaxial structure,

2D model, 2D model with heat sink, 3D rectangular model and 3D cylindrical

model development is discussed below;

6.2.1 High Speed Heterostructure Graded Gap Injector Gunn Diode

A state of art GaAs Gunn diode with a step graded AlGaAs hot electron injector

has been selected for modelling using SILVACO

. The typical device structure

is shown in Fig. 6.4 and is similar to that described in chapter 5.

6.2.2 Gunn Diode 2D Model

A 2D model of the structure was made in DevEdit

through definition of fourteen

regions shown in Fig. 6.5 and 6.6. The regions were defined using the click and

drag option provided by DevEdit

, before the individual material properties were

defined for each region.

Device Length in

0.0

9.625

0.0 0.38

Conduction Band

Fig.

77 GHz Gunn advanced step graded injector Gunn Diode s

tructure

0.25 um

Gold

Cathode

0.5 um

GaAs

n=5×10

Contact Layer

0.01 um

GaAs

undoped

Spacer

0.05 um

0.017

0.983

------------------

Step Graded Barrier

--------------------

0.32

0.68

0.01 um

GaAs undoped

Spacer

0.005um

GaAs

n=1×10

δ- doping

1.65 um

GaAs n=1×10

-3

Transit Region

0.5 um

GaAs

n=5×10

Contact Layer

6 um

GaAs

n=2×10

Substrate

0.25 um

Gold

Anode

100

Chapter 6 Gunn Diode Model developments

Transit Region

1-x

Heterostructure

Transit Region

Substrate

Buffer

GaAs Cap

Net Doping in /cm

Substrate

Device width in microns

Device length in microns

Device length in

microns

(a) (b)

Fig.

TonyPlot

of modelled device structure showing impurity doping p

rofile

(a) Device planar view (b) Device cross section view

GaAs Channel

005.0

101 ×

625.9

GaAs Spacer

01.0

AsGaAl

−

01.0

x 5 slices

(x= 0.017, 0.08, 16, 24, 32)

GaAs Spacer

01.0

GaAs Substrate

105 ×

GaAs Buffer Layer

5.0

105×

GaAs Transit Reg

65.1

101.1 ×

Gold Cathode

GaAs Cap

5.0

105×

Gold Anode

Fig.

Graded Gap Heterostructure Gun Diode

101

Chapter 6 Gunn Diode Model developments

Initially a default mesh was created throughout the model. An appropriate mesh

density then needs to be identified as too dense a mesh results in unfeasibly

slow device simulation and too coarse a mesh led to inaccurate simulation

results. In view of the importance of mesh density, it was optimized using various

options. In the beginning, the mesh refine option was used from the DevEdit

tool bar. Although a dense mesh was achieved the number of triangles exceeded

18,000 and as a result the DeckBuild

possessing time increased enormously.

The mesh was then refined in selected regions such as all the AlGaAs regions,

the doping spike and the transit region. Although this modified mesh provided

adequate simulations results, impurity mixing in the small regions such as the

doping spike and spacer were observed presumably due to lower number of

triangles leading to overlapping of the regional interfaces. This problem was

addressed by removing the mesh refine statement from the code and instead

using mesh constraint [83]. The meshes created using the refine and constraint

options are shown in Fig. 6.7 where it can be seen that the latter option provided

an improved and denser mesh whilst allowing for an acceptable simulation time.

The resulting conduction band diagram in the AlGaAs region for both meshes is

shown in Fig. 6.8. It can be seen that the mesh with constraints applied provided

Heterostructure

Substrate

Transit Region

Substrate

Dense Mesh

Less Dense Mesh

(a)

(b)

Fig.

Device structure with m

esh

dens

ity defined

(a) using

‘constraint’ (b) using ‘refine’, plotted in TonyPlot

102

Chapter 6 Gunn Diode Model developments

a large number of triangles even in the smallest region, thus yielding a more

realistic conduction band diagram.

The device structure can also be defined in DeckBuild

through the ATLAS

command line. The command line codes were written to create 3D rectangular

and cylindrical structures. The structures were created in an order of four distinct

set of commands. The initial mesh was defined with a set of horizontal (x.mesh)

and vertical (y.mesh) meshes with specific spacing between them. The mesh

spacing was defined using parameter spac. After specifying meshes, the regions

and materials were assigned. The electrodes and doping in the regions followed.

Finally the structure was saved and simulated in DeckBuild

During initial model development, convergence issues were experienced which

were attributed to meshes, material parameters and models definition. After

improving models and material parameters, the convergence problem was

narrowed down to number of meshes in regions such as the hot electron injector

(AlGaAs launcher and doping spike regions) and the transit region.

Subsequently, spacing between the meshes was reduced beyond SILVACO

recommended maximum number. It was achieved by using spac, which created

Conduction Band Energy (eV)

Less Number of Triangles

with ‘Refine Mesh’

Large Number of Triangles

with ‘Constraint Mesh’

Fig.

Conduction band c

omparative data

showing two models with course and

dense meshing in the step-graded launcher

103

Chapter 6 Gunn Diode Model developments

a very dense mesh especially in the hot electron injector and the transit region.

The number of meshes exceeded 100,000 against 18,000 limit imposed by the

software. Despite greater number of meshes the memory allocation failure error

message was not flagged and instead convergence issues were resolved. Thus

for all further model developments including sub-micron transit region length

models and with cylindrical geometry, ATLAS

command line was used to

define modelled device structures.

6.2.3 2D Model with Heat Sink

Giga, a self heating simulator available in SILVACO

, was used to simulate the

device’s lattice heating effects. Giga was used with the definition of heat sinks at

the cathode and anode respectively. Their dimensions were kept in accordance

with a fabricated device for 77 GHz autonomous cruise control systems at e2v

Technologies Plc. The device mesh structure with heat sinks is shown in Fig. 6.9.

The final DevEdit

structure file was then exported to DeckBuild

. The

DeckBuild

code comprised of two sections: the device structure section and a

Device Depth

40 µm heat sink

Cathode

Anode

Device Width

9.265 µm

Gunn Diode

10 µm heat sink

Fig. 6.9 Modelled device structure with heat sinks

104

Chapter 6 Gunn Diode Model developments

coding section which defines the material parameters, model definition, methods

and bias conditions. Bias was applied before the start of device simulation, which

was achieved using ATLAS

by initializing the device under zero biased

condition. The zero biased data obtained was then saved for subsequent I-V

simulations. The device DC response was determined to study the device I-V

characteristics. TonyPlot

was used to display the device simulation results.

The log files and device structure files, created by DeckBuild

, were saved and

subsequently plotted with TonyPlot

for analysis [82-84].

6.2.4 3D Rectangular Model Development

The 2D device structure was redefined in DeckBuild

with a third dimension.

The ‘z’ dimension was defined while keeping the device area similar to the

fabricated device. Fig. 6.10 – 6.11 shows the 3D device structure, where it can

be seen that device layers are in the y-axis and its width in the z-axis. Blaze3D (a

variant of Blaze2D which accounts for third dimension) was used to simulate the

3D device structure.

Fig. 6.10 3D rectangular model epitaxial structure

105

Chapter 6 Gunn Diode Model developments

6.2.5 3D Cylindrical Model Development

A novel 3D cylindrical model structure has also been developed in DevEdit3D

to accurately reflect the cylindrical geometry of a fabricated device. This

cylindrical model is shown in Fig. 6.12.

0 µm

dia

heat sink

Cathode

9.265 µm Gunn Diode

Anode

heat sink

Anode

Fig. 6.12 3D cylindrical modelled device structure with heat sinks

40 µm heat sink

Cathode

9.265 µm Gunn Diode

10 µm heat sink

Anode

Fig. 6.11 3D rectangular modelled device structure with heat sinks

106

Chapter 6 Gunn Diode Model developments

6.3 Initial Device Solution

The initial solution for a given model, on which subsequent simulation iterations

are based, is obtained at a bias condition of zero volts. This is because in

ATLAS

the initial values for internal potential and carrier concentration are

based on the device’s doping profile. Through use of the solve init statement the

initial response is calculated and saved as the device structure file shown in Fig.

6.13. This plot shows electron concentration through the device and on the left

side a display bar shows the other options that are available for display in the

device cross section view. By default the display bar only provides the doping

concentration. However, with the addition of the output statement after the

models statement in DeckBuild

, other options such as electric field, electron

mobility, electron temperature, electron velocity, conduction band and valence

band profile, and charge and current flow lines can also be plotted in the device

cross section view.

After determining initial device solution, DC simulations and device transient

response was obtained as discussed below;

Device length µm

Electron Concentration /cm

Transit Region

Buffer

Substrate

Contact Layer

Fig.

Device model structure cross section v

iew

plotted in TonyPlot

107

Chapter 6 Gunn Diode Model developments

6.3.1 DC Simulation Results

The DC solution was obtained in DeckBuild

using the solve statement. The

anode voltage was ramped from 0 to 4 volts with a step size of 0.1V. The file was

then saved and plotted to provide the I-V curve shown in Fig. 6.14, which shows

the start of the negative differential mobility region. The DC solution beyond 4V

could not be achieved due to convergence problems that were later, resolved for

transient simulations. However, it did provide the critical current at the onset of

NDR.

During reverse bias simulations the anode voltage was ramped from 0 to -4V

with a step size of -0.1V. The resulting I-V curve is also shown in Fig. 6.14 and

has been plotted on the positive axis for better visualization.

The modelled device’s asymmetrical I-V response indicates the successful

incorporation of the injector and matches published data and the basic theory

discussed previously.

6.3.2 Transient solutions

Transient solutions are used to obtain a time domain response such as

determining the device oscillating frequency in biased condition. In the Gunn

diode transient solutions for a linear ramp is obtained by specifying parameters

0 1 2 3 4

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Model Fwd 300K

Model Rev 300K

Fig. 6.14 Model forward-reverse asymmetrical I-V characteristics

108

Chapter 6 Gunn Diode Model developments

such as TSTART, RAMPTIME, TSTOP and TSTEP as shown in Fig. 6.15 [81].

The TSTART specifies the linear ramp start time, RAMPTIME specifies the time

that the linear ramp takes to achieve the final value, the solution is then stopped

at TSTOP and the initial step size is specified by TSTEP shown in Fig. 6.15.

The transient response is saved in the log file and subsequently, extract

statements are used to determine the maximum current value on y-axis and its

corresponding voltage on the x-axis. Finally the I-V curve for the transient

response is plotted using TonyPlot

6.3.3 Transient Response – free running oscillations

The 2D models with heat sinks and associated cavity time domain simulations

are discussed in chapter 8. The transient response presented here is that of a 77

GHz second harmonic Gunn 2D model without using the self heating simulator,

GIGA, and with no cavity. It was performed to extract the critical current, the

damping and final oscillating frequencies as shown in Fig. 6.16 - 6.18. The model

was simulated at 2V applied bias to determine the current at the onset of

negative differential resistance. The extracted current plot in Amps is shown in

Time in Pico sec

20 0 110 111 240

Simulation at 2V

TSTEP = 1×10

-12

Simulation at 4V

TSTEP = 1×10

-12

TSTOP

Ramp

Time

Vanode

Ramp

Time

130

Damped

Oscillations

Stable

Oscillations

Vanode

Fig.

evice model transient voltage ramp in ATLAS

109

Chapter 6 Gunn Diode Model developments

Fig. 6.16, which depends on the model width as discussed earlier. The first

damping oscillating transient response is shown in Fig. 6.17 and shows that the

domain is not fully formed resulting in the damping transient response which

eventually died down. The damping transient response although yielding very

high frequency resulting in a power that was very low. Therefore, it was

concluded that at the onset of negative differential resistance the applied

potential and hence the field is not enough to sustain the domain.

Current (A)

Transient Oscillations at 2V

Fig. 6.17 Transient response - damping frequency at 2V

Current (A)

Transient Simulation at 2V

Fig. 6.16 Transient response - critical current at 2V

110

Chapter 6 Gunn Diode Model developments

The applied bias was increased from 2 to 4 volts (representing a realistic device

operating bias) and the transient response was determined as shown in Fig.

6.18. It can be seen that the domain is quenched during the RF cycle due to a

decrease in terminal voltage below the sustaining voltage. A new domain at the

cathode does not nucleate until the terminal voltage rises above the threshold

value.

6.4 Free-running frequency of oscillation

It has been shown [12] that weak oscillations will occur in a device biased at

voltages below its turn on voltage and that these weak oscillations will be

extremely noisy at low bias voltages due to the excitation of numerous modes

within the transit region (and a reduction in the level of coherence). As the

external bias is increased the level of coherence also increases due to the larger

high-field domains reducing the overall internal electric field in the transit region

and so preventing multiple modes propagating. In addition, as the external bias

voltage increases the oscillation frequency decreases and the RF power

increases. At the threshold (turn on) voltage, the RF power locks onto a single

Current (A)

Transient Oscillations at 4V

Fig. 6.18 Transient response - stable oscillations at 4V

111

Chapter 6 Gunn Diode Model developments

frequency defined by the resonant oscillator circuit and a highly coherent output

is observed.

These effects are illustrated in Fig. 6.19 which shows the onset of oscillation,

region of weak oscillation, and turn on voltage of two GaAs Gunn devices. The

plot is taken from [12] and was produced using Monte Carlo simulations of the

free running frequency of GaAs Gunn devices with and without hot-electron

injection (it is noted that the power curve is based on measured values and that

the simulations did not include consideration of an oscillator circuit). This shows

that the free-running frequency of oscillation is reduced as the external field is

increased, meaning that the oscillation frequency of the simulated time-domain

response is typically expected to be higher than that measured from an oscillator

circuit. It also shows that the capability of a device to support sustained

oscillation can be established from simulation of the model at low bias voltages.

The majority of the time-domain simulations were therefore carried out at two

bias voltages: a low voltage (typically about 2 volts) to establish whether the

device was capable of producing sustained oscillations and, where possible, a

higher voltage (about 4 volts) to study the behaviour of the device under realistic

operating conditions.

Fig.

Monte Carlo simulation results

for GaAs Gunn diodes with and without hot

electron injection. It is noted that the power curve is taken from measurements of a typical

device with hot electron injection, and that an oscillator circuit was not included

in the

simulation [12].

112

Chapter 6 Gunn Diode Model developments

6.5 Conclusion

A step-by-step process has been discussed to develop a physical model in

SILVACO

VWF

environment using ATLAS

as the core tool. It was shown

that in the ATLAS

program flow, the order in which statements are defined is

crucial. The five groups of statements and their order was presented.

Development of the physical model was presented for a state of art GaAs Gunn

diode with a step graded AlGaAs hot electron injector. It includes 2D model, 2D

model with heat sink, 3D rectangular model and a novel 3D cylindrical model

developed for a 77GHz second harmonic graded gap injector Gunn diode.

It was shown that convergence issues were resolved by improving physical

models, selecting correct material parameters and refining meshes in regions

such as hot electron injector and transit region. The spacing between the

meshes was reduced beyond SILVACO

recommended number by using spac,

which created a very dense mesh. The number of meshes exceeded 100,000

against 18,000 limit imposed by the software. Despite greater number of meshes

the memory allocation failure error message was not flagged. For all further

model developments including sub-micron transit region length models and with

cylindrical geometry, ATLAS

command line was used with spac defined.

The modelled device initial solution, DC simulations and transient responses

were presented that are discussed in chapters 8 and 9. The modelled device’s

asymmetrical I-V response indicated the successful incorporation of the injector.

The transient response presented is that of a 77 GHz second harmonic Gunn 2D

model without using the self heating simulator, GIGA, and with no cavity. It was

performed to extract the critical current, the damping and final oscillating

frequencies. The model was simulated at 2V applied bias to determine the

current at the onset of negative differential resistance. The first damping

oscillating transient response, at 2V applied bias, showed that the domain was

not fully formed resulting in the damping transient response which eventually

died down. It was concluded that at the onset of negative differential resistance

the applied potential and hence the field is not enough to sustain the domain.

The applied bias was increased from 2 to 4 volts (representing a realistic device

operating bias) and the transient response was determined. It was shown that

113

Chapter 6 Gunn Diode Model developments

the domain was quenched during the RF cycle due to a decrease in terminal

voltage below the sustaining voltage.

Finally free running frequency of oscillation was discussed. It was shown that the

free-running frequency of oscillation is reduced as the external field is increased,

meaning that the oscillation frequency of the simulated time-domain response is

typically expected to be higher than that measured from an oscillator circuit. It

was also shown that the capability of a device to support sustained oscillation

could be established from simulation of the model at low bias voltages.

Chapter 7 Physical Models

7.1 Physical Models Used for Device Simulation

The accuracy of the simulated response depends on the choice of material

parameters and the selection of physical models (and their associated

parameters) used. The material and model parameters used were therefore

extensively researched from a variety of sources [3, 29, 65, 85, 86]. Where the

platform did not allow sufficient flexibility to accurately account for the

interdependency of parameters, appropriate C-language functions were written

and incorporated through the in-built C-interpreter [81, 83].

The analytic function based on the work of Caughey and Thomas [87] is used for

modelling field dependant mobility in the contact layers, spacer layers, AlGaAs

launcher and also the transit region for field strengths below the negative

differential mobility (NDM) threshold. For field strengths in the transit region

above the NDM threshold, the Barnes NDM model [88] is used. The Shockley-

Read-Hall (SRH) recombination model [89, 90] is applied in the nominally

undoped Al

(1-x)

As barrier. The SRH concentration-dependent lifetime model

[89] is used in the rest of the device to account for the varied silicon impurity

concentrations.

For local temperature, self-heating, lattice heat flow and heatsinking effects, Giga

[81], the ATLAS

non-isothermal device modeller module is used. This is based

on the models suggested by Wachutka [91] and accounts for Joule heating along

with heating and cooling due to carrier generation and recombination.

The models used for each material i.e. GaAs and AlGaAs are listed in table 7.1

and are discussed in detail [81]:

114

115

Chapter 7 Physical Models

Table 7.1 Physical models summary

7.2 Mobility Models Used

In the Gunn diode, as in other electron carrier devices, the electrons are

accelerated by the electric field. However, they lose momentum due to numerous

Material Physical Models

GaAs Mobility

Conmob – (Concentration Based Mobility)

Analytic – (Caughey Thomas)

Fldmob evsatmod=1 b.elec=2 – (Field Dependent

Mobility with NDR)

Recombination

Consrh – (SRH Concentration-Dependent Lifetime

model)

Carrier Statistics

and transport

Boltzmann – (Boltzmann Approximation)

Temperature=300 – (device lattice temperature at 0

bias)

AlGaAs Recombination

SRH – (Shockley-Read-Hall)

Mobility

Fldmob evsatmod=0 b.elec=2 – (Field Dependent

Mobility without NDR)

GaAs

and

AlGaAs

Carrier Statistics

and transport

Boltzmann – (Boltzmann Approximation)

Temperature=300 – (device lattice temperature at 0

bias)

hcte.el – (Hydrodynamic model for electrons)

GIGA

lat.temp - (Lattice Heat Flow Equation)

e.taur.var - (Electron energy relaxation time)

Heat

Sinks

Anode and

Cathode

GIGA - self heating simulator defined to simulate lattice

heating effects as a function of applied voltage.

116

Chapter 7 Physical Models

microscopic scattering processes. These microscopic phenomena are lumped

into macroscopic mobilities, which are included by the mobility and transport

equations during device modelling. The mobility equations used for the Gunn

diode model in the low and high electric field regions will be discussed

individually.

The carriers are almost in equilibrium with the lattice at low field values and so

the electron mobility

has a characteristic low-field value. This low-field value

depends on the phonon and impurity scattering that affects (decreasing its

value). However, the carriers are no longer in equilibrium at high electric fields

and are therefore subjected to a wider range of scattering processes. Therefore,

at high electric fields the drift velocity no longer increases linearly and becomes

saturated or decreases, which is termed as saturation velocity (

sat

). These

effects must obviously be accounted for using the appropriate models [81].

ATLAS

and its sub-engine Blaze provide various low-field and high-field

models that can be used depending on the device and its operation. The models

used for the simulations will now be discussed.

7.2.1 Low field Mobility Models

Suitable low field models for the simulation were carefully selected. Five

distinctive models and conditions are available in ATLAS

to define the low field

electrical mobility in GaAs and AlGaAs [81]:

•

The low-field mobility parameter ‘MUN’(cm

/V.s), can be defined from the

ATLAS

lookup table. The value is defined in the mobility statement for each

region.

•

The concentration based mobility model ‘CONMOB’ can be used. The values

for low field mobility at 300K are found from the ATLAS

look-up table for

the doping concentration in each region.

•

In order to relate the low field electron mobility with impurity concentration

and temperature, the ‘ANALYTIC’ and ‘ARORA’, may be used.

•

A carrier-carrier scattering model (CCSMOB), which relates the low field

mobility to both carrier concentration and temperature can be used.

117

Chapter 7 Physical Models

•

Using the unified low field mobility model (KLAASSEN). This model relates

the low field mobility to lattice, carrier-carrier, donor scattering, and

temperature.

7.2.2 AlGaAs Default Low field Mobility Model

For the un-doped Al

1-x

As regions the Constant Low Field Mobility model was

used which is independent of the doping concentration, carrier densities and

electric field. The electron mobility values in this model are determined using the

following formula:

MUN

UNn

−













300

……………………………...…………………(7.1)

where

is the lattice temperature in degrees Kelvin, M

is the undoped

mobility value in cm

/Vsec and T

MUN

is the temperature dependent coefficient

whose value is taken from the SILVACO

lookup table (T

MUN

=1.5) [81]. These

parameters are specified in the material statement as shown in table 7.2. The

simulations were performed at

300=

which resulted in an electron mobility

equal to the undoped mobility value (M

) defined for each Al concentration. The

temperature dependent co-efficient T

MUN

was ignored. The AlGaAs mobility

values (M

), which were defined separately for each AlGaAs region due to

dependability on the Al concentration, are shown in table 7.2 [81].

1-x

As Region Statement Code

x=0.017 mobility region=11 mun=4700

x=0.08 mobility region=10 mun=4300

x=0.16 mobility region=9 mun=3400

x=0.24 mobility region=8 mun=2400

x=0.32 mobility region=7 mun=800

Table 7.2 Mobility values for Al

1-x

As regions [52]

118

Chapter 7 Physical Models

7.2.3 GaAs Concentration Dependent Low Field Mobility Model

For a doped GaAs region the electron mobility depends on the impurity levels in

each region alongwith the operating temperature. The Concentration Based

Mobility conmob was therefore defined for all GaAs regions. The SILVACO

default library has built in values for all GaAs concentration values at 300K. And

this default data was used during the simulations. To take into account

temperature effects on the electron mobility the analytic model was used which

allowed the temperature to be varied from 77 to 450K.

7.2.4 GaAs Analytic Low Field Mobility Model

The Caughey Thomas Analytic Low Field Mobility model [87] was used for the

GaAs regions. The model was used to specify doping and temperature

dependent low field mobilities using the following equation:

CAUG

CAUGCAUG

CAUG

CRITN

CAUG

CAUGn

∆





































−

























300

300300

300

αβ

/V.sec…(7.2)

A description of the parameters used in equation 7.2 along with their values is

shown in table 7.3. The model is activated by defining both conmob and analytic

in the models statement.

The default mobility values for

CAUG

and

CAUG

were used for all GaAs regions

except the transit region, where the parameters were defined to achieve the

6300 cm

/V.sec [88], which was achieved through iterative simulations.

Optimised values for the mobilities at highest and lowest Al concentrations were

found as

940

CAUG

and

8685

CAUG

respectively, which were used for

subsequent simulations. The temperature dependent fitting parameters

CAUG

and

CAUG

were defined using the C-interpreter function. The default value

for

CAUG

∆

was kept as 1 to account for the increasing concentration effect on the

electron mobility [81].

119

Chapter 7 Physical Models

Parameter Description Value Units

CAUG

Mobility at highest concentration value,

depends on impurity doping

/Vsec

CAUG

Mobility at lowest concentration. Depends on

impurity doping

/Vsec

CAUG

Temperature dependent fitting parameters 0 Arbitrary

CAUG

Temperature dependent fitting parameters 0 Arbitrary

CAUG

Temperature dependent fitting parameters 0 Arbitrary

CAUG

∆

Temperature dependent fitting parameters 1 Arbitrary

CRITN

CAUG

Caughey Thomas critical concentration level

108.2 ×

-3

Impurity concentration defined for individual

region

- cm

-3

Temperature 300 K

Table 7.3 Caughey Thomas Analytic Low Field mobility model parameters (300K) [87]

7.2.5 Parallel Electric Field Dependent Mobility Model

Two types of electric field dependent mobility (fldmob) models are available in

ATLAS

. The models are named as the Standard Mobility model and Negative

Differential Mobility model, both of which contain appropriate default parameter

values for different materials. The models are defined by specifying evsatmod

along with the fldmob in the models statement. The evsatmod specifies which

parallel field dependent mobility model should be used for electrons, which are

defined as [81]:

•

evsatmod=0 allows the application of standard Mobility model

•

evsatmod=1 implements the GaAs negative differential mobility saturation

model.

120

Chapter 7 Physical Models

•

evsatmod=2 implements the simplified field dependent velocity model.

The standard electric field based mobility model is applied and the

temperature dependent mobility is not used.

The standard mobility model was used for AlGaAs step graded heterostructure

region. The standard mobility model is defined in terms of electron saturation

velocity, which is a function of applied electric field. The model is implemented in

ATLAS

using the following Caughey Thomas expression, which provides a

smooth transition between low-field and high field behaviour [87]:

satn

vEv

)(

























………………………………………..(7.3)

where,

satn

is the electron saturation velocity,

is a constant whose value for

both AlGaAs / GaAs and was taken as 1 from the SILVACO

reference tables.

The low field electron mobility,

, is determined from equations 7.1 and 7.2 for

AlGaAs and GaAs respectively. A similar equation for holes is also available but

has not been used, since holes do not play any role in the simulations due to

dopant use.

The Barnes Negative Differential Mobility model [88] was used in the transit

region at high electric filed. The electron saturation velocity, as a function of

electric field, is given by:

CRIT

satn

























)(

……………………..………………..(7.4)

where

satn

is the electron saturation velocity,

is the low field electron

mobility,

CRIT

is the critical electric field in

cmV /

is a constant equal to 2 for

GaAs and E is the electric field. The critical electric field for GaAs was defined as

3.4 10

CRIT

E = ×

121

Chapter 7 Physical Models

It should be noted here that for both standard and negative differential mobility

models, an empirical temperature dependent saturation velocity for GaAs is

implemented as:

Lsatn

102.1103.11 ×−×=

scm /

…………………..………….(7.5)

where,

is the temperature in degree Kelvin. For the device model simulated at

300K, a GaAs saturation velocity value of

scm /107.7

was used during device

simulations.

7.3 Carrier Generation – Recombination Models

The processes responsible for carrier generation and recombination fall under

following six main categories:

•

Phonon transitions

•

Photon transitions

•

Auger transitions

•

Surface recombination

•

Impact ionization

•

Tunnelling

In ATLAS

this six generation-recombination mechanisms are represented

using different models. For Gunn diode modelling, the Shockley-Read-Hall

(SRH) recombination (phonon Transition) and SRH Concentration-Dependent

Lifetime (Photon Transition) models have been used for AlGaAs and GaAs

regions respectively as listed in table 7.1.

7.3.1 Shockley-Read-Hall (SRH) Recombination

Phonon transition within a semiconductor is due to a trap or defect within its

forbidden gap. The theory of phonon transition was discovered first by Shockley

122

Chapter 7 Physical Models

and Read [89], and then later by Hall [90]. The Shockley-Read-Hall (SRH)

recombination model was used for the AlGaAs regions and implemented as:

























Ε−

























−

TRAP

ieN

TRAP

ieP

SRH

npn

expexp

ττ

……..(7.6)

where

and

are the electron and hole life times, which were defined for

each individual Al

1-x

As region. The values were calculated using the virtual

crystal approximation formula for both GaAs and AlAs. The values for these

materials were taken from the SILVACO

reference tables.

TRAP

indicates the

trap energy and intrinsic Fermi level difference in electron volts. Its default value

)0(

ETRAP

was used to correspond to efficient recombination of one trap layer

present in the material.

is the lattice temperature in degrees Kelvin. The

model was activated by defining srh in the model statement for the AlGaAs

region.

7.3.2 SRH Concentration-Dependent Lifetime model

The concentration based SRH model was used for the GaAs regions due to the

different impurity concentration level in each. The modified SRH model with

carrier life time as a function of doping concentration is implemented in ATLAS

[81, 92] as follows:

























Ε−

























−

TRAP

ien

TRAP

ieP

SRH

npn

expexp

ττ

…….… (7.7)

where

SRH

total

SRH

total

BPAP

























………………….………………………(7.8)

SRH

total

SRH

total

BNAN

























…………………..…………………….(7.9)

123

Chapter 7 Physical Models

Here, N is the carrier concentration level and the constant values AN, BN were

taken from the SILVACO

reference tables.

SRH

and

SRH

are the SRH

concentration levels for electrons and holes respectively.

total

is the total carrier

concentration level.

and

are the electron and hole life times as discussed

earlier. Most of these parameters are well defined for GaAs in the SILVACO

default library at 300K. And so the default values were used in the model [81].

7.4 Carrier Statistics and Transport

Within a semiconductor the thermal equilibrium of electrons at temperature

obeys Fermi-Dirac statistics. The electron occupation probability for a given

energy,

)(

, is given as [81]:













Ε−Ε

=Ε

exp1

)(

…………………………….…………….(7.10)

KT>>Ε−Ε

(non-degeneracy) then the equation can be approximated as:













Ε−Ε

=Ε

f exp)(

…….……………………………………….(7.11)

The above approximation is known as the Boltzmann statistic. The Boltzmann

approximation simplifies the calculations whilst yielding satisfactory results. It

was used in the model that minimized convergence issues as compared to the

Fermi-Dirac statistics. In ATLAS

the use of the Boltzmann statistic has been

set at its default implementation [81].

The transport models, Drift Diffusion, energy balance and hydrodynamic models

available in ATLAS

are discussed. In the Gunn diode model the carriers were

selected as electrons. The electrons equations are presented only.

124

Chapter 7 Physical Models

7.4.1 Drift Diffusion Model

The Drift Diffusion model is used as default in ATLAS

to model carrier

transport or current density. It applies approximations and simplifications to the

Boltzmann Transport Equation. It does not introduce any independent variable in

addition to electrostatic potential and the electron concentration. This model

although simple, ignores non-local transport effects such as velocity overshoot

and diffusion due to carrier temperature. Resultantly, its accuracy decreases for

sub micron devices. The current density



for electrons (

) is given as;

n n n

J q nE qD n

= + ∇

 

………..…………………………………….(7.12)

where

is the electron mobility,



is the local electric field in

cmV /

is the

diffusion coefficient, which is calculated using Einstein’s relationship as;

n n

……..………………………………………………….(7.13)

7.4.2 The Energy Balance and Hydrodynamic Transport Models

In ATLAS

, the energy balance and hydrodynamic transport models are

available as current density non-local models. These models uses additional

coupling of the current density to the carrier temperature. The Energy Balance

Transport Model is derived from the Boltzmann Transport Equation using

Stratton derivation [93, 94]. It is simplified by applying some assumptions into the

hydrodynamic model [95-97].

Three energy balance transport equations for electrons comprising of energy

balance equation, current density and energy flux are as follows;

( )

1 3

. .

n n

n n n

S J E W nT

q t

∗

∇ = − −

  

……...……………………….(7.14)

n n n n n

J qD n q n qnD T

µ ψ

= ∇ − ∇ + ∇



……………………………….(7.15)

n Bn n

S K Tn J T

 

= − ∇ −

 

 

  

…………………………….……….(7.16)

125

Chapter 7 Physical Models

The energy balance transport equations simplify into the hydrodynamic transport

equations by substituting

∗

and

. The hydrodynamic transport

equations are given as;

( )

1 3

. .

n n

S J E W nT

q t

∇ = − −

  

………………...……………….(7.17)

n n n n n

J qD n q n qnD T

µ ψ

= ∇ − ∇ + ∇



………………….…………….(7.18)

n Bn n n

S K T J T

 

= − ∇ −

 

 

  

……………………….…….……….(7.19)

where,



is the electron energy flux density

is the thermal diffusion coefficient (equation 7.13)

is the energy density loss rates for electrons (equation 7.29)

is the thermal conductivity of electrons given as;

Bn n n

K qn T

 

 

 

…………………………..………………….(7.20)

The hydrodynamic model provides realistic results and better convergence. It

was implemented in the Gunn diode model using parameter hcte.el on the

models statement.

7.5 GIGA

– Self Heating Simulator

Lattice heat flow and general thermal environment simulation in the model has

been implemented using GIGA

, a self heating simulator, which extends

ATLAS

capabilities in the DeckBuild

environment. GIGA

is based on the

models suggested by Wachutka [91], which precisely accounts for the Joule

heating and heating and cooling due to carrier generation and recombination.

Ambient temperature conditions, realistic heat-sinks and thermal impedances are

also accounted for. Additionally, all the material and transport parameters

126

Chapter 7 Physical Models

defined for each region are made lattice temperature dependent throughout the

model.

Statement Parameter Details

material hc.comp Heat Capacity model

tcon.power Thermal conductivity model

models lat.temp Lattice Heat Flow Equation

e.taur.var

Electron energy relaxation time

thermcontact ext.temper Heat sink temperature

Heat sink electrodes are defined

with dimensions

Table 7.4 GIGA

– self heating simulator parameters

7.5.1 The Lattice Heat Flow Equation

GIGA

adds the following lattice heat flow equation to the primary equations

that are solved by ATLAS

( )

C T H

∂

= ∇ ∇ +

∂

………………….………………………….(7.21)

where,

is the heat capacitance per volume

is the thermal conductivity

is the heat generation

is the local lattice temperature

The heat capacitance can also be defined as

C C

, where

is the density of

material and

is the specific heat. The lattice heat flow equation is included

during simulation by defining lat.temp parameter on the models statement.

127

Chapter 7 Physical Models

7.5.2 Heat Capacity

The temperature dependent heat capacity for defined regions in a structure is

modelled as [98];

( )

300 1

300

C T C C

T C

 

 

−

 

 

 

 

= +

 

 

 

 

 

 

(J/cm

/K) …………………(7.22)

where,

(

)

C T

is the temperature dependent heat capacity (J/cm

/K)

is the mass density (g/cm

)

300

is the specific value for heat at 300K (J/K Kg)

is the material dependent specific value for heat (J/K Kg)

is the material dependent fitting parameter

The values of

300

, and

for both GaAs and AlAs are listed in table 7.5

[86]. They were defined using the C-interpreter functions. The temperature

dependent heat capacity for each region was modelled by specifying the

parameter hc.comp on the material statement.

Parameter Material Units

GaAs AlAs

5.32 3.76 g/cm

300

322 441 J/K Kg

50 50 J/K Kg

1.6 1.2 -

Table 7.5 Heat Capacity parameters

128

Chapter 7 Physical Models

In case of ternary compound (Al

1-x

As),

(

)

C T

is calculated by linear

interpolation.

7.5.3 Thermal Conductivity

The value of thermal conductivity,

, for each region is defined by specifying the

parameter tcon.power on the material statement. It uses the following equation;

( . )

( )

300

T NPOW

T CONST

 

 

 

(W/cm.K) ………………………………….. (7.23)

where,

is the thermal conductivity

is the lattice temperature

T CONST

is the thermal conductivity at 300K

T NPOW

is the fitting parameter

The value of thermal conductivity at 300K (

T CONST

) and fitting parameter (

T NPOW

) are defined using the C-interpreter functions as listed in table 7.6 [34,

85, 86].

Parameter Material Units

GaAs AlAs

T CONST

0.46

0.8 W/cm.K

T NPOW

1.25 1.37 -

Table 7.6 Thermal Conductivity parameters

129

Chapter 7 Physical Models

In case of ternary compound (Al

1-x

As),

T CONST

and

T NPOW

are

calculated as;

( )

1 . .

. .

AlGaAs

AlAs GaAs bowing

C C

T CONST

x x

x com x com

T CONST T CONST C

− ×

 

−

+ +

 

 

(W/cm) …... (7.24)

(

)

1 . .

AlGaAs GaAs AlAs

C C C

T NPOW x com T NPOW x com T NPOW= − × + ×

(.K) ……... (7.25)

where,

x com

is the composition fraction

bowing

is the bowing factor (W/cm.K)

The bowing factor (

bowing

) compensates sudden change of thermal conductivity

due to changing composition fraction (

x com

). The

T NPOW

parameter is

calculated by linear interpolation using Vegards law since experimental data

other than 300K is currently not available [86].

7.5.4 Heat Generation

In GIGA

, the heat generation term H from equation 7.21 is set to zero for

insulators. For conductors it is given as;

( )

divV

(J/cm

/K) ………………………………………….. (7.26)

where,

is the voltage in mV/K

is the mass density in g/cm

In Gunn diode model hcte.el is specified on the models statement. It uses

hydrodynamic model for electrons only since holes do not play any role in the

carrier transport effects. The heat generation ‘H’ is given as;

n g

H W E U J E

= + +

 

(J/cm

/K) …………………………………. (7.27)

where,

130

Chapter 7 Physical Models

E U

is the recombination heat power

is the Joule heating term

J E

 

is the Peltier Thomson Heat power

The recombination heat power (

E U

) is the product of band gap energy and the

net generation-recombination rate

. The net generation-recombination rate is

defined as;

A A

SRH n p n p

U R R R G G

= + + − −

(/cm

.s) ………………….……….. (7.28)

where,

SRH

is the SRH recombination rate (equation 7.6 and 7.7)

A A

n p

R R

are the Auger recombination rates

n p

G G

are the impact ionization rates

The carriers are heated due to lattice temperature increase and they exchange

energy with the lattice using recombination and regeneration processes

(equation 7.28). Both these effects are taken into account by the Joule heating

term

which is defined as the energy density loss rate.

The energy density loss rate for electrons is given as;

(

)

( )

3 3

2 . 2

n L

n n SRH g n n

k T T

W n kT R E G R

taurn el

−

= + + −

(/cm

.s) ……….. (7.29)

where,

taurn el

is the electron energy relaxation time.

7.5.5 Electron Energy Relaxation Time

In ATLAS

no temperature dependent electron energy relaxation model is

available. The electron energy dependent model was used instead. It is activated

by using e.taur.var on the models statement. The model parameters listed in

131

Chapter 7 Physical Models

table 7.7 [85] are defined on the material statement. The electron energy

energy

is given as;

energy

n n

W kT

(eV) …………………………………………..….. (7.30)

The electron energy relaxation time is determined by quadratically varying values

with respect to electron energy

energy

, by using following relations;

. 1, . 1

. . 2, . 2

. 3, . 3

energy

e n

energy

TRE T W TRE W

taurn el TRE T W TRE W

TRE T W TRE W





= = =







(eV) ……………….. (7.31)

Parameter Value Units

. 1

TRE T

0.5

. 2

TRE T

1.24 ps

. 3

TRE T

1.74 ps

. 1

TRE W

0.06 eV

. 2

TRE W

0.3 eV

. 3

TRE W

4.5 eV

Table 7.7 Electron energy relaxation time parameters

7.5.6 Thermal Boundary Conditions

In ATLAS

, the thermal boundary condition provides total energy flux after the

lattice heat flow equation is solved. It is given by;

( )

boundary

thermal

J T Temp

= −

………..…………………….……... (7.31)

132

Chapter 7 Physical Models

where,

is the lattice temperature

Temp

is the temperature defined on the thermcontact statement

thermal

is the thermal resistance given as;

thermal

Alpha

………..………………………………….……. (7.31)

The parameter

Alpha

is defined on the thermcontact statement. The

thermcontact statement also defines the position of the thermal contact (heat

sink) and temperature

Temp

. In the Gunn diode model two thermal boundary

conditions were defined for Cathode and anode heat sinks respectively. The

parameter

2.5

Alpha

was defined.

7.6 C-Interpreter Functions

Some of the physical models used in ATLAS

can be modified using the built-in

C-language interpreter. C-interpreter functions are particularly useful to precisely

define material parameters not available in ATLAS

reference tables. The C-

interpreter functions containing analytic model descriptions which were written

are listed in table 7.8.

Function Model Description

f.bandcomp Band Gap Composition based band gap.

f.bandcomp parameter is defined in the material statement.

f.vsatn Saturation

Velocity

Composition and temperature dependent saturation

velocity.

f.vsatn parameter is defined in the material statement.

f.munsat Mobility Determines the temp dependent mobility. Set by fldmob

evsatmod=0.

f.enmun parameter is defined in the material statement.

f.enmun Mobility Determines temperature dependent low field mobility.

f.enmun parameter is defined in the mobility statement.

f.tcap GIGA Thermal capacity in Giga.

f.tcap parameter is defined in the material statement.

f.tcond GIGA Thermal conductivity in Giga.

f.tcond parameter is defined in the material statement.

Table 7.8 C-Interpreter Functions used in model development

133

Chapter 7 Physical Models

7.7 Conclusion

The accuracy of the simulated response depends on the choice of material

parameters and the selection of physical models (and their associated

parameters) used. The material and model parameters used were therefore

extensively researched from a variety of sources. They have been presented and

discussed, which were carefully selected for each layer depending on the

material and carrier concentration in the region. It included Mobility models,

Carrier Generation-recombination models, Carrier Statistics and Transport

models and GIGA

Self Heating Simulator. It was shown that the analytic

function based on the work of Caughey and Thomas is used for modelling field

dependant mobility in the contact layers, spacer layers, AlGaAs launcher and

also the transit region for field strengths below the negative differential mobility

(NDM) threshold. For field strengths in the transit region above the NDM

threshold, the Barnes NDM model is used. The Shockley-Read-Hall (SRH)

recombination model is applied in the nominally undoped Al

(1-x)

As barrier.

The SRH concentration-dependent lifetime model is used in the rest of the

device to account for the varied silicon impurity concentrations.

The ATLAS

non-isothermal device modeller module, Giga, used for local

temperature, self-heating, lattice heat flow and heatsinking effects have been

discussed. It is based on the models suggested by Wachutka and accounts for

Joule heating along with heating and cooling due to carrier generation and

recombination. Finally, it was shown that where the platform did not allow

sufficient flexibility to accurately account for the interdependency of parameters,

appropriate C-language functions were written and incorporated through the in-

built C-interpreter.

Chapter 8 Gunn diode Results: DC Analysis

8.1 Introduction

The simulated DC characteristics of the model are presented in this chapter.

They were used to study hot electron injection doping spike. The doping spike

carrier concentration optimum value was determined using a modelled-measured

approach.

In order to examine the accuracy of the developed model a comparison to

measured data needs to be made. A large amount of data from production and

engineering testing carried out at e2v Technologies Plc was made available for

this purpose. The available data included the characteristics of 77 GHz (1.65 µm

transit region device) devices with a range of epitaxial structures which were

measured over a range of temperatures. Also available was data from a new 125

GHz second harmonic device and 94 GHz fundamental device. A relatively

constant I-V current is

The Gunn diodes epitaxially grown at The University of Manchester were

processed / packaged at e2v technologies Plc. They were die-packaged and

etched to current (nominally 750-1200mA) to reduce heat generation and get

better heat sinking. Resultantly, relatively constant current levels were achieved

in the I-V relationships despite reducing device transit region length and

increasing its doping.

8.2 Advanced Step Graded Gunn Diode Modelled–Measured Results

The advanced 77 GHz second harmonic GaAs Gunn diodes discussed here

incorporate a step-graded AlGaAs hot-electron launcher, and are commercially

manufactured in volume by e2v Technologies (UK) Plc. for use in 77GHz

automotive Autonomous Cruise Control (ACC) systems. Tight specifications are

placed on parameters such as RF output power, applied voltage controlled

frequency tuning range and the oscillation frequency’s temperature dependence.

134

135

Chapter 8 Gunn diode Results: DC Analysis

These parameters are all highly dependant on hot-electron injector composition,

especially carrier concentration in the injector’s doping spike [2, 3].

The hot electron injector effectiveness was evaluated using the 77 GHz model.

The doping spike carrier concentration, DC I-V characteristics and doping spike

effects on I-V curves are discussed below.

8.2.1 Doping Spike Carrier Concentration

The hot electron injector, situated between the n

cathode contact layer and the

transit region, consists of two main components: a graded AlGaAs launcher

and a doping spike which is required to modify the forward-bias downstream

electric field in the transit region. As shown in fig. 8.1, in its absence a depletion

region is formed behind the launcher which prevents the nucleation of high-field

domains. With the increase of applied potential, the negative-field gradient rises

and as a result the depletion region grows forming a dead zone. The dead zone

at the start of transit region reduces the performance of the launcher. The dead

zone eventually diminishes as the applied voltage is increased.

The device model with

doping spike (cc

318

101

−

× cm

) resulted in the positive

electric field gradient, which keeps the electron concentration throughout the

1.5 2.0 2.5 3.0 3.5

1e13

1e14

1e15

1e16

1e17

1e18

1e19

Contact

Layer

Transit Region

AlGaAs Launcher

Electron Concentration (cm

-3

)

Device Distance (

Contact

Layer

Fig.

The effects of the doping spike (cc 1×10

) on electron concentration

in the transit region (——; spike present, -----; spike absent)

136

Chapter 8 Gunn diode Results: DC Analysis

transit region as shown in Fig. 8.1. Concomitantly it eliminated the depletion

region, thereby ensuring the elimination of the dead zone and leading to high

efficiency of the hot electron injector. This compares well to previous simulations

results obtained and published by another group using SILVACO

[10]. This

greatly increased confidence in the developed model.

8.2.2 DC I-V Characteristics of 77 GHz second Harmonic GaAs Gunn

A comparison of the simulated I-V characteristics and measured data at

temperature 300K is shown in Fig. 8.2. The measured results are from a

manufactured device, and supplied by ‘e2v Technologies Plc’. It can clearly be

seen that the simulated I-V characteristics in both forward and reverse bias

match well to those of an actual device thus validating the use of physical

models selected for the simulations. It is noted that in forward bias the simulated

response beyond three volts shows greater negative differential resistance than

the measured data. The higher negative differential is probably due to the ideal

heat sinks higher efficiency and lower resistance values used in the simulation.

8.3 Doping Spike Effects

To compliment experimental data and investigate the relationship between

doping spike carrier concentration and I

asym

, the model was simulated with

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Reverse IV Curves

Current (A)

Voltage (V)

Measured 1928A 300K

Model 1928A 300K

Forward

IV Curves

Fig.

Modelled and measured

forward

reverse

curves.

137

Chapter 8 Gunn diode Results: DC Analysis

varying carrier concentration (

18 3

2 10

−

318

101

−

× cm

17 3

7.5 10

−

, and

317

105

−

× cm

) shown in Fig. 8.3. It can be seen that as the doping spike carrier

concentration decreases the injector performance degrades and as a result the

asymmetry increases. Since the injector does not play any role being at the end

of the transit region during reverse bias condition, similar reverse I-V curve could

be seen in the Fig. 8.4.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Forward

bias

1.28

1.48

1.17

1.16

Current (A)

Voltage (V)

Spike cc = 2x10

-3

Spike cc =1x10

-3

Spike cc = 5x10

-3

Spike cc =2.5x10

-3

Reverse bias

Fig.

Effects of variation in doping spike carrier

concentration on simulated I

characteristics (numbers next to forward bias curves represent calculated asymmetry

values)

Electron Concentration with Increasing Spike at 0.3 volts

Device Distance along the length in microns

Electron Concentration (cm

)

Transit Region

Fig.

Electron concentration with changing doping spike carrier concentration (cc)

138

Chapter 8 Gunn diode Results: DC Analysis

Fig. 8.5 shows the I

asym

plotted verses doping spike carrier concentration. It can

be seen that although I

asym

decreases as the carrier concentration increases,

which agrees with the trend noted for the experimental data, there is a small

discrepancy between the absolute measured and simulated values. However

qualitatively, it shows the same trend and is remarkably reproduced. This

discrepancy was the result of inaccurate representation of the electric field and

current flow through the device inherent to the method used for scaling the 2D

model to obtain the correct device area. This has a much greater effect when

modelling the Gunn diode which is a ‘field driven’ device, than with voltage or

current driven devices for which SILVACO

was originally intended. These

results were published [99]. Cylindrical device geometry was defined to take into

account cylindrical geometry of the manufactured device.

A full 3D cylindrical model was developed and results compared to that of a 2D

model. This computationally expensive (tens times than a 2D model) full 3D

cylindrical model [100] was developed using SILVACO

ATLAS

version

5.12.1 in order to ensure that the modelled electric field across the device

realistically represented that in an actual device [100]. However, due to the

improved meshing capabilities of the since released version 5.12.32 of

SILVACO

ATLAS

, a cylindrical device geometry can now be defined and

0.0 0.5 1.0 1.5 2.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

Simulated

Measured

Current asymmetry

Doping spike carrier concentration (

−3

)

Curve fit to measured data

order exponential decay)

Fig.

Simulated and measured asymmetry values versus doping spike carrier

concentration

139

Chapter 8 Gunn diode Results: DC Analysis

applied to a 2D model. After verification that a 2D model using the new meshing

technique produced similar results to the full 3D model, the method was adopted

and a large reduction in computational time obtained for subsequent simulations.

8.4 Measured 77 GHz Devices Data Comparison with Modelled Devices

In order to further investigate the accuracy of the model, its simulated response

was compared to the measured results from devices with slightly different

epitaxial structures (the main differences are the doping spike carrier

concentrations) shown in table 8.1. Ambient temperature variation is also

included in both simulated and measured data.

Wafer Name

Doping Spike Carrier

Concentration (cm

-3

)

Transit Region

Length (µm)

Transit Region Carrier

Concentration (cm

-3

)

VMBE 1928A

1.65

1.1

VMBE 1900

1.65

1.1

VMBE 1909

7.5

1.65

1.1

Table 8.1 Measured devices with slightly different epitaxial structures – doping spike

carrier concentration variation

The results for each epitaxial structure will now be discussed.

8.4.1 Doping Spike Carrier Concentration 1

××

-3

Modelled Device

Results (VMBE 1928A)

The forward and reverse I-V curves of a measured device (VMBE 1928A) and

modelled device 1928A are shown in Fig. 8.6. Doping spike with carrier

concentration 1

-3

was used in both the modelled and measured devices.

Both modelled and measured data compared very well. The difference in

modelled forward I-V curve, showing higher negative differential resistivity, at

around 3V and onwards is thought due to the heat sinks higher efficiency.

140

Chapter 8 Gunn diode Results: DC Analysis

Modelled and measured temperature curves for VMBE 1928A devices are shown

in Fig. 8.7. The I-V curves were plotted for an ambient temperature of 300K and

323K respectively. Upon increasing the temperature to 323K the injector

effectiveness is reduced and as a result the modelled device’s current saturated

at a lower value which is in agreement with published theoretical data. The

effects of ambient temperature variation on the simulated and measured results

resulted in a close match. This further validates the developed model and use of

the GIGA simulator.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Measured VMBE 1928A 300K

Measured VMBE 1928A 323K

Modelled VMBE 1928A 300K

Modelled VMBE 1928A 323K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1928A Temperature forward

reverse

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Measured VMBE 1928A 300K

Modelled VMBE 1928A 300K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1928A Modelled and measured forward

reverse I

V curves

141

Chapter 8 Gunn diode Results: DC Analysis

8.4.2 Doping Spike Carrier Concentration 5

××

-3

Modelled Device

Results (VMBE 1900)

The spike doping carrier concentration was reduced from 1

-3

to 5

-3

. Both simulated and measured devices yielded the asymmetrical response

shown in Fig. 8.8. A good match can be seen between the modelled and

measured data especially in reverse bias conditions. Again the forward I-V curve

showing more negative differential resistance is attributed to the higher efficiency

of the ideal heat sinks used in the modelled device.

Fig. 8.9 shows the temperature simulation for VMBE 1900 with reduced spike

carrier concentration. Again the modelled reverse I-V curve matched closely to

the measured data.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Current (A)

Voltage (V)

Measured VMBE 1900 300K

Measured VMBE 1900 323K

Modelled VMBE 1900 300K

Modelled VMBE 1900 323K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1900

Temperature forward

reverse I

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Current (A)

Voltage (V)

Measured VMBE 1900 300K

Modelled VMBE 1900 300K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1900 Modelled and measured forward

reverse I

V curves

142

Chapter 8 Gunn diode Results: DC Analysis

8.4.3 Doping Spike Carrier Concentration 7.5

××

-3

Modelled Device

Results (VMBE 1909)

The spike doping carrier concentration was increased from 5

-3

7.5

-3

. Both simulated and measured devices yet again yielded the

asymmetrical response shown in Fig. 8.10. A good match can be seen between

the modelled and measured data especially in forward bias conditions.

Modelled and measured temperature curves for VMBE 1909 devices are shown

in Fig. 8.11. The effects of ambient temperature variation on the simulated and

measured results resulted in a close match, especially in forward bias conditions.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Current (A)

Voltage (V)

Modelled VMBE 1909 300K

Modelled VMBE 1909 323K

Measured VMBE 1909 300K

Measured VMBE 1909 323K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1909 Temperature forward

reverse I

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Current (A)

Voltage (V)

Modelled VMBE 1909 300K

Measured VMBE 1909 300K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1909 Modelled and measured forward

reverse I

V curves

143

Chapter 8 Gunn diode Results: DC Analysis

Although it differs slightly in the reverse I-V curves but the onset of Negative

Differential Resistance (NDR) and bias region where device provides NDR

remains similar. This further validates the developed model and use of the

GIGA

simulator.

Fig. 8.12 shows the I

asym

plotted verses doping spike carrier concentration for

three discussed devices including 1928A, 1900 and 1909. Both measured and

modelled devices provided decreasing current asymmetry values with the

increase in doping spike carrier concentration. The decrease in carrier

concentration is in accordance with the theory and indicates the increase in

injector efficiency. A close match between measured and modelled results is due

to simulating cylindrical geometry for the modelled devices. The cylindrical

geometry correctly simulated the electric field effects in the modelled devices.

The hot electron injector highest efficiency was achieved with doping spike

carrier concentration 1

-3

. Its efficiency reduced by decreasing doping

spike carrier concentration. However, increasing carrier concentration had

minimal effect. Therefore, doping spike carrier concentration 1

-3

was

used in both modelled and fabricated devices.

0.5 0.6 0.7 0.8 0.9 1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.0 x

−3

0.75 x

−3

Current asymmetry

Doping spike carrier concentration (10

−3

)

Measured Asymmetry

Modelled Asymmetry

0.5 x

−3

Fig.

Simulated and measured asymmetry values versus doping spike carrier

concentration.

144

Chapter 8 Gunn diode Results: DC Analysis

8.5 Higher Frequency Measured–Modelled DC I-V Characteristics

The entries listed in table 8.2 mostly correspond to material previously grown for

experimental studies in the development of D-band GaAs Gunn devices: as such

sample devices had already been fabricated and tested, with measured data

available. This is with the exception of XMBE189 (0.4 µm transit region length)

which was produced to test whether GaAs based Gunn oscillators operating at

sub-millimetre waves can be realized. It can be seen in table 8.2, that as the

transit region length was decreased, the length – carrier concentration product,

L.N

, was kept approximately constant.

Wafer Name

Doping Spike Carrier

Concentration (cm

-3

)

Transit Region

Length (µm)

Transit Region Carrier

Concentration (cm

-3

)

VMBE1901 1

1.65µm 1.1

VMBE1950 1

1.1µm 2

VMBE1897 1

0.9

m 2.5

VMBE1898 1

0.7

m 3.5

XMBE189 1

0.4

m 5.4

Table 8.2 Measured devices with slightly different epitaxial structures – transit

region length and its carrier concentration variation

The simulated DC characteristics for each device (table 8.2) are presented and

compared with measured results [101]. Ambient temperature variation is also

included in both simulated and measured data. These devices were die-

packaged and etched to current (nominally 750-1200mA) to reduce heat

generation and achieve better heat sinking

145

Chapter 8 Gunn diode Results: DC Analysis

8.5.1 1.65 µm Transit Region Device DC Results (VMBE 1901 – 77 GHz

Device)

The forward and reverse I-V curves of a measured device (VMBE 1901) and

modelled device 1901 are shown in Fig. 8.13. A doping spike with carrier

concentration 1

-3

was present in both the modelled and measured

devices. The complete reverse bias curve for a measured device could not be

obtained from e2v Technologies Plc. However, the available measured data

compared reasonably with the device model.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Current (A)

Voltage (V)

Measured VMBE 1901 300K

Measured VMBE 1901 323K

Modelled VMBE 1901 300K

Modelled VMBE 1901 323K

Forward

IV Curve

Reverse IV Curve

Fig.

VMBE 1901 Temperature

forward

and

reverse

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Current (A)

Voltage (V)

Measured VMBE 1901 300K

Modelled VMBE 1901 300K

Forward

IV Curve

Reverse IV Curve

Fig.

VMBE 1901

Forward

and

reverse

curves for a

GHz

2nd

harmonic

device at 300K

146

Chapter 8 Gunn diode Results: DC Analysis

Modelled and measured temperature curves for VMBE 1901 devices are shown

in Fig. 8.14. The I-V curves were plotted for an ambient temperature of 300K and

323K respectively. Upon increasing the temperature to 323K the injector

effectiveness reduced and as a result the modelled device’s current saturated at

a lower value which is in agreement with published theoretical data. The effects

of ambient temperature variation on the simulated and measured results resulted

in a close match. This further validates the developed model and use of the

GIGA

simulator.

8.5.2 1.1 µm Transit Region Device DC Results (VMBE 1950 - 125GHz

Device)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Measured VMBE 1950 300K

Measured VMBE 1950 323K

Modelled VMBE 1950 300K

Modelled VMBE 1950 323K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1950 Temperature

forward

and

reverse

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Measured VMBE 1950 300K

Modelled VMBE 1950 300K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1950

Forward

and

reverse

curves for a

62.5GHz

undamental

125GHz 2nd harmonic device at 300K

147

Chapter 8 Gunn diode Results: DC Analysis

125 GHz device comprised of a modified device structure. To model and

compare the new structure (VMBE 1950), the modelled 77 GHz structure’s

transit region was modified by reducing its length from 1.65 µm to 1.1 µm and its

doping was increased from 1.1

-3

to 2

-3

. The measured and

modelled data is shown in Fig. 8.15. A good match can be seen between the

modelled and measured data especially in forward bias low voltage conditions.

The forward and reverse I-V curves showing more negative differential

resistance is attributed to the higher efficiency of heat sinks used in the modelled

device.

125 GHz device model was simulated at increasing temperature. The results

indicated similar characteristics as were achieved for 77 GHz devices. The

temperature effects on the barrier effectiveness and current saturation correlated

to the published theory.

After successful development of a model for 125 GHz second harmonic device

the modelling work was extended to develop models for higher frequency

devices. Further models developed included VMBE 1897, VMBE 1898 and

XMBE 189.

8.5.3 0.9 µm Transit Region Device DC Results (VMBE 1897- 125 GHz

Device)

VMBE 1897 device comprised of a sub-micron transit region length device

structure. To model and compare the new structure (VMBE 1897), the modelled

77 GHz structure’s transit region was modified by reducing its length from 1.65

µm to 0.9 µm and its doping was increased from 1.1e16 cm

-3

to 2.5e16 cm

-3

The modelled and measured DC I-V curves again yielded a close match (Fig.

8.17).

The measured modelled results at increasing temperature provided a close

match (Fig. 8.18). The results indicated similar characteristics as were achieved

for larger transit region devices. The temperature effects on the barrier

effectiveness and current saturation correlated to the published theory.

148

Chapter 8 Gunn diode Results: DC Analysis

8.5.4 0.7 µm transit region device DC results (VMBE 1898 - 100GHz device)

The simulated and measured forward and reverse bias I-V characteristics of a

device with a 100 GHz fundamental frequency are shown in Fig. 8.19 where

again, an extremely strong match can be seen especially in the forward I-V

curves.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

Current (A)

Voltage (V)

Measured VMBE 1897 300K

Measured VMBE 1897 323K

Modelled VMBE 1897 300K

Modelled VMBE 1897 323K

Forward

IV Curve

Reverse IV Curve

Fig.

VMBE 1897 Temperature

forward

and

reverse

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

Current (A)

Voltage (V)

Measured VMBE 1897 300K

Modelled VMBE 1897 300K

Forward

IV Curve

Reverse IV Curve

Fig.

VMBE 1897

Forward

and

reverse

curves for a

62.5

GHz

undamental

125GHz 2nd harmonic device at 300K

149

Chapter 8 Gunn diode Results: DC Analysis

The measured modelled results at increasing temperature provided a close

match (Fig. 8.20). The results indicated similar characteristics as were achieved

for lower frequency devices. The temperature effects on the barrier effectiveness

and current saturation correlated to the published theory.

8.5.5 0.4 µm Transit Region Device DC Results (XMBE 189 – 200 GHz

Device)

Having gained sufficient experience and confidence in the models developed up

to 100 GHz fundamental frequency, the model was then extended to higher

frequency devices (~ 200GHz), thereby addressing experimentally unreached

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Measured VMBE 1898 300K

Measured VMBE 1898 323K

Modelled VMBE 1898 300K

Modelled VMBE 1898 323K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1898 Temperature

forward

and

reverse

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Current (A)

Voltage (V)

Measured VMBE 1898 300K

Modelled VMBE 1898 300K

Forward IV Curve

Reverse IV Curve

Fig.

VMBE 1898

Forward

and

reverse

curves for a

GHz

fundamental

device at 300K

150

Chapter 8 Gunn diode Results: DC Analysis

limits of operation for GaAs-based Gunn devices. Extensive modelling points out

to a transit region of 0.4 µm as providing the most promising asymmetric I-V

curves for a 200 GHz fundamental device as shown in Fig. 8.21. After

successive iterations simulation suggested that the optimal transit region carrier

concentration level was 5.4 x10

-3

and spike carrier concentration level was

1x10

-3

(Fig. 8.21) The device epitaxial structure was grown as discussed

before.

The device I-V measurements have provided an asymmetric response (Fig.

8.22). Thus a 200 GHz fundamental device has been made for the first time from

a GaAs/AlGaAs material system. It is achieved through predictive modelling and

by sequentially extending device’s operating frequency while studying its

underlying principles. The time-domain responses are presented in the next

chapter.

The measured modelled results at increasing temperature provided a close

match (Fig. 8.23). The results indicated similar characteristics as were achieved

for lower frequency devices.

Fig.

XMBE 189

Modelled forward

and

reverse

curves

Effects of variation in

doping spike carrier concentration (cc) on simulated I-

V characteristics (numbers next to

forward bias curves represent calculated asymmetry values)

. Transit region carrier

concentration was 5.4 x10

-3

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

Forward

1.73

1.57

1.52

1.46

Current (A)

Voltage (V)

Spike cc = 2.0x10

-3

Spike cc = 1.0x10

-3

Spike cc = 9.0x10

-3

Spike cc = 7.52x10

-3

Spike cc = 7.0x10

-3

Spike cc = 6.5x10

-3

Spike cc = 5.0x10

-3

1.26

1.30

Reverse

151

Chapter 8 Gunn diode Results: DC Analysis

8.6 Conclusion

The simulated DC I-V characteristics have been presented of a 2D physical

model developed using SILVACO

for a GaAs Gunn diode with a step graded

AlGaAs hot electron injector. They were used to study relationship between I

asym

and doping spike carrier concentration. The doping spike carrier concentration

optimum value 1x10

-3

was determined using modelled-measured approach.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Current (A)

Voltage (V)

Measured XMBE 189 300K

Measured XMBE 189 323K

Modelled XMBE 189 300K

Modelled XMBE 189 323K

Fig.

XMBE

189 Temperature

forward

and

reverse

V curves

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Current (A)

Voltage (V)

Measured XMBE 189 300K

Modelled XMBE 189 300K

Forward IV Curve

Reverse IV Curve

Fig.

XMBE 189

Modelled forward

and

reverse

curves for a 200

GHz

fundamental device at 300K

152

Chapter 8 Gunn diode Results: DC Analysis

Giga, a self heating simulator available in SILVACO

, was used to simulate the

device’s lattice heating effects. It was used with the definition of heat sinks at the

cathode and anode respectively.

The developed model comparison with measured data was shown. The

measured data over a range of temperatures included the DC characteristics of

77 GHz (1.65 µm transit region device) devices with a range of epitaxial

structures, 125 GHz second harmonic device, ~100 GHz fundamental device

and ~200 GHz fundamental device. A good match between measured and

simulated I-V curves validates the choice of physical models and parameters

used.

It was shown that the Gunn diodes were die-packaged and etched to current

(nominally 750-1200mA) to reduce heat generation and achieve better heat

sinking. Resultantly, relatively constant current levels were achieved in the I-V

relationships despite reducing device transit region length and increasing its

doping.

Chapter 9 Gunn diode Results: Time-domain analysis

9.1 Time-Domain (Transient) Solutions

Transient solutions were used to obtain a time dependent response such as

determining the device oscillating frequency in biased condition. These solutions

provided modelled device time-domain responses. The time-domain simulations

helped to analyse modelled device RF characteristics. The RF characteristics

such as output power and oscillation frequency are highly dependant on

interactions between the device and the resonant circuit in which it is mounted.

Wafer Name

Target frequency

(GHz, Harmonic)

Transit Region

Length (µm)

ansit Region Carrier

Concentration (cm

-3

)

VMBE1901 70-80GHz, 2

harmonic 1.65 µm

1.1

VMBE1950

62.5GHz, Fundamental /

125GHz, 2

harmonic

1.1 µm

VMBE1897

62.5GHz, Fundamental /

125GHz, 2

harmonic

0.9

m 2.5

VMBE1898 100GHz, Fundamental

0.7

m 3.5

XMBE189 NA

0.4

m 5.4

Table 9.1 Measured devices with slightly different epitaxial structures – transit

region length and its carrier concentration variation

9.2 Time-Domain Analysis of various Epitaxial Structures

The results of the time-domain simulations for each device listed in table 9.1 are

presented. These simulations were performed with the terminal impedance set to

Ω with the sole intention of establishing the limits of transit region length

153

154

Chapter 9 Gunn diode Results: Time-domain analysis

capable of supporting a steady-state (non-decaying) oscillation. This free-running

frequency is reduced when the device is embedded in a practical resonant

circuit: a full discussion of this concept was explained in chapter 3. The simulated

oscillation frequencies are therefore not expected to precisely match those

measured (which are defined by oscillator design), but to be slightly higher. In

addition, the RF power (calculated from the average bias voltage and the current

oscillation) cannot be expected to match that of a specific oscillator.

Nevertheless, these simulations allowed performing large-signal device

characterizations and comparing to measured results [102].

9.2.1 1.65 µm Transit Region Device Results (VMBE1901)

It can be seen that in Fig. 9.1, the simulated time-domain response for the device

with a 2.2V external bias, although sinusoidal in nature at a frequency of

approximately 45 GHz, appears to be extremely noisy. At a higher external bias

of 3.2V, the free-running oscillation in the time-domain response (Fig. 9.2) is now

at the lower frequency of 40 GHz, does not exhibit the noise seen at a 2.2V

external bias, and also exhibits a much higher peak-to-peak value (which

increases from 12 mA to 50 mA). This is in complete agreement with the

explanation given in chapter 6 section 6.4 regarding the behaviour of a free-

running Gunn device as the bias voltage is increased.

100 150 200 250 300 350 400

0.960

0.965

0.970

0.975

0.980

0.985

0.990

Anode Current (A)

Transient Time (ps)

45 GHz at 2.2V

Fig.

The

simulated fundamental frequency time

domain response of a device

fabricated from VMBE1901 (1.65 µm transit length) under a 2.2V external bias

155

Chapter 9 Gunn diode Results: Time-domain analysis

9.2.2 1.1 µm Transit Region Device Results (VMBE1950)

The time-domain response obtained and given in Fig. 9.3, shows a fundamental

free-running frequency of approximately 66 GHz. The fabricated device has been

shown to provide around 130 mW at 62 GHz operating in fundamental mode,

and 40 mW at 121.5 GHz in a second harmonic oscillator: the latter is the highest

reported power for a GaAs device at this frequency [103]. The results of the time-

100 150 200 250 300 350 400

0.760

0.765

0.770

0.775

0.780

0.785

Anode Current (A)

Transient Time (ps)

66 GHz at 2.5V

Fig.

The simulated fundamental frequency

time

domain response of a device

fabricated from VMBE1950 (1.1 µm transit length) at an external bias of 2.5V,

temperature

300K.

100 150 200 250 300 350 400

0.92

0.93

0.94

0.95

0.96

0.97

0.98

0.99

Anode Current (A)

Transient Time (ps)

40 GHz at 3.2V

Fig.

The simulated fundamental frequency time

domain response of a device

fabricated from VMBE1901 (1.65 µm transit length) under a 3.2 V external bias.

156

Chapter 9 Gunn diode Results: Time-domain analysis

domain simulations are therefore inline with those expected from measured

results. It is noted here that although the time-domain simulations of the device

were performed with an external bias of 2.5 V, unlike the result for the 1.65 µm

device at the same bias voltage, the trace shows no noise. This is thought to be

due to the shortened transit region leading to a higher applied field across the

device and so the formation of stronger oscillations for the same applied external

bias.

9.2.3 0.9 µm Transit Region Device Results (VMBE1897)

The time-domain simulations showed stable free running fundamental frequency

oscillations at 79 GHz as shown in Fig. 9.4. A fabricated device with this epitaxial

structure and geometry provided a maximum fundamental frequency power of

80-90 mW at 63 GHz, and around 10 mW at 123 GHz in a second harmonic

oscillator. The results are therefore again inline with theory and with those

obtained through measurement of a fabricated device.

9.2.4 0.7 µm Transit Region Device Results (VMBE1898)

The time-domain simulation results in Fig. 9.5, show stable oscillation at 106

GHz. The manufactured device yielded a maximum fundamental power of ~22

100 125 150 175 200 225 250 275 300

0.68

0.70

0.72

0.74

0.76

0.78

0.80

0.82

Anode Current (A)

Transient Time (ps)

79 GHz at 2.2V

Fig.

The simulated fundamental frequency time

domain response of a device

fabricated from VMBE1897 (0.9 µm transit length) at an external bias of 2.2V,

temperature

300K.

157

Chapter 9 Gunn diode Results: Time-domain analysis

mW at 94 GHz (that rolled off rapidly with frequency), the highest ever reported

for a GaAs Gunn device at this frequency: the device has not yet been tested at

second-harmonic and so no data is available. Again the time-domain solutions

are in agreement with the measured data.

A second time domain simulation was performed for this model at a higher

external bias voltage of 2.75V. A drop in the oscillation frequency to 99 GHz (Fig.

9.6) was observed along with an increase in the peak-to-peak current oscillation

from about 7.5 mA to about 100 mA. In addition, the oscillation observed was

Fig.

The simulated fundamental frequency time

domain response of a device

fabricated from VMBE1898 (0.7 µm transit length) at an external bias of 2.75 V

temperature 300K.

100 120 140 160 180 200 220 240 260 280

0.70

0.75

0.80

0.85

Anode Current (A)

Transient Time (ps)

99 GHz at 2.75V

100 125 150 175 200 225 250 275 300

0.780

0.785

0.790

0.795

0.800

Anode Current (A)

Transient Time (ps)

106 GHz at 1.7V

Fig.

The simulated fundamental frequency time

domain response of a device

fabricated from VMBE1898 (0.7 µm transit length) at an external bias of 1.7V,

temperature

300K.

158

Chapter 9 Gunn diode Results: Time-domain analysis

sinusoidal in nature and did not exhibit the deformation at the peaks observed in

Fig. 9.5 [104]. This again is in agreement with the expected behaviour for a

device outlined in chapter 6 section 6.4.

9.2.5 0.4 µm – 0.6 µm Transit Region Device Results (XMBE189)

Having gained sufficient confidence in the models with transit regions down to

0.7 µm from the excellent agreement between simulated and measured results,

the model was then extended to devices with even shorter transit regions.

Extensive DC modelling coupled with empirical data and device theory initially

pointed to a transit region of 0.4 µm as the minimum capable of supporting stable

oscillation [102]. It is noted that time-domain simulations were not available at the

time due to issues with computational solution convergence. After successive

iterations simulation suggested that the optimal transit region carrier

concentration level was 5.4 x10

-3

and spike carrier concentration level was

1x10

-3

The material was gown and devices fabricated and tested. The modelled and

measured DC I-V curves again matched very well as discussed in chapter 8.

However, although showing a negative differential resistance region in the

forward bias DC I-V characteristic, the fabricated device did not oscillate,

5 10 15 20 25 30

0.90

0.95

1.00

1.05

1.10

Anode Current (A)

Transient Time (ps)

Fig.

The simulated fundamental frequency time

domain response of a device

fabricated from XMBE189 (0.4 µm transit length) at an external bias of 3.5V,

temperature

300K.

159

Chapter 9 Gunn diode Results: Time-domain analysis

presumably due to the energy relaxation time of electrons in the GaAs being

larger than that required to support sustained oscillation in the device. This was

later confirmed with time-domain simulations of the device, the results of which

are shown in Fig. 9.7 where a decaying oscillation is observed. Subsequent

investigation showed that the shortest transit region length capable of supporting

sustained oscillation in this type of GaAs device is approximately 0.7 µm.

9.3 Time-Domain Response with a Resonant Cavity

The accurate physical modelling of Gunn diodes requires them to be mounted in

an oscillator circuit. The circuit provides DC power, and couples out the

generated RF power. Taking into account a particular oscillator in order to

compare modelled/measured results is challenging due to the interaction

between the diode and the circuit. The results presented here are for a model

that accurately computes the behaviour of an advanced Gunn diode mounted in

a cavity. The resonant cavity comprised of lumped-element LC parallel circuit,

resistive loss R is in series with the L.

Following previous successful development of physical models using

SILVACO

to predict the DC characteristics and time domain analysis of graded

gap Gunn diodes given its epitaxial structure and physical geometry, the

modelling work was extended to produce time-domain response of sub-

micrometer transit length device (0.7 µm transit region length) mounted in a

resonant cavity. The time-domain response allowed analyzing dipole domain

formation and transit region scaling effects on it.

9.3.1 Dipole Domain Formation

The dipole domains are nucleated in the transferred electron devices due to

doping inhomogeneities [9, 58]. The dipole nucleates from the cathode, grows to

a stable size and drift towards the anode. The dipole consists of an accumulation

and a depletion region with an associated electric field gradient. The electric

fields builds after the domain reaches the anode and a new dipole is formed. The

dipole domains cyclic formation results in current oscillations. This effect is

successfully modelled. Its formation in half cycle in the transit region (Fig. 9.8) is

160

Chapter 9 Gunn diode Results: Time-domain analysis

in agreement with [9, 58], which nucleate at the cathode, grows in size and drifts

towards the anode.

The dipole domain formation is shown in appendix C. The two dipole domain

formation cycles illustrates the complete RF cycle i.e. for one period whereas half

period is shown by one dipole domain cycle (Fig.9.8). The dipole domain

formation is cyclic and it repeats as a function of time to provide steady state RF

oscillations. In appendix C, the figure shows the increased carrier concentration

in the transit region as a function of time which shows nucleation, travel and

finally collapse of the dipole domain. In a complete cycle a total of 72 graphs

were saved and plotted in the TonyPlot

. The graphs were saved after every

0.158ps during time domain simulations. It took 9.36ps for the domain to

complete a RF cycle, thereby providing 106.8 GHz (

1 1

107

9.36

f GHz

= = 

). It is

in agreement with theoretical calculations, discussed in chapter 4 and chapter

seven section 7.2.5, (

7.6 10

108

0.7 10

f GHz

−

= =

∼

). The results are also in

agreement with experimental measurements.

0.2 0.3 0.4 0.5 0.6

8.60

8.62

8.64

8.66

8.68

8.70

8.72

8.74

8.76

8.78

Electron Concentration (1e17/cm3)

Transit Region Length (micro meter)

0.156ps

One domain total

nucleation time = 4.68ps

4.68ps

0.156ps

Fig.

patial dipole domain profiles in the transit region, for a 0.7

µm transit

length device. These are plotted at an applied bias of 3V at time, starting at 0.156ps

, in

one oscillation period T=4.68

ps. The domain first grows to a maximum size and then

drifts to the anode before it collapses there and a new domain nucleates.

161

Chapter 9 Gunn diode Results: Time-domain analysis

9.3.2 Modelled Device with Resonant Cavity Time-Domain Results

The modelled device has a predicted fundamental power of 55 mW at 94 GHz at

an applied bias of 3V (Fig. 9.9) when resistive losses of 0.9 Ohms were taken

into account (from matching of the DC I-V curves). The manufactured device, on

the other hand, delivered fundamental power of 22 mW at 94 GHz (applied bias

3V) and 21 mW at 96 GHz (applied bias 2.5V). These are the highest powers

ever reported for a GaAs Gunn operating in fundamental mode at these

frequencies. The modelled output power could be matched to the experimentally

observed ones when the resistive losses were increased by 1.1 Ohms to 1.8

Ohms. The higher modelled values are due to the ideal high quality factor Q

cavity assumed in the model, which clearly shows that more power can be

extracted from these devices by minimizing waveguide losses and device

package parasitic.

9.4 Conclusion

The results of the time-domain simulations for 77 GHz (1.65 µm transit region

device) devices with a range of epitaxial structures, 125 GHz second harmonic

device, ~100 GHz fundamental device and ~200 GHz fundamental device have

been presented. These simulations established the limits of transit region length

Fig.

Time

domain stable oscillation at 3

with C

diode

= 0.158

. The lumped

element LC parallel circuit values are L= 3.16 pH and C= 0.403

pF. The main loss R=0.9

Ohms is in series with the L. The current overshoot peaks are in agreement with [9]

which are due to electron experiencing energy and momentum inertial effects at high

frequencies.

60 70 80 90 100 110 120

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

Current (A)

Transient Time (ps)

0.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

Anode Internal Voltage (V)

voltage

current

current overshoot peaks

162

Chapter 9 Gunn diode Results: Time-domain analysis

capable of supporting a steady-state (non-decaying) oscillation. These

simulations allowed performing large-signal device characterizations and

comparing to measured results.

It was shown that that the shortest transit region length capable of supporting

sustained oscillation in GaAs device is approximately 0.7 µm (~100 GHz

fundamental device).

The modelling work was extended to produce time-domain response of sub-

micrometer transit length device (0.7 µm transit region length) mounted in a

resonant cavity. The resonant cavity designed in SILVACO

was presented that

comprised of lumped-element LC parallel circuit, resistive loss R is in series with

the L. The time-domain response allowed analyzing dipole domain formation and

transit region scaling effects on it.

It was shown that the dipole domains are nucleated in the transferred electron

devices due to doping inhomogeneities. Its formation in the transit region was

shown to be in agreement with published literature, which nucleates at the

cathode, grows in size and drifts towards the anode. In a 0.7 µm transit region

length device the two dipole cycles completed one RF time period thereby

providing ~107 GHz which was shown to be in agreement with theory and

measured results.

Chapter 10 Conclusion and future work

10.1 Conclusions

The ever increasing demand for THz devices and the huge market potential for

THz systems have led to an enormous amount of research in the area for a

number of years. Of all solid state millimetre wave oscillators, Gunn diodes have

the potential to be built into low cost, high-frequency and high output power

sources. These sources will ‘power’ future systems in the Terahertz (THz) region.

This region of the electromagnetic spectrum has attracted enormous attention

due to emerging applications in the field. These applications, including

automotive Autonomous Cruise Control (ACC), millimetre-wave imaging

systems, missile guidance, smart munitions and millimetre-wave radar systems

have made the requirement for physically-based models using model-experiment

approach a pressing need.

It was shown that a combination of Gunn and mixer diodes will make ideal power

source for future THz systems using a two-stage module with the initial

frequency source provided by the high frequency Gunn diodes. The output from

these diodes is then coupled into a multiplier module. The multiplier provides

higher frequencies by the generation of harmonics of the input signal by means

of a non-linear element. The non-linear element is provided by a Schottky diode

Varactor with low RC time constants to increase efficiency.

Design considerations of the microstrip and multiplier block were presented. The

semiconductor materials for such a complete device including modelling, design

and testing; focusing in depth on the Gunn diode module and Schottky diode

were also discussed.

A Gunn diode physical model has been developed using state of art software

SILVACO

. Thorough understanding of device physics was achieved to develop

the model. The model development included device physical structure building,

163

164

Chapter 10 Conclusion and future work

material properties specification, physical models definition and using

appropriate biasing conditions.

The physical models (and their associated parameters) used have been

presented and discussed. They were carefully selected for each layer, which

depended on the material and carrier concentration in the region. It included

Mobility models, Carrier Generation-recombination models, Carrier Statistics and

Transport models and GIGA

Self Heating Simulator. The analytic function

based on the work of Caughey and Thomas is used for modelling field

dependant mobility in the contact layers, spacer layers, AlGaAs launcher and

also the transit region for field strengths below the negative differential mobility

(NDM) threshold. For field strengths in the transit region above the NDM

threshold, the Barnes NDM model is used. The Shockley-Read-Hall (SRH)

recombination model is applied in the nominally undoped Al

(1-x)

As barrier.

The SRH concentration-dependent lifetime model is used in the rest of the

device to account for the various silicon impurity concentrations. The

hydrodynamic model which provides realistic results and better convergence was

implemented in the Gunn diode model to solve energy balance transport

equations. Finally, it is shown that where the platform did not allow sufficient

flexibility to accurately account for the interdependency of parameters,

appropriate C-language functions were written and incorporated through the in-

built C-interpreter

GIGA

, a self heating simulator available in SILVACO

, was used to simulate

the device local temperature, self-heating, lattice heat flow and heatsinking

effects. It was used with the definition of heat sinks at the cathode and anode

respectively. This is based on the models suggested by Wachutka and accounts

for Joule heating along with heating and cooling due to carrier generation and

recombination. The Wachutka models were discussed in detail.

A precise and accurate physical model was developed for determining device

performance of an advanced graded gap GaAs-AlGaAs Gunn diode. The initial

goal of the work was to reproduce experimental performances of a Gunn diode

commercially manufactured by e2v Technologies Plc. used in second harmonic

in 77 GHz ACC systems.

165

Chapter 10 Conclusion and future work

The simulations have confirmed basic device theory and measured data in that

the current asymmetry (I

asym

), can be used to evaluate injector performance, and

in particular, doping spike carrier concentrations. The doping spike carrier

concentration optimum value

318

101

−

× cm

was determined using modelled-

measured approach.

The model DC characteristics comparison with measured data was presented

over a range of temperatures. The measured data included the DC

characteristics of 77 GHz (1.65 µm transit region device) devices with a range of

epitaxial structures, 125 GHz second harmonic device, ~100 GHz fundamental

device and ~200 GHz fundamental device. A good match between measured

and simulated I-V curves validates the choice of physical models and parameters

used.

The results of the time-domain simulations for 77 GHz (1.65 µm transit region

device) devices with a range of epitaxial structures, 125 GHz second harmonic

device, ~100 GHz fundamental device and ~200 GHz fundamental device were

presented. These simulations established the limits of transit region length

capable of supporting a steady-state (non-decaying) oscillation. These

simulations allowed performing large-signal device characterizations and

comparing to measured results. It was shown that that the shortest transit region

length capable of supporting sustained oscillation in GaAs device is

approximately 0.7 µm (~100 GHz fundamental device).

The modelling work was extended to produce time-domain response of sub-

micrometer transit length device (0.7 µm transit region length) mounted in a

resonant cavity. The resonant cavity designed in SILVACO

was presented

which comprised of lumped-element LC parallel circuit, resistive loss R in series

with the L. The time-domain response allowed analyzing dipole domain formation

and transit region scaling effects on it.

10.2 Directions for Future Research

It was shown that the oscillator circuit effectively determines the oscillation

frequency and so in turn affects both the dynamics of the high-field domain within

166

Chapter 10 Conclusion and future work

the device’s transit region and therefore the dynamic reactance of the device

itself. It may not be accurate to simply extract a small-signal s-parameter model

of the Gunn device from the SILVACO

model and use this in conjunction with

the HFSS

model of the oscillator (i.e. terminate the diode port in the HFSS

simulation with a SILVACO

generated s-parameter matrix). This would not

enable the effects of the oscillator circuit to be taken into account when

computing a device’s time-domain response.

Two novel approaches are proposed to merge SILVACO

and HFSS

models.

It would aid in the development high power THz sources using multipliers. The

techniques include the lumped element model for SILVACO

and taking into

account Gunn diode domain growth and propagation in the HFSS

model.

These are discussed next.

10.2.1 Lumped Element Model for SILVACO

The equivalent frequency-domain behaviour of a second harmonic oscillator

circuit is simulated using HFSS

at the fundamental oscillation frequency and

the equivalent circuit parameters at this frequency extracted: this can then be

used to define a suitable circuit model for inclusion in the SILVACO

simulations. From simulation of the same HFSS

model at second harmonic

frequencies, the values of the equivalent circuit elements at this frequency can

be extracted and used in a lumped element equivalent circuit model to estimate

the conversion efficiency to second harmonic. This will allow estimation of the

second harmonic output power for a device mounted in the oscillator. This will

ultimately enable the effects of both variations in the semiconductor device and

the oscillator geometry to be evaluated simultaneously.

10.2.2 Taking into account Gunn Diode Domain Growth and Propagation

In order to account for the dynamic nature of the depletion region when

simulating the electromagnetic (EM) behaviour of the oscillator circuit (leading to

degradation in accuracy of the model as a whole) the growth of the high-field

domain will need to be adequately represented within the EM model. A proposed

167

Chapter 10 Conclusion and future work

technique to achieve this is to approximate the continuous process of domain

growth and propagation in the HFSS

model by treating it as a sequence of

discrete stages, each with an individual HFSS

model in which the length

and/or the position of the dielectric region is altered. The size, position and

material properties of the depletion region would again be estimated from an

initial SILVACO

result. A time-varying series of equivalent circuit parameters

could then be extracted, and using curve fitting, continuous functions

approximated for each of the oscillator circuit spice model parameters. Although

initially seeming labour intensive, this process could be simplified through

exploitation of the parametric sweep functionality of HFSS

. Suitable code

would then be written to update the values of the equivalent spice circuit model

used to represent the oscillator circuit in the SILVACO

time-domain

simulations, as the oscillation cycle progresses.

Appendix – A

Gunn diode models and material parameters

SILVACO

code for a 0.7µm transit region length Gunn Diode model is presented. The

statements with hash (#) are the remarks which are ignored during simulation code execution. A

line continuation operator ‘\’ shows that code continues in the next line. The C-Interpreter file

path is arbitrary.

go atlas

title "Gunn Diode model VMBE 1898 – 0.7µm transit region length "

set setpath=C:/Silvaco_PhD/00-Simulations/C-Interpreter

# SECTION 1: Mesh input

mesh auto cylindrical

x.mesh loc=0.0 spac=9.75

x.mesh loc=9.75 spac=9.75

x.mesh loc=19.5 spac=5.0

x.mesh loc=28.2 spac=5.0

x.mesh loc=37.0 spac=10.0

x.mesh loc=46.1 spac=10.0

y.mesh loc=0.0 spac=5.0

y.mesh loc=10.00 spac=1.0

y.mesh loc=10.250 spac=0.125

y.mesh loc=10.750 spac=0.05

y.mesh loc=10.7550 spac=0.001

y.mesh loc=10.7650 spac=0.005

y.mesh loc=10.7750 spac=0.005

y.mesh loc=10.7850 spac=0.005

y.mesh loc=10.7950 spac=0.005

y.mesh loc=10.8050 spac=0.005

y.mesh loc=10.8150 spac=0.005

y.mesh loc=10.82000 spac=0.005

y.mesh loc=11.52000 spac=0.01

y.mesh loc=12.02000 spac=0.1

y.mesh loc=20.02000 spac=0.5

y.mesh loc=20.27000 spac=0.125

168

169

Appendix - A Gunn diode models and material parameters

y.mesh loc=22.27000 spac=0.25

# SECTION 2: Structure Specification (regions, electrodes and delta doping)

# Regions

region num=1 name=air material=air x.min=0.0 x.max=19.5 y.min=10.000 y.max=22.2700

region num=2 name=contact_c material=GaAs x.min=19.5 x.max=28.2 y.min=10.2500 \

y.max=10.7500

region num=3 name=spacer material=GaAs x.min=19.5 x.max=28.2 y.min=10.7500 \

y.max=10.7550

region num=4 name=AlGaAs_1.7 material=AlGaAs x.comp=0.017 x.min=19.5 x.max=28.2 \

y.min=10.75500 y.max=10.76500

region num=5 name=AlGaAs_8 material=AlGaAs x.comp=0.08 x.min=19.5 x.max=28.2 \

y.min=10.76500 y.max=10.77500

region num=6 name=AlGaAs_16 material=AlGaAs x.comp=0.16 x.min=19.5 x.max=28.2 \

y.min=10.77500 y.max=10.78500

region num=7 name=AlGaAs_24 material=AlGaAs x.comp=0.24 x.min=19.5 x.max=28.2 \

y.min=10.78500 y.max=10.79500

region num=8 name=AlGaAs_32 material=AlGaAs x.comp=0.32 x.min=19.5 x.max=28.2 \

y.min=10.79500 y.max=10.80500

region num=9 name=spacer material=GaAs x.min=19.5 x.max=28.2 y.min=10.80500 \

y.max=10.81500

region num=10 name=spike material=GaAs x.min=19.5 x.max=28.2 y.min=10.81500 \

y.max=10.82000

region num=11 name=transit material=GaAs x.min=19.5 x.max=28.2 y.min=10.82000 \

y.max=11.52000

region num=12 name=contact_amaterial=GaAs x.min=19.5 x.max=28.2 y.min=11.52000 \

y.max=12.02000

region num=13 name=substrate material=GaAs x.min=19.5 x.max=28.2 y.min=12.02000 \

y.max=20.02000

region num=14 name=air material=air x.min=28.2 x.max=46.1 y.min=10.000 y.max=22.2700

# Electrodes

elec num=1 name=anode material=Gold x.min=19.5 x.max=28.2 y.min=20.02 y.max=20.27

elec num=2 name=cathode material=Gold x.min=19.5 x.max=28.2 y.min=10.0 y.max=10.25

elec num=3 name=anode material=Gold x.min=19.5 x.max=28.2 y.min=20.27 y.max=22.27

elec num=4 name=cathode material=Gold x.min=0.0 x.max=46.1 y.min= 0.0 y.max=10.00

# Delta doping

doping uniform n.type conc=5e18 Region=2

doping uniform n.type conc=1e18 Region=10

doping uniform n.type conc=3.5e16 Region=11

170

Appendix - A Gunn diode models and material parameters

doping uniform n.type conc=5e18 Region=12

doping uniform n.type conc=2e18 Region=13

# SECTION 3: Material Parameters Definitions

# Regional and Band Parameters (band gap)

material name=AlGaAs_1.7 eg300=1.445199 permittivity=13.1507 nc300=6.83e17 \

nv300=1.30e19

material name=AlGaAs_8 eg300=1.52376 permittivity=12.968 nc300=1.60e18 nv300=1.33e19

material name=AlGaAs_16 eg300=1.62352 permittivity=12.736 nc300=2.77e18 nv300=1.37e19

material name=AlGaAs_24 eg300=1.72328 permittivity=12.504 nc300=3.93e18 nv300=1.41e19

material name=AlGaAs_32 eg300=1.82304 permittivity=12.272 nc300=5.10e18 nv300=1.46e19

# Recombination Parameters (SRH)

material name=AlGaAs_1.7 taun0=4.93e-9 taup0=2e-8 NSRHN=4.92e16 NSRHP=4.92e16

material name=AlGaAs_8 taun0=4.68e-9 taup0=2e-8 NSRHN=4.60e16 NSRHP=4.60e16

material name=AlGaAs_16 taun0=4.36e-9 taup0=2e-8 NSRHN=4.20e16 NSRHP=4.20e16

material name=AlGaAs_24 taun0=4.04e-9 taup0=2e-8 NSRHN=3.80e16 NSRHP=3.80e16

material name=AlGaAs_32 taun0=3.72e-9 taup0=2e-8 NSRHN=3.40e16 NSRHP=3.40e16

# Mobility Parameters (caughey Thomas)

mobility name=substrate mu1n.caug=800 alphan.caug=-0.9 betan.caug=-2.2 \

gamman.caug=-3.1 deltan.caug=0.5 ncritn.caug=1e17 mu2n.caug=2300

mobility name=Contact_A mu1n.caug=800 alphan.caug=-0.9 betan.caug=-2.2 \

gamman.caug=3.1 deltan.caug=0.5 ncritn.caug=1e17 mu2n.caug=1350

mobility name=Transit mu1n.caug=800 alphan.caug=-0.9 betan.caug=-2.2 \

gamman.caug=-3.1 deltan.caug=0.5 ncritn.caug=1e17 mu2n.caug=6300

mobility name=Spike mu1n.caug=800 alphan.caug=-0.9 betan.caug=-2.2 \

gamman.caug=-3.1 deltan.caug=0.5 ncritn.caug=1e17 mu2n.caug=2650

mobility name=Contact_C mu1n.caug=800 alphan.caug=-0.9 betan.caug=-2.2 \

gamman.caug=3.1 deltan.caug=0.5 ncritn.caug=1e17 mu2n.caug=1350

mobility material=GaAs betan=1

mobility material=AlGaAs gamman=2

#Giga – Self Heating Simulator Model Parameters

## k(T) = (tc.const) / [(T/300)]^tc.npow

material tcon.power

material f.tcond="$setpath/tcond.lib"

# C Interpreter Functions

material material=AlGaAs f.bandcomp="$setpath/bandcompAl.lib"

171

Appendix - A Gunn diode models and material parameters

material material=GaAs f.bandcomp="$setpath/bandcomp.lib"

material material=AlGaAs f.vsatn="$setpath/vsatn_7e6.lib"

material material=GaAs f.vsatn="$setpath/vsatn_7e6.lib"

mobility material=AlGaAs f.enmun="$setpath/enmun.lib"

material material=GaAs f.munsat="$setpath/munsat_GaAs.lib"

material material=AlGaAs f.munsat="$setpath/munsat.lib"

#Thermal Contacts for GIGA

thermcontact number=1 name=Anode ext.temper=300

thermcontact number=2 name=Cathode ext.temper=300

# SECTION 3: Models Definitions

models material=GaAs consrh conmob analytic fldmob evsatmod=1 b.elec=2

models material=AlGaAs srh fldmob evsatmod=0 mobtem.simpl b.elec=2

models lat.temp joule.heat gr.heat

models material=AlGaAs cubic35

models hcte.el boltzmann temperature=300 print

# SECTION 4: Initial solution

# Method

method carriers=1 electrons

output e.field e.mobility e.temp e.velocity con.band val.band charge flowlines devdeg

method block newton itlimit=100 trap atrap=0.5 maxtrap=10 vsatmod.inc=0.001

# Contacts

contact all neutral

#contact name=anode neutral resistance=0.9 inductance=3.16e-12 capacitance=0.403e-12 \

# resonant

contact name=anode neutral resistance=0.9

contact name=cathode neutral

solve init

solve

save outfile=1898_P_0pt7Transit_298Ka.str

# SECTION 5a: DC I-V Simulations – Forward bias

log outf=1898_P_0pt7Transit_298K.log

solve vanode=0.0 vstep=0.1 vfinal=4.0 name=anode

save outfile=1898_P_0pt7Transit_Final_298K.str

log off

quit

172

Appendix - A Gunn diode models and material parameters

# SECTION 5b: DC I-V Simulations – Reverse bias

log outf=1898_N_0pt7Transit_298K.log

solve vanode=0.0 vstep=-0.1 vfinal=-4.0 name=anode

save outfile=1898_N_0pt7Transit_Final_298K.str

log off

quit

# SECTION 5c: Time Domain Simulations

solve init

save outfile=1898_ATLAST_HS.str

log outfile=1898_Trans_1_HS.log

solve vanode=2.5 ramptime=2.7e-11 tstop=2.7e-11 dt=1e-12

save outfile=1898_Trans_1_HS.str

# Extract I-V Characteristic above which the Negative Resistance begins

extract name="1898_Trans_HS.dat" curve(vint."anode", i."anode") outf="1898_Trans_HS.dat"

extract name="Icrit" max(i."anode")

extract name="UMax" max(vint."anode")

extract name="Ucrit" x.val from curve(vint."anode", i."anode") where y.val=$Icrit

extract name="Ecrit" $Ucrit/7e-5

# Time Domain Simulation at 2.5V Constant External Bias

log outfile=1898_Trans_2_HS.log

solve tstop=12e-11 dt=2e-11 outfile=1898_Trans_2_x0 master

log off

extract the oscillation frequency by measuring the time for 4 periods

extract init inf="1898_Trans_2_HS.log"

extract name="v_ave" ave(vint."anode")

extract name="t1" x.val from curve(time, vint."anode") where y.val=$"v_ave"

extract name="t2" x.val from curve(time, vint."anode") where y.val=$"v_ave" and val.occno=7

extract name="delta" ($"t2" - $"t1") / 3

extract name="f@2pt5V" 1 / $"delta"

# Ramp-up to 3.0 V External Bias

load inf=1898_Trans_1_HS.str master

solve previous

solve vanode=3 ramptime=2.7e-11 tstop=2.7e-11 dt=1e-12

173

Appendix - A Gunn diode models and material parameters

save outfile=1898_Trans_3_HS.str

# Time Domain Simulation at 3.0V Constant External Bias

log outfile=1898_Trans_3_HS.log

solve tstop=12e-11 dt=2e-11 outfile=1898_Trans_3_x0 master

log off

extract init inf="1898_Trans_3_HS.log"

extract name="v_ave" ave(vint."anode")

extract name="t1" x.val from curve(time,vint."anode") where y.val=$"v_ave"

extract name="t2" x.val from curve(time,vint."anode") where y.val=$"v_ave" and val.occno=7

extract name="delta" ($"t2" - $"t1") / 3

extract name="f@3V" 1 / $"delta"

quit

Appendix – B

Schottky Diode – SILVACO

code

SILVACO

code for a 14µm diameter Schottky diode is presented. The statements with hash

(#) are the remarks which are ignored during simulation code execution. A line continuation

operator ‘\’ shows that code continues in the next line. The C-Interpreter file path is arbitrary..

SILVACO

DeckBuild

code

go atlas

title "Schottky 4 multipliers 14um diameter"

# SECTION 1: Mesh input

mesh width = 120 auto

x.mesh loc=0.0 spac=0.5

x.mesh loc=1.5 spac=0.5

x.mesh loc=2.0 spac=0.25

x.mesh loc=2.5 spac=0.5

x.mesh loc=2.994 spac=0.01

x.mesh loc=2.995 spac=0.01

x.mesh loc=3.0 spac=0.01

x.mesh loc=3.5 spac=0.25

x.mesh loc=5.25 spac=0.15

y.mesh loc=-0.0605 spac=0.1

y.mesh loc=-0.033 spac=0.001

y.mesh loc=0.0 spac=0.001

y.mesh loc=0.35 spac=0.01

y.mesh loc=3.35 spac=0.2

y.mesh loc=6.35 spac=1.2

y.mesh loc=14.35 spac=2.0

# SECTION 2: Structure Specification (regions, electrodes and delta doping)

# Regions

region num=1 name="Cap" material=InGaAs y.min=-0.0605 y.max=-0.033 x.comp=0.2

166

174

175

Appendix - C Schottky Diode – SILVACO code

region num=2 name="Cap1" graded material=InGaAs y.min=-0.033 y.max=0.0 compx.top=0.2 \

compx.bottom=0.9999

region num=3 name="Supply" material=GaAs y.min=0.0 y.max=0.35 x.min=0.0 x.max=3.00

region num=4 name="Supply1" material=GaAs y.min=0.0 y.max=0.35 x.min=3.5 x.max=5.25

region num=5 name="Channel" material=GaAs y.min=0.35 y.max=3.35

region num=6 name="Buffer" material=GaAs y.min=3.35 y.max=6.35

region num=7 name="Subs" material=GaAs y.min=6.35 y.max=14.35

region num=8 material=Air y.min=-0.0605 y.max=0.0 x.min=2.995 x.max=5.25

region num=9 material=Air y.min=0.0 y.max=0.35 x.min=2.995 x.max=3.50

# Electrodes

elec num=2 name=anode material=gold x.min=3.5 x.max=5.25 y.min=0.0 y.max=0.0

elec num=1 name=cathode material=gold x.min=0.0 x.max=2.994 y.min=-0.0605 y.max=-0.0605

# Delta Doping

Doping uniform n.type conc=5e18 region=1

Doping uniform n.type conc=5e18 region=2

Doping uniform n.type conc=1.8e17 region=3

Doping uniform n.type conc=1.8e17 region=4

Doping uniform n.type conc=5e18 region=5

# SECTION 3: Material Parameters Definitions

## The physical parameters for InGaAs are defined in the following sub-sections. Default

## parameters are selected for GaAs

# Regional and Band Parameters (band gap)

material material=InGaAs f.bandcomp="C:/SILVACO_PhD/C-Interpreter/bandcomp_InGaAs.lib"

material region=2 align=0.34

material region=3 align=0.36

# Recombination Parameters (SRH)

material material=InGaAs taun0=3e-10 taup0=1e-8 NSRHN=9e17 NSRHP=8e17

# Saturation Velocities and Auger Recombination Parameters

material material=InGaAs vsatn=7.67e6 vsatp=6.01e6

material material=InGaAs f.vsatn="C:/SILVACO_PhD/C-Interpreter/vsat_InGa.lib"

material material=InGaAs augn=2e-27

176

Appendix - C Schottky Diode – SILVACO code

# Mobility Parameters (caughey Thomas)

mobility material=InGaAs mu1n.caug=2000 mu2n.caug=12000 mu1p.caug=050 mu2p.caug=400\

ncritn.caug=6.4e17 ncritp.caug=7.4e17

mobility material=InGaAs alphan.caug=0 alphap.caug=0 betan.caug=0 betap.caug=0 \

gamman.caug=0 gammap.caug=0 deltan.caug=0 deltap.caug=0

mobility material=InGaAs f.enmun="C:/SILVACO_PhD/C-Interpreter/enmun_In.lib"

material material=InGaAs f.munsat="C:/SILVACO_PhD/C-Interpreter/munsat_In.lib"

# SECTION 3: Models Definitions

models fldmob auger consrh fermidirac temperature=300 print

# SECTION 4: Initial Solution

output e.field e.mobility e.temp e.velocity con.band val.band charge flowlines

# Contacts

contact name=anode workfunc=4.815

contact name=cathode

method block newton itlimit=100 trap atrap=0.5 maxtrap=10 vsatmod.inc=0.001

solve init

save outf=Schottky_Ohmic-Cap_Rev-14um-DIA.str

# SECTION 4: AC solution for plotting a C-V curve

solve vanode=0

log outf=AC_Schottky_14um-DIA.log s.param inport=anode outport=anode y.param

solve vanode=0.0 vstep=-0.05 vfinal=-10.0 name=anode ac freq=1e9 fstep=10e9 nfsteps=9

save outf=AC_Schottky_Ohmic-ve14um-DIA.str

logoff

tonyplot AC_Schottky_14um-DIA.log

quit

Appendix – C

Dipole Domain Formation

1. One dipole formation showing the transit region electron concentration,

zoomed between 16.4 and 16.9 cm

-3

2. Two dipoles showing the transit region electron concentration, zoomed

between 16.4 and 16.9 cm

-3

Electron Concentration cm

177

178

Appendix - C Dipole Domain Formation

Electron

Concentration cm

References

1. Sze, S.M., High Speed Devices. 1990: New York ; Chichester : Wiley.

2. Hobson, G.S., The Gunn effect. 1974. Oxford : Clarendon Press.

3. Sze, S.M., Semiconductor Devices Physics and Technology. 2nd edition

2002, New York: NY Willey.

4. Montanari, S., Fabrication and Characterization of planar Gunn Diodes for

Monolithic Microwave Integrated Circuits, in RWTH. 2005, University of

Aachen.

5. Townsend, S.M., AlGaAs/GaAs Graded Gap Gunn Diodes for

Autonomous Intelligent Cruise Control Radar Applications, in EEE, M&N

Group. 1999, UMIST: Manchester.

6. Gallerano, G.P. Overview of Terahertz Radiation Sources. in Proceedings

of the 2004 FEL Conference. 2004.

7. Electronic Warfare and Radar Systems Engineering Handbook. 2 ed.

1997, Washington, D.C. 20361: Avionics Department. 1-298.

8. Shur, M. Terahertz technology: Devices and applications. in Proceedings

of ESSDERC 2005: 35th European Solid-State Device Research

Conference. 2005.

9. Tait, G.B. and Krowne, C.M., Efficient transferred electron device

simulation method for microwave and millimeter wave cad applications.

Solid-State Electronics, 1987. 30(10): p. 1025-1036.

10. Forster, A.A., Hot electron injector Gunn diode for advanced driver

assistance systems. Applied physics. A, Materials science & processing,

2007. 87(3): p. 545-58.

11. Montanari, S.S., High frequency investigation of graded gap injectors for

GaAs Gunn diodes. Solid-state electronics, 2005. 49(2): p. 245-50.

12. Couch, N.N.R., High-performance, graded AlGaAs injector, GaAs Gunn

diodes at 94 GHz. IEEE electron device letters, 1989. 10(7): p. 288-90.

13. N. Farrington, Carr, M., Sly, J. L., and Missous, M., Development of an on-

wafer test for rapid evaluation of doping spike carrier concentration levels

in commercially manufactured GaAs Gunn diodes for automotive radar

applications. in Interational Conferance on Compound Semiconductor

Manufacturing Technology (CSManTech’08). 2008. Chicago, USA.

14. Phillips, T.T.G., Submillimeter astronomy [heterodyne spectroscopy].

Proceedings of the IEEE, 1992. 80(11): p. 1662-78.

15. Tonouchi, M., Cutting-edge terahertz technology, in Nat Photon. 2007. p.

97-105.

16. Siegel, P.H., Terahertz technology. IEEE transactions on microwave

theory and techniques, 2002. 50(3): p. 910-28.

17. Siegel, P.H. THz applications for outer and inner space. in 17th

International Zurich Symposium on Electromagnetic Compatibility. 2006.

18. Carr, G., et al., High-power terahertz radiation from relativistic electrons.

Nature, 2002. 420(6912): p. 153-156.

19. Anwar, A. Staying the course in the GaAs industry. in International

Conference on Compound Semiconductor Manufacturing Technology.

2005.

179

180

References

20. Gunn, J., Microwave Oscillations of Current in III-V Semiconductors. Solid

State Communications, 1963. 1(4): p. 88-91.

21. Esaki, L., New Phenomenon in Narrow Germanium p-n Junctions.

Physical Review, 1958. 109(2): p. 603.

22. Chang, L., Esaki L., and Tsu, R., Resonant Tunneling in Semiconductor

Double Barriers. Applied physics letters, 1974. 24(12): p. 593-595.

23. Read, W., A proposed high-frequency, negative-resistance diode. The Bell

system technical journal, 1958. 37(2): p. 401-446.

24. Ruegg, H.W., A proposed punch-through microwave negative-resistance

diode. Electron Devices, IEEE Transactions on, 1968. 15(8): p. 577-585.

25. Okabe, T., et al., Bulk oscillation by tunnel injection. Electron Devices,

IEEE Transactions on, 1969. 16(2): p. 251-251.

26. Elta, M.E. and Haddad, G.I., High-Frequency Limitations of IMPATT,

MITATT, and TUNNETT Mode Devices. Microwave Theory and

Techniques, IEEE Transactions on, 1979. 27(5): p. 442-449.

27. Elta, M.E. and Haddad, G.I., Large-signal performance of microwave

transit-time devices in mixed tunneling and avalanche breakdown.

Electron Devices, IEEE Transactions on, 1979. 26(6): p. 941-948.

28. Gunn, J.J.B., Microwave oscillations of current in III-V semiconductors.

Solid State Communications, 1993. 88(11): p. 883-886.

29. Streetman, B.G., Solid state electronic devices. 4th ed. Prentice Hall

series in solid state physical electronics. 1995: Englewood Cliffs, N.J. :

Prentice Hall, c1995.

30. Townsend, S. Low resistance ohmic contacts to millimetre-wave graded

gap Gunn diode oscillators. in Workshop on High Performance Electron

Devices for Microwave and Optoelectronic Applications, EDMO. 1997.

31. e2v Technologies Plc, DC1279F-T 76–78GHz Gunn Diode Data sheet.

32. Brown, E. R., Soderstrom, J. R., Parker, C. D., Mahoney, L. J., Molvar, K.

M. and McGill, T. C., Oscillations up to 712 GHz in InAs/AlSb resonant

tunneling diodes. Applied Physics Letters, 1991. 58(20): p. 2291-2293.

33. Suzuki, S., Teranishi, A., Hinata, K., Asada, M., Sugiyama, H. and

Yokoyama, H., Fundamental oscillation up TO 831 GHz in GaInAs/AlAs

resonant tunneling diode. in IEEE International Conference on Indium

Phosphide & Related Materials, 2009. IPRM '09. . 2009.

34. Sze, S.M. and Kwok K. Ng, Physics of semiconductor devices. 3rd ed.

2006, New York: Wiley-Blackwell.

35. Jun-ichi, N, Plotka, P., Makabe, H. and Kurabayashi, T., GaAs TUNNETT

diodes oscillating at 430-655 GHz in CW fundamental mode. Microwave

and Wireless Components Letters, IEEE, 2005. 15(9): p. 597-599.

36. Elta, M., Fetterman, H., Macropoulos, W. and Lambert, J., 150 GHZ

GAAS MITATT SOURCE. Electron device letters, 1980. 1(6): p. 115-116.

37. Robert LaGuerra, D.F., Ed Rerisi, ABI (Automotive Radar/Lidar Systems)

A component level market analysis of radar, lidar, ultrasonic, and optics-

based automotive safety systems 2004.

38. Pilgrim, N., Macpherson, R., Khalid, A., Dunn, G. and Cumming, D.,

Multiple and broad frequency response Gunn diodes. Semiconductor

science and technology, 2009. 24(10): p. 105010.

39. Pilgrim, N. J., Khalid, A., Li, C., Dunn, G. M. and Cumming, D. R. S.,

Contact shaping in planar Gunn diodes. physica status solidi (c), 2010.

8(2): p. 313-315.

181

References

40. Perez, S., Gonzalez, T., Pardo, D. and Mateos, Terahertz gunn-like

oscillations in InGaAs/InAlAs planar diodes. Journal of applied physics,

2008. 103(9): p. 094516.

41. Huang, J., Yang, H., Tian, C., Dong, J., Zhang, H. and Guo, T., Design

and manufacture of planar GaAs Gunn diode for millimeter wave

application. Chinese Physics B, 2010. 19(12): p. 127203.

42. Li, C., Khalid, A., Pilgrim, N., Holland, M., Dunn, G. and Cumming, D.,

Novel Planar Gunn Diode Operating in Fundamental Mode up to 158

GHz. Journal of physics. Conference series, 2009. 193: p. 012029.

43. Crowe, T.T.W., Terahertz GaAs devices and circuits for heterodyne

receiver applications. International journal of high speed electronics, 1995.

6(1): p. 125-61.

44. Zhong, H. Inspection of Space Shuttle Insulation Foam Defects Using a

0.2 THz Gunn Diode Oscillator. in IEEE conference Digest of the 2004

Joint 29th International Conference on Infrared and Millimeter Waves and

12th International Conference on Terahertz Electronics. 2004.

45. Federici, J.J.F., THz imaging and sensing for security applications-

explosives, weapons and drugs. Semiconductor Science and Technology,

2005. 20(7): p. S266-80.

46. Leahy-Hoppa, M.M.R., Wideband terahertz spectroscopy of explosives.

Chemical Physics Letters, 2007. 434(4): p. 227-30.

47. Frost and Sullivan, The Market for Obstacle Sensing Technologies in

Passenger Cars and Commercial Vehicles in Europe. 2005.

48. Khlopov, G. and Schuenemann, K., Application of Millimetre Wave

Coherent Radar for Remote Measurement of Vibrations and

Displacements. in Conference Digest of the 2004 Joint 29th International

Conference on Infrared and Millimeter Waves and 12th International

Conference on Terahertz Electronics 2004: IEEE.

49. Barth, H. and Menzel, W., GaAs Gunn Oscillators Reach the 140 GHz

Range. in Microwave Symposium Digest, 1985 IEEE MTT-S International.

1985.

50. Neylon, S.S., State-of-the-art performance millimetre wave gallium

arsenide Gunn diodes using ballistically hot electron injectors. IEEE MTT-

S International Microwave Symposium Digest, 1989. 1: p. 519-522.

51. Beton, P.P.H., Use of n<sup>+</sup> spike doping regions as

nonequilibrium connectors. Electronics letters, 1988. 24(7): p. 434-435.

52. Kiuru, T., et al. Gunn oscillator modeling and second harmonic output

power optimization at 76 GHz. in Microwave Symposium Digest, 2008

IEEE MTT-S International. 2008.

53. Carlstrom, J.E., Plambeck, R.L., and D.D. Thornton, A Continuously

Tunable 65--15-GHz Gunn Oscillator. Microwave Theory and Techniques,

IEEE Transactions on, 1985. 33(7): p. 610-619.

54. Jones, S. H., Zybura, M. F., Carlstrom, J. E. and O'Brien, T. M., A 63-170

GHz second-harmonic operation of an InP transferred electron device.

Electron Devices, IEEE Transactions on, 1999. 46(1): p. 17-23.

55. Van Zyl, R., Perold, W., and Botha, R.. The Gunn-diode: fundamentals

and fabrication. in Communications and Signal Processing, 1998.

COMSIG '98. Proceedings of the 1998 South African Symposium on.

1998.

56. Bullman, J., Hobson, G.S, and Taylor, B.C., The Gunn-Hilsum Effect.

1979, New York: Academic Press Inc.

182

References

57. Wang, B., AbouElnour, A., and Schunemann, K., Determination of the

influence of doping profile on the performance of mm-wave gunn elements

by using an efficient physical simulator. International journal of infrared

and millimeter waves, 1996. 17(7): p. 1205-1218.

58. Butcher, P.P.N., Theory of stable domain propagation in the Gunn effect.

Physics letters, 1965. 19(7): p. 546-547.

59. Dunn, G. and Kearney M., A theoretical study of differing active region

doping profiles for W-band (75-110 GHz) InP Gunn diodes.

Semiconductor science and technology, 2003. 18(8): p. 794-802.

60. Teoh, Y., Dunn, G., Priestley, N. and Carr, M., Monte Carlo simulations of

asymmetry multiple transit region Gunn diodes. Semiconductor science

and technology, 2005. 20(5): p. 418-422.

61. Kamoua, R. and Eisele H., Theoretical and experimental comparison of

optimized doping profiles for high-performance InP Gunn devices at 220-

500 GHz, in 2003 Ieee Mtt-S International Microwave Symposium Digest,

Vols 1-3, H. Thal, Editor. 2003, Ieee: New York. p. 907-910.

62. Pilgrim, N., Khalid, A., Dunn, G. and Cumming, D., Gunn oscillations in

planar heterostructure diodes. Semiconductor science and technology,

2008. 23(7): p. 075013.

63. Bosch, B.G. and Engel, R.W.H., Gunn-effect electronics. 1975: London

(etc.) : Pitman.

64. Bulman, P.J., Transferred electron devices 1972: New York ; London :

Academic Press.

65. Navon, D.H., Semiconductor Microwave and Materials. 1987: HRW Int.

Edition.

66. Li, S.S., Semiconductor physical electronics. Microdevices. 1993 New

York : Plenum.

67. Couch, N.N.R., Hot electron injection in millimetre wave Gunn diodes.

Solid-state electronics, 1989. 32(12): p. 1685-8.

68. Ruch, J.J.G., Temperature dependence of the transport properties of

gallium arsenide determined by a Monte Carlo method. Journal of applied

physics, 1970. 41(9): p. 3843-9.

69. Couch, N.N.R., The use of linearly graded composition AlGaAs injectors

for intervalley transfer in GaAs: theory and experiment. Solid-state

electronics, 1988. 31(3): p. 613-16.

70. Lee, C.C., Monte Carlo comparison of heterojunction cathode Gunn

oscillators. Electronics letters, 1990. 26(7): p. 425-7.

71. Copeland, J.J.A., A new mode of operation for bulk negative resistance

oscillators. Proceedings of the IEEE, 1966. 54(10): p. 1479-1480.

72. Copeland, J.J.A., Characterization of bulk negative-resistance diode

behavior. IEEE transactions on electron devices, 1967. 14(9): p. 461-463.

73. Copeland, J., LSA Oscillator-Diode Theory. Journal of applied physics,

1967. 38(8): p. 3096.

74. Harrison, W.W.A., Universal linear-combination-of-atomic-orbitals

parameters for d-state solids. Physical review. B, Condensed matter,

1980. 21(8): p. 3214-21.

75. Greenwald, Z.Z., The effect of a high energy injection on the performance

of mm wave Gunn oscillators. Solid-state electronics, 1988. 31(7): p.

1211-14.

76. Missous, M., Conduction and Valence Band Offsets at the AlGaAs/GaAs

heterostructure interface. Properties of Aliminium Galium Aesenide, ed. S.

Adachi. 1991: IEE INSPEC. 325.

183

References

77. Missous, M., In situ, near-ideal epitaxial Al

1-x

As Schottky barriers

formed by molecular beam epitaxy. Journal of applied physics, 1990.

68(5): p. 2239-45.

78. Vurgaftman, I.I., Band parameters for III-V compound semiconductors and

their alloys. Journal of applied physics, 2001. 89(11): p. 5815-75.

79. Stock, J., Herstellung und Charakterisierung von GaAs Gunn-Dioden für

Anwendungen bei 77GHz. 2003, University of Aachen RWTH Aachen.

80. VWF Interactive Tools, Device Simulation Software. Vol. CA 95054. 2005,

Santa Clara: SILVACO International.

81. Atlas User’s Manual, Device Simulation Software. Vol. Version 5.10.R,.

2005, Santa Clara: SILVACO International.

82. DevEdit, User’s Manual. 2006, Santa Clara, CA 95054: SILVACO

International.

83. DeckBuild, User’s Manual. Vol.

Version 3.26.R

2005

Santa Clara, CA 95054

SILVACO International

84. TonyPlot, User’s Manual. Vol.

Version 3.4.0.R

2006

Santa Clara, CA 95054

SILVACO International

85. Palankovski, V. and Quay, R., Analysis and simulation of heterostructure

devices. 2004, Wien, Austria: Springer.

86. Palankovski, V., Schultheis, R., and Selberherr, S., Simulation of power

heterojunction bipolar transistors on gallium arsenide. Electron Devices,

IEEE Transactions on, 2001. 48(6): p. 1264-1269.

87. Caughey, D.M., Carrier mobilities in silicon empirically related to doping

and field. Proceedings of the IEEE, 1967. 55(12): p. 2192-2193.

88. Barnes, J.J.J., Finite-element simulation of GaAs MESFET's with laterall

doping profiles and submicron gates. IEEE transactions on electron

devices, 1976. 23(9): p. 1042-1048.

89. Shockley, W.W., Statistics of the recombinations of holes and electrons.

Physical review, 1952. 87: p. 835-842.

90. Hall, R.R.N., Electron-hole recombination in germanium. Physical review,

1952. 87: p. 387-387.

91. Wachutka, G.G.K., Rigorous thermodynamic treatment of heat generation

and conduction in semiconductor device modeling. IEEE transactions on

computer-aided design of integrated circuits and systems, 1990. 9(11): p.

1141-9.

92. Farahmand, M., Monte Carlo simulation of electron transport in the III-

nitride wurtzite phase materials system: Binaries and ternaries. IEEE

transactions on electron devices, 2001. 48(3): p. 535-542.

93. Stratton, R., Semiconductor current-flow equations (diffusion and

degeneracy). Electron Devices, IEEE Transactions on, 1972. 19(12): p.

1288-1292.

94. Stratton, R., Diffusion of Hot and Cold Electrons in Semiconductor

Barriers. Physical Review, 1962. 126(6): p. 2002.

95. Meinerzhagen, B., Bach, K. H., Bork, I. and Eng, W. L., A New Highly

Efficient Nonlinear Relaxation Scheme for Hydrodynamic MOS

Simulations. in Numerical Modeling of Processes and Devices for

Integrated Circuits, 1992. NUPAD IV. Workshop on. 1992.

96. Apanovich, Y., et al., Numerical simulation of submicrometer devices

including coupled nonlocal transport and nonisothermal effects. Electron

Devices, IEEE Transactions on, 1995. 42(5): p. 890-898.

97. Apanovich, Y., Blakey, P., Cottle, R., Lyumkis, E., Polsky, B., Shur, A. and

Tcherniaev, A., Steady-state and transient analysis of submicron devices

184

References

using energy balance and simplified hydrodynamic models. Computer-

Aided Design of Integrated Circuits and Systems, IEEE Transactions on,

1994. 13(6): p. 702-711.

98. Manual, S.I.A.U.s., Device Simulation Software. Vol. Software version

5.16.3.R. 2010, Santa Clara, CA 95054: SILVACO International.

99. Amir, F., Farrington, N., Tauqeer, T. and Missous, M., Physical Modelling

of a Step-Graded AlGaAs/GaAs Gunn Diode and Investigation of Hot

Electron Injector Performance. in Advanced Semiconductor Devices and

Microsystems, 2008. ASDAM 2008. International Conference on. 2008.

100. Amir, F., Farrington, N., Tauqeer, T. and Missous, M., Development of a

novel 3D physical model of an advanced mm-wave Gunn diode with hot-

electron injection, in 17th European Heterostructure Technology

Workshop, HETECH 2008 2008: Venice, Italy.

101. Amir, F., Mitchell, C., Farrington, N. and Missous, M., Advanced Gunn

Diode as High Power Terahertz Source for a Millimetre Wave High Power

Multiplier. in SPIE Europe Security + Defence 2009. 2009. Berliner

Congress Centre, Berlin, Germany, Berlin Germany: SPIE.

102. Amir, F., Farrington, N., Mitchell, C. and Missous, M., Time-domain

analysis of sub-micron transit region GaAs Gunn diodes for use in

terahertz frequency multiplication chains. in SPIE Europe Security +

Defence 2010, SD108 ‘Millimetre Wave and Terahertz Sensors and

Technology. 2010. Centre de Congrès Pierre Baudis, Toulouse, France:

SPIE.

103. Farrington, N. Norton, P., Carr, M., Sly, J. and Missous, M., A ruggedly

packaged D-Band GaAs Gunn diode with hot electron injection suitable

for volume manufacture. IEEE MTT-S International Microwave

Symposium digest, 2008. 1-4: p. 282-285.

104. Amir, F., Mitchell, C. and Missous, M., Development of advanced Gunn

diodes and Schottky multipliers for high power THz sources, in 8th

International Conference on Advanced Semiconductor Devices &

Microsystems (ASDAM 2010), 2010.