Semiconductor Module

User’s Guide
Semiconductor Module User’s Guide
© 1998–2019 COMSOL
Protected by patents listed on www.comsol.com/patents, and U.S. Patents 7,519,518; 7,596,474;
7,623,991; 8,457,932; 8,954,302; 9,098,106; 9,146,652; 9,323,503; 9,372,673; 9,454,625; and
10,019,544. Patents pending.
This Documentation and the Programs described herein are furnished under the COMSOL Software License
Agreement (www.comsol.com/comsol-license-agreement) and may be used or copied only under the terms
of the license agreement.
COMSOL, the COMSOL logo, COMSOL Multiphysics, COMSOL Desktop, COMSOL Compiler,
COMSOL Server, and LiveLink are either registered trademarks or trademarks of COMSOL AB. All other
trademarks are the property of their respective owners, and COMSOL AB and its subsidiaries and products
are not affiliated with, endorsed by, sponsored by, or supported by those trademark owners. For a list of such
trademark owners, see www.comsol.com/trademarks.
Version: COMSOL 5.5

Contact Information
Visit the Contact COMSOL page at www.comsol.com/contact to submit general
inquiries, contact Technical Support, or search for an address and phone number. You can
also visit the Worldwide Sales Offices page at www.comsol.com/contact/offices for
address and contact information.

If you need to contact Support, an online request form is located at the COMSOL Access
page at www.comsol.com/support/case. Other useful links include:

• Support Center: www.comsol.com/support

• Product Download: www.comsol.com/product-download
• Product Updates: www.comsol.com/support/updates
• COMSOL Blog: www.comsol.com/blogs
• Discussion Forum: www.comsol.com/community
• Events: www.comsol.com/events
• COMSOL Video Gallery: www.comsol.com/video
• Support Knowledge Base: www.comsol.com/support/knowledgebase

Part number: CM024101

C o n t e n t s

Chapter 1: Introduction

About the Semiconductor Module 12

Modeling Semiconductor Devices . . . . . . . . . . . . . . . . 12
What Can the Semiconductor Module Do? . . . . . . . . . . . . . 13
The Semiconductor Module Physics Interface Guide . . . . . . . . . 13
Common Physics Interface and Feature Settings and Nodes . . . . . . 15
The Semiconductor Module Study Capabilities by Physics Interface . . . . 16
The Semiconductor Materials Database . . . . . . . . . . . . . . 17
Where Do I Access the Documentation and Application Libraries? . . . . 18

Overview of the User’s Guide 22

Chapter 2: Modeling Guidelines

Physics for Semiconductor Modeling 24

Zero Reference of Electric Potential . . . . . . . . . . . . . . . 25
Units of Band Gap and Electron Affinity . . . . . . . . . . . . . . 25

Discretization and Formulation Options 27

Defining the Carrier Mobility 31

Doping 33
Using the Analytic Doping Model. . . . . . . . . . . . . . . . . 33
Using the Geometric Doping Model . . . . . . . . . . . . . . . 35
Visualizing the Dopant Distribution . . . . . . . . . . . . . . . . 38
Example of Multiple Superimposed Doping Profiles . . . . . . . . . . 38

Connecting to Electrical Circuits 42

About Connecting Electrical Circuits to Physics Interfaces . . . . . . . 42
Connecting Electrical Circuits Using Predefined Couplings . . . . . . . 43

CONTENTS |3
 Connecting Electrical Circuits by User-Defined Couplings . . . . . . . 43
Solving . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Postprocessing. . . . . . . . . . . . . . . . . . . . . . . . 44

Nonisothermal Models 46

SPICE Import and Export 48

SPICE Import . . . . . . . . . . . . . . . . . . . . . . . . 48
SPICE Export . . . . . . . . . . . . . . . . . . . . . . . . 49
Reference . . . . . . . . . . . . . . . . . . . . . . . . . 49

Small-Signal Analysis 50

Meshing 52

Solving 54

Quantum-Confined Systems 58

Chapter 3: Semiconductor Branch Interface

The Semiconductor Interface 60

Domain, Edge, Boundary, Pair, and Point Nodes for the
Semiconductor Interface . . . . . . . . . . . . . . . . . . 63
Semiconductor Material Model . . . . . . . . . . . . . . . . . 65
Optical Transitions . . . . . . . . . . . . . . . . . . . . . . 70
Indirect Optical Transitions . . . . . . . . . . . . . . . . . . . 74
Insulation . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Continuity/Heterojunction . . . . . . . . . . . . . . . . . . . 76
WKB Tunneling Model, Electrons . . . . . . . . . . . . . . . . 77
WKB Tunneling Model, Holes . . . . . . . . . . . . . . . . . . 79
Initial Values . . . . . . . . . . . . . . . . . . . . . . . . 79
User-Defined Mobility Model . . . . . . . . . . . . . . . . . . 79
Power Law Mobility Model (L). . . . . . . . . . . . . . . . . . 80
Fletcher Mobility Model (C). . . . . . . . . . . . . . . . . . . 81
Caughey-Thomas Mobility Model (E) . . . . . . . . . . . . . . . 83

4 | CONTENTS
Lombardi Surface Mobility Model (S) . . . . . . . . . . . . . . . 85
Arora Mobility Model (LI) . . . . . . . . . . . . . . . . . . . 87
Analytic Doping Model . . . . . . . . . . . . . . . . . . . . 88
Geometric Doping Model . . . . . . . . . . . . . . . . . . . 92
Boundary Selection for Doping Profile . . . . . . . . . . . . . . . 93
Analytic Trap Density . . . . . . . . . . . . . . . . . . . . . 93
Geometric Trap Density . . . . . . . . . . . . . . . . . . . . 95
Boundary Selection for Trap Density Profile . . . . . . . . . . . . 96
Trap-Assisted Recombination . . . . . . . . . . . . . . . . . . 96
Discrete Energy Level (Domains). . . . . . . . . . . . . . . . . 99
Continuous Energy Levels (Domains) . . . . . . . . . . . . . . 100
Auger Recombination . . . . . . . . . . . . . . . . . . . . 102
Direct Recombination . . . . . . . . . . . . . . . . . . . . 103
Impact Ionization Generation . . . . . . . . . . . . . . . . . 104
User-Defined Recombination . . . . . . . . . . . . . . . . . 105
User-Defined Generation . . . . . . . . . . . . . . . . . . 105
Metal Contact . . . . . . . . . . . . . . . . . . . . . . . 106
Thin Insulator Gate . . . . . . . . . . . . . . . . . . . . . 108
Trap-Assisted Surface Recombination . . . . . . . . . . . . . . 110
Discrete Energy Level (Boundaries) . . . . . . . . . . . . . . . 110
Continuous Energy Levels (Boundaries) . . . . . . . . . . . . . 111
Trap-Assisted Heterointerface Recombination . . . . . . . . . . 111
Discrete Energy Level (Heterointerfaces) . . . . . . . . . . . . 111
Continuous Energy Levels (Heterointerfaces) . . . . . . . . . . . 112
Floating Gate . . . . . . . . . . . . . . . . . . . . . . . 112
Insulator Interface . . . . . . . . . . . . . . . . . . . . . 113

Theory for the Semiconductor Interface 114

The Sommerfeld Model and the Density of States . . . . . . . . . 115
The Effect of Periodicity . . . . . . . . . . . . . . . . . . . 118
Electrons in a Perturbed Periodic Potential . . . . . . . . . . . . 133
The Semiclassical Model . . . . . . . . . . . . . . . . . . . 138
The Boltzmann Equation . . . . . . . . . . . . . . . . . . . 142
Doped Semiconductors . . . . . . . . . . . . . . . . . . . 144
Specifying the Dopant Distribution . . . . . . . . . . . . . . . 146
Equilibrium Carrier Concentrations. . . . . . . . . . . . . . . 151
The Semiconductor Equations . . . . . . . . . . . . . . . . . 155
Mobility Models . . . . . . . . . . . . . . . . . . . . . . 168

CONTENTS |5
 Traps . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Recombination and Generation . . . . . . . . . . . . . . . . 179
Incomplete Ionization . . . . . . . . . . . . . . . . . . . . 185
Band Gap Narrowing . . . . . . . . . . . . . . . . . . . . 186
Optical Transitions . . . . . . . . . . . . . . . . . . . . . 188
Boundary Conditions . . . . . . . . . . . . . . . . . . . . 205
Metal Contacts . . . . . . . . . . . . . . . . . . . . . . 206
Thin Insulating Gates . . . . . . . . . . . . . . . . . . . . 213
Continuity/Heterojunction . . . . . . . . . . . . . . . . . . 216
Boundary Conditions for Charge Conservation . . . . . . . . . . 219
Boundary Conditions for the Density-Gradient Formulation . . . . . 220
Tunneling. . . . . . . . . . . . . . . . . . . . . . . . . 220
References for the Semiconductor Interface . . . . . . . . . . . 223

The Semiconductor Optoelectronics, Beam Envelopes Interface

227
Semiconductor-Electromagnetic Waves Coupling . . . . . . . . . 229

The Semiconductor Optoelectronics, Frequency Domain

Interface 230

Chapter 4: The AC/DC Branch

The Electrostatics Interface 234

Domain, Boundary, Edge, Point, and Pair Nodes for the
Electrostatics Interface . . . . . . . . . . . . . . . . . . 236
Charge Conservation . . . . . . . . . . . . . . . . . . . . 238
Conduction Loss (Time-Harmonic) . . . . . . . . . . . . . . . 239
Initial Values . . . . . . . . . . . . . . . . . . . . . . . 240
Space Charge Density . . . . . . . . . . . . . . . . . . . . 240
Zero Charge . . . . . . . . . . . . . . . . . . . . . . . 241
Ground . . . . . . . . . . . . . . . . . . . . . . . . . 241
Electric Potential . . . . . . . . . . . . . . . . . . . . . . 242
Surface Charge Density . . . . . . . . . . . . . . . . . . . 242
External Surface Charge Accumulation . . . . . . . . . . . . . 243
Electric Displacement Field . . . . . . . . . . . . . . . . . . 243

6 | CONTENTS
Periodic Condition . . . . . . . . . . . . . . . . . . . . . 244
Thin Low Permittivity Gap . . . . . . . . . . . . . . . . . . 245
Dielectric Shielding . . . . . . . . . . . . . . . . . . . . . 245
Terminal . . . . . . . . . . . . . . . . . . . . . . . . . 246
Floating Potential . . . . . . . . . . . . . . . . . . . . . . 248
Distributed Capacitance . . . . . . . . . . . . . . . . . . . 250
Line Charge . . . . . . . . . . . . . . . . . . . . . . . . 250
Line Charge (on Axis) . . . . . . . . . . . . . . . . . . . . 251
Line Charge (Out-of-Plane) . . . . . . . . . . . . . . . . . . 251
Point Charge . . . . . . . . . . . . . . . . . . . . . . . 252
Point Charge (on Axis) . . . . . . . . . . . . . . . . . . . 253
Change Cross Section . . . . . . . . . . . . . . . . . . . . 253
Change Thickness (Out-of-Plane). . . . . . . . . . . . . . . . 254
Charge Conservation, Piezoelectric . . . . . . . . . . . . . . . 254
Electrostatic Point Dipole . . . . . . . . . . . . . . . . . . 255

The Electrical Circuit Interface 256

Ground Node . . . . . . . . . . . . . . . . . . . . . . . 258
Voltmeter . . . . . . . . . . . . . . . . . . . . . . . . 258
Ampère Meter. . . . . . . . . . . . . . . . . . . . . . . 259
Resistor . . . . . . . . . . . . . . . . . . . . . . . . . 259
Capacitor. . . . . . . . . . . . . . . . . . . . . . . . . 259
Inductor . . . . . . . . . . . . . . . . . . . . . . . . . 260
Voltage Source . . . . . . . . . . . . . . . . . . . . . . 260
Current Source . . . . . . . . . . . . . . . . . . . . . . 261
Voltage-Controlled Voltage Source . . . . . . . . . . . . . . . 262
Voltage-Controlled Current Source. . . . . . . . . . . . . . . 263
Current-Controlled Voltage Source. . . . . . . . . . . . . . . 263
Current-Controlled Current Source . . . . . . . . . . . . . . 264
Subcircuit Definition . . . . . . . . . . . . . . . . . . . . 264
Subcircuit Instance . . . . . . . . . . . . . . . . . . . . . 265
NPN BJT and PNP BJT. . . . . . . . . . . . . . . . . . . . 265
n-Channel MOSFET and p-Channel MOSFET . . . . . . . . . . . 266
Mutual Inductance . . . . . . . . . . . . . . . . . . . . . 267
Transformer . . . . . . . . . . . . . . . . . . . . . . . 267
Diode . . . . . . . . . . . . . . . . . . . . . . . . . . 268
External I vs. U . . . . . . . . . . . . . . . . . . . . . . 268
External U vs. I . . . . . . . . . . . . . . . . . . . . . . 269

CONTENTS |7
 External I-Terminal . . . . . . . . . . . . . . . . . . . . . 270
SPICE Circuit Import . . . . . . . . . . . . . . . . . . . . 271
SPICE Circuit Export . . . . . . . . . . . . . . . . . . . . 271

Theory for the Electrostatics Interface 273

Charge Relaxation Theory . . . . . . . . . . . . . . . . . . 273
Electrostatics Equations . . . . . . . . . . . . . . . . . . . 276
The Electrostatics Interface in Time Dependent or Frequency
Domain Studies . . . . . . . . . . . . . . . . . . . . . 277

Theory for the Electrical Circuit Interface 278

Electric Circuit Modeling and the Semiconductor Device Models. . . . 278
Bipolar Transistors . . . . . . . . . . . . . . . . . . . . . 279
MOSFET Transistors . . . . . . . . . . . . . . . . . . . . 282
Diode . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Reference for the Electrical Circuit Interface . . . . . . . . . . . 287

Chapter 5: Schrödinger Equation Interfaces

The Schrödinger Equation Interface 290

Domain and Boundary Nodes for the Schrödinger Equation Interface . . 291
Effective Mass . . . . . . . . . . . . . . . . . . . . . . . 292
Electron Potential Energy. . . . . . . . . . . . . . . . . . . 292
Initial Values . . . . . . . . . . . . . . . . . . . . . . . 294
Zero Flux . . . . . . . . . . . . . . . . . . . . . . . . 294
Zero Probability . . . . . . . . . . . . . . . . . . . . . . 294
Periodic Condition . . . . . . . . . . . . . . . . . . . . . 294
Open Boundary . . . . . . . . . . . . . . . . . . . . . . 294
Coupling Operators. . . . . . . . . . . . . . . . . . . . . 296

Theory for the Schrödinger Equation Interface 297

Sign Convention and Energy Units . . . . . . . . . . . . . . . 297
Units of Wave Functions . . . . . . . . . . . . . . . . . . . 298
Sign Convention of the Electron Potential Energy . . . . . . . . . 298
References for the Schrödinger Equation Interface . . . . . . . . . 298

8 | CONTENTS
 The Schrödinger-Poisson Equation Multiphysics Interface 299
Schrödinger-Poisson Coupling . . . . . . . . . . . . . . . . . 301
Schrödinger-Poisson Study Step . . . . . . . . . . . . . . . . 303

Chapter 6: Glossary

Glossary of Terms 306

CONTENTS |9
10 | C O N T E N T S
 1

Introduction

This guide describes the Semiconductor Module, an add-on package for the
COMSOL Multiphysics® modeling environment. The Semiconductor Module
provides physics interfaces for modeling semiconductor devices as well as additional
related functionality.

This chapter introduces the capabilities of this module with a summary of the
physics interfaces. We will also show you where to find documentation and model
examples. The last section is a brief overview with links to each chapter in this user’s
guide.

In this chapter:

• About the Semiconductor Module

• Overview of the User’s Guide

11
 About the Semiconductor Module
These topics are included in this section:

• Modeling Semiconductor Devices

• What Can the Semiconductor Module Do?
• The Semiconductor Module Physics Interface Guide
• Common Physics Interface and Feature Settings and Nodes
• The Semiconductor Module Study Capabilities by Physics Interface
• The Semiconductor Materials Database
• Where Do I Access the Documentation and Application Libraries?

The Physics Interfaces and Building a COMSOL Multiphysics Model in

the COMSOL Multiphysics Reference Manual.

Modeling Semiconductor Devices

The invention of the bipolar transistor at Bell labs in the late 1940s and the
subsequent development of field-effect devices and integrated circuits led to many of
the important technological advancements of the second half of the 20th century. The
widespread adoption of integrated circuits and the continuing efforts to improve the
technology has motivated significant efforts to develop detailed models of devices. The
semiconductor equations are highly nonlinear, so it is very difficult to develop detailed
analytic solutions to them, and consequently numerical modeling has been important
in the development of the technology.

Multiphysics effects can often have important influences on semiconductor device

performance. Device processing often occurs at high temperatures and consequently
stresses can be introduced into the materials. Furthermore, high power devices can
generate a significant amount of heat. Optoelectronic devices and ion sensors are
inherently multiphysics. The Semiconductor Module enables device level modeling on
the COMSOL Multiphysics platform and helps easily create customized simulations
involving multiple physical effects.

12 | CHAPTER 1: INTRODUCTION
What Can the Semiconductor Module Do?
The Semiconductor Module is a collection of interfaces and predefined models for
COMSOL Multiphysics, which can be used to model semiconductor devices.

It includes a dedicated physics interface to model semiconductor devices with the

drift-diffusion equations, enhanced capabilities for modeling electrostatics, and an
Electrical Circuits interface (with a SPICE import facility) that can be coupled to the
device level models.

To include the effect of quantum confinement in the conventional drift-diffusion

equations, the computationally efficient density-gradient formulation is available.

The Semiconductor Module also provides dedicated interfaces to solve the

single-particle Schrödinger Equation and Schrödinger-Poisson Equation, for general
quantum mechanical problems as well as quantum-confined systems such as quantum
wells, wires, and dots.

The included Applications Libraries contains a suite of examples that demonstrate how
to use the physics interfaces for modeling a range of practical devices. These models
provide step-by-step instructions demonstrating how to use the physics interface to
model practical devices.

The Semiconductor Module is well documented and includes a detailed outline of the
underlying theory on which it is based. The software is uniquely transparent because
the model equations are always visible. Furthermore, you have complete freedom to
include phenomena that are not predefined in the Module with easy-to-enter
user-defined expressions and equations.

The Semiconductor Module Physics Interface Guide

The Semiconductor Module extends the functionality of the physics interfaces of the
base package for COMSOL Multiphysics. The details of the physics interfaces and
study types for the Semiconductor Module are listed in the table below.

In the COMSOL Multiphysics Reference Manual:

• Studies and Solvers

• The Physics Interfaces
• For a list of all the core physics interfaces included with a COMSOL
Multiphysics license, see Physics Interface Guide.

ABOUT THE SEMICONDUCTOR MODULE | 13

 PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

AC/DC

Electrical Circuit cir Not space stationary; frequency

dependent domain; time dependent;
small signal analysis,
frequency domain
Electrostatics1 es all dimensions stationary; time dependent;
stationary source sweep;
eigenfrequency; frequency
domain; small signal
analysis, frequency domain
Thermal Stress, Layered — 3D stationary; time dependent
Shell2, 4
Layered Shell lshell 3D stationary; eigenfrequency;
eigenfrequency,
prestressed; mode analysis;
time dependent; time
dependent, modal; time
dependent, prestressed,
modal; time dependent,
modal reduced-order
model; frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; frequency domain,
AWE reduced-order
model; random vibration
(PSD); linear buckling
Piezoelectricity, Layered — 3D stationary; eigenfrequency;
Shell2, 3 time dependent; frequency
domain

14 | CHAPTER 1: INTRODUCTION
PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Semiconductor

Semiconductor semi all dimensions small-signal analysis,

frequency domain;
stationary; time dependent;
semiconductor equilibrium
Semiconductor — 3D, 2D, and frequency-stationary;
Optoelectronics, Beam 2D frequency-transient;
Envelopes2 axisymmetric small-signal analysis,
frequency domain
Semiconductor — 3D, 2D, and frequency-stationary;
Optoelectronics, 2D frequency-transient;
Frequency Domain2 axisymmetric small-signal analysis,
frequency domain
Schrödinger Equation schr all dimensions eigenvalue; stationary; time
dependent
Schrödinger-Poisson — all dimensions Schrödinger-Poisson
Equation
This physics interface is a predefined multiphysics coupling that automatically adds all the
physics interfaces and coupling features required.
Requires the addition of the AC/DC Module.

1 This physics interface is included with the core COMSOL package but has added
functionality for this module.
2
Requires both the Wave Optics Module and the Semiconductor Module.

Common Physics Interface and Feature Settings and Nodes

There are several common settings and sections available for the physics interfaces and
feature nodes. Some of these sections also have similar settings or are implemented in
the same way no matter the physics interface or feature being used. There are also some
physics feature nodes that display in COMSOL Multiphysics.

ABOUT THE SEMICONDUCTOR MODULE | 15

 In each module’s documentation, such as this one, only unique or extra information
is included; standard information and procedures are centralized in the COMSOL
Multiphysics Reference Manual.

In the COMSOL Multiphysics Reference Manual see Table 2-4 for

links to common sections and Table 2-5 to common feature nodes.
You can also search for information: press F1 to open the Help
window or Ctrl+F1 to open the Documentation window.

The Semiconductor Module Study Capabilities by Physics Interface

Table 1-1 lists the physics interfaces specific to semiconductor modeling and this
module. It includes the physical quantities solved for and the standard abbreviation
each one uses. The physical quantities in these interfaces are:

• The electric scalar potential, V

• The electron concentration, N
• The hole concentration, P
• The wave function, psi

Studies and Solvers in the COMSOL Multiphysics Reference Manual.

TABLE 1-1: SEMICONDUCTOR MODULE INTERFACE DEPENDENT VARIABLES AND PRESET STUDY OPTIONS

PHYSICS INTERFACE NAME DEPENDENT PRESET STUDY OPTIONS

VARIABLES
FREQUENCY STATIONARY
SMALL SIGNAL ANALYSIS,

FREQUENCY TRANSIENT

SCHRÖDINGER-POISSON
FREQUENCY-DOMAIN
FREQUENCY DOMAIN
EIGENFREQUENCY
SEMICONDUCTOR

TIME DEPENDENT
STATIONARY

EQUILIBRIUM

EIGENVALUE

AC/DC

Electrostatics es V √ √ √ √ √

16 | CHAPTER 1: INTRODUCTION
TABLE 1-1: SEMICONDUCTOR MODULE INTERFACE DEPENDENT VARIABLES AND PRESET STUDY OPTIONS

PHYSICS INTERFACE NAME DEPENDENT PRESET STUDY OPTIONS

VARIABLES

FREQUENCY STATIONARY
SMALL SIGNAL ANALYSIS,

FREQUENCY TRANSIENT

SCHRÖDINGER-POISSON
FREQUENCY-DOMAIN
FREQUENCY DOMAIN
EIGENFREQUENCY
SEMICONDUCTOR

TIME DEPENDENT
STATIONARY

EQUILIBRIUM

EIGENVALUE
Electrical Circuit cir none √ √ √
SEMICONDUCTOR

Semiconductor semi V, N, P √ √ √ √
Semiconductor — V, N, P √ √ √
Optoelectronics, E1, E2
Beam Envelopes1
Semiconductor — V, N, P √ √ √
Optoelectronics, E
Frequency
Domain1
Schrödinger schr psi √ √ √
Equation
Schrödinger-Poiss schrp V, psi √
on Equation
1
Requires both the Wave Optics Module and the Semiconductor Module. These
are also multiphysics interfaces.

The Semiconductor Materials Database

The Semiconductor Module includes an additional Semiconductors material database
with material properties appropriate for several materials.

For detailed information about materials and the Semiconductor

Materials Database, see Materials in the COMSOL Multiphysics
Reference Manual.

ABOUT THE SEMICONDUCTOR MODULE | 17

 Where Do I Access the Documentation and Application Libraries?
A number of internet resources have more information about COMSOL, including
licensing and technical information. The electronic documentation, topic-based (or
context-based) help, and the application libraries are all accessed through the
COMSOL Desktop.

If you are reading the documentation as a PDF file on your computer,

the blue links do not work to open an application or content
referenced in a different guide. However, if you are using the Help
system in COMSOL Multiphysics, these links work to open other
modules, application examples, and documentation sets.

THE DOCUMENTATION AND ONLINE HELP

The COMSOL Multiphysics Reference Manual describes the core physics interfaces
and functionality included with the COMSOL Multiphysics license. This book also has
instructions about how to use COMSOL Multiphysics and how to access the
electronic Documentation and Help content.

Opening Topic-Based Help

The Help window is useful as it is connected to the features in the COMSOL Desktop.
To learn more about a node in the Model Builder, or a window on the Desktop, click
to highlight a node or window, then press F1 to open the Help window, which then
displays information about that feature (or click a node in the Model Builder followed
by the Help button ( ). This is called topic-based (or context) help.

To open the Help window:

• In the Model Builder, Application Builder, or Physics Builder click a node or

window and then press F1.
• On any toolbar (for example, Home, Definitions, or Geometry), hover the
mouse over a button (for example, Add Physics or Build All) and then
press F1.
• From the File menu, click Help ( ).
• In the upper-right corner of the COMSOL Desktop, click the Help ( )
button.

18 | CHAPTER 1: INTRODUCTION
 To open the Help window:

• In the Model Builder or Physics Builder click a node or window and then
press F1.
• On the main toolbar, click the Help ( ) button.
• From the main menu, select Help>Help.

Opening the Documentation Window

To open the Documentation window:

• Press Ctrl+F1.
• From the File menu select Help>Documentation ( ).

To open the Documentation window:

• Press Ctrl+F1.
• On the main toolbar, click the Documentation ( ) button.
• From the main menu, select Help>Documentation.

THE APPLICATION LIBRARIES WINDOW

Each model or application includes documentation with the theoretical background
and step-by-step instructions to create a model or application. The models and
applications are available in COMSOL Multiphysics as MPH files that you can open
for further investigation. You can use the step-by-step instructions and the actual
models as templates for your own modeling. In most models, SI units are used to
describe the relevant properties, parameters, and dimensions, but other unit systems
are available.

Once the Application Libraries window is opened, you can search by name or browse
under a module folder name. Click to view a summary of the model or application and
its properties, including options to open it or its associated PDF document.

The Application Libraries Window in the COMSOL Multiphysics

Reference Manual.

ABOUT THE SEMICONDUCTOR MODULE | 19

 Opening the Application Libraries Window
To open the Application Libraries window ( ):

• From the Home toolbar, Windows menu, click ( ) Applications

Libraries.
• From the File menu select Application Libraries.

To include the latest versions of model examples, from the File>Help

menu, select ( ) Update COMSOL Application Library.

Select Application Libraries from the main File> or Windows> menus.

To include the latest versions of model examples, from the Help menu
select ( ) Update COMSOL Application Library.

CONTACTING COMSOL BY EMAIL

For general product information, contact COMSOL at [email protected].

COMSOL ACCESS AND TECHNICAL SUPPORT

To receive technical support from COMSOL for the COMSOL products, please
contact your local COMSOL representative or send your questions to
[email protected]. An automatic notification and a case number are sent to you by
email. You can also access technical support, software updates, license information, and
other resources by registering for a COMSOL Access account.

COMSOL ONLINE RESOURCES

COMSOL website www.comsol.com

Contact COMSOL www.comsol.com/contact
COMSOL Access www.comsol.com/access
Support Center www.comsol.com/support
Product Download www.comsol.com/product-download
Product Updates www.comsol.com/support/updates
COMSOL Blog www.comsol.com/blogs
Discussion Forum www.comsol.com/community
Events www.comsol.com/events

20 | CHAPTER 1: INTRODUCTION
COMSOL Video Gallery www.comsol.com/video
Support Knowledge Base www.comsol.com/support/knowledgebase

ABOUT THE SEMICONDUCTOR MODULE | 21

 Overview of the User’s Guide
The Semiconductor Module User’s Guide gets you started with semiconductor
modeling using COMSOL Multiphysics. The information in this user’s guide is
specific to the Semiconductor Module. Instructions on how to use COMSOL
Multiphysics in general are included in the COMSOL Multiphysics Reference
Manual.

As detailed in the section Where Do I Access the Documentation and

Application Libraries? this information can also be searched from the
COMSOL Multiphysics software Help menu.

TABLE OF CONTENTS AND INDEX

To help you navigate through this guide, see the Contents and Index.

MODELING SEMICONDUCTORS
The Modeling Guidelines chapter discusses topics including the Physics for
Semiconductor Modeling, Connecting to Electrical Circuits, Defining the Carrier
Mobility, Meshing, Solving, Quantum-Confined Systems, and many others.

THE SEMICONDUCTOR INTERFACE

The Semiconductor Branch Interface chapter includes physics feature information and
describes the theory relating to The Semiconductor Interface. Topics include The
Sommerfeld Model and the Density of States, Electrons in a Perturbed Periodic
Potential, Equilibrium Carrier Concentrations, Band Gap Narrowing, and many
others.

THE AC/DC BRANCH INTERFACES

The AC/DC Branch chapter includes physics feature information for The
Electrostatics Interface and The Electrical Circuit Interface as well as theory is
discussed.

THE SCHRÖDINGER EQUATION INTERFACE

The Schrödinger Equation Interfaces chapter describes the physics features and usage
of The Schrödinger Equation Interface.

22 | CHAPTER 1: INTRODUCTION
 2

Modeling Guidelines

In this chapter:

• Physics for Semiconductor Modeling

• Discretization and Formulation Options
• Defining the Carrier Mobility
• Doping
• Connecting to Electrical Circuits
• Nonisothermal Models
• SPICE Import and Export
• Small-Signal Analysis
• Meshing
• Solving
• Quantum-Confined Systems

23
 Physics for Semiconductor Modeling
The physics of semiconductor devices is highly dependent on the size of the device.
Over the last 50 years logic devices have been progressively miniaturized, enabling
improvements in speed and reductions in power consumption. Nonetheless
semiconductors are so ubiquitous that devices of larger sizes such as photodiodes and
power transistors are still widely used for many applications.

The Semiconductor physics interface was initially directed at larger scale devices (with
length scales of 100s of nm or more), which can be modeled by a conventional
drift-diffusion approach using partial differential equations.

More recently, the density-gradient theory has been added to the Semiconductor
interface as a computationally efficient option to include the effect of quantum
confinement in the drift-diffusion method, pushing its applicability to nanometer
length scales.

The Schrödinger Equation Interface and The Schrödinger-Poisson Equation

Multiphysics Interface have also been added to the Semiconductor Module to provide
more tools for analyzing quantum systems.

It is useful to note some of the important assumptions implicit to the drift-diffusion

type of approaches:

• The relaxation-time approximation is used to describe the scattering process. This

is a much simplified form of the scattering probability, which is elastic and isotropic.
• Magnetic fields are not included in the model.
• The carrier temperature is assumed to be equal to the lattice temperature and,
consequently, the diffusion of hot carriers is not properly described.
• The energy bands are assumed to be parabolic. In reality the band structure is
significantly altered in a complex manner in the vicinity of free surfaces or grain
boundaries.
• Velocity overshoot, and other complex time-dependent conductivity phenomena,
are not included in the model.

In addition to these intrinsic assumptions, The Semiconductor Interface allows you to

make additional assumptions to simplify the solution process:

• For nondegenerate semiconductors it is possible to assume a Maxwell-Boltzmann

distribution for the carrier energies at a given temperature, which reduces the

24 | CHAPTER 2: MODELING GUIDELINES

 nonlinearity of the semiconductor equations. If degenerate semiconductors are
present within the model, or at lower temperatures, it is necessary to use
Fermi-Dirac statistics.
• In majority carrier devices, it is often only necessary to solve for one of the carrier
concentrations (the majority carrier). The minority carrier concentration is usually
unimportant for the device operation and can be estimated by assuming the mass
action law.

By default Maxwell-Boltzmann statistics are assumed by the Semiconductor interface

and the interface explicitly solves for both the electron and hole concentrations.

The Semiconductor interface solves Poisson’s equation in conjunction with continuity

equations for the charge carriers. An important length scale to consider when
modeling electrostatic fields in the presence of mobile carriers is the Debye length:

k B Tε 0 ε r
Ld = ---------------------
-
2
q N

where kB is Boltzmann’s constant, T is the lattice temperature, ε0 is the permittivity of

free space, εr is the relative permittivity of the semiconductor, N is the total
concentration of the electrons and holes, and q is the electron charge. The Debye
length is the length scale over which the electric field decays in the presence of mobile
carriers; it is important to resolve this length scale with the mesh in semiconductor
models.

Theory for the Semiconductor Interface

Zero Reference of Electric Potential

The electric potential variable implemented in the Semiconductor interface uses the
vacuum level as its zero reference. This is why, for example for a piece of grounded
intrinsic silicon, the electric potential is about -4.6 V, not 0 V.

Units of Band Gap and Electron Affinity

In the graphical user interface, the input for the band gap is actually “band-gap divided
by the elementary charge”, so it takes on the unit of electric potential but has the same

PHYSICS FOR SEMICONDUCTOR MODELING | 25

 numerical value as the band gap energy in eV. The same applies to the input for the
electron affinity.

26 | CHAPTER 2: MODELING GUIDELINES

Discretization and Formulation
O pti o ns
The Semiconductor Interface includes quasi-Fermi level, density-gradient, linear, and
logarithmic finite element formulations and a finite volume formulation. The
formulation used is selected in the Discretization section since the shape functions that
can be used are directly related to the formulation employed.

The finite volume formulation uses constant shape functions, whilst the finite element
formulations can use either linear or quadratic shape functions. In the different
formulations the carrier concentration dependent variables (by default Ne and Ph)
represent different quantities. In the linear finite element and finite volume
formulations Ne = N and Ph = P, where N is the electron concentration and P is the
hole concentration. For the logarithmic finite element formulation Ne = ln(N) and
Ph = ln(P). For the quasi-Fermi level and density-gradient formulation, the
quasi-Fermi levels for the electrons and holes are the dependent variables.

To change the formulation, first expand the Discretization section. Then under
Discretization select a Formulation (as in Figure 2-1). Each formulation has advantages
and disadvantages.

In the COMSOL Multiphysics Reference Manual see Table 2-4 for links
to common sections (such as Discretization) and Table 2-5 to common
feature nodes. You can also search for information: press F1 to open the
Help window or Ctrl+F1 to open the Documentation window.

DISCRETIZATION AND FORMULATION OPTIONS | 27

 Figure 2-1: The Discretization section. Choose between a finite volume or a linear or
quadratic finite element, quasi-Fermi level, or density-gradient formulation.

The finite volume discretization inherently conserves current. Consequently, it usually

provides the most accurate result for the current density of the charge carriers. This is
why it is the default setting. In order to enhance numerical stability, a
Scharfetter-Gummel upwinding scheme is used for the charge carrier equations.
Poisson’s equation is discretized using a centered difference scheme. When doing
multiphysics simulations it is important to realize that the shape functions are constant.
Consequently, fluxes cannot be evaluated using spatial derivatives of the dependent
variables (for example, expressions such as d(V,x) evaluate to zero as V is represented
by a constant shape function within each element). Flux quantities such as fields and
currents can be evaluated and used in equations (both within the Semiconductor
interface and in other physics interfaces) if the predefined variables from Table 2-1 are
used in expressions:
TABLE 2-1: SEMICONDUCTOR PREDEFINED VARIABLES

NAME VARIABLE

Electric displacement field semi.D semi.Dx, semi.Dy, semi.Dz)

Electron current density semi.Jn (semi.Jnx, semi.Jny, semi.Jnz)
Hole current density semi.Jp (semi.Jpx, semi.Jpy, semi.Jpz)
Electron drift current semi.Jn_drift (semi.Jn_driftx,
density semi.Jn_drifty, semi.Jn_driftz)

Hole drift current density semi.Jp_drift (semi.Jp_driftx,

semi.Jp_drifty, semi.Jp_driftz)
Electron diffusion current semi.Jn_diff (semi.Jn_diffx, semi.Jn_diffy,
density semi.Jn_diffz)

Hole diffusion current semi.Jp_diff (semi.Jp_diffx, semi.Jp_diffy,

density semi.Jp_diffz)

28 | CHAPTER 2: MODELING GUIDELINES

TABLE 2-1: SEMICONDUCTOR PREDEFINED VARIABLES

NAME VARIABLE

Electron thermal diffusion semi.Jn_th (semi.Jn_thx, semi.Jn_thy,

current density semi.Jn_thz)

Hole thermal diffusion semi.Jp_th (semi.Jp_thx, semi.Jp_thy,

current density semi.Jp_thz)

Any variables that involve expressions directly derived from the variables in Table 2-1
can also be used in expressions, for example, the electric field, semi.E, or the total
current, semi.J.

The finite element formulation typically solves faster than the finite volume
formulation. One reason is that, for an identical mesh, the finite element method with
linear shape functions typically results in fewer degrees of freedom. In 2D, for
triangular mesh elements, the number of degrees of freedom for the finite element
method with linear shape functions is approximately half that for a finite volume
discretization. Coupling to other physics interfaces is straightforward and variables can
be differentiated using the d operator. The finite element method is an energy
conserving method and thus current conservation is not implicit in the technique.
Current conservation for the linear formulation is poor and this formulation is
provided primarily for reasons of backward compatibility. Current conservation in the
log formulation is much better but still not as good as the finite volume method. In
order to help with numerical stability, a Galerkin least-squares stabilization method is
included. This method usually enhances the ability to achieve a converged solution,
particularly when using the linear formulation. However, it can be preferable to disable
the stabilization, since the additional numerical diffusion that the technique introduces
can produce slightly unphysical results. As a result of the reduced gradients in the
dependent variables obtained when using the log formulation, stabilization is often not
required when using this technique.

It is also possible to solve a problem with the stabilization active and then
use this solution as the initial condition in a separate study, where the
stabilization is inactive. To turn the stabilization on or off, click the Show
More Options button ( ) and select Stabilization in the Show More
Options dialog box. Then click to select the Streamline diffusion check box
or click to clear.

DISCRETIZATION AND FORMULATION OPTIONS | 29

 In the COMSOL Multiphysics Reference Manual:

• Numerical Stabilization
• Introduction to Solvers and Studies

The quasi-Fermi level formulation uses the quasi-Fermi level as the dependent variable,
instead of the carrier concentration. It is advantageous in some cases, for example for
systems with wide band gaps or at very low temperatures.

The density-gradient formulation provides a computationally efficient option to

include the effect of quantum confinement in the drift-diffusion method, pushing its
applicability to nanometer length scales. It also uses the quasi-Fermi level as the
dependent variable.

30 | CHAPTER 2: MODELING GUIDELINES

Defining the Carrier Mobility
Realistic models for the carrier mobility are important to simulate semiconductor
devices with a drift diffusion approach. The mobility is limited by scattering of the
carriers within the material. Phenomena that can affect the mobility include:

• Phonons (L): thermally generated acoustic waves traveling through the crystal
(scattering by phonons is frequently referred to as lattice scattering, which is
somewhat confusing as the lattice itself never scatters the carriers unless perturbed
from perfect periodicity).
• Ionized impurities (I): these begin to have an effect at doping levels above
approximately 1015 cm−3 at room temperature.
• Carrier-carrier scattering (C)
• Neutral impurity scattering (N): this is important only at low temperatures, typically
below 77 K.
• High field velocity saturation (E).
• Surface scattering (S): this includes effects such as interface charges, scattering by
surface phonons, and so on. It is important in field effect devices, such as field effect
transistors.

The Semiconductor Module includes several predefined mobility models, as well as

user-defined models. These models are added as subnodes to the Semiconductor
Material Model node.

The letters in the above list (for example, L for phonons) are used to
identify which effects a given predefined mobility model incorporates and
appears in the name of the feature.

Both user-defined and predefined mobility models can be combined in arbitrary ways.
Each mobility model defines an output electron and hole mobility. If appropriate
(some filtering occurs in the predefined mobility models to prevent inappropriate
combinations), the output mobility can be used as an input to other mobility models.
User-defined mobility models accept any of the predefined mobility models as inputs
and can be used as inputs for all the predefined mobility models that require an input.
The model used within the simulation is selected for electrons and holes by changing
the Electron mobility and the Hole mobility settings in the Mobility Model section of the

DEFINING THE CARRIER MOBILITY | 31

 Semiconductor Material Model node, which by default uses a constant mobility
obtained from the material properties.

It is important to understand that each type of mobility model subnode

only defines mobility variables for electrons and holes that can be used by
other models, or by the Semiconductor Material Model parent node. The
mobility models for electrons and holes actually used in the simulation are
determined by the selections or settings in the Semiconductor Material
Model node (the parent node), which do not change when additional
mobility model subnodes are added. In order to add a mobility model to
a simulation, it is necessary to both add the sequence of mobility model
subnodes to the Model Builder, and then to select the required final
mobility for the electrons and holes in the Semiconductor Material Model
parent node.

In order to implement field dependent mobility models it is usually

necessary to add additional dependent variables for the field components
used in the model (typically the electric field components parallel to or
perpendicular to the current components). This is the approach taken in
the Caughey-Thomas Mobility Model (E) and Lombardi Surface
Mobility Model (S). Consequently, creating such mobility models is only
recommended for advanced users.

• Mobility Models
• User-Defined Mobility Model
• Power Law Mobility Model (L)
• Arora Mobility Model (LI)
• Fletcher Mobility Model (C)

32 | CHAPTER 2: MODELING GUIDELINES

Doping
Specifying the doping distribution is critical to model semiconductor devices.
COMSOL multiphysics includes a range of tools which enable any doping profile to
be realized. This functionality is provided by the Analytic Doping Model and
Geometric Doping Model.

In this section:

• Using the Analytic Doping Model

• Using the Geometric Doping Model
• Visualizing the Dopant Distribution
• Example of Multiple Superimposed Doping Profiles.

Using the Analytic Doping Model

The Analytic Doping Model makes it possible to express doping profiles as a function
of the local coordinate system, which can be rotated with respect to the global
coordinate system if required. This is suitable to achieve profiles that are convenient to
define in relation to the coordinate axes. The dopant distribution can be defined using
either the user defined or box methods.

ANALYTIC DOPING MODEL: USER DEFINED

A user-defined doping distribution can be used to specify a constant dopant density in
a domain. Alternately a spatially varying doping profile can be specified; the doping
profile is defined using the functions available within COMSOL Multiphysics. Any
analytic expression, written in terms of the local coordinate system, is permitted.
Figure 2-2 shows an example of a user-defined profile consisting of a Gaussian decay
away from the lower boundary.

Doping distributions can also be imported from external files. This is useful when the
required distribution cannot be defined analytically; for example, if the doping profile
is output from an external diffusion simulation.

To import a doping profile using an interpolation function:

1 From the Definitions toolbar, click Interpolation ( ).
2 On the Settings window for Interpolation select File as the Data source.
3 Specify the Filename and the Number of arguments.

DOPING | 33
 4 Once an interpolation function is defined, it can be used in an Analytical Doping
Model feature. This is achieved by selecting a User defined dopant distribution and
calling the interpolation function with the appropriate arguments.

Interpolation in the COMSOL Multiphysics Reference Manual.

Figure 2-2: Example of a user-defined dopant distribution created using the Analytic
Doping Model feature. The Analytic Doping Model feature is the appropriate doping
feature as this distribution is trivial to express as a function of the coordinate axes.

ANALYTIC DOPING MODEL: BOX

The box method allows a block-shaped region of constant doping to be defined, along
with a decay profile away from the region. This is useful for approximating some
physical doping techniques, such as diffusion processes, which distribute dopants away
from regions of high concentration resulting in characteristic decay profiles. The
location of the region is defined by specifying either the corner or center coordinate
using the global coordinate system. If a local rotated coordinate system is used within
the feature, the orientation of the block rotates around this specified coordinate. A
profile is selected from the Dopant profile away from uniform region list. The decay

34 | CHAPTER 2: MODELING GUIDELINES

length scale of the profile is controlled by either specifying a control length directly or
by specifying a junction depth. The junction depth specifies the distance, from the
boundary of the uniformly doped region, where the dopant concentration is equal to
the value specified in the Background doping concentration input. This distance can be
specified independently for each axis in the component geometry by selecting Specify
different length scales for each direction. Figure 2-3 is an example of a doping profile
created using this method. A rectangular region of uniform doping is specified in the
top left of the geometry, and a Gaussian profile is selected away from this region with
different junction depths in the x and y directions.

Figure 2-3: Example of a box doping distribution created using the Analytic Doping
Model feature. This distribution is defined as a rectangle of constant doping located in the
top left of the domain with a Gaussian decay profile away from the boundaries of this
region. The region of uniform doping in the top left is outlined with a blue box.

Using the Geometric Doping Model

The Geometric Doping Model enables doping profiles to be expressed as a function of
the distance from selected boundaries. This is suitable for cases where the profile
wanted is more conveniently expressed in terms of the geometry than the coordinate
axes. The boundaries from which the distance is calculated using the Boundary

DOPING | 35
 Selection for Doping Profile node. The form of the profile is selected from the Dopant
profile away from the boundary list. This list contains preset functions that enable
several common profiles to be easily defined. A User defined option is also available that
enables any expression to be manually entered.

GEOMETRIC DOPING MODEL: USER DEFINED

A user-defined profile can be used to specify any dopant distribution that is expressed
as a function of the distance from boundaries in the geometry. This distance is available
as semi.gdm#.D within COMSOL, where # indicates the number of the Geometric
Doping Model feature. Figure 2-4 is an example doping profile that is created with the
expression 1e16[1/cm^3]*exp(-(semi.gdm1.D/0.5[um])^2).

Figure 2-4: Example of a user-defined dopant distribution created using the Geometric
Doping Model feature. The dopant concentration decays as a Gaussian away from the
lower boundary. The Geometric Doping Model feature is the appropriate doping feature
as it is more convenient to define this distribution in terms of the distance from the curved
boundary than as a function of the coordinate system.

GEOMETRIC DOPING MODEL: PRESET PROFILES

The Geometric Doping Model makes Gaussian, linear, and error function profiles
available as options in the Dopant profile away from the boundary list. When a preset
option is selected, the dopant concentration at the boundary can be input along with
a parameter to set the decay length scale of the profile. The decay length scale of the

36 | CHAPTER 2: MODELING GUIDELINES

profile is controlled by specifying a control length directly or by specifying a junction
depth. The junction depth specifies the distance from the selected boundaries at which
the dopant concentration is equal to the value specified in the Background doping
concentration user input.

Figure 2-5 is an example profile that is created using a combination of an Analytic

Doping Model and a Geometric Doping Model. The dopant concentration is constant
within the top left domain, which is bounded by a curve. Away from this constant
region the concentration decays with a Gaussian profile. The constant region is
specified using a user-defined profile in an Analytic Doping Model. The Gaussian
profile is achieved using a preset profile in a Geometric Doping Model with the curve
as the selected boundary. This example shows how the Geometric Doping Model
allows curves in the geometry to be easily incorporated into the dopant distribution,
without the need to express the curves analytically in terms of the coordinate system.

Figure 2-5: Example of a profile that uses the Geometric Doping Model feature to
accommodate a curved boundary. An Analytic Doping Model feature is used to create a
region of constant doping in the top left that is bounded by the curve. This curve is used as
the selected boundary in the Geometric Doping Model feature to generate a Gaussian
decay profile away from the constant region.

DOPING | 37
 Visualizing the Dopant Distribution
To visualize the distribution, on the Study toolbar, click Get Initial Value ( ). Then
add a plot group and display the corresponding dopant density. The signed doping
concentration, semi.Nd-semi.Na (positive for net donor doping and negative for net
acceptor doping), gives the net doping and can be useful for visualizing the effect of
different doping steps. If multiple doping features are superimposed, it is possible to
visualize the total dopant concentration or to display the contribution from individual
doping features.

Study and Computing the Initial Values in the COMSOL Multiphysics

Reference Manual.

The semi.Nd-semi.Na expression shows the total net doping due to the sum of all
doping features. To see the distribution from a single doping feature the
corresponding node tag must be included in the plot variable.

To select the node labels to display, on the Model Builder toolbar click
Model Builder Node Label .

The node tags are shown in curly brackets next to the nodes in the Model Builder.
Analytic Doping Model features are tagged as adm# and Geometric Doping Model features
are tagged as gdm#, where # corresponds to the number of the feature. The dopant
concentration from each doping feature is available as semi.<tag>.Na for acceptor
doping or semi.<tag>.Nd for donor doping. For example, semi.adm1.Nd for a
donor distribution from the first Analytic Doping Model added to the component.

Example of Multiple Superimposed Doping Profiles

This example uses multiple Gaussian profiles to demonstrate how to create an intricate
doping distribution using a combination of several doping features. Figure 2-6 shows
the geometry along with a line graph of the dopant concentrations. The geometry
shows the location of the contacts and the line cut along which the line graph is
plotted.

To create the doping profile three Analytic Doping Model nodes and one Geometric
Doping Model node are used. The line graph shows the total resultant dopant
concentration as well as the individual contribution from each node.

38 | CHAPTER 2: MODELING GUIDELINES

 Care should be taken when specifying multiple doping profiles in the same
domains as the resultant dopant concentration is the sum of all coincident
doping models.

Figure 2-6: Top: Geometry in which the example dopant concentration is created. Bottom:
Line graphs of the total doping distribution and the individual contributions from each
of the steps in the instructions. The numbered circles highlight the overlap points of the
different distributions due to the junction depth specified in the corresponding step.

DOPING | 39
 To create the doping profile:

The following example is similar to the model Bipolar Transistor, which

is available in the Semiconductor Applications Libraries
(Semiconductor_Module/Devices/bipolar_transistor).

1 Constant background: a constant n-type doping of 1017 cm−3 is added using an

Analytic Doping Model that specifies a User defined profile with constant Donor
Concentration.
2 Top surface: a Box dopant distribution that decays away from the top surface with a
Gaussian profile is added using a second Analytical Doping Model. The uniformly
doped region is set to cover the entire width of the model at the top surface but to
have zero depth, thus the profile decays away from the surface.
- Acceptor doping is applied with a maximum concentration of 1019 cm−3.
- The donor concentration defined in the previous step, semi.adm1.Nd, is selected
for the Background doping concentration.
- The Junction depth is set to 2 μm below the top surface. Since the background
doping is of a different type, semi.adm2.Na has equal magnitude to
semi.adm1.Nd at this depth, as indicated by point 2 shown in Figure 2-6. Thus
the overall contribution to semi.Nd-semi.Na due to the first two distributions
is zero at a depth of 2 μm. This is an example of doping distribution into a
constant background distribution of the opposite type.
3 Emitter contact: a Geometric Doping Model is used to define a Gaussian profile which
decays away from the emitter contact.
- The boundary that represents the emitter contact is selected in Boundary Selection
for Doping Profile node and a Gaussian Dopant profile away from the boundary is
selected.
- The Junction depth is set to 1.5 μm.
- The Background doping concentration must account for the effect of both of the
two preceding steps. To achieve this, choose a User defined background
concentration and set it to semi.adm2.Na-semi.adm1.Nd. This causes the
resulting distribution, semi.gdm1.Nd, to be equal magnitude to semi.adm2.Na

40 | CHAPTER 2: MODELING GUIDELINES

 at the desired junction depth of 1.5 μm, as can be seen at point 3 labeled in
Figure 2-6.
- This is an example that demonstrates the need to carefully consider the
cumulative effects of coincident dopant distributions when placing junction
depths.
4 Collector contact: another Box distribution is defined with a third Analytic Doping
Model. This is used to create a Gaussian profile that decays away from the collector
contact. The uniformly doped region is set to cover the entire width of the model
at the bottom surface but to have zero depth.
- Donor doping is applied with a maximum concentration of 1022 cm−3.
- The Junction depth is set to 0.2 μm, which produces a junction at a depth of
4.8 μm from the top surface.
- In this case the constant donor concentration defined in step 1 (1017 cm−3),
semi.adm1.Nd, is selected for the Background doping concentration. Since donors
are being doped into donors the concentration due to this final distribution
(semi.adm3.Nd) is 2·1017 cm−3 at the junction depth, which is double the
background level as indicated a point 4 shown in Figure 2-6. This is because at
the junction depth semi.adm3.Nd has equal magnitude to the constant
background and, because it is the same doping type, the sum of the contribution
from both distributions results in double the concentration. This is an example of
doping one type into a constant background of the same type.

DOPING | 41
 Connecting to Electrical Circuits
In this section:

• About Connecting Electrical Circuits to Physics Interfaces

• Connecting Electrical Circuits Using Predefined Couplings
• Connecting Electrical Circuits by User-Defined Couplings
• Solving
• Postprocessing

About Connecting Electrical Circuits to Physics Interfaces

This section describes the various ways electrical circuits can be connected to other
physics interfaces in COMSOL Multiphysics. If you are not familiar with circuit
modeling, it is recommended that you review the Theory for the Electrical Circuit
Interface.

In general electrical circuits connect to other physics interfaces via one or more of three
special circuit features:

• External I vs. U
• External U vs. I
• External I-Terminal

These features either accept a voltage measurement from the connecting noncircuit
physics interface and return a current from a or the Electrical Circuit interface or the
other way around.

The “External” features are considered “ideal” current or voltage sources

by the Electrical Circuit interface. Hence, you cannot connect them
directly in parallel (voltage sources) or in series (current sources) with
other ideal sources. This results in the error message The DAE is
structurally inconsistent. A workaround is to provide a suitable parallel
or series resistor, which can be tuned to minimize its influence on the
results.

42 | CHAPTER 2: MODELING GUIDELINES

Connecting Electrical Circuits Using Predefined Couplings
In addition to these circuit features, interfaces in the AC/DC Module, RF Module,
MEMS Module, Plasma Module, and Semiconductor Module (the modules that
include the Electrical Circuit interface) also contain features that provide couplings to
the Electrical Circuit interface by accepting a voltage or a current from one of the
specific circuit features (External I vs. U, External U vs. I, and External I-Terminal).

This coupling is typically activated when:

• A choice is made in the Settings window for the noncircuit physics interface feature,
which then announces (that is, includes) the coupling to the Electrical Circuit
interface. Its voltage is then included to make it visible to the connecting circuit
feature.
• A voltage that has been announced (that is, included) is selected in a feature node’s
Settings window.

These circuit connections are supported in Terminals.

Connecting Electrical Circuits by User-Defined Couplings

A more general way to connect a physics interface to the Electrical Circuit interface is
to:

• Apply the voltage or current from the connecting “External” circuit feature as an
excitation in the noncircuit physics interface.
• Define your own voltage or current measurement in the noncircuit physics interface
using variables, coupling operators and so forth.
• In the Settings window for the Electrical Circuit interface feature, selecting the
User-defined option and entering the name of the variable or expression using
coupling operators defined in the previous step.

DETERMINING A CURRENT OR VOLTAGE VARIABLE NAME

To determine a current or voltage variable name, look at the Dependent Variables node
under the Study node. To do this:

1 In the Model Builder, right-click the Study node and select Show Default Solver.

CONNECTING TO ELECTRICAL CIRCUITS | 43

 2 Expand the Solver>Dependent Variables node and click the state node, in this
example, Current through device R1 (comp1.currents). The variable name is shown in
the Settings window for State.

Typically, voltage variables are named cir.Xn_v and current variables

cir.Xn_i, where n is the “External” device number — 1, 2, and so on.

Solving

Some modeling errors lead to the error message The DAE is structurally
inconsistent being displayed when solving. This error typically occurs
from having an open current loop, from connecting voltage sources in
parallel, or connecting current sources in series.

In this respect, the predefined coupling features are also treated as (ideal)
voltage or current sources. The remedy is to close current loops and to
connect resistors in series with voltage sources or in parallel with current
sources.

Postprocessing
The Electrical Circuits interface, unlike most of the other physics interfaces, solves for
a relatively large number of global dependent variables (such as voltages and currents),

44 | CHAPTER 2: MODELING GUIDELINES

instead of solving for a few space-varying fields (such as temperature or displacement).
For this reason, the Electrical Circuit interface does not provide default plots when
computing a study.

The physics interface defines a number of variables that can be used in postprocessing.
All variables defined by the Electrical Circuit interface are of a global scope, and can
be evaluated in a Global Evaluation node (under Derived Values). In addition, the time
evolution or dependency on a parameter can be plotted in a Global plot (under a 1D
Plot Group node).

The physics interface defines a Node voltage variable for each electrical node in the
circuit, with name cir.v_name, where cir is the physics interface Label and <name>
is the node Name. For each two-pin component, the physics interface also defines
variables containing the voltage across it and the current flowing through it.

In the COMSOL Multiphysics Reference Manual:

• Derived Values, Evaluation Groups, and Tables and Global Evaluation

• Plot Groups and Plots and Global

CONNECTING TO ELECTRICAL CIRCUITS | 45

 Nonisothermal Models
The Semiconductor interface is designed to model nonisothermal devices. In order to
model heat transfer in the device, an additional Heat Transfer in Solids interface should
be added to the model. This physics interface can be used to model the heat transfer
within the domain. Appropriate thermal boundary conditions must be added in
addition to a heat source term that is coupled to the Semiconductor interface.

To add the heat source term, from the Physics toolbar, Domain menu,
select Heat Source. Then click the Heat Source 1 node and select Total heat
source (semi) from the list of coupling variables in the General source Q list
(note that if several heat sources or Semiconductor interfaces are added
then the name of the heat source feature and the semiconductor heat
source coupling variable is incremented accordingly). To couple the
temperature computed by a heat transfer interface back into the
Semiconductor interface, go to the Model Inputs section of the
Semiconductor Material Model 1 node. Under Temperature, T, select
Temperature (ht).

Currently COMSOL implements a simplified heat source term incorporating only

Joule heating and the heat source due to recombination (see Ref. 26 for a detailed
discussion of the appropriate heat source term to be used in semiconductor modeling).
The heat source term Q is given by:

Q = E ⋅ J + ( E g + 3k B T )U

where U is the total recombination rate computed from summing all the
recombination mechanisms. U is computed from the mean of the total electron and
hole recombination rates U=(Un+Up)/2, which are almost always equal except in
advanced models.

Additional terms can be added to the heat source manually in the finite element
method (the existing terms are available as the variable semi.Q_tot).

It is not straightforward to add these terms for the finite volume method,
as they involve the derivatives of the dependent variables, which must be
computed using appropriate expressions.

46 | CHAPTER 2: MODELING GUIDELINES

Some of the boundary conditions for the Semiconductor interface require an
equilibrium reference temperature to be defined in order to correctly define the
potentials with respect to the equilibrium Fermi energy.

To change this reference temperature click the Show More Options button
( ) and select Advanced Physics Options. Then adjust the Temperature
reference for energy levels (T0) setting in the Temperature Reference section
of the Semiconductor interface Settings window. The reference
temperature should be within the range of validity of any
temperature-dependent material properties used in the model.

When solving nonisothermal semiconductor models, a segregated solver

should be used, with the temperature degree of freedom included in a
separate segregated group from the carrier densities and the potential.

To add a segregated solver, click the study and from the Study toolbar,
click Show Default Solver ( ) (this is not necessary if the Solver
Configurations node is already visible in the Model Builder). Expand the
Solver Configurations node until you can see either the Stationary Solver 1
or the Time-Dependent Solver 1 node (depending on whether the study is
stationary or time-dependent). Right-click the node and select Segregated.
Expand the resulting Segregated 1 node and select the Segregated Step
node. Then in the General settings section, delete the Temperature
(mod1.T) variable from the list. Right-click the Segregated 1 node and add
a Segregated Step. In the Settings window for Segregated Step 1, add
Temperature (mod1.T) to the list.

The node numbers given above can be altered if multiple studies or study
steps are present in the model.

In the COMSOL Multiphysics Reference Manual:

• Introduction to Solvers and Studies

• Segregated and Segregated Step

NONISOTHERMAL MODELS | 47
 SPICE Import and Export
SPICE Import
The circuit definition in COMSOL Multiphysics adheres to the SPICE format
developed at the University of California, Berkeley (Ref. 1). SPICE netlists can be
imported and the corresponding circuit nodes are generated in the COMSOL
Multiphysics model. Most circuit simulators can export to this format or some version
of it.

The Electrical Circuit interface supports the following device models:

TABLE 2-2: SUPPORTED SPICE DEVICE MODELS

STATEMENT DEVICE MODEL

R Resistor
C Capacitor
L Inductor
V Voltage Source
I Current Source
E Voltage-Controlled Voltage Source
F Current-Controlled Current Source
G Voltage-Controlled Current Source
H Current-Controlled Voltage Source
D Diode
Q NPN BJT and PNP BJT
M n-Channel MOSFET and p-Channel MOSFET
X Subcircuit Instance

Statements corresponding to multiple devices are resolved by parsing the associated

.model statement. The physics interface also supports the .subckt statement, which
is represented in COMSOL by a Subcircuit Definition node, and the .include
statement. SPICE commands are interpreted case-insensitively. The statement defining
each device is also interpreted as the Device name.

According to SPICE specification, the first line in the netlist file is assumed to be the
title of the netlist and it is ignored by the parser.

48 | CHAPTER 2: MODELING GUIDELINES

SPICE Export
The SPICE Export functionality creates a SPICE netlist file containing a description of
the circuit represented by the physics interface. This functionality can be accessed from
the physics interface context menu (right-click the physics interface node and select
Export SPICE Netlist). After specifying a file name, the circuit is exported and messages
from the export process display in the Messages window. During the export process, a
series of operations are performed:

• In order to avoid conflicts, each component must be identified by a unique Device

name. If one or more components have the same device name, the export operation
fails and an error message is displayed. All characters in a Device name that are not
letters, digits or underscores are replaced by underscores.
• According to the SPICE specification, each circuit must have a node with name 0,
which is assumed to be the only ground node. When exporting a circuit, any node
with name 0 that is not connected to a Ground component is exported with a
different node name. All nodes that are connected to a Ground components are
exported as a merged node with name 0. The Messages window shows a log message
if these operations are performed, showing the name of the renamed or merged
nodes.
• All characters in node names that are not letters, digits or underscores are replaced
by underscores.
• Some components (most notably, the External components used to couple to other
physics interfaces) cannot be exported to a SPICE netlist. These components are
ignored during the export process, and a message is shown in the Messages window.
Note that this can change the exported circuit, since some components are then
missing.
• Subcircuit definitions are added as .subckt statements in the netlist.
Semiconductor devices (such as MOSFETs, BJTs, and diodes) are exported as a
SPICE device with a corresponding .model statement.

The title of the exported netlist file is the model’s filename, and the time, date, and
version of COMSOL Multiphysics is added as a comment in the netlist file.

Reference
1. http://bwrc.eecs.berkeley.edu/Classes/IcBook/SPICE/

SPICE IMPORT AND EXPORT | 49

 S m a ll- S i g na l A n al ysi s
COMSOL Multiphysics includes sophisticated and very general tools for small-signal
analysis, which are available in the Semiconductor Module. Small signals are applied to
features such as source, drain, or gate contacts by using the Harmonic Perturbation
subfeature (a subnode in the Model Builder tree structure). This feature should be
used in conjunction with the Small-Signal Analysis, Frequency Domain study. The
Small-Signal Analysis, Frequency Domain study includes Stationary and Frequency
Domain, Perturbation study steps. The Stationary study step is used to compute the
linearization point for the equation system, which is the solution for the DC operating
point in the absence of any AC signals. The Frequency Domain, Perturbation study
step then computes the response of the system to small AC deviations from the
linearization point, correctly accounting for nonlinearities in the equation system.

To specify an AC voltage (with a DC offset) applied, for example, to the Thin Insulator
Gate feature, the DC voltage should be specified as usual by typing it into the Voltage
user input in the Terminal section of the settings for the feature. Next add a Harmonic
Perturbation subnode to the Thin Insulator Gate node.

The Harmonic Perturbation subnode is available from the context menu

(right-click the parent node) or from the Physics toolbar in the Contextual
group.

The amplitude of the AC signal is typed into this Harmonic Perturbation feature. By
default, the solution from a Small-Signal Analysis, Frequency Domain study contains
both the Static (or DC) solution and the Harmonic Perturbation (or AC solution).
When evaluating expressions in result analysis you can choose which part of the
solution is displayed in the Evaluate expression for list (a range of other options are also
available, which are described in the link below).

When evaluating the Harmonic Perturbation it is important that the Compute

differential check box is selected so that COMSOL Multiphysics differentiates the
solution at the linearizion point when evaluating the expression. The solutions to
Harmonic Perturbation studies are in general complex valued, with the argument
representing the phase of the signal. By default, the real part of the solution is plotted
in graphs. Use the imag operator to obtain the imaginary part of a variable or an
expression. To drive the system with several signals at different phases, enter
complex-valued numbers into the Harmonic Perturbation subnode.

50 | CHAPTER 2: MODELING GUIDELINES

Care must be taken when computing quantities such as the
transconductance and output conductance for a MOSFET. For example,
if the drain is a Metal Contact with terminal name 1, then evaluating the
global terminal current semi.I0_1 for the Harmonic Perturbation (with
the Compute differential check box selected) gives the (complex)
perturbation in the drain current, δIdrain. The output conductance is
obtained by taking the absolute value of the perturbation and dividing by
the amplitude of the voltage perturbation applied (for example 0.01[V]).
Thus the output conductance (dIdrain/dVdrain) is obtained by evaluating
the quantity: abs(semi.I0_1)/0.01[V]. It is not possible to evaluate
the output conductance by evaluating the perturbation part of the
expression: abs(semi.I0_1)/abs(semi.V0_1) because COMSOL
Multiphysics linearizes the expression entered, computing the quantity:
δIdrain/Vdrain-(Idrain/Vdrain2)δVdrain.

Alternatively, for complicated expressions, use the lindev operator, for

example: lindev(semi.I0_2)/lindev(semi.V0_2). In this case, the
Compute differential check box must be cleared.

In the COMSOL Multiphysics Reference Manual:

• For details on the small-signal analysis study type see Frequency

Domain Perturbation Study Step in the Studies and Solvers section.
• For details on postprocessing small signals see Results Analysis and
Plots

SMALL-SIGNAL ANALYSIS | 51
 M e s hi ng
The mesh is an important component of any numerical model of a semiconductor
device. When assessing a numerical solution it is always important to ensure that the
results do not change significantly when the mesh is refined; this is sometimes referred
to as grid independence. Different meshing strategies produce optimum results with
the finite element and finite volume formulations of the semiconductor equations. The
Semiconductor Module includes default mesh suggestions that automatically refine
the mesh based on the physics features selected. These mesh suggestions are usually
appropriate for both the finite element and the finite volume methods. In some cases
manually tuning the default mesh improves performance and helps with the solution
process.

The finite volume formulation works best in 2D when triangular elements are used. In
3D a swept mesh is currently required (tetrahedral elements are not currently
supported for finite volumes in 3D). In this case it is best to mesh with a triangular
mesh in the plane of the wafer surface and to sweep the mesh into the wafer. This is
particularly important for gate contacts. Figure 2-7 shows some of the techniques that
can be used when meshing gates for the finite volume discretization in 3D.

The finite element method (with either the log or linear formulation) works well with
all the types of mesh elements available in COMSOL. It is also possible to use adaptive
meshing to refine the mesh selectively in regions where the error in the solution is
greatest. Adaptive mesh refinement can be added to a Stationary or Time Dependent
study step by selecting the Adaptation and error estimates drop down menu in the
Adaptation and error estimates section.

For details on the settings for adaptive meshing, see Adaptive Mesh
Refinement in the COMSOL Multiphysics Reference Manual.

For finite elements it is particularly important to assess the dependence of the global
current conservation on the mesh — this can give a good indication of the overall
accuracy of the solution. If current conservation is still poor with a refined mesh
tightening the solver tolerances can help (see Solving).

52 | CHAPTER 2: MODELING GUIDELINES

Figure 2-7: Example of a swept mesh appropriate for a gate contact in 3D. The mesh is
swept in the direction perpendicular to the gate. The source (left), drain (right), and gate
(center) contacts are highlighted in red. The mesh is refined in between the contacts, where
the Gaussian profile drops off. A geometric sequence with an element ratio of 15 is used to
sweep the mesh in the z direction.

For both finite elements and finite volumes, increasing the mesh density in the vicinity
of junctions is important. Note that the Size feature in the mesh sequence can be
applied to boundaries, edges, and points, and if the settings are adjusted appropriately
this feature can be used to produce a local refinement of the mesh.

For more details on setting up a mesh sequence see the Meshing chapter
in the COMSOL Multiphysics Reference Manual.

MESHING | 53
 Solving
The equations solved by the Semiconductor interface are highly nonlinear and are
consequently difficult to solve in the absence of good initial conditions. The default
initial conditions are good approximate guesses for the carrier concentration in the
absence of applied voltages or currents. For stationary models when currents flow it is
often necessary to ramp up the currents or voltages slowly to the desired operating
point within the solver. To achieve this, ensure that the currents and voltages in the
model are controlled by parameters and then use the Auxiliary sweep settings in the
Study Extensions section of the Stationary study step. Then set up the voltage and
current parameters so that these are swept up from zero. Note that you select the
parameter on which the continuation solver should be used in the Run continuation for
drop down menu. When performing multiple sweeps it is desirable to set the Reuse
solution for previous step setting to Auto. Note also that the solution from one study
can be used as the starting point for a second study by altering the settings in the Values
of Dependent Variables section of the Stationary study step. To do this, select the Initial
values of variables solved for check box, select Solution for the Method, and then select
the appropriate solution from the Study list.

In addition to ramping up currents and voltages it is also possible to gradually

introduce contributions to the equation system using the continuation solver. This is
particularly useful when problems are highly nonlinear. When this approach is
appropriate a Continuation Settings section is present in some of the semiconductor
feature settings. By default No continuation is selected, which means that the equation
contribution is as specified. For User defined an input for the Continuation parameter
appears and this should be set to a parameter with values between 0 and 1 to determine
the scaling of the equation contribution. Adding the name of a parameter to this
setting and ramping it from 0 to 1 in the continuation solver will gradually introduce
the contribution into the equation system. By choosing Use interface continuation
parameter the continuation parameter is linked to the value of the interface-level
continuation parameter specified in the Continuation Settings section of the
Semiconductor interface’s Settings window. This enables several features to be ramped
up simultaneously. Finally it is possible to ramp up the dopant concentration, starting
from a small fraction of that specified (or even zero) and gradually increasing the
doping. By default the Continuation Settings for the doping features link to an
interface-level doping continuation parameter that can be set up in the Continuation
Settings section of the Semiconductor interface’s Settings window. In turn this can be

54 | CHAPTER 2: MODELING GUIDELINES

linked to the interface level continuation parameter or specified separately. For the
finite volume method the doping continuation parameter should be started at a small
value (for example, 10−6) rather than zero, as the continuation solver does not handle
the transition from no doping to finite doping when this numerical method is used.
The doping continuation can also be disabled or defined within each individual
feature.

The nanowire_traps tutorial in the Application Libraries

(Semiconductor Module>Transistors>nanowire_traps) provides a good
example on using the continuation solver (through the Auxiliary sweep
and the Continuation parameter as mentioned above) to ramp up a
number of different nonlinear contributions to the model.

The Semiconductor Equilibrium study step can often provide a good initial condition for
subsequent ramping up of applied voltages.

The heterojunction_1d tutorial in the Application Libraries

(Semiconductor Module>Verification_Examples>heterojunction_1d)
provides a set of standard techniques to achieve convergence for
stationary (steady-state) studies.

For time-dependent models it is recommended that the simulation is started in the

manner described previously using a known solution to an appropriate stationary
problem. The time-dependent effects can then be turned on during the simulation.

The pin_forward_recovery and pin_reverse_recovery tutorials in

the Application Libraries (Semiconductor Module>Device Building
Blocks>pin_forward_recovery and pin_reverse_recovery) provide good
examples on using the equilibrium or stationary solution as the initial
condition for the time-dependent study step.

In many cases the time-dependent solver performs better than the continuation solver
for nonlinear stationary problems. By changing the Equation setting in the
Semiconductor interface Settings window from the default (Study controlled) to
Stationary, you can force the physics interface to use the stationary equation form even
though the time dependent solver is used. Then parameters can be defined as functions

SOLVING | 55
 of time, so that the applied currents and voltage, as well as nonlinear equation
contributions and the doping, can be gradually turned on.

Using the correct solver tolerance for a problem is also important. The nonlinear solver
tolerances are set up differently in 1D, 2D, and 3D to achieve a balance between
solution speed and accuracy. For minority carrier devices it might be necessary to
tighten the solver tolerances to ensure that the solution is converged to within
sufficient accuracy. For a Stationary study the solver tolerance is adjusted in the
Stationary Solver node of the Solver Configurations branch (if this branch is not visible
click the study and from the Study toolbar, click Show Default Solver ( )). The Relative
tolerance setting determines the fractional accuracy that the carrier concentrations are
solved to. The Time-Dependent Solver has both relative and absolute tolerance
settings. The relative tolerance setting is available on the Time Dependent study step in
the Study Settings section. The absolute tolerance is set in the Time-Dependent Solver
node on the Solver Configurations branch.

When modeling nonisothermal semiconductor devices, the solver settings must be

adjusted so that the temperature degree of freedom is in a separate segregated group
from the carrier densities and the potential. To do this, click the study and from the
Study toolbar, click Show Default Solver ( ) (this step is not necessary if the Solver
Configurations node is already visible in the model tree). Then expand the Solver
Configurations node until you can see either the Stationary Solver 1 or the
Time-Dependent Solver 1 node (depending on whether the study is stationary or
time-dependent). Right-click this node and select Segregated. Expand the resulting
Segregated 1 node and select the Segregated Step node. Then in the General settings for
this node, delete the Temperature (mod1.T) variable from the list. Next right-click the
Segregated 1 node and select Segregated Step. In the settings for the resulting
Segregated Step 1 node, add Temperature (mod1.T) to the list. Note that the node
numbers given above can be altered if multiple studies or study steps are present in the
model.

For coupled optoelectronic models, it is better to first solve only the wave optics part,
and then solve the fully coupled problem in the following study step.

56 | CHAPTER 2: MODELING GUIDELINES

In the COMSOL Multiphysics Reference Manual:

• Introduction to Solvers and Studies

• Study and Study Step Types
• Computing a Solution
• Solution Operation Nodes and Solvers

SOLVING | 57
 Quantum-Confined Systems
The Schrödinger Equation Interface enables the simulation of quantum-confined
systems such as quantum wells, wires, and dots. The single-particle Schrödinger
Equation is solved for the electron or hole wave function under the assumption of the
envelope function approximation.

The Schrödinger-Poisson Equation Multiphysics Interface creates a bidirectional

coupling between an Electrostatics physics interface and a Schrödinger Equation physics
interface, to model charge carriers in quantum-confined systems. The electric potential
from the Electrostatics contributes to the potential energy term in the Schrödinger
Equation. A statistically weighted sum of the probability densities from the eigenstates
of the Schrödinger Equation contributes to the space charge density in the
Electrostatics. All spatial dimensions (1D, 1D axial symmetry, 2D, 2D axial symmetry,
and 3D) are supported. A dedicated Schrödinger-Poisson Study Step is available for
automated generation of self-consistent iterations in the solver sequence.

For faster engineering investigations of devices affected by quantum confinement, the

density-gradient formulation (under Discretization) of The Semiconductor Interface
can be used.

58 | CHAPTER 2: MODELING GUIDELINES

 3

Semiconductor Branch Interface

This chapter describes the physics interfaces found under the Semiconductor
branch ( ) (except the Schrödinger Equation interface, which is covered in a
separate chapter.)

In this chapter:

• The Semiconductor Interface

• Theory for the Semiconductor Interface
• The Semiconductor Optoelectronics, Beam Envelopes Interface
• The Semiconductor Optoelectronics, Frequency Domain Interface

59
 The Semiconductor Interface
The Semiconductor (semi) interface ( ), found under the Semiconductor branch ( )
when adding a physics interface, solves Poisson’s equation for the electric potential and
the drift-diffusion equations for electrons and holes in a semiconductor material. The
default domain feature is the Semiconductor Material Model, which adds these equations
to the domain, solving for the electric potential and dependent variables related to the
electron and hole concentrations.

When this physics interface is added, these default nodes are also added to the Model
Builder — Semiconductor Material Model, Insulation, Zero Charge, and Initial Values.
Then, from the Physics toolbar, add other nodes that implement, for example,
boundary conditions and Generation-Recombination models. You can also right-click
Semiconductor to select physics features from the context menu.

SETTINGS
The Label is the default physics interface name.

The Name is used primarily as a scope prefix for variables defined by the physics
interface. Refer to such physics interface variables in expressions using the pattern
<name>.<variable_name>. In order to distinguish between variables belonging to
different physics interfaces, the name string must be unique. Only letters, numbers, and
underscores (_) are permitted in the Name field. The first character must be a letter.

The default Name (for the first physics interface in the model) is semi.

THICKNESS (1D/2D MODELS)

Enter the Out-of-plane thickness d (SI unit: m) (2D components), or Cross-section area
A (SI unit: m2) (1D components), or Vertical height d (SI unit: m) (1D axisymmetric
components).

MODEL PROPERTIES
Use Model properties to set the carrier statistics and solution variables in the model.

60 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Carrier Statistics
Select an option from the Carrier statistics list — Maxwell-Boltzmann (the default) or
Fermi-Dirac.

• Maxwell-Boltzmann statistics apply when both the electron and hole quasi-Fermi
levels are within the band gap and at least several kBT away from the band edges.
• Fermi-Dirac statistics are required to simulate degenerate semiconductors.
Fermi-Dirac statistics should be used when one or more of the quasi-Fermi levels
are close to a band edge or even within the band.

Solution
Select an option from the Solution list — Electrons and holes (the default) to solve for
both, or Majority carriers only to solve the drift diffusion equations for only one of the
carriers, computing the concentration of the other carrier by means of the mass action
law: np=ni2. For Majority carriers only also select the Majority carriers — Electrons (the
default) or Holes.

CONTINUATION SETTINGS
Enter the Interface continuation parameter Cp (dimensionless). The default is 1.

Select a Doping and trap density continuation — No continuation (the default), Use
interface continuation parameter, or User defined. For User defined enter a value for the
Doping and trap density continuation parameter Cp (dimensionless). The default is 1.

REFERENCE TEMPERATURE
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options in the Show More Options dialog box. Enter a Reference temperature for
energy levels T0 (SI unit: K). The default is 293.15 K.

The reference temperature is the temperature at which the equilibrium

Fermi level is defined. The equilibrium Fermi level is the Fermi-level of
the material when there are no applied voltages or currents and when the
temperature is uniform throughout the model at the reference
temperature. All potentials within the band diagram (for example, the
electron and hole Fermi levels and the conduction and valence bands) are
measured with respect to the equilibrium Fermi-level (which corresponds
to 0 eV).

THE SEMICONDUCTOR INTERFACE | 61

 DISCRETIZATION
Use this section to change the discretization of the semiconductor equations.

Select a Formulation — Finite volume (constant shape function) (the default), Finite
element, log formulation (linear shape function), Finite element, log formulation
(quadratic shape function), Finite element (linear shape function), Finite element
(quadratic shape function), Finite element quasi Fermi level (quadratic shape function),
Finite element quasi Fermi level (linear shape function), Finite element density-gradient
(quadratic shape function), or Finite element density-gradient (linear shape function).

The Finite volume discretization is the default, and provides better current
conservation in most cases. The Finite element, log formulation uses fewer degrees of
freedom (with a linear shape function) and can be more suitable for certain types of
problem. The two Finite element discretization settings are provided mainly for
backward compatibility and are not recommended for most purposes. The quasi-Fermi
level formulation can be suitable for some systems, such as those with wide band gap
or at very low temperatures. The density-gradient formulation provides a
computationally efficient method to include the effect of quantum confinement in the
conventional drift-diffusion equations.

Also specify the Value type when using splitting of complex variables — Real or Complex
(the default).

STABILIZATION
To display the section: Click the Show More Options button ( ) and select Stabilization
in the Show More Options dialog box. Then under Discretization, select one of these
options to further define this section: Finite element, log formulation (linear shape
function), Finite element, log formulation (quadratic shape function), Finite element
(linear), or Finite element (quadratic).

Streamline diffusion is on by default when the Discretization is set to Finite element.

When Streamline diffusion is active additional contributions to the equation system are
added which improve the numerical stability. This, however, can add unphysical
artificial diffusion to the problem, which can perturb the carrier densities if they take
very small values. While streamline diffusion is usually required for the Finite element
(linear/quadratic) discretizations, the finite element, log formulation (linear/
quadratic shape function) discretizations usually work without it.

62 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

OPTIMAL AVERAGING FOR THE DIFFUSION ENHANCEMENT FACTORS GN
GP
To display the section click the Show More Options button ( ) and select Advanced
Physics Options. This section is only applicable and visible for the Fermi-Dirac statistics
with the Finite volume discretization. By selecting the check box, the computed
current density is most consistent with the limiting case of equilibrium condition (zero
current density).

DEPENDENT VARIABLES
The dependent variable (field variable) is for the Electric potential V, Electron
concentration Ne, and/or Hole concentration Ph depending on the selected Solution
(Electron and holes or Majority carrier only) under Model Properties. For the
quasi-Fermi level formulation and the density-gradient formulation, the dependent
variable for the electrons or holes is their quasi-Fermi level instead of their
concentration. In addition, for the density-gradient formulation, the Slotboom
variables are added to the list of dependent variables.

The name of the dependent variable can be changed but the names of fields and
dependent variables must be unique within a model.

• Discretization and Formulation Options

• Domain, Edge, Boundary, Pair, and Point Nodes for the
Semiconductor Interface
• Theory for the Semiconductor Interface

Domain, Edge, Boundary, Pair, and Point Nodes for the

Semiconductor Interface
The Semiconductor Interface has these domain, edge, boundary, pair, and point nodes
available from the Physics ribbon toolbar (Windows users), Physics context menu (Mac
or Linux users), or by right-clicking the main physics interface node to access the
context menu (all users).

In general, to add a node, go to the Physics toolbar, no matter what

operating system you are using. Subnodes are available by clicking the
parent node and selecting it from the Attributes menu.

THE SEMICONDUCTOR INTERFACE | 63

 In the COMSOL Multiphysics Reference Manual see Table 2-4 for links
to common sections and Table 2-5 to common feature nodes. You can
also search for information: press F1 to open the Help window or Ctrl+F1
to open the Documentation window.

• The Semiconductor Interface

• Theory for the Semiconductor Interface

These nodes and subnodes are described in this section and listed in alphabetical order:

• Analytic Doping Model • Geometric Doping Model

• Analytic Trap Density • Geometric Trap Density
• Arora Mobility Model (LI) • Impact Ionization Generation
• Auger Recombination • Indirect Optical Transitions
• Boundary Selection for Doping Profile • Initial Values
• Boundary Selection for Trap Density • Insulation
Profile • Insulator Interface
• Caughey-Thomas Mobility Model (E) • Lombardi Surface Mobility Model (S)
• Continuity/Heterojunction • Metal Contact
• Continuous Energy Levels • Optical Transitions
(Boundaries)
• Power Law Mobility Model (L)
• Continuous Energy Levels (Domains)
• Semiconductor Material Model
• Continuous Energy Levels
• Thin Insulator Gate
(Heterointerfaces)
• Trap-Assisted Recombination
• Direct Recombination
• Trap-Assisted Heterointerface
• Discrete Energy Level (Boundaries)
Recombination
• Discrete Energy Level (Domains)
• Trap-Assisted Surface Recombination
• Discrete Energy Level
• User-Defined Generation
(Heterointerfaces)
• User-Defined Mobility Model
• Fletcher Mobility Model (C)
• User-Defined Recombination
• Floating Gate
• WKB Tunneling Model, Electrons
• WKB Tunneling Model, Holes

64 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

These nodes and subnodes are described for The Electrostatics Interface:

• Change Cross Section • Line Charge

• Change Thickness (Out-of-Plane) • Line Charge (on Axis)
• Charge Conservation • Line Charge (Out-of-Plane)
• Distributed Capacitance • Periodic Condition
• Electric Displacement Field • Point Charge
• Electric Potential • Point Charge (on Axis)
• External Surface Charge • Space Charge Density
Accumulation • Surface Charge Density
• Floating Potential • Terminal
• Ground • Zero Charge

Semiconductor Material Model

Use the Semiconductor Material Model to set the physical and transport properties of the
semiconducting material. The following subnodes are available from the context menu
(right-click the parent node) or from the Physics toolbar, Attributes menu.

• User-Defined Mobility Model

• Power Law Mobility Model (L)
• Arora Mobility Model (LI)
• Fletcher Mobility Model (C)
• Caughey-Thomas Mobility Model (E)
• Lombardi Surface Mobility Model (S)

See the theory in The Semiclassical Model.

MATERIAL PROPERTIES
For each of the following, the default is taken From material. For User defined enter a
value or expression in the text field.

• Relative permittivity εr (dimensionless). For User defined, also select Isotropic,

Diagonal, Symmetric, or Full.

THE SEMICONDUCTOR INTERFACE | 65

 • Band gap Eg,0 (SI unit: V). The band gap is the energy difference between the
conduction and valence band at equilibrium (independent of doping). The input is
actually the band gap energy divided by the elementary charge. Therefore it takes
on the unit of electric potential, and its numerical value is the same as the band gap
energy in eV. The default is 1.12 V.
• Electron affinity χ0 (SI unit: V). The electron affinity is the difference in energy
between the vacuum level and the conduction band at equilibrium. The input is
actually the electron affinity divided by the elementary charge. Therefore it takes on
the unit of electric potential, and its numerical value is the same as the electron
affinity in eV. The default is 4.05 V.
• Effective density of states, valence band Nv (SI unit: 1/m3). The default is
2.66·1019 1/cm3.
• Effective density of states, conduction band Nc (SI unit: 1/m3). The default is
2.86·1019 1/cm3.

MATERIAL PROPERTIES, DENSITY-GRADIENT

This section appears only when one of the density-gradient formulation options are
selected under the Discretization section.

• Electron effective mass, density-gradient meDG (SI unit: kg). The default is
me_const.
• Hole effective mass, density-gradient mhDG (SI unit: kg). The default is me_const.

SEMICONDUCTOR EQUILIBRIUM STUDY SETTINGS

To display this section, which only applies for the Semiconductor Equilibrium study step,
click the Show More Options button ( ) and select Advanced Physics Options.

The default Equilibrium condition is Bias voltage, suitable for situations where there is a
well-defined bias electric potential for the domain, for example, given by a metal
contact. Enter a Bias voltage V0,bias (SI unit: V). The default is 0 V. All metal contacts
attached to the domain will be set to the specified value of V0,bias for the Semiconductor
Equilibrium study step.

The other option for Equilibrium condition is Total charge, suitable for situations where
there is a well-defined total charge for the domain, for example, if the domain is
insulated on all its exterior boundaries by dielectric material and/or thin insulator gate.
Enter a Total charge Q0,tot (SI unit: C). The default is 0 C. In this case, the electric
potential of the system still needs to be constrained, for example, by a Terminal or an
Electric Potential boundary condition attached to a dielectric domain, or a Thin

66 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Insulator Gate boundary condition attached to the Semiconductor Material Model
domain. If there are also metal contacts attached to the semiconductor domain, then
the bias voltage on those contacts will be set to a value that is consistent with the
specified total charge Q0,tot.

If there are two separate Semiconductor Material Model nodes, one assigned
to each of two adjacent domains, then care must be taken when specifying
the above settings, to ensure that the two domains are at the same
equilibrium electric potential, for example, set the Bias voltage of the two
domains to the same value.

MOBILITY MODEL
These settings determine the mobility values actually used by the Semiconductor
interface. The defaults take constant values From material. For User defined manually
define the electron and hole mobilities.

• Electron mobility μn (SI unit: m2/(V⋅s)).

• Hole mobility μp (SI unit: m2/(V⋅s)).

If mobility model subfeatures are added to the material model it is possible to select
the output of these mobility models to be used as the mobility by changing the feature
input from the appropriate model.

See the theory in Mobility Models.

BAND GAP NARROWING

Band gap narrowing effects occur at high doping levels. To apply band gap narrowing,
select an option from the Band gap narrowing list — None (the default), Slotboom,
Jain-Roulston model, or User defined. Enter values or expressions for the following:

The Band Gap Narrowing theory section also describes these options.

Slotboom
This option applies Slotboom’s empirical model for the band gap narrowing. The
model computes the band gap narrowing according to the equation:

THE SEMICONDUCTOR INTERFACE | 67

  NI NI 2
1
ΔE g = E ref  ln  ----------- + ln  ----------- + ---
 N ref N ref 2

where NI = Nd + Na and the other parameters are material properties. The fraction of
the band gap narrowing taken up by the conduction band is defined directly as a
material property by default.

For each of the following model properties the default takes values From material, or
for User defined enter a different value or expression. There are additional options also
available for each.

• Conduction band fraction α (dimensionless). The default is 0.5.

• Band gap narrowing reference energy Eref (SI unit: V). The default is 0.00692 V.
• Band gap narrowing reference concentration Nref (SI unit: 1/m3). The default is
1.3·1017 1/cm3.

Jain-Roulston Model
This option applies the physics-based model of Jain and Roulston for the band gap
narrowing. Coefficients for the model are available for a wide range of III-V materials,
as well as for silicon and germanium. The band gap narrowing is given by:

 1
---
1
---
1
---
  Nd  3  Nd  4  Nd  2
 A n  ----------- + B n  ----------- + C n  ----------- Nd ≥ Na
 N ref N ref N ref
ΔE g = 
 1
---
1
---
1
---
  Na  3  N a 
4
 N a 
2
----------- -----------
 A p  N  + Bp  N  + Cp  N  ----------- Nd < Na
 ref ref ref

where An, Ap, Bn, Bp, Cn, and Cp are material properties (with the same units as the
band gap) and Nref is a reference doping level (SI unit: 1/m3). The fraction of the
band gap narrowing taken up by the conduction band is defined directly as a material
property by default.

For each of the following properties the default takes values From material, or for User
defined enter a different value or expression. There are additional options available for
each.

• Conduction band fraction α (dimensionless). The default is 0.5.

• Jain-Roulston coefficient (n-type), A (An) (SI unit: V). The default is 10e-9 V.
• Jain-Roulston coefficient (n-type), B (Bn) (SI unit: V). The default is 3e-7 V.

68 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

• Jain-Roulston coefficient (n-type), C (Cn) (SI unit: V). The default is 10e-12 V.
• Jain-Roulston coefficient (p-type), A (Ap) (SI unit: V). The default is 10e-9 V.
• Jain-Roulston coefficient (p-type), B (Bp) (SI unit: V). The default is 3e-7 V.
• Jain-Roulston coefficient (p-type), C (Cp) (SI unit: V). The default is 10e-12 V.
• Band gap narrowing reference concentration Nref (SI unit: 1/m3). The default is
1.3·1017 1/cm3.

User Defined
Enter a value or expression for Band gap narrowing voltage ΔEg (SI unit: V). The default
is 0 V, which represents the amount of band narrowing. Usually a function of the
dopant concentrations should be specified.

Enter a value or expression for Conduction band fraction α (dimensionless). In general,

the change in the band-gap energy can be taken up by a decrease in the conduction
band energy and an increase in the valence band energy. The parameter α determines
what fraction of the change is taken up by the conduction band. When band-gap
narrowing is active Ec→Ec−αΔEg and Ec→Ec+(1−α)ΔEg. The default value of α is 1.

DOPANT IONIZATION
This section determines the ionization of the donors and acceptors. Specify the Dopant
ionization — Complete ionization (the default) or Incomplete ionization. For Incomplete
ionization, select an Ionization model — Standard (the default) or User defined.

For Standard enter values or expressions for the following:

• Relative donor energy (below conduction band) ΔEd (SI unit: V). The default is
0.05 V.
• Relative acceptor energy (above valence band) ΔEd (SI unit: V). The default is 0.05 V.
• Donor degeneracy factor gd (dimensionless). The default is 2.
• Acceptor degeneracy factor gd (dimensionless). The default is 4.

For User defined enter values or expressions for the following:

• Donor ionization fraction Nd+/Nd (dimensionless). The default is 1.

• Acceptor ionization fraction Na–/Na (dimensionless). The default is 1.

Incomplete ionization cannot be used together with the mobility models.

THE SEMICONDUCTOR INTERFACE | 69

 See the theory in Incomplete Ionization.

Optical Transitions
The Optical Transitions feature adds stimulated and spontaneous emission generation/
recombination rates to the semiconductor and computes the corresponding changes
in the complex refractive index or relative permittivity.

OPTICAL TRANSITIONS
Select a Frequency domain representation — Extra dimension (the default) or No extra
dimension. This setting determines whether an extra dimension is added to represent
the frequency domain. The use of an extra dimension typically makes the problem
faster to solve, but it means that additional memory is required. An extra dimension
also allows quantities such as the stimulated emission power to be visualized as a
function of frequency.

Select a Transitions model — Direct band gap model (the default) or User defined.

Direct Band Gap Model

The Direct band gap model uses a two-band model to represent a direct band gap
semiconductor, as described in Optical Transitions in the Theory for the
Semiconductor Interface.

Select the Spontaneous emission check box to compute the recombination due to
spontaneous emission. Select the Stimulated absorption and emission check box to add
the stimulated emission generation or recombination terms. Both check boxes are
selected by default.

User Defined
The User defined transitions model allows arbitrary specification of the recombination
and generation rates due to spontaneous and stimulated emission, and can also be used
to compute the Kramers-Kronig integral for the stimulated emission contribution to
the change in the real part of the dielectric constant or the refractive index.

For this model enter the:

• Net generation rate from stimulated processes, Gstim (SI unit: 1/(m3·s)).

70 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

• Net generation rate from stimulated processes, reference material, Grefstim (SI unit:
1/(m3·s)). This rate is used when performing the Kramers–Kronig integral to
compute the change in the real part of the dielectric constant or refractive index.
• Spontaneous emission recombination rate, Rspon (SI unit: 1/(m3·s)).

TRANSITION MATRIX ELEMENT

This section is available when the Direct band gap model is selected as the Transitions
model.

Select how to define the Matrix element — From spontaneous lifetime (the default),
Momentum matrix element, Dipole matrix element, or Kane 4-band model.

Based on the options chosen enter:

• The spontaneous lifetime, τspon (SI unit: s). The default is 2 ns.
• The momentum matrix element, |M12av|2, (SI unit: kg2⋅m2/s2). The default is
0 kg2⋅m2/s2.
• The dipole matrix element, |μ12av|2, (SI unit: m2⋅s2⋅A2). The default is 0 m2⋅s2⋅A2.

For the Kane 4-band model enter the Kane 4-band model parameters:

• Electron effective mass: Select From band structure properties (the default, defined in
the corresponding section), From density of states (converted from Nc, which is
defined on the Semiconductor Material Model domain feature), or User defined. For
User defined enter the ratio of the effective mass to the electron mass me*/me
(dimensionless). The default is 0.053.
• Band gap: Select From band structure properties (defined in the corresponding
section), From semiconductor material model (defined in the Semiconductor Material
Model node), or User defined. For User defined enter the band gap Eg (SI unit: V).
The default is 0.341 V.
• Valence band spin orbital splitting Δ (SI unit: V). The default is 0.341 V.

The matrix element can be dependent on the excitation angular frequency

if desired. Any function of the angular frequency can be used, but the
angular frequency should take the form: semi.ot1.omega (for interface
tag semi and feature tag ot1).

THE SEMICONDUCTOR INTERFACE | 71

 BAND STRUCTURE PROPERTIES
This section is available when the Direct band gap model is selected for the Transitions
model.

• Select an Electron effective mass me* — From density of states (the default, which is
converted from Nc, defined on the Semiconductor Material Model domain feature),
or User defined. For User defined enter the ratio of the electron effective mass to the
electron mass me*/me (dimensionless). The default is 0.063.
• Select a Hole effective mass mh* — From density of states (the default, which is
converted from Nv, defined on the Semiconductor Material Model domain feature),
or User defined. For User defined enter the ratio of the hole effective mass to the
electron mass mh*/me (dimensionless). The default is 0.51.
• Select a Band gap Eg — From semiconductor material model (defined in the material
properties section of the Semiconductor Material Model domain feature) or User
defined. For User defined enter the band gap Eg (SI unit: V). The default is 1.424 V.

OPTICAL EXCITATION FREQUENCY

This section is available when the excitation frequency is needed to calculate the
stimulated emission rate or the matrix element and when the feature is not coupled to
an Electromagnetic Waves interface.

Select an Excitation frequency f0 — From study (the default), User-defined frequency, or

User-defined wavelength.

• For User-defined frequency, enter the excitation frequency, f0 (SI unit: Hz). The
default is 344 GHz.
• For User-defined wavelength, enter the wavelength, λ0 (SI unit: m). The default is
870 nm.

OPTICAL INTENSITY
This section is available in the absence of a multiphysics coupling when the Direct band
gap model is selected for the Transitions model and when the Stimulated absorption and
emission check box is selected. It is used to define the intensity of the optical field that
is driving the stimulated emission. Enter the Electric field norm, E0 (SI unit: V/m). The
default is 500 V/m.

OPTICAL MATERIAL PROPERTIES

Select the Electric displacement field model — Refractive index (the default) or Relative
permittivity.

72 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Refractive Index
For Refractive index the default Refractive index n and Refractive index, imaginary part
k take values From material. For User defined select Isotropic (for Diagonal the average
of the diagonal elements are taken).

The time-harmonic sign convention used in COMSOL Multiphysics

requires that a lossy material has a negative imaginary part of the refractive
index.

Relative Permittivity
The default Relative permittivity (real part) ε′ and Relative permittivity (imaginary part)
ε″ take values From material. For User defined select Isotropic (for Diagonal the average
of the diagonal elements are taken).

The time-harmonic sign convention used in COMSOL requires that a

lossy material has a negative imaginary part of the relative permittivity.

OPTICAL OUTPUTS
Select the Compute change in real part of susceptibility check box to compute the
change in the real part of the susceptibility (or permittivity/refractive index) using a
Kramers–Kronig integral.

INTEGRAL PROPERTIES
The settings in this section determine the range and resolution of the numerical
scheme used to compute integrals over the frequency domain.

The Frequency domain integral discretization factor (dimensionless) is an integer related

to the discretization of the of the frequency domain. A higher number indicates a finer
mesh in the extra dimension (if present) or smaller steps in the numerical integration
(if no extra dimension is used). The default is 6.

The Frequency domain integral lower limit specifies the lower limit used in frequency
domain integrals in units of ħω0 (if Stimulated absorption and emission is enabled, or
user defined is chosen — ω0 is the excitation angular frequency) or in units of the band
gap (if only spontaneous emission is chosen). The lower limit should be less than the
band gap. The default is 0.5.

The Frequency domain integral upper limit specifies the upper limit used in frequency
domain integrals in units of ħω0 (if Stimulated absorption and emission is enabled, or

THE SEMICONDUCTOR INTERFACE | 73

 user defined is chosen — ω0 is the excitation angular frequency) or in units of the band
gap (if only spontaneous emission is chosen). The upper limit should be chosen so that
the processes of interest are negligible above it. The default is 3.

The Kramers-Kronig integral pole region width specifies the total width of the pole
region around the excitation frequency in units of ħω0 (if Stimulated absorption and
emission is enabled, or user defined is chosen — ω0 is the excitation angular frequency)
or in units of the band gap (if only spontaneous emission is chosen). It should be
chosen so that the change in the susceptibility or the change in the absorption
coefficient is linear to a good accuracy across the region. The default is 0.01.

Indirect Optical Transitions

The Indirect Optical Transitions feature adds an empirical absorption model for silicon
to enable easy simulation of optical absorption in photovoltaic devices. The model uses
the data from Ref. 44 to obtain the frequency-dependent real and imaginary
components of the refractive index of silicon. Alternately, you can choose to specify a
user-defined absorption via either supplying an absorption coefficient or the imaginary
part of the refractive index — this enables the feature to be used for other materials for
which this data is known. The photogeneration due to absorbed photons is then added
into the model via a term that contributes to the carrier generation rate. If a
multiphysics coupling between the Semiconductor interface and an Electromagnetic
Waves interface is active, the refractive index data from the Indirect Optical Transitions
feature is automatically used by the Electromagnetic Waves interface for calculating
wave propagation. Also, the frequency and magnitude of the electromagnetic wave,
which are required in order to calculate the absorption rate, are obtained automatically
from the Electromagnetic Waves interface. If multiphysics coupling is not used, these
quantities can be specified directly within the Indirect Optical Transitions feature.

TRANSITIONS MODEL
Select a Transitions model — Empirical silicon absorption (Green and Keevers) (the
default) or User-defined absorption.

• Empirical silicon absorption (Green and Keevers): This option automatically obtains
the frequency-dependent real and imaginary components of the refractive index and
computes the absorption and associate carrier generation. This model is most
appropriate for intrinsic silicon at temperatures around 300 K.
• User-defined absorption: This option enables the user to specify the absorption. The
photogeneration rate is then computed and the extra carriers are added to the model

74 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 automatically via an extra generation term. You can either supply the imaginary part
of the refractive index or the absorption coefficient.

SPECIFY ABSORPTION
This section is available when User-defined absorption is selected as the transitions
model.

Select an option from the Specify absorption list — Absorption coefficient (the default)
or Refractive index. Then enter the following as required:

• Imaginary part of refractive index n (dimensionless).

• Absorption coefficient α (SI unit: 1/m).
• Real part of the refractive index k (dimensionless).

OPTICAL EXCITATION FREQUENCY

This section is available if a multiphysics coupling is not used. It enables the frequency
of the electromagnetic wave to be specified. If the Empirical silicon absorption (Green
and Keevers) transition model is selected, this frequency is used to look up the
corresponding refractive index data. For both transition model options the frequency
is used in calculating the photogeneration rate.

Select an Excitation frequency — From study (the default), User-defined frequency, or

User-defined wavelength.

• For User-defined frequency, enter the excitation frequency, f0 (SI unit: Hz). The
default is 375 GHz.
• For User-defined wavelength, enter the wavelength, λ0 (SI unit: m). The default is
800 nm.

OPTICAL INTENSITY
This section is available if a multiphysics coupling is not used. It enables the magnitude
of the electromagnetic wave to be specified.

Enter the Electric field norm, E0 (SI unit: V/m).

Insulation
The Insulation node is the default boundary condition. It specifies the zero normal flux
condition at the selected boundaries, that is, setting the normal component of the
electron current, hole current, and electric displacement field to zero.

THE SEMICONDUCTOR INTERFACE | 75

 DENSITY-GRADIENT
This section appears only when one of the density-gradient formulation options are
selected under the Discretization section.

Formulation
• Zero density gradient as the name suggested.
• Potential barrier adds the quantum confinement effect of an abrupt potential barrier
at the boundary. Enter the carrier effective masses (SI unit: kg) the defaults are
me_const. Enter the Barrier heights (SI unit: V) the defaults are 0.1[V].

The functionality of the Surface traps check box in versions prior to 5.4
has been replaced and expanded by the Trap-Assisted Surface
Recombination boundary condition.

Continuity/Heterojunction
The Continuity/Heterojunction node is added by default to all interior boundaries in a
model. When the material properties are continuous across the boundary the feature
has no effect. If there is a discontinuity in the material properties across the boundary
then the feature includes the equations required to model a heterojunction.

CONTINUATION SETTINGS
These settings are the same as for Fletcher Mobility Model (C).

The Continuation Settings do not scale the equation contributions for the
Continuous Quasi-Fermi level option.

HETEROJUNCTION
Select a Continuity model — Continuous quasi-Fermi levels (the default), Thermionic
emission, or User-defined junction current. The Continuous quasi-Fermi levels model
enforces continuity of both quasi-Fermi levels at the boundary. The Thermionic
emission model adds thermionic emission of the carriers over the boundary to the
model. The User-defined junction current option allows the user to specify the normal
electron and hole current densities across the heterojunction.

76 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

EXTRA CURRENT CONTRIBUTION
This section appears if the Continuity model is set to Thermionic emission (See the
section Heterojunction above).

Select an option for the Extra electron current drop-down menu:

1 No extra current (the default)

2 WKB tunneling model Computes the extra current contribution using the WKB
approximation for tunneling. For this option to work, the user needs to manually
add a WKB Tunneling Model, Electrons subnode to specify the tunneling
parameters.
3 User-defined extra current fraction Computes the extra current contribution using
the user-defined extra current fraction entered into the input field Extra current
fraction, electrons that appears when this option is selected.
Select an option for the Extra hole current drop-down menu, which has a set of options
analogous to the ones for electrons described above. For the option WKB tunneling
model to work, remember to add a WKB Tunneling Model, Holes subnode.

• Solving in the Modeling Guidelines chapter.

• Boundary Conditions and Continuity/Heterojunction

WKB Tunneling Model, Electrons

This subnode is available when the option WKB tunneling model is selected for a
heterojunction or Schottky contact. Add this subnode to compute the contribution to
the electron current density from tunneling, using the WKB approximation.

POTENTIAL BARRIER DOMAIN SELECTION

Select a domain adjacent to the boundary selection of this subnode, where the
potential barrier is located and electrons tunnel across the domain.

When drawing the geometry for the device model, consider adding a thin
layer of domain adjacent to the heterojunction or Schottky contact, where
you plan to use the WKB tunneling feature.

THE SEMICONDUCTOR INTERFACE | 77

 OPPOSITE BOUNDARY SELECTION
Select a boundary on the opposite side of the Potential Barrier Domain Selection from
the boundary selection of this subnode. These two boundary selections are each at the
upstream and downstream of the electric field line. In other words, if an electron
tunnel into the boundary selection of the subnode, it emerges out of the Opposite
Boundary Selection.

WKB TUNNELING MODEL, ELECTRONS

In this section, enter the electron effective mass and one or more coordinates
depending on the space dimension of the model.

• For Effective mass, enter the electron effective mass m (SI unit: kg). The default is
the free electron mass me_const.
• For Field line coordinate, enter an appropriate variable flc (SI unit: m) that is
(approximately) proportional to the arc length along the electric field line across the
potential barrier domain selection. The default is the first coordinate of the global
coordinate system. The variable should be valid within the potential barrier domain
selection and its isosurfaces should be approximately parallel to the equipotential
surfaces in the domain.

In general a Curvilinear Coordinates physics interface can be used to

compute the (approximated) electric field line coordinate, or each of the
boundary coordinates below. See the Application Libraries example
Semiconductor_Module/Verification_Examples/heterojunction_tunneling

• For 2D and 3D components, enter Boundary coordinate bndc (SI unit: m). The
default is the second coordinate of the global coordinate system. See below for more
details.
• For 2D axisymmetry and 3D components, enter Boundary coordinate 2 bndc2 (SI
unit: m). The default is the Z coordinate of the global coordinate system. For this
input and the input for bndc above, enter an appropriate variable that can be used
to form the boundary coordinate system of the boundary selection of the subnode.
The variable should be valid within the potential barrier domain selection and its
isosurfaces should be approximately parallel to the electric field lines in the domain.

For the formulation used in the WKB Tunneling Model subnodes, see the
theory section WKB Tunneling Model.

78 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

WKB Tunneling Model, Holes
Add this subnode to compute the contribution to the hole current density from
tunneling, using the WKB approximation.

The settings of this subnode is analogous to the ones for the subnode WKB Tunneling
Model, Electrons described above.

Initial Values
The Initial Values node provides initial values for the electric potential V, electron
concentration, and the hole concentration that serve as an initial condition for a
transient simulation or as an initial guess for the nonlinear solver in other studies.

INITIAL VALUES
Select an option from the Specify initial values list — Automatic (the default),
Equilibrium (carriers and potential), Equilibrium (carriers), Intrinsic (carriers and
potential), Intrinsic (carriers), or User defined.

The Automatic setting varies based on the Discretization settings for the
physics interface. If Finite volume (constant shape function) is selected,
then equilibrium conditions are used. If any of the Finite element
discretizations are selected, then the intrinsic concentrations are used.

Then depending on the choice, enter a value or expression or keep the defaults as
follows:

• For Equilibrium (carriers) or Intrinsic (carriers) the default Electric potential V

(SI unit: V) is −4 V. The carrier concentrations are automatically set to the
equilibrium or intrinsic values.
• For User defined the default Electric potential V (SI unit: V) is −4 V, the default
Electron concentration N (SI unit: 1/m3) is semi.n_init, and the default Hole
concentration P (SI unit: 1/m3) is semi.p_init. For quasi-Fermi level formulation,
the default user-defined quasi-Fermi levels are 0 V.

User-Defined Mobility Model

The User-Defined Mobility Model subnode is available from the context menu
(right-click the Semiconductor Material Model parent node) or from the Physics

THE SEMICONDUCTOR INTERFACE | 79

 toolbar, Attributes menu. This can be used to define a mobility model using
expressions based on, for example, the temperature and the dopant concentrations.

DOMAIN SELECTION

Usually the mobility model selection should not be changed from that of
the parent Semiconductor Material Model.

USER-DEFINED MOBILITY MODEL

Enter a value in the Electron input mobility μn,in (SI unit: m2/(V⋅s)) text field. This
parameter is used to define an input mobility that can be edited in the Electron
mobility text field. The default Electron input mobility is for silicon at equilibrium, that
is, 1448 cm2/(V⋅s). Any electron input mobility from other mobility models added to
the Semiconductor Material Model can be chosen from the list.

Enter a value in the Hole input mobility μp,in (SI unit: m2/(V⋅s)) text field. This
parameter is used to define an input mobility that can be edited in the Hole mobility
field. The default Hole input mobility is for silicon at equilibrium, that is, 473 cm2/
(V⋅s). Any hole input mobility from other mobility models added to the
Semiconductor Material Model can be chosen from the list.

The Electron mobility μ0,n (SI unit: m2/(V⋅s)) field defines the output electron
mobility. The default value is taken from Electron input mobility. Any valid
mathematical expression can be used to modify the default value.

The Hole mobility μ0,p (SI unit: m2/(V⋅s)) field defines the output hole mobility. The
default value is taken from Hole input mobility. Any valid mathematical expression can
be used to modify the default value.

Mobility Models and Theory for the User-defined Mobility Model

Power Law Mobility Model (L)

The Power Law Mobility Model (L) subnode is available from the context menu
(right-click the Semiconductor Material Model parent node) or from the Physics
toolbar, Attributes menu. This model incorporates the effect of lattice or phonon
scattering into the electron and hole mobilities for the material (the ‘L’ given in

80 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

brackets indicates lattice scattering). The equations for the mobility model are shown
in the Equation section.

DOMAIN SELECTION

Usually the mobility model selection should not be changed from that of
the parent Semiconductor Material Model.

GENERAL PARAMETERS
The default Reference temperature Tref (SI unit: K) is taken From material. For User
defined enter a value in the text field. The default is 300 K.

ELECTRON PARAMETERS
For each of the following properties the default takes values From material or for User
defined enter a different value or expression in the text field.

• Electron mobility reference μn,0 (SI unit: m2/(V⋅s)). The default value is for silicon
(1448 cm2/(V-s)). This value represents the zero field mobility at equilibrium.
• Electron exponent αn (dimensionless). The default for silicon is 2.33.

HOLE PARAMETERS
For each of the following properties the default takes values From material or for User
defined enter a different value or expression in the text field.

• Hole mobility reference μp,0 (SI unit: m2/(V⋅s)). The default value is for silicon
(473 cm2/(V⋅s)). This value represents the zero field mobility at equilibrium.
• Hole exponent αp (dimensionless). The default value is for silicon (2.23).

Mobility Models and Theory for the Power Law Mobility Model (L)

Fletcher Mobility Model (C)

The Fletcher Mobility Model (C) subnode is available from the context menu (right-click
the Semiconductor Material Model parent node) or from the Physics toolbar,
Attributes menu. Use this to add carrier-carrier scattering to an existing mobility model
(or to a constant input mobility). The ‘C’ in brackets indicates the model includes
carrier-carrier scattering only. It accepts input mobilities of type L or LI, as well as

THE SEMICONDUCTOR INTERFACE | 81

 user-defined input mobilities. The set of equations defining the Fletcher mobility
model is shown in the Equation section.

DOMAIN SELECTION

Usually the mobility model selection should not be changed from that of
the parent Semiconductor Material Model.

CONTINUATION SETTINGS
The continuation settings enable the equation contributions for the feature to be
gradually introduced into the model.

The Continuation type defaults to No continuation, which means that the equation
contribution is added in the usual way.

The continuity feature allows the equation contribution to be weighted by an external

parameter: the value should be between 0 and 1. Then you can sweep this parameter
with the continuity solver.

Selecting User defined allows the parameter to be specified within the feature or select
Use interface level continuation parameter to specify the parameter on the physics
interface level.

• For User defined enter a Continuation parameter Cp (dimensionless) between 0 and

1 to determine the scaling of the equation contribution. Generally the parameter is
ramped from 0 to 1 in a continuation study.
• For Use interface continuation parameter, the continuation parameter is linked to the
value of the continuation parameter specified in the Continuation Settings section
for The Semiconductor Interface. This enables several features to be introduced
simultaneously into the equation system with a single setting controlling it all.

Solving

INPUT MOBILITIES
The Electron input mobility μn,in (SI unit: m2/(V⋅s)) parameter is used to define an
input mobility to which carrier-carrier scattering is added. The default value is User
defined and is a room temperature mobility for silicon, that is, 1448 cm2/(V⋅s).

82 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

The Hole input mobility μp,in (SI unit: m2/(V⋅s)) parameter is used to define an input
mobility to which carrier-carrier scattering is added. The default value is User defined,
and is for silicon at equilibrium, that is, 473 cm2/(V⋅s).

Other mobility models of type L (Power Law Mobility Model (L)) or LI

(Arora Mobility Model (LI) as well as User-Defined Mobility Model
input mobilities can be used as the Electron input mobility and Hole input
mobility.

GENERAL PARAMETERS
For each of the following properties the default takes values From material or for User
defined enter a different value or expression in the text field.

• Reference temperature Tref (SI unit: K). The default is 300 K.

• Fletcher mobility coefficient1 F1 (SI unit: 1/(m.V⋅s)). The default is for silicon
(1.04e21 1/(cm-V⋅s)).
• Fletcher mobility coefficient2 F2 (SI unit: 1/(m2)). The default is for silicon
(7.45e12 1/cm-2).

Mobility Models and Theory for the Fletcher Mobility Model (C)

Caughey-Thomas Mobility Model (E)

The Caughey-Thomas Mobility Model (E) subnode is available from the context menu
(right-click the Semiconductor Material Model parent node) or from the Physics
toolbar, Attributes menu. It adds a high field correction to an existing mobility model.
The ‘E’ in brackets indicates the model includes high field effects only. It accepts input
mobilities of type L, I, C, or S as well as user-defined input mobilities. The set of
equations defining the Caughey-Thomas mobility model is shown in the Equation
section.

DOMAIN SELECTION

Usually the mobility model selection should not be changed from that of
the parent Semiconductor Material Model.

THE SEMICONDUCTOR INTERFACE | 83

 CONTINUATION SETTINGS
These settings are the same as for Fletcher Mobility Model (C).

The continuation solver should be used in most cases with the

Caughey-Thomas mobility model.

Solving

INPUT MOBILITIES
The Electron input mobility μn,in (SI unit: m2/(V⋅s)) parameter is used to define the
input mobility to which high field velocity saturation is added. The default is User
defined for silicon at equilibrium, 1448 cm2/(V⋅s).

The Hole input mobility μp,in (SI unit: m2/(V⋅s)) defines the input mobility to which
high field velocity saturation is added. The default is User defined for silicon at
equilibrium, 473 cm2/(V⋅s).

Any other Electron input mobility and Hole input mobility from mobility
models of the Semiconductor Material Model can be chosen.

GENERAL PARAMETERS
The default Reference temperature Tref (SI unit: K) is taken From material. For User
defined enter a different value in the text field. The default is 300 K.

ELECTRON PARAMETERS
The default for each of the following is taken From material. For User defined enter a
different value in the text field.

• Electron alpha coefficient αn,0 (dimensionless). The default is 1.11.

• Electron saturation velocity vn,0 (SI unit: m/s). The default is 1·107 cm/s.
• Electron alpha exponent βn,1 (dimensionless). The default is 0.66.
• Electron velocity saturation exponent βn,2 (dimensionless). The default is −0.87.

HOLE PARAMETERS
• Hole alpha coefficient αp,0 (dimensionless). The default is 1.21.

84 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

• Hole saturation velocity vp,0 (SI unit: m/s). The default is 8.37·106 cm/s.
• Hole alpha exponent βp,1 (dimensionless). The default is 0.17.
• Hole velocity saturation exponent βp,2 (dimensionless). The default is −0.52.

Mobility Models and Theory for the Caughey-Thomas Mobility Model

(E)

Lombardi Surface Mobility Model (S)

The Lombardi Surface Mobility Model (S) subnode is available from the context menu
(right-click the Semiconductor Material Model parent node) or from the Physics
toolbar, Attributes menu. It changes an input mobility to account for surface effects,
particularly surface roughness and scattering due to acoustic phonons. The ‘S’ in
brackets indicates that the model includes surface effects only. The mobility model
accepts input mobilities of type L, LI, or C as well as user-defined input mobilities.

The set of equations defining the Lombardi Surface mobility model is shown in the
Equation section.

DOMAIN SELECTION

Usually the mobility model selection should not be changed from that of
the parent Semiconductor Material Model.

CONTINUATION SETTINGS
These settings are the same as for Fletcher Mobility Model (C).

The continuation solver should be used in most cases with the Lombardi
Surface mobility model.

Solving

THE SEMICONDUCTOR INTERFACE | 85

 Lombardi Surface Mobility shows how to use the continuity solver and
how to set up the solver in the most efficient manner when using this
mobility model. Application Library path: Semiconductor_Module/Devices/
lombardi_surface_mobility

INPUT MOBILITIES
The Electron input mobility μn,in (SI unit: m2/(V⋅s)) parameter is used to define an
input mobility. The default is User defined for silicon at equilibrium, 1448 cm2/(V⋅s).

The Hole input mobility μp,in (SI unit: m2/(V⋅s)) defines an input mobility. The default
is User defined for silicon at equilibrium 473 cm2/(V⋅s).

Other mobility models of type L (Power Law Mobility Model (L)), LI

(Arora Mobility Model (LI), C Fletcher Mobility Model (C), or
User-Defined Mobility Model input mobilities can be used as the Electron
input mobility and Hole input mobility.

GENERAL PARAMETERS
The default for each of the following is taken From material. For User defined enter a
different value or expression in the text field.

• Reference temperature Tref (SI unit: K). The default is 1 K.

• Electric field reference Eref (SI unit: V/m). The default is 1 V/cm.
• Doping concentration reference Nref (SI unit: 1/m3). The default is 1 1/cm3.

ELECTRON PARAMETERS
The default for each of the following is taken From material. For User defined enter a
different value or expression in the text field.

• Electron delta coefficient δn(SI unit: V/s). The default is 5.82·1014 V/s.
• Electron mobility reference μn,1(SI unit: m2/(V⋅s)). The default is 4.75·107 cm2/
(V⋅s).
• Electron mobility reference μn,2(SI unit: m2/(V⋅s)). The default is 1.74·105 cm2/
(V⋅s).
• Electron alpha coefficient αn(SI unit: 1). The default is 0.125.

HOLE PARAMETERS
• Hole delta coefficient δp (SI unit: V/s). The default is 2.05·1014 V/s.

86 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

• Hole mobility reference μp,1(SI unit: m2/(V⋅s)). The default is 9.93·107 cm2/(V⋅s)
• Hole mobility reference μp,2(SI unit: m2/(V⋅s)). The default is 8.84·105 cm2/(V⋅s)
• Hole alpha coefficient αp(SI unit: 1). The default is 0.0317.

ADVANCED
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options. Enter a Current density activation threshold Jtol (SI unit: A/m2). The
default is 1·106 A/m2.

Mobility Models and Theory for the Lombardi Surface Mobility Model
(S)

Arora Mobility Model (LI)

The Arora Mobility Model (LI) subnode is available from the context menu (right-click
the Semiconductor Material Model parent node) or from the Physics toolbar,
Attributes menu. Use it to simulate the effect of phonon/lattice and impurity
scattering on the electron and hole mobilities in the semiconducting material (the ‘L’
in brackets indicates that the model includes lattice scattering, and the ‘I’ indicates that
the model includes impurity scattering). The Arora mobility model is empirical and
involves many parameters. The set of equations defining the model is shown in the
Equation section.

DOMAIN SELECTION

Usually the mobility model selection should not be changed from that of
the parent Semiconductor Material Model.

GENERAL PARAMETERS
For each of the following properties the default takes values From material, or for
User defined enter a different value or expression in the text field.

• Reference temperature Tref (SI unit: K). The default is 300 K.

• Alpha coefficient α0 (dimensionless). The default is for silicon 0.88.
• Mobility reference minimum exponent β1 (dimensionless). The default is for
silicon −0.57.

THE SEMICONDUCTOR INTERFACE | 87

 • Mobility reference exponent β2 (dimensionless). The default is for silicon −2.33.
• Impurity concentration reference exponent β3 (dimensionless). The default is for
silicon 2.4.
• Alpha coefficient exponent β4 (dimensionless). The default is for silicon −0.146.

ELECTRON PARAMETERS
For each of the following properties the default takes values From material, or for
User defined enter a different value or expression in the text field.

• Electron mobility reference μn,0ref (SI unit: m2/(V⋅s)). The default value is for
silicon (1252 cm2/(V⋅s)).
• Electron mobility minimum reference μn,refmin (SI unit: m2/(V⋅s)). The default value
is for silicon (88 cm2/(V⋅s)).
• Electron reference impurity concentration Nn,ref0 (SI unit: 1/m3). The default value
is for silicon (1.26e17 1/cm3).

HOLE PARAMETERS
For each of the following properties the default takes values From material, or for
User defined enter a different value or expression in the text field.

• Hole mobility reference μp,0ref (SI unit: m2/(V⋅s). The default value is for silicon
(407 cm2/(V⋅s).
• Hole mobility minimum reference μp,refmin (SI unit: m2/(V⋅s)). The default value is
for silicon (54.3 cm2/(V⋅s)).
• Hole reference impurity concentration Np,ref0 (SI unit: 1/m3). The default value is
for silicon (2.35e17 1/cm3).

Mobility Models and Theory for the Arora Mobility Model (LI)

Analytic Doping Model

Use the Analytic Doping Model to set the doping type and concentration within a
Semiconductor Material Model domain. The concentration is defined as a function of
the local coordinate system, making this doping method suitable for profiles that are
convenient to express in terms of the coordinate axes.

88 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

CONTINUATION SETTINGS
In some cases it is desirable to increase the doping from zero or a low value to that
required in the simulation during the solution process. This can improve the
convergence of the simulation significantly. The continuity settings allow the dopant
densities to be weighted by an external parameter, whose value should be between 0
and 1.

• By default the Continuation type is set to Use interface continuation parameter. This
means that the continuity settings are determined by the Doping and trap density
continuation in the Continuation Settings section for The Semiconductor Interface.
• To introduce the doping from a specific feature independently of the other features,
select User defined and manually enter the Continuation parameter. This should be
set to a dimensionless parameter that is swept from zero to 1 using the continuation
solver.
• To disable the continuation for a particular doping feature select No continuation for
the Continuation type.

For the finite volume method it is recommended that a zero value of the
parameter is avoided. Instead, start with a small value (for example, 1e-6)
and increase the value of the parameter gradually to 1. The continuation
solver usually does not manage the transition from zero to finite doping
when the finite volume method is used.

Solving

DISTRIBUTION
Choose the type of dopant distribution to apply on the selected domains from the
Distribution list: User defined (the default) to enable any expression to be used to define
the dopant distribution, or Box to create a block-shaped region of uniform doping
away from which the concentration decays to a background level according to a
specified profile. The Profile of the distribution away from the uniform region is
selected from a list of preset functions.

IMPURITY
Choose an Impurity type to set the species of the dopants — Acceptor doping (p-type)
(the default) or Donor doping (n-type). Once the impurity type is selected, enter a value

THE SEMICONDUCTOR INTERFACE | 89

 or expression for the Acceptor concentration NA0 or Donor concentration ND0
(SI unit: 1/m3).

For a User defined Distribution input, you can use any expression in terms of the
coordinate system. This expression then defines the dopant concentration through the
selected domains (if a constant value is entered this simply assigns a uniform
concentration to all selected domains). For a Box distribution, a constant value must
be entered and this sets the concentration within the region of uniform doping. The
default value for both distributions is a constant value of 1·1016 cm3.

UNIFORM REGION
This section is available when Box is selected under Distribution. It is used to define the
location and size of the region of uniform doping.

Set the position of the region using the Base list. Select Corner (the default) or Center.

The Base position r0 (SI unit: m) vector sets the location of the region. When Corner is
selected it describes the coordinate of the lower left front corner of the region, and
when Center is selected it describes the coordinate of the center of the region. The
default value is at the origin of the coordinate system.

It is possible to set the height, width, or depth of the region to zero. In

2D, for example, setting the height of the region to zero results in a line
of constant doping away from which the profile function decays.

The extent of the region is defined with the Width, Depth, and Height inputs. Only the
inputs relevant to the dimension of the component to which the feature belongs are
displayed, for example, in a 2D component only the Width and Depth are available.

• Width W (SI unit: m). This is the extent of the region in the direction parallel to the
local x-axes.
• Depth D (SI unit: m). This is the extent of the region in the direction parallel to the
local y-axes.
• Height H (SI unit: m). This is the extent of the region in the direction parallel to the
local z-axes.

The default value is 0 for each dimension.

90 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

PROFILE
This section is available when Box is selected under Distribution. Select a Profile away
from uniform region — Gaussian (the default), Linear, or Error function (erf).

Then select an option from the Specify profile length scale list — Junction depth
(available for all profiles), Decay length (available for a Gaussian), Gradient (available for
Linear), or Argument factor (available for Error function (erf)).

By default, the Specify different length scales for each direction check box is not selected.
It is possible to specify a different profile length scale in each coordinate direction
when this check box is selected.

Junction Depth
For Junction depth enter a value for the Junction depth dj (SI unit: m, default value is
1 μm) to specify the location at which the profile concentration drops to equal the
Background doping concentration Nb (SI unit: 1/m3). The default background
concentration is User defined with a doping concentration of 1·1015 1/cm3.
Alternately, the doping distribution from another Analytic Doping Model or
Geometric Doping Model can be used as the background concentration.

Each profile has a profile-specific parameter to directly control the length scale of the
decay.

Decay Length — for Gaussian profile

For Decay length enter a value for the Decay length lj of the Gaussian decay. (SI unit: m,
default value is 1 μm).

Gradient — for Linear profile

For Gradient enter a value for the Gradient Ngrad of the Linear profile (SI unit: 1/m4,
default value is −1·1020 1/cm4).

Argument Factor — for Error function (erf) profile

For Argument factor enter a value for the Argument factor m (SI unit: 1/m, default
value is 1 (1/μm)) to specify the factor by which the coordinate values are multiplied
before being operated on by the erf function.

As when the Junction depth option is used, by default the Specify different length scales
for each direction check box is not selected. Select it to specify a different profile length
scale in each coordinate direction.

THE SEMICONDUCTOR INTERFACE | 91

 • Doped Semiconductors
• Doping and Using the Analytic Doping Model
• Specifying the Dopant Distribution and Theory for the Analytic
Doping Model

Geometric Doping Model

Use the Geometric Doping Model to define a dopant distribution in terms of the distance
from selected boundaries. This is convenient when specifying distributions which are
not easily expressed in terms of the coordinate system. For example, if a profile is
defined in terms of the distance from a surface then it may not need to have its
expression modified if the geometry is altered. This feature also allows any user defined
decay profile away from the selected boundaries, which is suitable for cases where none
of the preset decay functions available in the Analytic Doping Model are appropriate.

A default Boundary Selection for Doping Profile is added where the boundaries from
which the distance is calculated can be specified.

CONTINUATION SETTINGS
These settings are the same as for Analytic Doping Model.

DISTRIBUTION
Select an option from the Profile away from the boundary list — Gaussian (the default),
Linear, Error function (erf), or User defined.

IMPURITY
For Gaussian, Linear, or Error function (erf) profiles, select the Impurity type — Acceptor
doping (the default) or Donor doping. Then specify the maximum doping concentration
of the distribution; this is the concentration of the dopants at the selected boundaries.

• For Acceptor doping specify the Acceptor concentration at boundary NA0 (SI unit: 1/
m3) of the electron acceptor impurities. The default value is 1·1016 1/cm3.
• For Donor doping specify the Donor concentration at boundary ND0 (SI unit: 1/m3)
of the electron donor impurities. The default value is 1·1016 1/cm3.

For User defined profiles select the Dopant type — Acceptor doping (the default) or
Donor doping. Then enter an expression for Define profile as a function of the distance,

92 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

D, from the boundaries N(D) (SI unit 1/m3). The distance from the selected
boundaries is available as variable semi.gdm#.D, which can be used in an expression to
define the doping profile. The default expression gives an example of a Gaussian profile
defined using this variable.

PROFILE
For the Gaussian, Linear, and Error function (erf) profiles it is necessary to specify the
length scale of the profile decay. The Profile section allows the length scale of the profile
to be controlled. As with the Analytic Doping Model, the length scale can be
controlled either by entering a junction depth or via a profile-specific length scale
parameter. See Analytic Doping Model, under Profile for the settings for more details.

• Doped Semiconductors
• Doping and Using the Geometric Doping Model
• Specifying the Dopant Distribution and Theory for the Geometric
Doping Model

Boundary Selection for Doping Profile

A default Boundary Selection for Doping Profile is added to the Geometric Doping
Model. This node enables the boundaries from which the distance is calculated to be
selected. Any internal or external boundary that bounds, or is within, the domains to
which the corresponding Geometric Doping Model is applied can be selected.

• Doped Semiconductors
• Doping
• Specifying the Dopant Distribution

Analytic Trap Density

Use the Analytic Trap Density node to specify the density of traps within the domain.
This feature enables a density to be defined by an expression written in terms of the
local coordinate system. Alternately, it enables a region of uniform trap density which
decays into a background density with one of three preset profile functions to be
created. The trap density can be used in conjunction with a Trap-Assisted

THE SEMICONDUCTOR INTERFACE | 93

 Recombination feature to specify recombination via trapping sites whose density varies
in space. The available options are similar to the Analytic Doping Model feature.

CONTINUATION SETTINGS
In some cases it is desirable to increase the trapping from zero or a low value to that
required in the simulation during the solution process. This can improve the
convergence of the problem significantly. The continuity settings allow the trap
densities to be weighted by an external parameter whose value should be between 0
and 1. By default the Continuation type is set to Use interface continuation parameter.
This means that the continuity settings are determined by the Doping and trap density
continuation in the Continuation Settings section for The Semiconductor Interface. To
introduce the trapping from a specific feature independently of the other features,
select User defined and manually enter the Continuation parameter. This should be set
to a dimensionless parameter that is swept from zero to 1 using the continuation solver.

To disable the continuation for a particular trapping feature select No continuation for
the Continuation type. The effect of the traps can also be ramped on using the
continuation feature within the trap-assisted recombination feature.

Solving

DISTRIBUTION
Select a Distribution — User defined (the default) or Box.

IMPURITY
Select an Impurity type to give the conductivity type of the selected semiconductor
material — Donor traps (the default), Acceptor traps, Neutral electron traps, or Neutral
hole traps.

The Impurity type setting has no effect if a neutral level is used within the
Trap-Assisted Recombination feature(s).

For any selection, enter a Trap density Nt0 (SI unit: 1/m3). The default value is
1·1016 1/cm3.

94 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

UNIFORM REGION
This section is available when Box is selected under Distribution. See Analytic Doping
Model for these settings.

PROFILE
This section is available when Box is selected under Distribution. Select a Profile away
from uniform region — Gaussian (the default), Linear, or Error function (erf).

By default, the Specify different length scales for each direction check box is not selected.
It is possible to specify a different profile length scales in each coordinate direction
when this check box is selected.

Based on the selected profile above, enter one of the following:

• Decay length ld (SI unit: m, default value is 1 μm) (Gaussian profiles).

• Gradient Ngrad(SI unit: 1/m4, default value is −1·1020 1/cm4) (Linear profiles).
• Argument factor m (SI unit: 1/m, default value is 1 (1/μm)) to specify the factor by
which the coordinate values are multiplied before being operated on by the erf()
function (Error function (erf) profiles).

Geometric Trap Density

Use the Geometric Trap Density node to specify the density of traps within the domain.
This feature enables the trap density to be entered as a function of the distance from
selected boundaries. In addition to enabling a fully user-defined trap density to be
entered, three preset density profiles are also available. The trap density can be used in
conjunction with a Trap-Assisted Recombination feature to specify recombination via
trapping sites whose density varies in space. The available options are similar to the
Geometric Doping Model feature.

CONTINUATION SETTINGS
These settings are the same as for Analytic Doping Model.

DISTRIBUTION
Select an option from the Profile away from the boundary list — Gaussian (the default),
Linear, Error function (erf), or User defined.

THE SEMICONDUCTOR INTERFACE | 95

 IMPURITY
Select an Impurity type to give the conductivity type of the selected semiconductor
material — Donor traps (the default), Acceptor traps, Neutral electron traps, or Neutral
hole traps.

The Impurity type setting has no effect if a neutral level is used within the
Trap-Assisted Recombination feature(s).

For Gaussian, Linear, or Error function (erf) distribution profiles, enter a Trap density at
boundary Nt0 (SI unit: 1/m3). The default value is 1·1016 1/cm3.

For a User defined profile, enter a value to Define profile as a function of the distance, D,
from the boundaries N(D) (SI unit 1/m3).

PROFILE
This section is available for Gaussian, Linear, or Error function (erf) distribution profiles.
Based on the profile selection enter one of the following:

• Decay length ld (SI unit: m, default value is 1 μm) (Gaussian profiles).

• Gradient Ngrad(SI unit: 1/m4, default value is −1·1020 1/cm4) (Linear profiles).
• Argument factor m (SI unit: 1/m, default value is 1 (1/μm)) to specify the factor by
which the coordinate values are multiplied before being operated on by the erf()
function (Error function (erf) profiles).

Boundary Selection for Trap Density Profile

A default Boundary Selection for Trap Density Profile is added to the Geometric Trap
Density. This node enables the boundaries from which the distance is calculated to be
selected. Any internal or external boundary that bounds, or is within, the domains to
which the corresponding Geometric Trap Density is applied can be selected.

Trap-Assisted Recombination

This feature is an expanded version of the original Shockley-Read-Hall

Recombination feature, with new options to allow for more detailed
modeling of traps.

96 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Use the Trap-Assisted Recombination model to set the electron and hole recombination
rates in indirect band-gap semiconductors such as silicon under low electric fields.
Select this option from the Generation-Recombination submenu.

Recombination and Generation and Theory for Trap-Assisted

Recombination: Shockley-Read-Hall Recombination

TRAP-ASSISTED RECOMBINATION
Select a Domain trapping model:

• Shockley-Read-Hall model (the default). This option corresponds to the original

Shockley-Read-Hall model for steady state recombination via states located at the
midgap.
• Explicit trap distribution. This option allows the specification of a number of discrete
traps or a continuous density of trap states at energies within the band-gap. Degrees
of freedom are added to the equation system to represent the occupancy of traps at
a particular energy.

When Explicit trap distribution is selected as the Domain trapping model,

the Discrete Energy Level (Domains) subnode is available from the
context menu (right-click the parent node) or from the Physics toolbar,
Attributes menu.

SHOCKLEY-READ-HALL RECOMBINATION
This section is available when Shockley-Read-Hall model is selected as the Domain
trapping model:

The Shockley-Read-Hall recombination rate is defined as:

2
np – n i ,mod
R n = R p = ------------------------------------------------------------
τp ( n + n1 ) + τn ( p + p1 )

with

E g – ΔE g
n i ,mod = γ n γ p N c0 N v0 exp  – ------------------------
 2V th 

THE SEMICONDUCTOR INTERFACE | 97

  E g – ΔE g  ΔE t
n 1 = γ n N c0 N v0 exp  – ------------------------ exp  ----------
 2V th   V th 

 E g – ΔE g  ΔE t
p 1 = γ p γ p N c0 N v0 exp  – ------------------------ exp  ----------
 2V th   V th 

where γn and γp are the electron and hole degeneracy factors, Nc,0 and Nv,0 are the
effective densities of states for the conduction and valence band, Eg is the band gap
and ΔEg the band gap narrowing energies (SI unit: V) — scaled by the electron charge,
q. Vth=kBT/q, where kB is Boltzmann’s constant and T is the temperature. The
parameters τn and τp are carrier lifetimes (SI unit: s) and Et is the trap energy level (SI
unit: V) — scaled by the electron charge.

Enter values for the following:

• The Electron lifetime, SRH τn (SI unit: s) is taken From material. For User defined enter
a value in the text field. The default is 10 μs.
• The Hole lifetime, SRHτp (SI unit: s) is taken From material. For User defined enter a
value in the text field. The default is 10 μs.
• Enter a value for Energy difference between the defect level and the intrinsic level ΔEt
(SI unit: V) in the text field. The default is 0 V.

TRAPPING
This section is available when Explicit trap distribution is selected as the Domain trapping
model.

By default, Specify the discrete levels only is selected. Or select Specify continuous and/
or discrete levels.

When Explicit trap distribution is selected as the Domain trapping model,

and Specify continuous and/or discrete levels is selected above, the
Continuous Energy Levels (Domains) subnode is available from the
context menu (right-click the parent node) or from the Physics toolbar,
Attributes menu.

For either choice:

98 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

• The Specify trap species check box is not selected by default. In this case the software
does not distinguish between different types of trap and the charge due to the traps
is determined by the neutral energy level.
• To set the Neutral energy level, select — Midgap (the default), From valence band edge,
From conduction band edge, or Relative. Then enter the applicable information:
- Neutral level for traps (available when From valence band edge or From conduction
band edge is selected) E0 (SI unit: V). The default is 0 V.
- Neutral level for traps (between 0 and 1, 0.5 is midgap) (available when Relative is
selected) E0 (dimensionless). The default is 0.5.

Discrete Energy Level (Domains)

The Discrete Energy Level subnode is available from the context menu (right-click the
parent node) or from the Physics toolbar, Attributes menu when Explicit trap
distribution is selected as the Domain trapping model for the Trap-Assisted
Recombination node. Use this feature to specify a single discrete energy level for the
traps.

TRAP TYPE
This section is available when the Specify trap species check box is selected for the
parent node. Different trap types have a different charge when they are occupied or
unoccupied. Select a Trap type:

• Donor traps. Neutral when unionized (occupied by electron) and positively charged
when ionized (unoccupied by electron).
• Acceptor traps. Neutral when unionized (occupied by hole) and negatively charged
when ionized (unoccupied by hole; that is, filled by electron).
• Neutral electron traps. Negatively charged when occupied (electron trapped) and
neutral when unoccupied (no electron trapped).
• Neutral hole traps. Positively charged when occupied (hole trapped) and neutral
when unoccupied (no hole trapped).

TRAPS
When there is no Trap type defined, enter a Trap number density Nt (SI unit: 1/m3).
The default is 1·1012 1/cm3.

When the Trap type is defined, enter one of the following:

• Donor trap density Nt,d (SI unit: 1/m3). The default is 1·1012 1/cm3.

THE SEMICONDUCTOR INTERFACE | 99

 • Acceptor trap density Nt,a (SI unit: 1/m3). The default is 1·1012 1/cm3.
• Neutral electron trap density Nt,ne (SI unit: 1/m3). The default is 1·1012 1/cm3.
• Neutral hole trap density Nt,nh (SI unit: 1/m3). The default is 1·1012 1/cm3.

For any option, select an Impurity energy level — Midgap (the default), From valence
band edge, From conduction band edge, or Relative. Then enter the applicable
information:

• Impurity energy level (available when From valence band edge or From conduction
band edge is selected) Et,0 (SI unit: V). The default is 0 V.
• Impurity energy level (between 0 and 1, 0 being the valence band) (available when
Relative is selected) Et,0 (dimensionless). The default is 0.5.

Enter the Ground-state degeneracy factor gd (dimensionless). The default is 1.

Enter the Trap occupancy initial value ft,init (dimensionless). The default is 0.5.

CARRIER CAPTURE
Choose how to define the Probability of electron capture (Cn) — From cross section (the
default) or User defined.

• For From cross section, enter an Averaged cross section for electron capture <σn>
(SI unit: m2; the default is 1·1018 1/cm2) and Electron thermal velocity Vthn
(SI unit: m/s; the default is 1.562·107 m/s).
• For User defined, enter Cn (SI unit: m3/s). The default is 2.042·10-11 cm3/s.

Define the Probability of hole capture (Cp) — From cross section (the default) or User
defined.

• For From cross section, enter an Averaged cross section for hole capture <σp>
(SI unit: m2; the default is 1·1018 1/cm2) and Hole thermal velocity Vthp (SI unit:
m/s; the default is 1.562·107 m/s).
• For User defined enter Cp (SI unit: m3/s). The default is 1.562·10-11 cm3/s.

Continuous Energy Levels (Domains)

The Continuous Energy Level subnode is available from the context menu (right-click
the Trap-Assisted Recombination parent node) or from the Physics toolbar, Attributes
menu when Explicit trap distribution is selected as the Domain trapping model, and
Specify continuous and/or discrete levels is selected under Trapping. Use this node to
define a density of trap states.

100 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

TRAP TYPE
This section is available when the Specify trap species check box is selected for the
parent node. See Discrete Energy Level (Domains).

TRAPS
Select a Trap density distribution — Gaussian (the default), Rectangle, Exponential, or
User defined. Then enter the applicable information:

• Width (for Gaussian distribution) σ (SI unit: V). The default is 0.333 V.
• Width (for rectangle distribution) Ewidth(SI unit: V). The default is 0 V.
• Size of transition zone (for rectangle distribution) ΔEtran(SI unit: V). The default is
0.05 V. This value determines the width of the smoothing employed on the step
function for the rectangle distribution.
• Damping coefficient (for exponential distribution) d (SI unit: V). The default is 0.1 V.
• Density of trap states (for user defined) gt(E) (SI unit: s3⋅A/(m5⋅kg)). The default is
0 s3⋅A/(m5⋅kg).

Enter a Trap number density Nt (SI unit: 1/m3). The default is 1·1012 1/cm3.

When the Trap type is defined, enter one of the following:

• Donor trap density Nt,d (SI unit: 1/m3). The default is 1·1012 1/cm3.
• Acceptor trap density Nt,a (SI unit: 1/m3). The default is 1·1012 1/cm3.
• Neutral electron trap density Nt,ne (SI unit: 1/m3). The default is 1·1012 1/cm3.
• Neutral hole trap density Nt,nh (SI unit: 1/m3). The default is 1·1012 1/cm3.

This quantity represents the total trap density, or the integrated density of states:

E t, max
Nt = E t, min
g t ( E ) dE

where g (E) is the density of trap states, and Et,min and Et,max define the range of the
energy (see user inputs below), with the default being the energy levels corresponding
to the conduction and valence band edges in the absence of band gap narrowing. The
energy range is truncated within the default range so that there is no trap level within
the conduction band or the valence band.

THE SEMICONDUCTOR INTERFACE | 101

 Select a Distribution centerpoint — Midgap (the default), From valence band edge, From
conduction band edge, or Relative. Then enter the applicable information:

• Distribution centerpoint Et,0 (SI unit: V). The default is 0 V.

• Distribution centerpoint (between 0 and 1, 0 being the valence band) Et,0
(dimensionless). The default is 0.5.

Enter the Ground-state degeneracy factor gd (dimensionless). The default is 1.

Enter the Trap occupancy initial value ft,init (dimensionless). The default is 0.5.

Enter the Continuous energy discretization, number of mesh points Nxdelem

(dimensionless). The default is 25.

Enter the Continuous energy discretization, min energy Et,min (SI unit: V). The default
is semi.Ev0, the valence band edge in the absence of band gap narrowing.

Enter the Continuous energy discretization, max energy Et,max (SI unit: V). The default
is semi.Ec0, the conduction band edge in the absence of band gap narrowing.

CARRIER CAPTURE
These settings are the same as for Discrete Energy Level (Domains).

Auger Recombination
Use the Auger Recombination model to set the electron and hole recombination rates
at high nonequilibrium carrier densities in semiconducting materials. Select this option
from the Generation-Recombination submenu.

Recombination and Generation and Theory for Auger Recombination

The Auger recombination rate is defined as:

2
R n = R p = ( C n n + C p p ) ( np – n i ,mod )

with

 E g – ΔE g
n i ,mod = γ n γ p N c0 N v0 exp  – ------------------------
 2V th 

102 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

where γnand γp are the electron and hole degeneracy factors, Nc,0 and Nv,0 are the
effective densities of states for the conduction and valence band, Eg is the band-gap
(SI unit: V), and ΔEg the band gap narrowing (SI unit: V). The parameters Cn and Cp
are material constants (SI unit: m6/s).

AUGER RECOMBINATION
The electron lifetime Auger recombination factor, electrons Cn (SI unit: m6/s) is taken
From material. For User defined, enter a value in the text field, the default value is for
silicon, 2.8·10−31 cm6/s.

The electron lifetime Auger recombination factor, holes Cp (SI unit: m6/s) is taken From
material. For User defined, enter a value in the text field, the default value is for silicon,
9.9·10−32 cm6/s.

Direct Recombination
Use the Direct Recombination to set the recombination rate in direct band-gap
semiconductor materials such as gallium arsenide. Select this option from the
Generation-Recombination submenu.

Recombination and Generation and Theory for the Direct

Recombination Feature

The recombination rate for both electrons and holes is set using the following
equation:

2
R n = R p = C ( np – n i ,mod )

with

 E g – ΔE g
n i ,mod = γ n γ p N c0 N v0 exp  – ------------------------
 2V th 

where γnand γp are the electron and hole degeneracy factors, Nc,0 and Nv,0 are the
effective densities of states for the conduction and valence band, Eg is the band-gap
and ΔEg the band-gap narrowing (SI unit: V).

THE SEMICONDUCTOR INTERFACE | 103

 DIRECT RECOMBINATION
The default Direct recombination factor C (SI unit: m3/s) is taken From material. For
User defined enter a value in the text field. The default value is 0 m3/s.

Impact Ionization Generation

Impact Ionization Generation occurs when, for example, an energetic electron
undergoes a collision in which it loses sufficient energy to promote an electron in the
valence band to the conduction band, resulting in an additional electron hole pair
being produced. It is the mechanism responsible for avalanche breakdown. Select this
option from the Generation-Recombination submenu.

Recombination and Generation and Theory for Impact Ionization

The carrier generation rate due to impact ionization is given by:

II II αn αp
R n = R p = – ------ J n – ------ J p
q q

The values of αn and αp can be user-defined or can be related to the material properties
in the following manner:

b n ( 1 + d n ( T – T ref ) ) 2
α n = a n ( 1 + c n ( T – T ref ) )E || ,n exp  –  ------------------------------------------------------ 
  E || ,n  
b p ( 1 + d p ( T – T ref ) ) 2
α p = a p ( 1 + c p ( T – T ref ) )E || ,p exp  –  ----------------------------------------------------- 
E || ,p

where E || ,n and E || ,p are the components of the electric field parallel to the electron
and hole currents respectively and Tref, an, ap, bn, bp, dn, and dp are material
properties.

IMPACT IONIZATION GENERATION

The default Impact Ionization model is Okuto-Crowell model. For User defined, enter a
desired value or expression in the input field for the values of αn and αp.

For the Okuto-Crowell model, the default Temperature reference Tref is 300 K.

104 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

For the Okuto-Crowell model, the defaults for the following are all taken From material.
For User defined, enter a different value or expression in the text field.

• a factor, electrons, impact ionization an (SI unit: 1/V). The default is 0.426 (1/V).
• a factor, holes, impact ionization ap (SI unit: 1/V). The default is 0.243 (1/V).
• b factor, electrons, impact ionization bn (SI unit: V/m). The default is 4.81·105 V/
cm.
• b factor, holes, impact ionization bp (SI unit: V/m). The default is 6.53·105 V/cm.
• c factor, electrons, impact ionization cn (SI unit: 1/K). The default is 3.05·10-4 (1/
K).
• c factor, holes, impact ionization cp (SI unit: 1/K). The default is 5.35·10-4 (1/K).
• d factor, electrons, impact ionization dn (SI unit: 1/K). The default is 6.86·10-4 (1/
K).
• d factor, holes, impact ionization dp (SI unit: 1/K). The default is 5.67·10-4 (1/K).

User-Defined Recombination
Use the User-Defined Recombination feature to manually set a recombination rate for
electrons and holes in the semiconducting material. Select this option from the
Generation-Recombination submenu.

USER DEFINED RECOMBINATION

Enter the following recombination rates:

• User-defined recombination rate, electrons Rn,0 (SI unit: 1/(m3s)).

• User-defined recombination rate, holes Rp,0 (SI unit: 1/(m3s)).

Recombination and Generation and User-Defined Recombination

User-Defined Generation
Use the User-Defined Generation feature to manually set a generation rate for electrons
and holes in the semiconducting material. Select this option from the
Generation-Recombination submenu.

THE SEMICONDUCTOR INTERFACE | 105

 USER DEFINED GENERATION
Enter the following generation rates:

• User-defined generation rate, electrons Gn,0 (SI unit: 1/(m3s)).

• User-defined generation rate, holes Gp,0 (SI unit: 1/(m3s)).

Recombination and Generation and User-Defined Generation

Metal Contact
The Metal Contact node is used for modeling metal-semiconductor contacts.

The Harmonic Perturbation subnode (it is of the exclusive type) is available from the
context menu (right-click the parent node) or on the Physics toolbar, click the
Attributes menu and select Harmonic Perturbation.

TERMINAL
Specify the terminal properties. To indicate which boundaries belong to the same
terminal, enter the same name in the Terminal name field. The Terminal name should
be numeric for sweeps to work properly.

Select a Terminal type — Voltage (the default), Current, or Circuit. Select:

• Voltage to enter an electric potential V0 (SI unit: V). The default is 0 V.

• Current to enter a current I0 (SI unit: A). The default is zero current, corresponding
to an open circuit.
• Circuit (current) to specify a terminal connected to an external circuit, where the
external circuit acts as a current source for the semiconductor model.
• Circuit (voltage) to specify a terminal connected to an external circuit, where the
external circuit acts as a voltage source for the semiconductor model.
• Power to enter a current P0 (SI unit: W). The default is zero power, corresponding
to an open circuit.

106 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

CONTACT TYPE
Select a Type — Ideal ohmic (the default) or Ideal Schottky.

• Use the Ideal ohmic option for a nonrectifying metal-semiconductor junction, that
is a contact with negligible resistance relative to the total resistance of the modeled
semiconductor device.
• Use the Ideal Schottky option for a simple rectifying metal-semiconductor junction,
that is, when the current-voltage characteristics at the interface depend on the
potential barrier formed at the junction. For this option, surface recombination
effects and surface charge densities associated with surface traps can be included in
the model by adding Trap-Assisted Surface Recombination boundary conditions to
the same boundary selection as the Metal Contact boundary condition.

CONTACT PROPERTIES
This section is available if Ideal Schottky is selected as the contact Type.

Choose between an Ideal (the default) and User defined definition of the Barrier height.

For the Ideal barrier height, the default Metal work function Φ (SI unit: V) is 4.5 V. The
metal work function is the difference in energy between the vacuum level and the
conduction band at equilibrium in the metal in contact with the semiconductor. The
ideal barrier height is given by the difference between the metal work function and the
electron affinity of the semiconductor.

For the User defined barrier height, enter a value for the Barrier height ΦB (SI unit: V).
In this case, the metal work function is not used.

THERMIONIC CURRENTS
This section is available if Ideal Schottky is selected as the contact Type.

Select the Thermionic currents — Richardson’s coefficients (the default) or Surface

recombination velocities.

For Richardson’s coefficients enter an:

• Effective Richardson constant for electrons An*(SI unit: A/(m2⋅K2)). The default
constant is 110 A/(K⋅cm)2 for silicon.
• Effective Richardson constant for holes Ap* (SI unit: A/(m2⋅K2)). The default
constant is 90 A/(K⋅cm)2 for silicon.

THE SEMICONDUCTOR INTERFACE | 107

 For Surface recombination velocities enter a:

• Surface recombination velocity, electrons Vs,n (SI unit: m/s). The default is
21605 m/s for silicon.
• Surface recombination velocity, holes Vs,p (SI unit: m/s). The default is 19006 m/s
for silicon.

EXTRA CURRENT CONTRIBUTION

This section is available if Ideal Schottky is selected as the contact Type.

The settings are the same as described in this section Extra Current Contribution.

CONSTRAINT SETTINGS
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options.

• Boundary Conditions and Metal Contacts

• Small-Signal Analysis
• Harmonic Perturbation, Prestressed Analysis, and Small-Signal
Analysis in the COMSOL Multiphysics Reference Manual.

Thin Insulator Gate

Use the Thin Insulator Gate node to model a thin insulating material between the
semiconductor and a metal. The layer has the thickness dins and the relative
permittivity εins. The indices ins and s refer r to the insulator and semiconductor side
of the boundary, respectively.

The Harmonic Perturbation subnode (it is of the exclusive type) is available from the
context menu (right-click the parent node) or on the Physics toolbar, click the
Attributes menu and select Harmonic Perturbation.

• Small-Signal Analysis
• Harmonic Perturbation — Exclusive and Contributing Nodes in the
COMSOL Multiphysics Reference Manual.

108 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

CONTINUATION SETTINGS
These settings are the same as for Fletcher Mobility Model (C).

Solving

TERMINAL
Specify the terminal properties. To indicate which boundaries belong to the same
terminal, enter the same name in the Terminal name field. The Terminal name should
be numeric for sweeps to work properly.

Select a Terminal type — Voltage (the default), Charge, or Circuit. Select:

• Voltage to enter an electric potential V0 (SI unit: V). The default is 0 V.

• Charge to specify a charge Q0 (SI unit: C). This is the charge on the contact side of
the parallel plate capacitor. The default is 0 V.
• Circuit to specify a terminal connected to an external circuit.

It is important to realize that the Charge setting does not specify an

interface charge at the boundary between the semiconductor and the
insulator — it specifies the charge on the top conductor of the gate
capacitor.

GATE CONTACT
Specify the insulator dimensions and permittivity as well as the metal work function.

• Oxide relative permittivity εins (dimensionless). The default relative permittivity is 1.

• Oxide thickness dins (SI unit: m). The default thickness is 0.1 μm.
• Metal work function Φ (SI unit: V). The default barrier height is 4.1 V. The metal
work function is the difference in energy between the vacuum level and the
conduction band at equilibrium in the metal in contact with the semiconductor

THE SEMICONDUCTOR INTERFACE | 109

 DENSITY-GRADIENT
This section appears only when one of the density-gradient formulation options are
selected under the Discretization section. See the identically named section
Density-Gradient for details.

The functionality of the Surface traps check box in versions prior to 5.4
has been replaced and expanded by the Trap-Assisted Surface
Recombination boundary condition.

• Boundary Conditions and Thin Insulating Gates

• Harmonic Perturbation, Prestressed Analysis, and Small-Signal
Analysis in the COMSOL Multiphysics Reference Manual.

Trap-Assisted Surface Recombination

The Trap-Assisted Surface Recombination boundary condition adds contributions to the
surface recombination rate and the surface charge density associated with the surface
traps for the following boundary conditions: Insulation, Thin Insulator Gate, Insulator
Interface, and Metal Contact (Ideal Schottky type only). This feature replaces and
expands the functionality of the Surface traps check box seen in versions prior to 5.4.
Both the Shockley-Read-Hall model and the Explicit trap distribution options are
available. All the settings are the same as its domain condition counterpart
Trap-Assisted Recombination.

Discrete Energy Level (Boundaries)

The Discrete Energy Level subnode is available from the context menu (right-click the
parent node) or from the Physics toolbar, Attributes menu when the Trapping model
drop-down menu is set to the Explicit trap distribution option for the Trap-Assisted
Surface Recombination boundary condition.

The settings for this node are the same as for Discrete Energy Level
(Domains). The only difference is that the units for Trap number density
are 1/m2 instead of 1/m3.

110 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Continuous Energy Levels (Boundaries)
The Continuous Energy Level subnode is available from the context menu (right-click
the parent node) or from the Physics toolbar, Attributes menu when the Trapping model
drop-down menu is set to the Explicit trap distribution option for the Trap-Assisted
Surface Recombination boundary condition.

The settings for this node are the same as for Continuous Energy Levels
(Domains). The only difference is that the units for Trap number density
are 1/m2 instead of 1/m3.

Trap-Assisted Heterointerface Recombination

The Trap-Assisted Heterointerface Recombination boundary condition adds
contributions to the interface recombination rate and the interface charge density
associated with the interface traps for the Continuity/Heterojunction boundary
condition (not applicable for the Continuous quasi-Fermi levels option). All the settings
are the same as its surface counterpart Trap-Assisted Surface Recombination, except
for the Trapping model option, only the Explicit trap distribution option is available,
since there is no simple formula like the Shockley-Read-Hall model in the Trap-Assisted
Recombination and Trap-Assisted Surface Recombination features, due to the dual
contributions to the recombination rates from both sides of the heterojunction.

Discrete Energy Level (Heterointerfaces)

The Discrete Energy Level subnode is available from the context menu (right-click the
parent node) or from the Physics toolbar, Attributes menu for the Trap-Assisted
Heterointerface Recombination boundary condition.

The settings for this node are the same as for Discrete Energy Level
(Boundaries). The only difference is that now there are two Carrier
Capture sections, one for each side of the heterojunction.

THE SEMICONDUCTOR INTERFACE | 111

 Continuous Energy Levels (Heterointerfaces)
The Continuous Energy Level subnode is available from the context menu (right-click
the parent node) or from the Physics toolbar, Attributes menu for the Trap-Assisted
Heterointerface Recombination boundary condition.

The settings for this node are the same as for Continuous Energy Levels
(Boundaries). The only difference is thatnow there are two Carrier
Capture sections, one for each side of the heterojunction.

Floating Gate
Use the Floating Gate feature to model a gate contact to a dielectric domain (not a
semiconductor domain), which is unconnected to any voltage source, or which is
connected to a circuit with a floating potential.

In combination with the Insulator Interface feature that connects the dielectric
domain and an adjacent semiconductor domain, the Floating Gate can model both the
charging and discharging of a floating gate by tunneling from the semiconductor
domain through the dielectric domain.

Add this feature from the boundary level Electrostatics submenu.

FLOATING GATE
Select an External connection — None (the default) or Circuit (to connect the terminal
to an electrical circuit).

Select a Tunnel current Itun (SI unit: A) from the list — None, User defined, or choose
a tunnel current announced by one of the Insulator Tunneling features in the model.
Current injected into the gate accumulates as charge.

Enter a Contact work function Φc (SI unit: V).

Use the Initial charge Qinit (SI unit: C) setting to determine the initial charge on the
gate in a transient simulation, or the charge on the gate in a stationary simulation.

112 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Insulator Interface
The Insulator Interface feature is automatically applied to the boundaries at the
interface between Semiconductor Material Model domains and Charge Conservation
(dielectric) domains. A number of options are available to modify the behavior of the
boundary.

The functionality of the Surface traps check box in versions prior to 5.4
has been replaced and expanded by the Trap-Assisted Surface
Recombination boundary condition.

TUNNELING
Select a Tunneling type — None (the default), Fowler-Nordheim tunneling, or User
defined.

FOWLER-NORDHEIM TUNNELING
In combination with the Floating Gate feature, use this option to model both the
charging and discharging of a floating gate by Fowler-Nordheim tunneling from the
domain.

For Fowler-Nordheim tunneling, select an option from the Fowler-Nordheim tunneling list
— Electrons (the default), Holes, or Electrons and holes. This determines the species for
which the tunnel current is computed. Based on this selection, enter:

• Electron Fowler-Nordheim coefficients AnFN (SI unit: A/V2) and BnFN (SI unit: V/
m).
• Hole Fowler-Nordheim coefficients ApFN (SI unit: A/V2) and BpFN (SI unit: V/m).

USER DEFINED TUNNELING

For User defined, enter an Electron tunnel current density JnUD (SI unit: A/m2). Enter
a Hole tunnel current density JpUD (SI unit: A/m2) to define a user defined tunnel
current expression.

DENSITY-GRADIENT
This section appears only when one of the density-gradient formulation options are
selected under the Discretization section. See the identically named section
Density-Gradient for details.

THE SEMICONDUCTOR INTERFACE | 113

 Theory for the Semiconductor
Interface
Detailed introductions to the physics of semiconductors are available in Refs 1 to 5.
Ref. 6 is an introduction to semiconductor device modeling using the drift-diffusion
method used by COMSOL Multiphysics. Ref. 45 provides a comprehensive review of
the density-gradient (DG) theory which adds the effect of quantum confinement to
the drift-diffusion method in a computationally efficient fashion. The
DG-Confinement theory is implemented by COMSOL and the corresponding
equations are pointed out when they differ from the conventional drift-diffusion
method.

A brief summary of the physics of semiconductors is included here, to a level sufficient

to derive the drift-diffusion equations and to develop a basic understanding of
semiconductor band structures. The derivation of the density-gradient theory is not
included. See Ref. 45 if interested.

To completely describe the dynamics of electrons within a solid, the many body
Schrödinger equation must be solved in the periodic structure defined by the crystal
structure of the solid. In practice this is not possible and approximations must be made.
Solid state physicists have devised a number of methods to solve simplified forms of
this equation and these methods have been validated through experiment (see, for
example, chapters 9 to 11 of Ref. 1 and chapters 4 and 5 of Ref. 2). A starting point
for many of these methods is to consider only the motion of the electrons through an
essentially stationary lattice of nuclei (known as the adiabatic approximation). Then
the many electron wave function is simplified into a form in which it reduces to a set
of one electron wave functions. The effect of the nuclei and of electron-electron
interactions can be incorporated into these one electron models by modifications to
the effective potential. Models within the so-called one-electron approximation have
been very successful at predicting the properties of semiconductors and semiconductor
transport. Although the one electron model explained in this section seems very
simplistic (particularly when considering the complexity of electron-electron and
electron-ion interactions), in practice this model can be used to develop a very detailed
understanding of transport in semiconductors.

114 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

In this section:

• The Effect of Periodicity • Incomplete Ionization

• Electrons in a Perturbed Periodic • Band Gap Narrowing
Potential • Optical Transitions
• The Semiclassical Model • Boundary Conditions
• The Boltzmann Equation • Metal Contacts
• Doped Semiconductors • Thin Insulating Gates
• Specifying the Dopant Distribution • Continuity/Heterojunction
• Equilibrium Carrier Concentrations • Boundary Conditions for Charge
• The Semiconductor Equations Conservation
• Mobility Models • Tunneling
• Traps • References for the Semiconductor
• The Sommerfeld Model and the Interface
Density of States • Semiconductor-Electromagnetic
• Recombination and Generation Waves Coupling

Modeling Guidelines

The Sommerfeld Model and the Density of States

Sommerfeld was the first person to successfully apply quantum mechanics to the
physics of transport in solids. The Sommerfeld model considers the physics of
independent electrons in a large potential well. The starting point for this model is the
time-independent Schrödinger equation for an independent electron:

_
h 2
– --------- ∇ ψ = Eψ (3-1)
2m
_
where h is Planck’s constant divided by 2π and m is the electron mass. The
time-independent Schrödinger equation takes the form of an eigenvalue equation. The
modulus squared of the (complex) eigenfunctions (ψ) or wave functions that solve the
equation represent the probability that an electron in the corresponding state can be
found at a given position p(r) (that is, p(r) = ⏐ψ⏐2). The corresponding eigenvalue E

THEORY FOR THE SEMICONDUCTOR INTERFACE | 115

 for the state gives the energy associated with the state. For a detailed introduction to
quantum mechanics see Ref. 7 and Ref. 8.

Since it is expected that the solid is periodic, the equation is solved on a cube of side
L assuming the periodic boundary condition:

ψ ( x + L, y + L, z + L ) = ψ ( x, y, z ) (3-2)

The solutions of Equation 3-1 and Equation 3-2 are plane waves of the form:

1
ψ ( x, y, z ) = ------------
1⁄2
- exp ( i [ k x x + k y y + k z z ] ) (3-3)
Ω

where Ω = L3 is the volume of the solid (which appears in the equation to correctly
normalize the wave function) and:

2πn x
k x = -------------
L
2πn y
k y = -------------
L
2πn z
k z = -------------
L

where nx, ny, and nz are integers.

Since these states are all periodic, it is convenient to label them by means of the wave
number k. Think of a k-space populated by these states in a regular cubic grid. As a
consequence of the Pauli exclusion principle each state can hold two electrons (one
spin up and one spin down).

The density of states (g(k) = dns/ dξk), that is, the number of states (ns) per unit
volume of k-space (ξk) for unit volume of the material, is given by:

1
g ( k ) = ---------3
4π

Substituting Equation 3-3 into Equation 3-1 gives the energy of the particle in a given
state:

_2 2 2 2 _2 2
h ( kx + ky + kz ) h k
E k = ----------------------------------------------------- = ---------------- (3-4)
2m 2m

116 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

At zero temperature electrons fill up the states with the lowest energy first, filling the
grid so as to minimize the total energy. The surface dividing the filled states from the
empty states — known as the Fermi surface — is spherical for large numbers of
electrons as a result of Equation 3-4. For an electron number density n at zero
temperature the magnitude of the wave vector corresponding to the states at the Fermi
surface (kF,0) is given by:

4 3 n- 3
--- πk F ,0 = ----------- = 4π n
3 g( k)

so that:

2 1⁄3
k F ,0 = ( 3π n )

At finite temperatures and at equilibrium, the principles of statistical mechanics (see for
example Ref. 10) give the mean occupancy of the states (f0 (k)) as:

1
f 0 ( k ) = ------------------------------------------------------------------------ (3-5)
1 + exp [ ( E k – E F ) ⁄ ( k B T ) ]

where EF is the Fermi energy or the chemical potential. At finite temperatures EF is

determined by the requirement that the total number of electrons per unit volume is
equal to n:

∞ ∞
g( E)
n = 0 f0 ( k )g ( E ) dE = 0 1-----------------------------------------------------------------------
+ exp [ ( E k – E F ) ⁄ ( k B T ) ]
- dE (3-6)

where g(E) = dns/ dE is the energy density of states, given by:

2
1 2m  3 ⁄ 2 1 ⁄ 2
dE = ---------2  ---------
k dk mk -
g ( E ) dE = g ( k )dξ k = -----2- -------- dE = --------------- E dE (3-7)
_  _ 2
π dE
2 2
π h 2π h

In deriving Equation 3-7, the E-k relationship (Equation 3-4) was used to evaluate the
derivative and to convert from k to E.

The density of states varies with the square root of the energy.

These results form the basis of the Sommerfeld model and are useful in this discussion.
Using this simple model, you can predict the thermal and electrical properties of some
metals with reasonable accuracy (see Ref. 1 and Ref. 2 for details). However, the

THEORY FOR THE SEMICONDUCTOR INTERFACE | 117

 model does not explain the existence of insulators or semiconductors because it is
missing important information about the microscopic periodicity of the material. The
Effect of Periodicity section describes how this periodicity can be treated.

The Effect of Periodicity

THE LATTICE AND THE RECIPROCAL LATTICE

The 3D periodicity of crystalline materials is conventionally described by a set of lattice
vectors and a basis (a specific pattern of atoms) associated with each lattice point.
Consider the case of a “primitive lattice”, that is, a crystal made up of a single atom,
where the atomic locations are coincident with the lattice points. Such a lattice does
not require a basis (or more formally the basis is a single atom located at (0,0,0)). The
set of lattice vectors R can be written as:

R = n1 a1 + n2 a2 + n3 a3 (3-8)

where n1, n2, and n3 are integers (taking all values between –∞ and ∞) and a1, a2, and
a3 are the lattice vectors. For a primitive lattice, the unit cell is the parallelepiped
constructed from the vectors a1, a2, and a3.

A useful way to represent the lattice is by means of an array of delta functions. A

physical quantity of interest (for example the electric potential) can then be
represented by the convolution of the variation in the potential within a single unit cell
of the lattice with the delta function array. This approach is easiest to understand
in 1D, when the delta function array is known as a Dirac comb. A 1D lattice can be
represented as:

n=∞

 δ ( x – na )
n = –∞

where a is the lattice parameter. This periodic function can be represented by a Fourier
series of the form:

n=∞ n=∞

--1- e 2πinx ⁄ a
 δ ( x – na ) =  a
(3-9)
n = –∞ n = –∞

118 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Of particular importance to understand semiconductor transport is the concept of the
reciprocal lattice. This is the lattice produced by taking the Fourier transform of the
real space lattice. For the Dirac comb this is given by:

n=∞  n=∞ n=∞

 
   δ ( x – na )e 
ikx ikna
FT  δ ( x – na )  = dx = e
 
 n = –∞  n = –∞ n = –∞

n=∞

------ δ  k – 2πn
-----------
2π
=  a  a 
n = –∞

where the final step follows from Equation 3-9. The reciprocal lattice is another Dirac
comb with spacing proportional to the reciprocal of the real space lattice.

In 3D the lattice can be represented as δ (r–R) where the summation over all
combinations of lattice vectors is implied by the use of the set of vectors R. The Fourier
transform of the lattice is:

n=∞ 
  ik ⋅ r ik ⋅ R
FT 

 δ(r – R)  =

 δ ( r – R )e dr = e
 n = –∞ 
3
( 2π )
= -------------- δ ( k – K∗ )
Ω

where Ω is the volume of the real space unit cell (Ω = a1 ⋅ (a2 × a3)) and K*is the set
of reciprocal lattice vectors given by:

K∗ = n 1 b 1 + n 2 b 2 + n 3 b 3

where n1, n2, and n3 are integers (taking all values between –∞ and ∞) and b1, b2, and
b3 are the reciprocal lattice vectors given by:

2π
b 1 = ------ a 2 × a 3
Ω
2π
b 2 = ------ a 3 × a 1 (3-10)
Ω
2π
b 3 = ------ a 1 × a 2
Ω

THEORY FOR THE SEMICONDUCTOR INTERFACE | 119

 Understanding the reciprocal lattice in terms of Fourier transforms as described above
is useful. The Heisenberg uncertainty principle can be seen to be related to the
properties of the Fourier transform (in the time domain: ΔfΔt≈1). Also the effect of a
lattice basis can be straightforwardly introduced by taking the convolution of the basis
with the lattice in real space. The convolution theorem then tells us that the result in
reciprocal space (or k-space) is the product of the reciprocal lattice and the Fourier
transform of the basis. The main effect of the basis is to modulate the amplitude (and
phase) of the reciprocal lattice points. When zeros in the Fourier transform of the basis
coincide with reciprocal lattice points, the basis leads to the elimination of these points
in the reciprocal lattice. In X-ray imaging experiments, which sample the reciprocal
lattice by crystal diffraction, this is referred to as extinction.

Importantly, any physical quantity with a periodicity that matches that of the lattice can
straightforwardly be represented as the convolution of some function with the
reciprocal lattice in real space, or as a modulating function for the reciprocal lattice in
k-space, changing the amplitudes of each of the lattice points in k-space. Since each of
these points represents a single harmonic component of the quantity of interest this
construction can be thought of as a representation of a three dimensional Fourier
series. For example, consider a periodic potential V(r) = V(r + R). V(r), which can be
written in the form:

– i K∗ ⋅ r
V(r ) =  VK∗ e (3-11)
K∗

where the summation occurs over all the reciprocal lattice vectors K*. The reciprocal
lattice is therefore a representation of the Fourier components required in 3D to
represent a function with the periodicity of the lattice.

BLOCH FUNCTIONS
The Sommerfeld model did not attempt to include the effective electric potential of
the crystal or the other electrons. It is clear that the effective potential must be periodic
with the same periodicity as the crystal lattice. The periodicity of the lattice has
important consequences for the electrons. First, since the problem has periodic
symmetry, observable physical quantities must also have periodic symmetry. As a
consequence ⏐ψ⏐ must be periodic, so that, using the notation developed previously:

ψ( r + R ) = ψ(r )

equivalently:

120 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 iθ ( R )
ψ ( r + R ) = ψ ( r )e (3-12)

The translational symmetry of the lattice imposes additional restrictions on the form
of θ(R). For two lattice vectors, RA and RA + RB, translational symmetry implies:

θ ( RA + RB ) – θ ( RA ) = θ ( RB )

which leads to the requirement that θ depends linearly on the three integers, n1, n2,
and n3, which specify R (see Equation 3-8):

θ ( RA + RB ) = θ ( RA ) + θ ( RB )

θ can therefore be written in the form:

θ = n1 c1 + n2 c2 + n3 c3

for constants c1, c2, and c3. Writing k = c1b1 + c2b1 + c3b1 leads to the requirement
θ = k⋅R and Equation 3-12 then becomes:

ik ⋅ R
ψ ( r + R ) = ψ ( r )e (3-13)

If ψ (r) is written in the form:

ik ⋅ r
ψ k ( r ) = u k ( r )e (3-14)

substituting into Equation 3-13 shows that uk(r+R) = uk(r). Equation 3-14, along
with the periodicity requirement on uk(r), is known as Bloch’s theorem. It is extremely
useful as it allows the wave function corresponding to a particular k vector to be
expanded in a Fourier series of the same form as that of the potential (Equation 3-11):

–i K ⋅ r
uk =  CK e
K

Wave functions which satisfy Bloch’s theorem are frequently referred to as Bloch
functions.

INDEPENDENT ELECTRONS IN A PERIODIC POTENTIAL

The time-independent Schrödinger equation can now be written using the periodic
expansion for both the potential and the Bloch wave function:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 121

 _2 
h i(k – K) ⋅ r – i G ⋅ r  i ( k – K ) ⋅ r
 CK e  VG e  CK e
2
--------- ∇ +  
2m   
K G K
i(k – K) ⋅ r
= E  CK e
K

where G is used for the potential reciprocal lattice vector to distinguish it from K.
Simplifying this equation gives:

_2
h 2 –i K ⋅ r –i ( G + K ) ⋅ r
---------
2m  CK ( k – K ) e +  VG CK e
K G, K
–i K ⋅ r
= E  CK e
K

To obtain the equation for the coefficients of the sum, premultiply the sum by

i K' ⋅ r
e

and integrate over all space. This gives:

_2
h 2
---------
2m K'
C ( k – K' ) +  VG CK' – G = EC
K'
(3-15)
G

Equation 3-15 is valid for any periodic potential, small or large.

THE NEARLY FREE ELECTRON MODEL

To obtain the small potential limit of Equation 3-15, consider the case of a single
sinusoidal potential component with a small amplitude:

i K1 ⋅ R –i K1 ⋅ R
V ( r ) = 2V 1 cos ( K 1 ⋅ R ) = V 1 ( e +e )

Since the periodicity of the potential is one dimensional, it is only necessary to consider
Fourier components in the direction of K1 in the expansion of uk(r), which can
therefore be written in the form:

– ip K 1 ⋅ r
uk =  Cp e
p

122 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

and Equation 3-15 takes the form:

_2
h
--------- 2
C ( k – pK 1 ) + V 1 ( C p – 1 + C p + 1 ) = EC p (3-16)
2m p

In the limit V1 = 0, p = 0, and Equation 3-16 recovers the form of the Sommerfeld
E-k relationship, Equation 3-4. However, for V1 = 0, additional solutions now also
exist for nonzero values of p, which take the form of a set of parabolas with origins
shifted by p times the reciprocal lattice vector K1. As a result of the periodicity of the
lattice, E has become multivalued for a given k as shown in Figure 3-1. The E-k
relationship is periodic, with a single repeating unit contained within the dashed lines
(shown at ±K1/2). Since the parts of the plot outside the dashed lines consist of
repeated information, Ek plots are conventionally drawn showing only the region
between the dashed lines. This region is called the Brillouin zone and in 3D it consists
of all the points in k-space closer to one particular reciprocal lattice point K than to
any other point.

-2K1 -K1 -K1/2 0 K1/2 K1 2K1

k·K1

Figure 3-1: Diagram illustrating how the energy becomes multivalued for the case of a
periodic potential.

Writing Equation 3-16 explicitly for the case V1 = 0 gives:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 123

 _2
h 2
E = --------- ( k – pK 1 ) (3-17)
2m

Figure 3-1 shows that on the planes half between the lattice vectors (defined by the
equations k·K1 = qK1/2, for integers q, and indicated by the dashed line), the energy
associated with two different values of p, for example p1 and p2, can be equal. Note
that p1 and p2 are associated with two (different) periodic components of the wave
function or with two different Ek curves in the figure. Away from these planes, the
energy associated with each periodic component of the wave function differs from all
the other components.

Next consider the energy on the plane k·K1 = qK1/2, at the point corresponding to
p1 = 0 and p2 = 1 as Equation 3-17 changes into Equation 3-16 by a slow increase of
V1 from zero. Close to the plane it is expected that initially, as V1 is increased, a
solution exists in which C0 and C1 are significant but all other coefficients are
extremely small; that is, a solution in which only the p1 = 0 and p2 = 1 components of
the wave function play a significant role. Making the assumption that the other
coefficients are zero reduces the set of Equation 3-16 to just two equations:

_2
C 0  E – --------- ( k ⊥ – k || ) = V 1 C 1
h 2 2
 2m 
_2
C 1  E – --------- ( k ⊥ + ( k || – K 1 ) ) = V 1 C 0
h 2 2
 2m 

where the vector k has been decomposed into components parallel (k||) and
perpendicular (k⊥) to K1. Taking the product of these two equations leads to a
quadratic equation that can be solved to obtain the value of E. It is most convenient
to use the following nondimensional variables when solving the equation:

2 k⊥  2
- –  -------------
8mE
ε = ----------------------
_2 2  K 
h K1 1

8mV 1
U = ----------------------
_2 -
2
h K1
k ||
δ = 1 – 2 ----------
K1

A little algebra then leads to the equation:

2 2 2 2 2
ε – 2 ( 1 – δ )ε + ( 1 + δ ) ( 1 – δ ) – U = 0

124 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

that has solutions:

2 2 2
ε = 1 + δ ± U + 4δ (3-18)

In the case U = 0, k⊥= 0, this equation reproduces the two parabolas centered on 0
and K1 (Figure 3-1). A nonzero value of k⊥ simply shifts the parabolas to greater
energies. At the edge of the Brillouin zone (indicated by the dashed line), where the
two parabolas cross, δ is zero, and as k|| decreases, δ increases, reaching a maximum of
1 at the origin. When a small but finite value of U is introduced (for example U < 0.1)
there is little effect on the curve away from the Brillouin zone edge, but as δ becomes
comparable to, or less than, U its effect becomes significant. At the edge of the
Brillouin zone the two Ek curves (corresponding to two different Fourier components
of the wave function) no longer cross but are separated by a small gap such that ε=1±U.
This corresponds to a gap between the lowest two Ek curves of magnitude 2V1.

The form of Equation 3-16 makes it clear what the effect of other harmonics in the
periodic potential would be. For a potential of the form:

– ih K 1 ⋅ r
V(r ) =  Vh e
h

Equation 3-16 becomes:

_2
h
--------- 2
C ( k – pK 1 ) + V h ( C p – h + C p + h ) = EC p
2m p

The effect of Vh is to couple the equations involving Cp and Cp±h, and it therefore
perturbs apart the parabolas centered on pK1 and on (p ± h)K1. Provided Vh is small,
its effect is only significant near the edge of the Brillouin zone. Thus, for a more
general periodic potential with higher harmonics, the Ekrelationship takes the form
shown in Figure 3-2.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 125

 E

-2K1 -K1 -K1/2 0 K1/2 K1 2K1

k·K1

Figure 3-2: Diagram showing the effect of a periodic potential on the E-k relationship. The
repeated zone scheme is used.

The Ek diagram has now changed so that certain energies are forbidden. The allowed
states exist within bands of permitted energies, with band gaps separating them.
Figure 3-2 can be redrawn in various ways. In the figure the repeated zone scheme is
shown. This scheme highlights the periodicity of the lattice and makes clear the
concept of energy bands and band gaps. An alternative is to show only the nth band
in the nth Brillouin zone (known as the extended band scheme). This approach
produces results that look similar to the equivalent (parabolic) plot for the Sommerfeld
model, with gaps appearing in the curve at the edges of the Brillouin zones. Practically,
the more compact reduced zone scheme is usually employed, which shows only the
information in the first Brillouin zone (between the dashed lines). When the band
structure information of real materials is displayed in this form, it is typical to show the

126 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

bands along several connected lines within the first Brillouin zone and to use the
reduced zone scheme.

There are now several states corresponding to a given wave number k

with several associated energies. It is conventional to label the individual
states with a band number, n, to distinguish between them. Thus the
wave function Ψnk with energy Enk corresponds to the state in the nth
band with wave vector k (note that in the above argument, the band
index n corresponds to the harmonic p, which dominates in the Bloch
function away from the Brillouin zone edges).

Before considering the effect of a potential varying in three dimensions, it is useful to

visualize Equation 3-18 in a different way. Figure 3-3 shows two surfaces of constant
energy for the solution to this equation (the nondimensionalized form of the equation
is not used for the plot). In the free electron model the constant energy surfaces are
spheres, centered on the origin. The periodicity of the structure means that there are
now spheres centered on each of the reciprocal lattice points. The effect of the periodic
potential is to split apart the spherical surfaces at the points where they would have
overlapped, to form a set of nonintersecting surfaces.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 127

 k⊥

k||

-K1/2 K1/2

Figure 3-3: Constant energy contours produced by a one-dimensional periodic potential.

Two contours are shown with slightly different energy levels.

In the more general case where the potential varies periodically in all three dimensions,
exactly the same arguments apply, provided that the potential coefficients are small,
(that is VG << h2⏐G2⏐/8m). Where the spheres centered on the different lattice
points K and K– G intersect, the effect of the potential is to cause the kind of
remapping shown in Figure 3-2 and Figure 3-3 — where the two surfaces intersect it
splits apart instead. The nearly free electron energy surfaces can therefore be
constructed by drawing spheres of equal radius centered on each lattice point K and
rejoining them in this manner where they intersect, to form a set of nonintersecting
surfaces. Although this procedure sounds simple, in practice rather complicated energy
surfaces result from the procedure. Ref. 2 considers the example of a simple cubic
material in detail, and Ref. 1 shows several examples of constant energy surfaces for
different lattices.

Considering the approximations made in deriving the nearly free electron model, it is
quite remarkable that the model agrees so well with the measured Ek surfaces of many
real materials, particularly considering that the true potential is expected to vary rapidly
in the vicinity of the atomic cores. The reason for the success of the nearly free electron

128 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

theory is related to many-body effects. In materials where it is possible to divide the
electrons into tightly bound core electrons and weakly bound valence electrons, the
core electrons and the ions can be replaced by a pseudo-ion with a weakly varying
pseudopotential, surrounded by the outermost valence electrons. The wave function
for the valence states must be orthogonal to that of the core states and the
pseudopotential is constructed to ensure that this is the case. The resulting potential
varies much more slowly than the true potential as a result of a Pauli repulsion effect.
This effect repels the valence electrons away from the core states in order to ensure that
the corresponding wave functions are orthogonal. The nearly free electron model can
then be applied to real materials if the pseudopotential, rather than the true potential,
is used.

THE DENSITY OF STATES IN A PERIODIC POTENTIAL

To compute the density of states in k-space the periodic boundary condition is applied,
in this case on a crystal made up of Nc = N1× N2× N3 unit cells in the a1, a2, and a3
directions. Using the periodic boundary condition for the wave function gives the
equation:

ψ ( r + Ni ai ) = ψ ( r )

Equation 3-13 adds the additional requirements:

iN i ( k ⋅ a i )
ψ ( r + N i a i ) = ψ ( r )e

Thus:

iN i ( k ⋅ a i )
e = 1

and the allowed values of k are:

n1 n2 n3
k = ------- b 1 + ------- b 2 + ------- b 3
N1 N2 N3

where n1, n2, and n3 are integers. The reciprocal space volume per allowed k-vector
is given by:

b1 ⋅ b2 × b3 ξ BZ
δξ = ----------------------------- = ----------
N1 N2 N3 Nc

where ξBZ is the volume of the Brillouin zone in k-space. Consequently, the number
of allowed wave vectors in a single Brillouin zone is equal to the number of unit cells

THEORY FOR THE SEMICONDUCTOR INTERFACE | 129

 in the crystal. Since each state can accommodate two electrons of opposite spin, filling
one band over the entire Brillouin zone corresponds to a crystal with two valence
electrons per unit cell. The density of states in k-space is given by:

2N c 1
g ( k ) = --------------- = ---------3 (3-19)
Ωξ BZ 4π

where the volume of the Brillouin zone (ξBZ=8π3/Ωu where Ωu is the unit cell
volume) has been calculated explicitly using Equation 3-10 and vector algebra (note
also that Ω = NcΩu is the volume of the crystal itself). This result is identical to that
obtained by the Sommerfeld model.

The available states in the crystal are filled up in the same way as in the Sommerfeld
model. The occupancy of the states is still given by Equation 3-5 and the Fermi level
is defined by Equation 3-6. For a metal, the Fermi surface geometry in a periodic
potential reflects the equipotential surfaces of the band structure at the Fermi energy
(for example in the previous section surfaces like those shown in Figure 3-3). In
semiconductors and insulators the Fermi energy lies within the band gap so there is no
clear Fermi surface. However, in semiconductors the Fermi function slightly overlaps
the band above (below) the Fermi level, known as the conduction band (or the
valence band), and the states near the bottom (top) of the band have a low but
significant probability of being occupied (unoccupied). Since it is these states that lead
to the conductivity of semiconductors (see The Semiclassical Model), it is worth
considering what form the density of states takes at the very edge of a band.

Consider first the minimum of a conduction band. Choose a new coordinate system
such that the Taylor series for the E-k relationship expanded about the band minimum
up to second order takes the form:

_2 _2 _2
h 2 h 2 h 2
E ( k' ) = E c + -------------- k' x + -------------- k' y + -------------- k' z (3-20)
2m x∗ 2m y∗ 2m z∗

Here the constants m∗x, m∗y, and m∗z are associated with the k′x, k′y, and k′z terms in
the series, respectively. The reason for choosing this form of the (arbitrary) constant
becomes apparent below. The coordinate system for the vectors k′ has its origin at the
minimum of the band in k-space and is aligned so that Equation 3-20 applies in the
form given. There are no first-order terms in the expansion since it is at a minimum in
the E(k') relationship (that is, it is at the bottom of a band), which represents the
equation of an ellipsoidal surface. Close to the band edge the constant energy surfaces
are therefore ellipsoids, with a semi-axis in the k′x direction given by

130 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 1- 1⁄2
_ ( 2m x∗ ( E – E c ) )
k' E ,x = ------
h

and similarly for the k′y and k′z directions. A given constant energy surface contains a
volume given by

4 4π -
V E = --- πk E ,x k E ,x k E ,x = ------------ ∗ ∗ ∗ 1⁄2 3⁄2
3 _ 3 ( 8m x m y m z ) ( E – E c )
3h

The number of states enclosed by the constant energy surfaces is therefore

VE 1 1⁄2
- ( 8m x∗ m y∗ m z∗ ) ( E – E c ) 3 ⁄ 2
N c ( E ) = ---------3 = -------------------
2 _ 3
4π 3π h

and the density of states gc (E) is given by:

dN c ( E ) 1 2m∗ 3 ⁄ 2
g c ( E ) = -------------------- = ---------2  -----------
1⁄2
dE _ 2- ( E – Ec ) (3-21)
2π h

This result is identical in form to Equation 3-7, except that the mass has been replaced
by an effective mass, m*=(m∗x m∗y m∗z)1/3 (the constants in Equation 3-20 are
named in a manner consistent with this result).

The same argument can be applied to the top of the valence band (using a new axis
system, k′′) leading to the similar result:

3⁄2
1  2m h∗ 1⁄2
g v ( E ) = ---------2  --------------
_ 2 - ( Ev – E ) (3-22)
2π  h 

ALTERNATIVE REPRESENTATIONS OF THE WAVE FUNCTION: WANNIER

FUNCTIONS
At this point it is useful to consider an alternative representation of the wave function,
known as the Wannier function. Wannier functions, Wn(r − R), are wave packets of
Bloch functions (Ψnk(r)) that are localized at a particular lattice vector R. These are
defined in the following way:

1 – ik ⋅ R
W n ( r – R ) = ---------
N  ψnk ( r )e (3-23)
k

where N is the number of unit cells in the crystal. The Wannier functions are
orthogonal since:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 131

 
3
W n ( r – R )∗ W n ( r – R' ) d r
crystal
1 i ( k ⋅ R – k' ⋅ R' ) ∗
 
3
= ---- e ψ nk ( r ) ψ nk' ( r ) d r
N
k ,k' crystal
1 i ( k ⋅ R – k' ⋅ R' ) 1 ik ⋅ ( R – R' )
= ----
N e δ ( k – k' ) = ----
N e = δ ( R – R' )
k ,k' k

A single Bloch state can be represented in the form:

1 ik ⋅ R
ψ n k ( r ) = ---------
N  Wn ( r – R )e (3-24)
R

Wannier functions are useful to represent tightly bound, localized states, since the
spatial extent is limited. The functions themselves are not unique, since the phase of
the Bloch states Ψnk(r) is arbitrary. However, it is possible to define a maximally
localized Wannier function, which gives an intuitive picture of the bonding in the solid.

Wannier functions are important because these form the basis of alternative approaches
to computing the band structures of solids. For example, in the tight binding
approximation (TBA) it is assumed that the Wannier function is an atomic orbital,
enabling the wave function to be constructed directly from Equation 3-24. Instead of
using a single atomic wave function, one can employ a linear combination of them,
resulting in the linear combination of atomic orbitals (LCAO) approach.

SILICON BAND STRUCTURE AND DENSITY OF STATES

The band structure of silicon is illustrated in Figure 3-4. Although the
three-dimensional band structure is considerably more complex than the simple
picture described, many of the principles described are appropriate. Silicon is an
indirect band-gap semiconductor, which means that the bottom of the conduction
bands occur at a different point in k-space to the top of the valence bands. The valence
band maxima occur at the center of the Brillouin zone. The conduction band minima
occur at approximately 4/5 of the distance from the zone center to its edge along the
kx, ky, and kz axes.

Considering first the conduction bands, there are six symmetry equivalent minima in
the locations shown in Figure 3-4. Physically the form of the energy density of states
is important to determine the transport properties of the semiconductor. All six of the
conduction band minima are equivalent and consequently the contributions to the
density of states can be added. The constant energy surfaces near the band minima are

132 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

close to being ellipsoidal; Equation 3-21 gives a good description of the energy density
of states. The transport properties of the band can be characterized by a single effective
mass without any loss of accuracy in the model.

kz

contour shown
on right X
Γ ky
1.13eV kx

L
0.04 eV

L Γ X

Figure 3-4: Band structure of silicon. Left: Schematic E-k diagram showing energy along
the directions indicated by the dotted lines in the figure to the right. This part of the figure
is equivalent to Figure 3-2. Right: Constant energy surfaces (shown in gray)
corresponding to the contour indicated in the figure on the left. The first Brillouin zone is
also shown. This part of the figure is equivalent to a 3D version of Figure 3-3.

There are two coincident valence band maxima located at point Γ. An additional
valence band, with a slightly lower maximum energy (produced by spin-orbit coupling,
see Ref. 1) is also located at this point. Each of these bands has a different effective
mass associated with it. It is common to represent the effect of the three valence bands
with an average density of states so that Equation 3-22 is assumed. Strictly speaking
this assumption is less accurate for holes than it is for electrons, as a result of the
different energies associated with the band minima.

This discussion motivates the adoption of the so-called one-band model, in which a
single valence and conduction band is considered in the transport model. The
one-band model can be applied to many practical semiconducting materials.

Electrons in a Perturbed Periodic Potential

Semiconductor devices are almost never homogeneous, and consequently it is
necessary to consider the effect of perturbations to the periodic potential. First

THEORY FOR THE SEMICONDUCTOR INTERFACE | 133

 consider solutions of the Schrödinger equation for the unperturbed problem, with a
Hamiltonian, H0:

0 0
H 0 ψ k ( r ) = E 0 ( k )ψ k ( r )

For simplicity, the band number n is dropped from the wave function in
the above equation.

The perturbed problem has a set of solutions in the form:

Hψ m ( r ) = E m ψ m ( r ) (3-25)

where H=H0+H1.

The perturbed wave functions do not necessarily have a constant wave

vector, so the subscript k is dropped and a different quantum number, m
replaces it. In many practical applications H1 varies slowly on the length
scale of the lattice and this assumption is subsequently made.

The approach taken in Ref. 9 (and originally derived by Wannier) is inspired by

Equation 3-24. The unperturbed wave function can be written in the form:

1 ik ⋅ R
ψ k ( r ) = ---------
N  W ( r – R )e
R

where W(r-R) is the Wannier function. Solutions of the form:

ψm ( r ) =  Ψm ( R )W ( r – R ) (3-26)
R

are sought. Here Ψm(R) is a function that is employed to weight the Wannier
functions in an expansion of the perturbed wave function. In the limit H1→0,
Ψm(R)→exp(ik⋅R)/√N.
Substituting Equation 3-26 into Equation 3-25 gives:

( H0 + H1 )  Ψm ( R )W ( r – R ) = Em  Ψm ( R )W ( r – R ) (3-27)
R R

This equation is premultiplied by W*(r-R′) and integrated over the crystal:

134 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

  
  Ψm ( R )W ( r – R ) d r
3
W∗ ( r – R' )H 0 

crystal R

  Ψm ( R )W ( r – R ) d r
3
+ W∗ ( r – R' )H 1 (3-28)
crystal R
 
  Ψm ( R )W ( r – R ) d r
3
= W∗ ( r – R' )E m 

crystal R

Consider first the right-hand side of Equation 3-28:

 
  Ψm ( R )W ( r – R ) d r  δ ( R – R' )
3
W∗ ( r – R' )E m  = Em Ψm ( R )

crystal R R
= Em Ψm ( R )

where both the orthogonality of the Wannier functions is employed.

The term in H1 on the left-hand side of Equation 3-28 gives:

 
  Ψm ( R )W ( r – R ) d r
3
W∗ ( r – R' )H 1 

crystal R

= H1  Ψm ( R )δ ( R – R' ) = H 1 Ψ m ( R' )
R

where the fact that the Wannier function is localized around R or R′ and the
assumption that H1 is a constant on this length scale is employed.

Finally the H0 term is considered. Using the definition of the Wannier function given
in Equation 3-23, this term is written as:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 135

 1  0∗ k' ⋅ R'  – ik ⋅ R 3
  ψk'  Ψm ( R )  ψk ( r )e
0
----  ( r )e  H0  d r
N    
crystal k' R k

1  0∗ k' ⋅ R'  – ik ⋅ R 3
   Ψm ( R )  H0 ψk ( r )e
0
= ----  ψ k' ( r )e  d r
N   
crystal k' R k

1  0∗ k' ⋅ R'  – ik ⋅ R 3
   Ψm ( R )  E0 ( k )ψk ( r )e
0
= ----  ψ k' ( r )e   d r(3-29)
N   
crystal k' R k

1 – ik' ⋅ ( R – R' )
= ----
N  Ψm ( R )  E0 ( k )e
R k'
1 ik' ⋅ R''
=  Ψm ( R' – R'' )  ----N E0 ( k )e
R'' k'

where the orthogonality of the unperturbed wave functions is used and the dummy
variable, R′′=R′-R is defined. Next we note that ground state energy, E0(k), is
periodic in k-space (see for example Figure 3-2), and can be written as a Fourier series
in the form:

– iR ⋅ k
E0 ( k ) =  ER e (3-30)
R

where R is the set of real space lattice vectors (real space forms a reciprocal space for
k-space). The coefficients of the series ER are given by:

1 – iR ⋅ k 3 1 iR ⋅ k
E R = ----
Ω  E0 ( k )e d k =  ----N E0 ( k )e (3-31)
BZ k

where the integral over the Brillouin zone (with volume Ω) is replaced with a
summation over the individual k-states in the zone for consistency with the notation
employed in this section. Note that there are N states in the Brillouin zone, as shown
in The Density of States in a Periodic Potential. Recognizing that the final term in
Equation 3-29 takes the same form as the definition of EK* given in Equation 3-31, it
is possible to write Equation 3-29 in the form:

136 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 1  0∗ k' ⋅ R'  – ik ⋅ R 3
  ψk'  Ψm ( R )  ψk ( r )e
0
----   ( r )e  H0   d r
N     
crystal k' R k

=  Ψm ( R' – R'' )ER''

R''

Finally, expanding Ψm(R′ − R′′) in a Taylor series about the point R′′ gives:

1
Ψ m ( R' – R'' ) = Ψ m ( R' ) – R'' ⋅ ∇Ψ m ( R' ) + ----- ( R'' ⋅ ∇ ) ( R'' ⋅ ∇ )Ψ m ( R' ) + …
2!
= exp ( – R'' ⋅ ∇ )Ψ m ( R' )

where the final step introduces an operator based on the Taylor series expansion of the
exponential function. Equation 3-29 can be written as:

1  0∗ k' ⋅ R'  – ik ⋅ R 3
  ψk'  Ψm ( R )  ψk ( r )e
0
----   ( r )e  H0   d r
N     
crystal k' R k
(3-32)
=  ER'' exp ( –R'' ⋅ ∇ )Ψm ( R' )
R''

Next an operator E0(-i∇) is introduced that results from replacing every instance of k
in the function E0(k) with -i∇. Comparing Equation 3-32 with equation
Equation 3-30 shows that:

1  0∗ k' ⋅ R'  – ik ⋅ R 3
  ψk'  Ψm ( R )  ψk ( r )e
0
----   ( r )e  H0   d r
N     
crystal k' R k

= E 0  --- ∇ Ψ m ( R' )
1
i 

Assembling the terms derived above into Equation 3-28 (with R′→r) gives the
following equation for Ψm(r):

( E 0 ( – i ∇ ) + H 1 )Ψ m ( r ) = E m Ψ m ( r ) (3-33)

Equation 3-33 is an equation for Ψm(r), similar in form to the Schrödinger equation,
with the perturbing potential H1 appearing as the potential energy and the operator
E0(-i∇) replacing the kinetic energy operator. This equation can be used to derive the
semiclassical model, which in turn determines the transport properties of electrons.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 137

 The Semiclassical Model
Consider the motion of a wave packet, constructed using the functions Ψm(r) which
modulate the Wannier functions. The wave function evolves according to
Equation 3-33. Note that the perturbing potential, H1, is a function of position r and
E0(-i∇) can be straightforwardly written as a function of the quantum mechanical
_
momentum operator p=-i h ∇. The equivalent classical Hamiltonian is
_ _
H=E0(p)+H1(r) (where E0 is written as a function of p= h k). The quantity h k,
which is analogous to the classical momentum in the Hamiltonian, is often referred to
as the crystal momentum of the wave packet. The classical Hamiltonian results in
classical particles that move according to the equations:

dr ∂H ∂E
-------- = --------- = ---------0-
dt ∂p ∂p
dp ∂H ∂H
--------- = – --------- = – ----------1-
dt ∂r ∂r

According to Ehrenfest’s theory (see Ref. 7 for a detailed discussion), the center of
gravity of the wave packet moves in the same way as the corresponding classical
Hamiltonian. Consequently the motion of a wave packet with associated charge -q, in
a perturbing Hamiltonian of the form H1 = −qV moves according to the equations:

∂r ∂E ( k )
1- ----------------
v ( k ) = ------ = ------
_ (3-34)
∂t h ∂k

_ ∂k ∂V
h ------- = q ------- = – qE = F (3-35)
∂t ∂r

Here v(k) is the velocity of the wave packet, and F is the Lorentz force acting on it.

These equations are referred to as the semiclassical model (in the absence of magnetic
fields). When magnetic fields are present Equation 3-35 takes the form:

_ ∂k
h ------- = – q ( E + v ( k ) × B )
∂t

(Ref. 1 provides references that derive this equation in full. Note also that
Equation 3-34 is also modified in modern solid state theory to include an additional
term due to the Berry curvature of the band. This term is usually zero for

138 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

semiconducting materials of interest). The model is extremely successful to describe a
range of practical transport phenomena.

The instantaneous velocity and momentum of a single electron fluctuates

rapidly as it passes through the periodic potential of the crystal, so the
crystal momentum is not directly related to the actual electron
momentum and the wave packet velocity should be interpreted as an
average drift velocity.

The semiclassical model is valid for wave packets that are localized to within greater
than approximately ten lattice spacings. This restriction is apparent from the previous
_
analysis but is also implied by the Heisenberg uncertainty principle Δx≈ h /Δp≈1/kf
where kf is the approximate magnitude of wave vectors at the Fermi surface. It would
be unreasonable to use more than approximately 10% of the available states in a single
wave packet. The wave packet is therefore large compared to the lattice size, and so
sees only the average effect of the lattice.

Additionally, the physical size of the wave packets must be small compared to the
length scale of electric or magnetic field variation. This is why the model is called
semiclassical; the external fields are treated classically, but the periodic potential of the
lattice is not. There is a further restriction on the local validity of the semiclassical
model, which results from the fact that in the limit of zero periodic potential the
classical limit should be recovered. In the classical limit the electron momentum and
the crystal momentum become identical. At a given point in k-space the semiclassical
equations are valid provided:

2
Eg ( k )
ea E « -------------------
Ef
2
e B Eg ( k )
---------- « -------------------
m Ef

where a is of the order of the lattice constant, Ef is the Fermi energy, and Eg(k) is the
energy difference to the nearest energy in a different band at the specified k-space
point. In semiconductors the first condition is violated during electric breakdown,
when electrons can make an interband transition driven by a large electric field. A
similar phenomenon known as magnetic breakdown can occur in high magnetic fields.
Although the semiclassical model really describes the transport of wave packets,

THEORY FOR THE SEMICONDUCTOR INTERFACE | 139

 conventionally the transport of electrons is discussed in the literature. This convention
is therefore adopted from here onward.

The semiclassical model has important consequences for charge transport in materials.
One of the more surprising consequences is that a filled band is completely inert. Since
the semiclassical model does not allow intraband transitions, the electrons can only
move between states within a band. Each state has a particular velocity associated with
it (given by Equation 3-34) and electrons move between states according to
Equation 3-35. In the reduced zone scheme, electrons that pass out of the Brillouin
zone immediately reappear unchanged (except for a relabeling of the k-vector) at the
opposite face. Summing the drift velocity over all the states in the band produces
exactly the same result in the presence of applied fields as without them, since the band
structure remains unchanged and since all the states are still occupied (see Ref. 1 for a
formal derivation that integrates over all the states in a band). This result explains the
existence of metals, insulators, and semiconductors. Metals are materials without a full
band, in which an applied electric field can produce a large current. Insulators have a
set of full bands and the Fermi energy lies within the band gap. The band gap is
sufficiently large that the occupancy of the bands below the Fermi level is essentially
one for all states at temperatures of interest. Similarly the band above the gap is
essentially unoccupied in an insulator. Semiconductors are materials in which the
Fermi level lies within the band gap, but in which the variation in the Fermi function
overlaps the edges of adjacent bands at temperatures of interest, so that not all the
states in the bands above (or below) the Fermi level are unoccupied (or occupied). This
is why semiconductors usually have a higher resistance than metals since only a
relatively small number of states in the band are unoccupied. The physics of
semiconductor transport is determined by the band structure close to the top or the
bottom of the bands adjacent to the Fermi level.

In 3D the band gap must extend across all directions in k-space to

produce a semiconductor or insulator.

Before considering how to compute the transport properties of bands that are not full,
it is worth considering the implications of Equation 3-34 and Equation 3-35 for the
relationship between force and electron velocity. In order to do so, the velocity must
be related to the wave vector k. Therefore:

2
∂v ∂v ∂k 1 ∂ E ( k )-
------ = ------- ------- = ---------
_ 2 ------------------ F (3-36)
∂t ∂k ∂t h ∂k
2

140 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 _
Note that both ∂v/∂k and h -2∂2E(k)/∂k2 are rank 2 tensors (matrices).
Equation 3-36 defines the effective masses introduced in Equation 3-20.

In general, the effective mass matrix is defined as:

2 2 2
E ( k )- ∂------------------
∂------------------ E ( k )- ∂------------------
E ( k )-
2 ∂k ∂k ∂k ∂k
∂k x x y x z

–1 1 ∂2 E ( k ) ∂2 E ( k ) ∂2 E ( k )
m∗ = ---------
_ 2 ------------------- ------------------
- -------------------
h ∂k y ∂k x ∂k y ∂k y ∂k z
2

2 2 2
∂------------------
E ( k )- ∂------------------
E ( k )- ∂------------------
E ( k )-
∂k z ∂k x ∂k z ∂k y ∂k 2
z

such that F = m* v. In Equation 3-20 the coordinate system for k' was chosen such
that the m* matrix has zero for the off-diagonal terms. For an isotropic band m* is
proportional to the identity matrix. In this instance, near the top of a band, it is easy
to see that its value could be negative. Instead of thinking in terms of negative mass, it
is more conventional to reverse the sign on the force and to consider a wave packet
corresponding to a particle with positive charge and mass (this is known as a hole).

Finally consider the form of Equation 3-33 in the case where the band structure takes
the form given by Equation 3-20. For simplicity consider the case of an isotropic
effective mass so that:

_2
h 2
E ( k ) = E 0 + ------------ k
2m∗

Equation 3-33 now takes the form:

_2
 – -----------
h - 2
∇ + E 0 + H 1 Ψ m ( r ) = E m Ψ m ( r ) (3-37)
 2m∗ 

which is identical in form to the Schrödinger equation, except that the effective mass
rather than the electron mass appears in the equation system. It is important to
remember that for Equation 3-37 to apply the E–k relationship must be defined in an
appropriate coordinate system. Note, however, that for the case of an anisotropic
effective mass there are different coefficients for the derivatives in different directions
in the Laplacian operator. Equation 3-37 also applies near the top of a band, so it is
relevant for holes.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 141

 The Boltzmann Equation

TRANSPORT PROPERTIES AND HOLES

To compute the transport properties of a material without full bands it is necessary to
know the occupancy of the states within each of the partially filled bands. Under the
influence of an applied force the distribution becomes distorted from that obtained at
thermal equilibrium (Equation 3-5) and additionally can vary in space (for example in
the presence of nonuniform electric fields, temperatures, or material properties) and
time. The distribution function (f (m, r, t)) is related to the number of electrons
(δn(k, r) in a volume element δ3k of k-space and δ3r of real space at point k, r in the
following way:

1 3 3
δn ( k ,r ) = ---------3 f ( k ,r ,t )δ kδ r
4π

Transport properties can be derived directly from the distribution function, for
example the current density is given by:

q q 1- ∂E (k)
 
3 3
J n ( r ,t ) = – ---------3 v ( k )f ( k ,r ,t ) d k = – ---------3 ------
_ ---------------- f ( k ,r ,t ) d k (3-38)
4π 4π h ∂k

Here the integral is performed over the first Brillouin zone in k-space. Since f (k, r, t)
gives information about the electron occupancy of states, at a given instant in time
Equation 3-38 can be rewritten in the following form:

q

3
J ( r ,t ) = – ---------3 v(k )d k
4π occupied

Noting that:

  
3 3 3
v(k )d k = v( k)d k + v( k)d k = 0
band occupied unoccupied

it is clear that the current density can be written in the form:

q q
 
3 3
J p ( r ,t ) = ---------3 v ( k ) d k = ---------3 v ( k ) ( 1 – f ( k ,r ,t ) ) d k (3-39)
4π unoccupied 4π

Equation 3-39 is useful when computing the current contribution from a band that is
nearly full, as only a small fraction of the band needs to be considered in the integral.
The current is exactly the same as that which would be produced if the unoccupied

142 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

electron states were occupied by positively charged particles (referred to as holes) and
the occupied electron states were empty. The hole occupancy factor, f h(k, r, t), is
simply given by:

f h ( k ,r ,t ) = 1 – f ( k ,r ,t )

and holes respond within the semiclassical equations as if they had a charge opposite
to that of electrons.

THE BOLTZMANN EQUATION

Equation 3-38 shows that the time- and space-dependence of electric currents in a
semiconductor is determined by the distribution function f (k, r, t). Within a volume
element δ3kδ3r there are f (k, r, t)δ3kδ3r/4π3 electrons. In the presence of an
external force the electrons move through real space at a velocity v (k) (given by
Equation 3-34) and through k-space at a velocity ∂k/∂t (from Equation 3-35). At a
time f (k, r) the electrons that occupy the volume element are those which were at k
− δk, r − δr, at time t − δt. Therefore:

f ( k ,r ,t ) = f ( k – δk ,r – δr ,t – δt )
∂k ∂r ∂f
f ( k ,r ,t ) = f ( k ,r ,t ) –  ------- ⋅ ∇ k f + ------ ⋅ ∇ r f + ----- δt
 ∂t ∂t ∂t

Here the subscript to the gradient term indicates that derivatives are taken with respect
to the coordinates indicated. Therefore, in the absence of collisions:

∂f- ∂k ∂r
---- = – ------- ⋅ ∇ k f – ------ ⋅ ∇ r f
∂t ∂t ∂t

Note that the assumption has been made that the size of the volume element (δ3kδ3r)
remains unchanged in time — Liouville’s theorem asserts that it is (see Ref. 1). In
practice electrons collide with defects in the lattice (impurities) and with lattice
distortions caused by mechanical waves in the lattice (phonons). The electrons do not
collide with the lattice itself as the effect of the lattice is already incorporated into the
Bloch functions. However, any deviation from perfect periodicity produce collisions.
The effect of collisions is to add additional time dependence to the equation system,
so the full Boltzmann equation is given by:

∂f ∂k ∂r ∂f
----- = – ------- ⋅ ∇ f – ------ ⋅ ∇ r f + -----
∂t ∂t k ∂t ∂t coll

Using Equation 3-35 and substituting ∂r/∂t = v gives the final form of the Boltzmann
equation for electrons:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 143

 ∂f- 1- ∂f
---- _ F ⋅ ∇ f – v ⋅ ∇ r f + -----
= – ------ (3-40)
∂t h k ∂t coll

The hole distribution function can be shown to obey a similar Boltzmann equation:

∂f 1- ∂f h
-------h- = – ------
_ F h ⋅ ∇ f h – v h ⋅ ∇ r f h + --------
∂t h k ∂t coll

These equations are difficult to solve and consequently approximations are usually
made to simplify them significantly.

Doped Semiconductors
In practical applications it is common to modify the properties of the semiconductor
to add impurity atoms or dopants in a controlled manner. Using silicon as an example
(with a chemical valency of 4), impurity atoms such as phosphorous or arsenic (with a
valency of 5), have the effect of adding extra electrons to the conduction band. Such
impurities are called donors. Similarly impurities such as boron and gallium (with a
valency of 3) can capture an electron from the valence band, creating holes. These
impurities are called acceptors.

Consider the case of a donor atom. If the atom loses its additional electron, it has
enough remaining electrons to fit into the lattice of the crystal. However the atom is
now positively charged, and consequently a perturbing coulomb potential exists
around it. Equation 3-37 can be used to compute the states that the additional
electron can occupy within this potential. The problem becomes that of the hydrogen
atom, except that the medium typically has a large relative dielectric permittivity (εr≈12
for silicon) and the electron has an effective mass m* that is typically less than the
electron mass. The solutions to the corresponding Schrödinger equation are swollen
orbits (typically 2.5–5.0 nm in radius) similar to that of a hydrogen atom. The reduced
mass and larger dielectric constant cause a reduction in the effective ionization energy
of the electron so that it is significantly less than that of a hydrogen atom — donor
ionization levels are typically 10s of meV. The energy of the donor states is
consequently just below that of the conduction band as shown in Figure 3-5. Similarly,
the energy of the acceptor states lies just above the valence band (also in the figure).
Figure 3-5 shows an n-type semiconductor. This is a material with significantly more
donors than acceptors so that the majority carriers are electrons.

144 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 E

Ec
Ed

Ea
Ev

g(E)
f(E,T=0)
f(E,T>0)

Figure 3-5: Density of states and Fermi function for an n-type semiconductor. The Fermi
function is shown at zero temperature and at a finite temperature.

Since donor and acceptor states cannot hold two electrons (unlike states within the
band), the carrier statistics are modified slightly. The occupancy of the states in
equilibrium is determined from statistical mechanics as:

+
Nd 1
------- = ---------------------------------------------------
- (3-41)
Nd Ef – Ed
1 + g d exp  -------------------
kB T

where gd is the donor degeneracy factor, which is 2 if the conduction band minimum
is nondegenerate but which varies when the degeneracy of the donor levels is altered
by the band structure having degenerate conduction bands. Similarly, the occupancy
of the acceptor states is given by:

–
Na 1
- = ---------------------------------------------------
--------- (3-42)
Na Ea – Ef
1 + g a exp  -------------------
 kB T 

where ga is the acceptor degeneracy factor.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 145

 In equilibrium the value of the Fermi level is determined self-consistently from the
charge neutrality condition, along with the requirement that the number of holes and
electrons is determined from the Fermi function:

+ –
Nd – Na + p – n = 0

n =  gc ( E )f ( E ) dE
p =  gv ( E )fh ( E ) dE
Here gc(E) is the density of states in the conduction band, gv(E) is the density of states
in the valence band, and f(E) is the (equilibrium) Fermi function.

Specifying the Dopant Distribution

The distribution of dopants within the semiconductor can be specified using the
Analytic Doping Model and Geometric Doping Model features. The Analytic Doping
Model enables distributions to be defined in terms of the coordinate system; the
Geometric Doping Model enables definitions in terms of the distance from selected
boundaries in the geometry.

Doping

THEORY FOR THE ANALYTIC DOPING MODEL

The Analytic Doping Model has two options to specify the doping concentration in
terms of the coordinate system — User defined or Box. It supports the use of a local
coordinate system that is rotated relative to the global coordinates.

User Defined Profiles

For user defined profiles any arbitrary dopant concentration can be created with
user-defined functions or imported data (which can be specified using an interpolation
function). Any expression written in terms of the local coordinate system can be used
to define the distribution. Use of the COMSOL Multiphysics built-in functions
enables the creation of intricate distributions.

Box Profiles
The Box profile enables a box-shaped region of uniform doping to be defined away
from which the dopant concentration decays with one of three preset functions. The

146 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

location of the region of constant doping is defined by specifying either the corner or
center coordinate using the global coordinate system. The size of the region is defined
by specifying its extent in each of the directions parallel to the local coordinate axis. In
3D the height, width, and depth are supplied to create a block; in 2D the width and
depth are supplied to create a rectangle; and in 1D the width is supplied to create a
line. It is possible to set the constant region to have zero extend in a given direction
to, for example, create a plane of constant doping in a 3D geometry. The orientation
of the constant region is aligned with the local coordinate axes. If a rotated local
coordinate system is used the constant region rotates around its specified corner or
center coordinate, which is expressed using the global coordinate system.

The distribution outside of the constant region decays with one of three preset
functions: Gaussian, Linear, or Error Function. These functions are defined using
Ramp functions with unity gradient that begin at the edges of the constant region and
which continue throughout the remainder of the domain. Thus there are two Ramp
functions for each geometry axis. Figure 3-6 shows the four Ramp functions used to
create a 2D Box profile. Note how each Ramp function is zero inside the constant
region, and that the unity gradient outside the region creates and effective coordinate
axis for each direction that is zeroed at the region boundary.

In 3D, the Gaussian distribution is given by

– 2 + 2 – 2 + 2 – 2 + 2
  rx   rx   ry   ry   rz   rz  
N d, a = N 0 exp  –  ------ +  ---------- +  ------ +  ---------- +  ------ +  ----------  (3-43)
  x l  x l l
 y  y l l
 z  lz  

where Nd,a is the donor or acceptor concentration, N0 is the concentration inside the
uniformly doped region, li is the Decay length in the i direction, and rx-, rx+, ry-, ry+,
rz-, and rz+ are the Ramp functions.

In models with less than three dimensions the unused Ramp functions are
omitted from this expression. Also, 2D-axisymmetric profiles are
obtained in the same manner but using ramp functions rr-, rr+, rz-, and
rz+.

Functions and Ramp in the COMSOL Multiphysics Reference Manual.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 147

 The decay length can be specified directly or calculated from a specified junction depth
via:

d j, i
l i = ---------------------------
N
ln  ------0- 
Nb

where dj,i is the junction depth in the i direction, and Nb is the background doping
concentration that can be entered directly or taken from the output of another doping
feature in the model. By default dj,x=dj,y=dj,z=dj, such that lx=ly=lz. To use different
decay lengths in different directions select the Specify different length scales for each
direction check box under the Profile section.

The linear distribution is given by

– + – + – +
N a, d = N 0 + g x ( r x + r x ) + g y ( r y + r y ) + g z ( r z + r z ) (3-44)

where gi is the gradient in i direction. The gradient can be supplied directly or

calculated from a specified junction depth via:

Nb – N0
g i = ---------------------
d j, i

By default the gradient is the same in all directions, however it can be set to be
direction dependent using the Specify different length scales for each direction check
box. Note that a negative dopant distribution is not physical, so the concentration is
set to zero in regions where Equation 3-44 gives Na,d<0.

The Error Function distribution is given by

– + – + – +
N a, d = N 0 [ 1 – erf ( m x ( r x + r x ) + m y ( r y + r y ) + m z ( r z + r z ) ) ]

where mx is an argument factor which controls the length scale of the profile. The
argument factor can be entered directly or calculated from a specified junction depth
via:

Nb
m i = --------- erfinv  1 – -------
1
d j, i  N 0

By default the argument factor is the same in all directions, however it can be set to be
direction dependent using the Specify different length scales for each direction check
box.

148 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 r x- rx+

ry- ry+

Figure 3-6: Ramp functions used to specify box doping profiles on a rectangular domain
in 2D. The height out of the plane shows the magnitude of the ramp function (not to scale).
Top: Individual ramp functions for the x direction. Middle: Individual ramp functions
for the y direction. Bottom: Composite of all the ramp functions, showing the region of
constant doping (highlighted in red).

THEORY FOR THE GEOMETRIC DOPING MODEL

The Geometric Doping Model feature enables doping distributions to be defined in
terms of the distance from selected boundaries in the geometry. This is convenient
when working with geometries with intricate shapes that would be challenging to

THEORY FOR THE SEMICONDUCTOR INTERFACE | 149

 describe analytically using the coordinate system. The boundaries from which the
distance is calculated are selected using the Boundary Selection for Doping Profile
node. Any boundary that bounds, or is within, the domains to which the
corresponding Geometric Doping Model feature is applied can be selected. The form
of the distribution can be selected from a range of preset functions or a user defined
expression can be supplied.

User Defined Profiles

When a user defined profile is selected any arbitrary distribution that can be written in
terms of the distance from the selected boundaries can be defined. This distance is
available as the variable semi.gdm#.D, where # corresponds to the number of the
Geometric Doping Model feature.

Preset Profiles
The other selections in the Dopant profile away from the boundary list allow either
Gaussian, Linear, or Error Function profiles to be generated. These profiles are defined
in terms of the distance, D, from the selected boundaries as described below.

The Gaussian profile is given by

D 2
N a, d = N 0 exp  –  ---- 
  l 

where Na,d are the concentration of the acceptors or donors, N0 is the concentration
of dopants at the selected boundaries, and l is the decay length of the Gaussian
function. The decay length can be entered directly or can be calculated from a specified
junction depth, dj, via:

dj
l = ---------------------------
N
ln  ------0- 
Nb

where Nb is a specified background doping concentration that can either be directly

defined or taken from the output of another doping model feature.

The Linear profile is given by

N a, d = N 0 + gD (3-45)

where g is the gradient, which can be entered directly or calculated from a specified
junction depth via

150 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 Nb – N0
g = ---------------------
dj

Note that, as a negative concentration is not physical, the dopant concentration is set
to zero in regions where Equation 3-45 gives Na,d<0.

The Error Function profile is given by

N a, d = N 0 ( 1 – erf ( mD ) )

where m is an argument factor which controls the length scale of the profile. The
argument factor can be defined directly or calculated from a specified junction depth
via:

Nb
m = ----- erfinv  1 – -------
1
dj  N 0

USING ANALYTIC AND GEOMETRIC DOPING FEATURES TOGETHER

A doping distribution can be created by combining the functionality of the two doping
model features. Often it is desirable to specify a domain of constant doping using a user
defined profile in the Analytic Doping Model feature, and then to set the profile away
from this region using the Geometric Doping Model feature. This approach has the
advantage that the constant user defined profile can be assigned to a domain of any
shape, thus removing the requirement for the uniformly doped region to be
block-shaped that is imposed by a Box profile. The use of the Geometric Doping
Model to specify the profile away from the constant domain has the advantage that it
can accommodate a constant region that has curved boundaries. This is because the
Geometric Doping Model profile only depends on the distance from the selected
boundaries, rather than on the coordinate system.

Equilibrium Carrier Concentrations

The carrier concentrations at equilibrium are given by the equations:

gc ( E )
n =  1------------------------------------------------------------------
+ exp [ ( E – E f ) ⁄ ( k B T ) ]
dE

gv ( E )
p =  1---------------------------------------------------------------------
+ exp [ ( E f – E h ) ⁄ ( k B T ) ]
dE

THEORY FOR THE SEMICONDUCTOR INTERFACE | 151

 The variables W and Wh are defined such that E = Ec + W and Eh = Ev −Wh. The
density of states for electrons and holes then becomes (from Equation 3-21 and
Equation 3-22):

1 2m∗ 3 ⁄ 2 1 ⁄ 2
g c ( E ) = ---------2  -----------
- W
 _ 2
2π h
3⁄2
1  2m h∗ 1⁄2
g v ( E ) = ---------2  --------------
_ 2 - Wh
2π  h 

So the electron and hole densities are written in the form:

∞ 1⁄2
1 2m∗- 3 ⁄ 2 -----------
n = ---  ------------
2 W
4 πh _ 2
π
1⁄2  --------------------------------------------------------------------------------
0 1 + exp [ ( W + E c – E f ) ⁄ ( k B T ) ]
- dW

3⁄2 1⁄2
1  2m h∗- 2 ∞ Wh
p = ---  --------------
4 π h _ 2

-----------
π
1⁄2  0
-----------------------------------------------------------------------------------
1 + exp [ ( W h + Ef – Ec ) ⁄ ( kB T ) ]
- dW h

The following equations are obtained with the substitutions ε=W/(kBT):

3⁄2
 m∗ k B T 2 ∞
ε
1⁄2
n = 2  -------------------
 2π h 2 
_ - -----------
π
1⁄2 0 ----------------------------------------------------------------------------
1 + exp [ ε – ( E f – E c ) ⁄ ( k B T ) ]
- dε

3⁄2 1⁄2
 m h∗ k B T 2 ∞ εh
p = 2  ----------------------
 2π h 2 
_ - -----------
π
1⁄2  --------------------------------------------------------------------------------
0 1 + exp [ ε h – ( E v – E f ) ⁄ ( k B T ) ]
- dε h

The jth order Fermi-Dirac integral (Fj(η)) is defined in the following manner:

∞ j
1 ε η
F j ( η ) = --------------------
Γ(j + 1) 0 ------------------------------------
1 + exp [ ε – η ]
- dε = – Li j + 1(– e ) (3-46)

where Γ is the gamma function and Li the polylogarithm (see Ref. 15 for a brief review
on the properties of the Fermi-Dirac integral). Note that Γ(3/2)= π(1/2)/2. The
electron and hole densities can be written in the compact forms:

Ec – Ef
n = N c F 1 ⁄ 2  – ------------------
 kB T 
(3-47)
Ef – Ev
p = N v F 1 ⁄ 2  – -------------------
 kB T 

where:

152 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 3⁄2
 m∗ k B T
N c = 2  -------------------
_ -
 2π h 2 
(3-48)
3⁄2
 m h∗ k B T
N v = 2  ----------------------
_ - 
 2π h 2 

For the density-gradient formulation (Ref. 45), Equation 3-47 is replaced by one
similar to Equation 3-53 below.

One of the properties of the Fermi-Dirac integral is that Fj(η)→eη as η→ −∞ (this

result applies for all j). In semiconductors this limit is known as the nondegenerate
limit and is often applicable in the active region of semiconductor devices. In order to
emphasize the nondegenerate limit, Equation 3-47 is rewritten in the form:

Ec – Ef
n = N c γ n exp  – ------------------
 kB T 
(3-49)
Ef – Ev
p = N v γ p exp  – -------------------
 kB T 

where:

Ec – Ef
F 1 ⁄ 2  – ------------------
 kB T 
γn = ----------------------------------------
Ec – Ef
exp  – ------------------
 kB T 
(3-50)
Ef – Ev
F 1 ⁄ 2  – -------------------
 kB T 
γ p = ----------------------------------------
Ef – Ev
exp  – -------------------
 kB T 

In the nondegenerate limit, the Fermi-Dirac distribution reverts to the

Maxwell-Boltzmann distribution and γn and γp are 1. By default, COMSOL
Multiphysics uses Maxwell-Boltzmann statistics for the carriers, with γn and γp set to 1
in Equation 3-49, irrespective of the Fermi level. When Fermi-Dirac statistics are
selected Equation 3-50 is used to define γn and γp.

In the Semiconductor Equilibrium study step, the above equations are used to express
the carrier concentrations as functions of the electric potential, and the electric
potential is solved via the resulting partial differential equations.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 153

 Away from equilibrium the above equations can still be applied but instead of using
the Fermi level, the quasi-Fermi level is employed in Equation 3-41 and
Equation 3-42 (Ref. 11). For a detailed description of the origin of the quasi-Fermi
levels for electrons and holes see The Semiconductor Equations. The final results are:

E fn – E c E c – E fn
n = N c F 1 ⁄ 2  --------------------- = N c γ n exp  – ---------------------
kB T kB T
(3-51)
E v – E fp E fp – E v
p = N v F 1 ⁄ 2  ---------------------- = N v γ p exp  – ----------------------
 kB T   kB T 

and:

E c – E fn
F 1 ⁄ 2  – ---------------------
 kB T 
γn = -------------------------------------------
E c – E fn
exp  – ---------------------
 kB T 
(3-52)
E fp – E v
F 1 ⁄ 2  – ----------------------
 kB T 
γ p = -------------------------------------------
E fp – E v
exp  – ----------------------
kB T

The formulas above are used to compute the carrier concentrations as functions of the
quasi-Fermi levels, when the quasi-Fermi level formulation is used. In other
formulations where the carrier concentrations are solved for, the formulas above are
used to compute the quasi-Fermi levels as functions of the carrier concentrations.

For the density-gradient formulation (Ref. 45), the formulas above are modified with
additional contributions from the quantum potentials VnDG and VpDG (SI unit: V):

DG
 E fn – E c + qV n 
n = N c F 1 ⁄ 2  --------------------------------------------
 kB T 
(3-53)
DG
 E v – E fp + qV p 
p = N v F 1 ⁄ 2  --------------------------------------------
 kB T 

where the quantum potentials VnDG and VpDG are defined in terms of the density
gradients:

154 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 n DG
∇ ⋅ ( b n ∇ n ) = ------- V n
2
(3-54)
p DG
∇ ⋅ ( b p ∇ p ) = ------- V p
2

The density-gradient coefficients bn and bp (SI unit: V m^2) are given by the inverse
of the density-gradient effective mass tensors mn and mp (kg):

_2 –1
h
b n = ---------- m n
12q
(3-55)
_2 –1
h
b p = ---------- m p
12q

where hbar is the Planck constant.

The Semiconductor Equations

NONUNIFORM BAND STRUCTURE AND THE REFERENCE ENERGY

Most practical simulations deal with a band structure that varies in space — it is
therefore necessary to define quantities such as the band energies with respect to a
reference energy. In order to make the definition of boundary conditions simpler, the
reference energy chosen is the Fermi energy in an equilibrium state when no potentials
are applied to any of the boundaries in the system and when there are no thermal
gradients in the system (note that in COMSOL Multiphysics, the reference
temperature used to define the equilibrium Fermi energy can be changed in the
Semiconductor interface’s Settings window under Reference Temperature). This is
illustrated in Figure 3-7 for an isothermal, abrupt p-n junction (that is a boundary
between a p- and n-doped region of a semiconductor with a constant doping level on
the two sides of the device). In equilibrium the Fermi energy is well-defined and is
constant throughout space (since there are no gradients in the Fermi level/chemical
potential). As a result in the immediate vicinity of the junction, the bands bend to
accommodate a constant Fermi energy. In this region of band bending the Fermi level
lies close to the center of the gap and consequently the carriers are depleted — this is
known as the depletion region. A space charge layer is associated with the depletion
region since the charges of the ionized donors and acceptors are no longer
compensated by free carriers. It is the space charge layer that generates an electric field
and the corresponding potential gradient that results in the bending of the vacuum
level, E0, and the conduction band and valence band edges (Ec and Ev, respectively).

THEORY FOR THE SEMICONDUCTOR INTERFACE | 155

 This results in a self-consistent picture of the band structure in the vicinity of the
junction. As a result of the bending of the vacuum level and bands, a built-in potential,
Vbi, develops across the junction. When an additional potential difference is applied to
the p-side of the junction, Vp, the junction is in a condition known as reverse bias. The
Fermi energy is no longer well-defined in the vicinity of the depletion region, but the
equilibrium Fermi level, Ef0 is still used as a reference potential. Away from the
depletion region the semiconductor is close to equilibrium and the concept of the
Fermi level can still be applied. On the p-side of the junction, the Fermi energy shifts
up to a value Ef1, which differs from Ef0 by qVp where q is the electron charge.

In the most general case the electron affinity, qχ = E0 − Ec, and the band gap, Eg = Ec –
Ev, can vary with position. However, the values can be considered to be material
properties. This means that the conduction band edge cannot be parallel to the valence
band edge or to the vacuum level (this is not shown in Figure 3-7).

156 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 Equilibrium Vp=0
depletion region

E0
qVbi
E0
qψ qψ qψ
Ec
qVbi
Ec
Ef0 0
Ev
Eg

Ev
nonequilibrium Vp>0
depletion region

E0
qVp

qVbi

Ec E0
qVp

qψ qψ qVbi qψ
Ef1 qVp
Ev Ec
Ef0 0
Eg

Ev
p-type n-type

Figure 3-7: Diagram showing the variation in the band structure in a p-n diode. The
horizontal axis represents position in the device and the vertical axis indicates energy. The
vacuum level E0, the bottom of the conduction band Ec, and the top of the valence band
Ev, are shown. Top: The band structure when there is no applied potential. A built-in
potential, Vbi, develops across the junction as a result of the space charge layers associated
with the depletion region. The Fermi energy Ef0, is constant throughout the device. Bottom:
A potential Vp is applied to the p-type side of the junction, resulting in a reverse bias. Away
from the junction, where the material is close to equilibrium, a second Fermi level, Ef1,
can be defined in the p-type region–shifted by the applied potential.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 157

 QUASI-FERMI LEVELS
Consider the motion of an individual electron moving in the curved conduction band
(see Figure 3-8). In between collisions the energy of the electron remains
approximately constant (although the force accelerates the electron, its velocity only
increases slightly in between collisions). The energy of the electron measured with
respect to the band edge, W, increases as a result of band bending. The total energy
associated with an electron is given by:

E ( k ,r ) = E c ( r ) + W ( k ,r )

Similarly the energy of a hole is defined with respect to the band edge in the following
manner:

E h ( k ,r ) = E v ( r ) + W h ( k ,r )

Collision

W
Collision

Ec

Figure 3-8: Motion of an individual electron within a curved band.

It is assumed that scattering processes within the band cause the electrons (or holes)
to reach a collective equilibrium on time scales that are small compared to the
simulation time scales (if highly nonequilibrium/hot electron effects are important
within the band, then it is necessary to account for the deviation of the Fermi function
from a Fermi-Dirac distribution by solving an additional energy equation — currently
this is not possible in the Semiconductor interface). Since the relaxation time for
electrons within the conduction band is much less than the corresponding relaxation

158 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

time for transitions across the band gap, the electron and hole distributions can have
different associated Fermi levels (Efn for electrons and Efp for holes). In
nonequilibrium circumstances (for example in the presence of an electric field or a
thermal gradient) these quasi-Fermi levels are not coincident. The electron and hole
quasi-Fermi levels typically lie within the band gap and consequently the electrons are
scattered to states close to the band edge as they travel through space, as shown in
Figure 3-8.

SIMPLIFYING THE BOLTZMANN EQUATION

As discussed in The Semiclassical Model, the evolution of the distribution function for
electrons and holes is governed by the Boltzmann equation (see Equation 3-40). Since
this equation is difficult to solve, it is common to make simplifying approximations in
its solution. Ultimately these approximations produce the drift-diffusion equations.
The analysis presented here is based on Ref. 12 and Ref. 13.

The first step taken when simplifying the Boltzmann equation is usually to make the
relaxation-time approximation, which assumes that the collision terms in the
Boltzmann equations take the form:

∂f- f – f 0 ( E , Efn , T )
---- = – -----------------------------------------
∂t coll τ( W)

h
∂f f h – f 0 ( E h , Efn , T )
-------h- = – ------------------------------------------------
∂t coll τh ( Wh )

Where f0 (E, Efn, T) is the Fermi function for electrons and f0h (E, Efp, T) is the Fermi
function for holes. This is essentially the simplest form of the collision term that returns
the electrons to the Fermi-Dirac distribution desired. In practice this assumption is a
significant simplification; Ref. 1 and Ref. 2 provide more detailed discussions of
collision mechanisms in real solids.

From Hamilton’s equations the spatial gradient of the electron total energy is equal to
the rate of change of crystal momentum, which in turn is related to the applied force
(Equation 3-35 in The Semiclassical Model), so that:

F = – ∇ E = – ∇ E c – ∇W

F h = – ∇ E h = ∇E v – ∇W h

Using these results the Boltzmann equations become

THEORY FOR THE SEMICONDUCTOR INTERFACE | 159

 ∂f- ------
1 f – f0
---- – _- ( ∇E c + ∇W ) ⋅ ∇ k f + v ⋅ ∇ r f = – ------------
∂t h τ

h
∂f 1- fh – f0
-------h- + ------
_ ( ∇E v – ∇W h ) ⋅ ∇ f h + v h ⋅ ∇ r f h = – ---------------
-
∂t h k τh

For compactness, the explicit dependence of f0 and τ is dropped. Considering first the
electron density, the assumption that the electron distribution function is close to its
quasi-equilibrium state allows the distribution function to be written in the form:

f = f0 + f1

where f1 << f0. The spatial gradient terms are dominated by terms involving the
gradient of f0, so the f1 terms are neglected in comparison to these. Likewise the term
df/dt is also small since the deviation from the quasi-equilibrium is small and the f0
term varies on time scales much slower than the collision time if the temperature is a
function of time. The Boltzmann equation can be expressed in the approximate form:

1- f
_ ( ∇E c + ∇W ) ⋅ ∇ f 0 = – ----1
v ⋅ ∇ r f 0 – ------ (3-56)
h k τ

Next the gradients in f0=1/(1+exp[(E-Efn)/(kbT)]) are expressed in terms of

parameters related to the band structure, using the chain rule:

∂f 0 _ ∂f 0
∇ k f 0 = -------- ∇ k E = h v -------- (3-57)
∂E ∂E

where the last step follows from the semiclassical definition of velocity (Equation 3-34
in The Semiclassical Model). From the chain rule the spatial gradient of f0 is given by:

∂f 0 ∂f 0 ∂f 0
∇f 0 = -------- ∇E + ------------ ∇E fn + -------- ∇T
∂E ∂E fn ∂T

Since f0=f0([E-Efn]/T):

1- ∂f ∂f 0
1- -----------
-------0- = – ---
E – E fn ∂f 0
--- - = – ------------------
- --------
T ∂E T ∂E fn T
2 ∂E

therefore:

160 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 ∂f 0  E – E fn 
∇f 0 = --------  ∇E – ∇E fn – ------------------
- ∇T (3-58)
∂E  T
2


Substituting Equation 3-58 and Equation 3-57 into Equation 3-56, using the
definition E = Ec + W and rearranging gives:

∂f 0  E – E fn 
f 1 = – τ --------  v ⋅ ∇E fn + ------------------
- v ⋅ ∇T
∂E  T
2


By definition the number of electrons is given by:

1 1
 
3 3
n = ---------3 fg ( k ) d k = ---------3 f 0 g ( k ) d k
4π 4π

The final equality holds because f1 is an odd function when integrated over the region
of k-space in the vicinity of a band edge and f0 is an even function (assuming, without
loss of generality, that the origin of k-space at the band minimum). To see why this is
the case, consider Equation 3-38 in The Semiclassical Model (shown below for
convenience).

q q ∂E ( k )
1- ----------------
 
3 3
J n ( r ,t ) = – ---------3 v ( k )f ( k ,r ,t ) d k = – ---------3 ------
_ f ( k ,r ,t ) d k
4π 4π h ∂k

In equilibrium (f = f0) no current flows so the integrand must be an odd function

(since it is not zero over all k-space). Since the origin is at a band minimum, ∂E(k)/
∂k is odd so f0 must be even. Deviations from equilibrium produce a current without
changing the total number of electrons, so f1 must be an odd function. The electron
current density can be written in the form:

∂f 0 T 3
J n ( r ,t ) = – ---------3 v ( k )f 1 d k = ---------3  τ -------- vv d k ⋅ ∇E fn
q q
 
3
 ∂E
4π 4π
E c – E fn ∂f 0 T 3 1 ∂f 0 
 
T 3
+ --------------------- τ -------- vv d k ⋅ ∇T + ---- τ -------- Wvv d k ⋅ ∇T 
T
2 ∂E T ∂E 

Note that the definition E=Ec+W is used to split the energy terms.

The rigid band assumption is made next. This asserts that even when the bands bend,
the functional form of W in k-space (measured with respect to the band edge) is
unchanged. Thus W = W(k). Given this assumption the quantities inside the integrals
are dependent only on the local band structure, except for the df0/dE term.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 161

 Next define the quantities:

q ∂f 0 T 3
4π n ∂E
3 
μ n = – ------------- τ -------- vv d k (3-59)

q ∂f 0

T 3
Q n = – ------------- τ -------- Wvv d k (3-60)
4π n
3 ∂E

To obtain:

n
J n ( r ,t ) = nμ n ∇E fn + ---- ( ( E c – E fn )μ n + qQ n )∇T (3-61)
T

Qn and μn are tensor quantities (although most semiconducting materials are cubic) so
that the corresponding tensors are diagonal with identical elements and consequently
can be represented by means of a scalar (cubic materials are assumed in the
Semiconductor interface). Careful examination of Equation 3-59 and Equation 3-60
shows that these cannot straightforwardly be considered material constants because
both depend on the quasi-Fermi level through the quantities df0/dE and 1/n. In the
nondegenerate limit, the quasi-Fermi level dependence of these two quantities cancels
out, since:

1/n∝exp[-Efn/(kBT)] and ∂f0/∂E∝exp[-Efn/(kBT)]

These quantities can therefore only strictly be considered material constants in the
nondegenerate limit. In the degenerate limit it is, however, possible to relate Qn to the
mobility for specific models for the relaxation time (see below). In principle a mobility
model could be used, which is dependent on the electron quasi-Fermi level (or the
electron density) in the degenerate limit.

Exactly the same arguments are applied for holes, leading to the equation for the hole
current:

p
J p ( r ,t ) = – p μ p ∇E fp – ---- ( ( E v – E fp )μ p + qQ p )∇T (3-62)
T

Here, the symbols introduced with the subscript change n→p have the same
definitions as the electron quantities except that the relevant integrals are over the hole
band.

Equation 3-61 and Equation 3-62 describe the evolution of the quasi-Fermi levels
within a semiconductor. It is possible to formulate the whole equation system so that
the hole and electron quasi-Fermi levels are the dependent variables. This is

162 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

implemented in some of the discretization options available in the Semiconductor
Module (the quasi-Fermi level formulation and the density-gradient formulation).
Before deriving the more familiar drift-diffusion equations it is useful to derive the
relationship between Qn and μn, and Qp and μp.

R E L A T I N G Qn T O μn
For isotropic materials within the rigid band approximation, Equation 3-20 can be
written as:

_2
h 2 2 2
W ( k ) = E ( k ,r ) – E c ( r ) = --------- ( k x + k y + k z ) (3-63)
m∗

Since Ec (r) is independent of k the derivatives with respect to E can be replaced by

derivatives with respect to W, enabling the (scalar) mobility to be written as:

q ∂f 0  ∂W ( k ) 2 3

_ 2 3 - τ --------
μ n = – -----------------------
4h π n ∂W
------------------
 ∂k x  d k

_
where the semiclassical result v=(1/ h )(∂E(k)/∂k) is used and the xx element of the
mobility tensor is evaluated to compute the scalar (the yy and zz elements produce the
same result). From Equation 3-63:

_2
∂W ( k )- 2 h kx
----------------- = -------------------
∂k x m∗

The integral in the mobility expression can be transformed to spherical polar

coordinates using the definitions:

k x = k sin ( θ ) cos ( φ )
k y = k sin ( θ ) sin ( φ )
k x = k cos ( θ )

and a functional form can be assumed for the relaxation time:

r
τ ( W ) = τ0 W (3-64)

In Equation 3-64 r = −1/2 corresponds to acoustic phonon scattering and r = 3/2 is

appropriate for ionized impurity scattering (Ref. 14). In general r can be considered a
function of temperature. In COMSOL Multiphysics, r is assumed to be −1/2.

The mobility integral can now be written as:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 163

 qτ 0 r ∂f 0  k sin ( θ ) cos ( φ ) 2 2

4h π n
_ 2 3 - W --------
μ n = – ----------------------- 
∂W 
-------------------------------------- k sin ( θ ) dk dθ dφ
m∗ 

To transform the integral over k to an integral over W note that W = h2 k2 /(2m∗) so

_
that dk=(m*/2 h 2)1/2W -1/2dW. Writing each quantity in the integral as a function
of W and rearranging gives:

qτ 0 m∗ ∞
r + 3 ⁄ 2 ∂f 0
π 2π
0 W 0 sin 0
3 2
μ n = – ---------------------------
_3 3- -------- dW ( θ ) dθ cos ( φ ) dφ
2h π n ∂W

Evaluating the angular integrals and integrating the energy integral by parts gives:

qτ 0 8m∗  3 ∞
r+1⁄2
μ n = ------------------------
3h π n
_ 3 2 -  r + --2- 0 W f 0 dW (3-65)

Where the result:

r+3⁄2 ∞
[W f0 ]W = 0 = 0

is used. Equation 3-65 can be rewritten in the form:

qτ 0 8m∗  3 E c – E fn
Γ ( r + 3 ⁄ 2 )F r + 1 ⁄ 2  – ---------------------
r+3⁄2
μ n = ------------------------
_ 3 2 -  r + --2- ( k B T ) (3-66)
 kB T 
3h π n

where the definition of the Fermi-Dirac function (Equation 3-46) and of f0

(Equation 3-5) is used.

Following the same procedure for Qn gives:

τ 0 8m∗  5 E c – E fn
Γ ( r + 5 ⁄ 2 )F r + 3 ⁄ 2  – ---------------------
r+5⁄2
Q n = -----------------------
_ 3 2 -  r + --2- ( k B T ) (3-67)
 kB T 
3h π n

Equation 3-66 and Equation 3-67 show that Qn can be related to μn:

E c – E fn
F r + 3 ⁄ 2  – ---------------------
kB T  kB T
= ----------- r + --- -------------------------------------------------- μ n
5
Qn (3-68)
q  2  E c – E fn
F r + 1 ⁄ 2  – ---------------------
 kB T 

where the result Γ (j + 1) = jΓ(j) is used.

164 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Similarly Qp is related to μp in exactly the same manner:

E fp – E v
F r + 3 ⁄ 2  – ----------------------
kB T   kB T 
Q p = -----------  r + --- --------------------------------------------------- μ p
5
(3-69)
q 2 E fp – E v
F r + 1 ⁄ 2  – ----------------------
kB T

In COMSOL Qn and Qp are computed from μn and μp using Equation 3-68 and
Equation 3-69 with r=−1/2. Note that in the nondegenerate limit:

kB T
Q n = -----------  r + --- μ n
5
q  2
kB T
Q p = -----------  r + --- μ p
5
q  2

THE DRIFT-DIFFUSION EQUATIONS

Having related Qn and Qp to the corresponding mobilities the next task is to derive
equations relating the current to the carrier concentrations from Equation 3-61 and
Equation 3-62. Once again the case of electrons is considered in detail and the results
for holes are similar.

Inverting Equation 3-51 gives the following equation for the electron quasi-Fermi
level:

E fn = E c + k B TF 1 ⁄ 2  -------
–1 n
(3-70)
Nc

where we have defined the inverse Fermi-Dirac integral (F-11/2(α)=η implies that
α=F1/2(η)). Let α=n/Nc and η=F-11/2(α). To obtain the drift-diffusion equations
Equation 3-70 is substituted into Equation 3-61. Note that Nc=Nc(T) so that η=F-11/
2(n/Nc)=η(n,T). To compute the current the gradient of the quasi-Fermi level is
required:

∂η ∂η
∇E fn = ∇E c + ∇( k B Tη ) = ∇E c + ηk B ∇T + k B T ------- ∇T + k B T ------ ∇n
∂T ∂n

Since α=n/Nc and Nc∝T3/2 (Equation 3-48):

∂η ∂α dη α dη
------ = ------- ------- = --- -------
∂n ∂n dα n dα
(3-71)
∂η- ∂α dη 3 α dη
------ = ------- ------- = – --- ---- -------
∂T ∂T dα 2 T dα

THEORY FOR THE SEMICONDUCTOR INTERFACE | 165

 In order to evaluate ∂η/∂α the result dFn(η)/dη=Fn-1(η) is required (see Ref. 15 for
details). Given this result:

dF 1 ⁄ 2 ( η )
- = F –1 ⁄ 2 ( η ) = dα
------------------------ -------
dη dη

so that:

dη 1
------- = ----------------------- (3-72)
dα F–1 ⁄ 2 ( η )

Substituting Equation 3-71 and Equation 3-72 into Equation 3-46 and then using
α = F1/2(η) gives:

3 F1 ⁄ 2 ( η ) F1 ⁄ 2 ( η ) kB T
∇E fn = ∇E c + ηk B ∇T – --- ----------------------- k B ∇T + ----------------------- ----------- ∇n (3-73)
2 F –1 ⁄ 2 ( η ) F–1 ⁄ 2 ( η ) n

Substituting Equation 3-73 into Equation 3-61 gives:

F1 ⁄ 2 ( η ) F1 ⁄ 2 ( η )
J n ( r ,t ) = nμ n ∇E c + μ n k B T ----------------------- ∇n + ----  qQ n – --- μ n k B T ----------------------- ∇T
n 3
F–1 ⁄ 2 ( η ) T 2 F –1 ⁄ 2 ( η )

where the result η = (Efn − Ec)/(kBT) is used and follows from Equation 3-70.

Defining the function:

F1 ⁄ 2 ( η ) α –1
G ( α ) = ----------------------- = ----------------------------------------- η = F1 ⁄ 2 ( α )
F–1 ⁄ 2 ( η ) –1
F–1 ⁄ 2 ( F1 ⁄ 2 ( α ) )

This result can be written in the form:

nq
J n ( r ,t ) = nμ n ∇E c + μ n k B TG ( n ⁄ N c ) ∇n + ------- D n ,th ∇T (3-74)
T

Following the same argument for the hole current gives:

pq
J p ( r ,t ) = pμ p ∇E v – μ p k B TG ( p ⁄ N v ) ∇p – ------- D p ,th ∇T (3-75)
T

The thermal diffusion coefficients Dn,th and Dp,th are defined as:

166 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 3 kB T
D n ,th = Q n – --- ----------- G ( n ⁄ N c )μ n
2 q
3 kB T
D p ,th = Q p – --- ----------- G ( p ⁄ N v )μ p
2 q

In the nondegenerate limit G(α)→1 and the following equations are obtained:

J n ( r ,t ) = nμ n ∇E c + μ n k B T ∇n + ( r + 1 )nμ n k B ∇T
J p ( r ,t ) = pμ p ∇E v – μ p k B T ∇p – ( r + 1 )pμ p k B ∇T

For a relaxation time dominated by phonon scattering (currently assumed by the

COMSOL software) r = −1/2.

POISSON’S EQUATION AND THE CONTINUITY EQUATIONS

Equation 3-74 and Equation 3-75 define the hole and electron currents used by
COMSOL Multiphysics in the Semiconductor interface. To solve a model the
Semiconductor interface uses these definitions in combination with Poisson’s equation
and the current continuity equations.

Poisson’s equation takes the form:

+ –
∇ ⋅ ( – ε 0 ε r ∇V ) = q ( p – n + N d – Na ) (3-76)

and the current continuity equations are given by:

∂n 1
------ = --- ( ∇ ⋅ J n ) – U n
∂t q
(3-77)
∂p
------ = – 1--- ( ∇ ⋅ J p ) – U p
∂t q

where Un=ΣRn,i-ΣGn,i is the net electron recombination rate from all generation
(Gn,i) and recombination mechanisms (Rn,i). Similarly, Up is the net hole
recombination rate from all generation (Gp,i) and recombination mechanisms (Rp,i).
Note that in most circumstances Un=Up. Both of these equations follow directly from
Maxwell’s equations (see Ref. 6).

The Semiconductor interface solves Equations 3-76 and 3-77. The current densities Jn
and Jp are given by Equation 3-74 and Equation 3-75, if the charge carrier dependent
variables are the carrier concentrations or the log of the concentrations. For the
quasi-Fermi level formulation and the density-gradient formulation, Equation 3-61
and Equation 3-62 are used for the current densities and the charge carrier dependent

THEORY FOR THE SEMICONDUCTOR INTERFACE | 167

 variables are the quasi-Fermi levels. For the density-gradient formulation, there are
also two additional dependent variables φn and φp (so called Slotboom variables, SI
unit: V), defined in terms of the carrier concentrations:

qφ n
n = exp  -----------
 k B T
(3-78)
qφ p
p = exp  -----------
 k B T

These additional dependent variables are used for solving Equation 3-53,
Equation 3-54 and Equation 3-55 together with the semiconductor equations.

Mobility Models
In The Semiconductor Equations section, a closed-form expression was computed for
the electron and hole mobilities by making the relaxation-time approximation and by
assuming a particular form for the scattering time as a function of energy. Any
mechanism that disturbs the perfect periodicity of the lattice can produce scattering of
the carriers. Such mechanisms include:

• Phonons (L): Thermally generated acoustic waves traveling through the crystal
(scattering by phonons is frequently referred to as lattice scattering, which is
somewhat confusing as the lattice itself never scatters the carriers unless perturbed
from perfect periodicity).
• Ionized impurities (I): These begin to have an effect at doping levels above
approximately 1015 cm-3 at room temperature.
• Carrier-carrier scattering (C).
• Neutral impurity scattering (N): This is important at low temperatures only,
typically below 77K.
• High field velocity saturation (E).
• Surface scattering (S): This includes effects such as interface charges, scattering by
surface phonons, and so on. It is important in field effect devices, such as field effect
transistors.

In the Semiconductor interface, the letters in brackets in the list above are
used to label the effects included in a specific mobility model (the
convention used in Ref. 6 is adopted).

168 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

In practice the mobility is typically computed using analytic functions with or without
a rigorous physical motivation and designed to fit the experimental data to a good
degree of accuracy. Often, mobility models are designed to address one particular
effect (for example high field velocity saturation) and require other mobility models as
a basis (for example a model incorporating scattering due to phonons and impurities).
As an example, high field effects are often incorporated into a model by defining the
high field mobility as a function of the mobility due to phonon and impurity scattering.
In other cases the scattering due to different mechanisms is combined using
Matthiessen’s rule — an approximate method for combining mobilities that, while not
rigorously correct, is frequently employed to produce empirical fits to experimental
data in the context of mobility models. Matthiessen’s rule combines mobilities in the
following manner:

1- 1 1
--------- = ------ + ------
μ tot μ1 μ2

An example of the manner in which mobility models are often combined is:

μ tot = μ E ( μ S ( μ LI ) ) (3-79)

where μE, the mobility including high field effects, is a function of μs (the surface
mobility), which in turn is a function of the mobility model for phonon (or lattice) and
impurity scattering (μLI).

COMSOL Multiphysics uses a general mechanism to combine both user-defined and

predefined mobility models that accommodate combinations of the form given in
Equation 3-79. Mobility models are added as subnodes to the Semiconductor Material
Model node. If the mobility model requires an input mobility, this is selected from the
available mobilities appropriate for this model (individual selections are required for
both the electron and hole mobility inputs). Multiple user-defined mobility models can
be added and these can be used as inputs to predefined or other user-defined mobility
models. The user-defined mobility models are always available as inputs in predefined
mobility models (if the model requires an input). Within this system it is possible to
mix user-defined mobility models with models predefined by COMSOL Multiphysics
in a very general manner. The model used within the simulation is selected for
electrons and holes by changing the electron mobility and the hole mobility settings in

THEORY FOR THE SEMICONDUCTOR INTERFACE | 169

 the Mobility Model section of the Semiconductor Material Model node, which by
default uses a constant mobility obtained from the material properties.

It is important to understand that each type of mobility model subnode

only defines mobility variables for electrons and holes that can be used by
other models, or by the Semiconductor Material Model node. The
mobility models for electrons and holes actually used in the simulation are
determined by the selections or settings in the Semiconductor Material
Model node (the parent node) which do not change when additional
mobility models are added. In order to add a mobility model to a
simulation it is therefore necessary to both add the sequence of mobility
model subnodes to the model tree and then to select the desired final
mobility for both electrons and holes in the Semiconductor Material
Model node.

The following sections describe the predefined mobility models currently available in
the Semiconductor interface.

THEORY FOR THE POWER LAW MOBILITY MODEL (L)

The Power Law Mobility Model (L) is a simple mobility model and requires no input
as it includes scattering due to phonons. The electron (μn,pl) and hole (μp,pl)
mobilities are determined by the equations:

T –αn T –αp
μ n ,pl = μ 0 ,n  ---------- μ p ,pl = μ 0 ,p  ----------
 T ref  T ref

where T is the lattice temperature and μp,pl, μp,pl, αn, αp, and Tref are material
properties. For silicon the values of the material properties are taken from Ref. 16.

THEORY FOR THE ARORA MOBILITY MODEL (LI)

The empirical Arora Mobility Model (LI) includes both phonon and impurity
scattering. The electron (μn,ar) and hole (μp,ar) mobilities are determined by the
equations:

170 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 μ 0 ,n μ 0 ,p
μ n ,ar = μ min ,n + ------------------------------α- μ p ,ar = μ min ,p + ------------------------------α-
1 +  ------------ 1 +  ------------
N N
 N 0, n  N 0, p

T β1 T β1
μ min ,n = μ min ,n  ---------- μ min ,p = μ min ,p  ----------
ref ref
T ref T ref
T β2 ref T β 2
μ 0 ,n = μ 0 ,n  ---------- μ 0 ,p = μ 0 ,p  ----------
ref
T ref T ref
T β3 ref T β 3
N 0 ,n = N 0 ,n  ---------- N 0 ,p = N 0 ,p  ----------
ref
 T ref  T ref
β4
α = α 0 ,n  ----------
ref T – +
N = Na + Nd
 T ref

where T is the lattice temperature, Na− is the ionized acceptor concentration, and Nd+
is the ionized donor concentration. All the other parameters are material properties.
For silicon the values of the material properties are taken from Ref. 16.

THEORY FOR THE FLETCHER MOBILITY MODEL (C)

The Fletcher Mobility Model (C) adds carrier-carrier scattering to an existing mobility
model (or to a constant input mobility). It accepts input mobilities of type L or LI, as
well as user-defined input mobilities.

The model uses Matthiessen’s rule to combine the input mobility with a carrier-carrier
scattering mobility term that is identical for electrons and holes. The model is based
on Ref. 17. The electron (μn,ar) and hole (μp,ar) mobilities are determined by the
equations:

1- 1 1 1- 1 1
---------- = ------------ + -------- ---------- = ------------ + --------
μ n ,fl μ in ,n μ cc μ p ,fl μ in ,p μ cc
T - 3 ⁄ 2
 ---------
 T ref F1
μ cc = --------------------------------------------------------------------------------------------
-
1⁄2  T 2
ln  1 +  ---------- ( np ) F 2
–1 ⁄ 3
( np )
T ref

where T is the lattice temperature, μin,n and μin,p are the electron and hole input
mobilities, n is the electron concentration, and p is the hole concentration. F1 (SI
unit: s2A/(m3kg)), F2 (SI unit: 1/m2), and Tref are material properties. For silicon
the values of the material properties are taken from Ref. 18.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 171

 THEORY FOR THE LOMBARDI SURFACE MOBILITY MODEL (S)
The Lombardi Surface Mobility Model (S) adds surface scattering resulting from
surface acoustic phonons and from surface roughness. Mobility contributions
corresponding to these effects are combined with the input mobility using
Matthiessen’s rule. The model accepts input mobilities of type L, LI, or C as well as
user defined input mobilities. The model is based on Ref. 19. The electron (μn,lo) and
hole (μp,lo) mobilities are determined by the following equations:

1 - ------------
1 1 1 1-
----------- 1 1 1
----------- = + ------------- + ------------ = ------------ + ------------- + ------------
μ n ,lo μ in ,n μ ac ,n μ sr ,n μ p ,lo μ in ,p μ ac ,p μ sr ,p
N αn N αp
μ 2 ,n  ----------- μ 2 ,p  -----------
μ 1 ,n  N ref μ 1 ,p  N ref
μ ac ,n = ----------------- + ------------------------------------------
- μ ac ,p = ----------------- + ------------------------------------------
E ⊥ ,n  E ⊥ ,n 1 ⁄ 3  T  E ⊥ ,p E ⊥ ,p
1⁄3
 ---------
T 
------------ ------------ ---------- ----------- ----------- -
 E ref   E ref   T ref  E ref   E ref   T ref
δn δp
μ sr ,n = -----------
2
- μ sr ,p = -----------
2
E ⊥ ,n E ⊥ ,p
– +
N = Na + Nd

where T is the lattice temperature, μin,n and μin,p are the electron and hole input
mobilities, Na- is the ionized acceptor concentration, Nd+ is the ionized donor
concentration, E ⊥, n is the component of the electric field perpendicular to the
electron current and E ⊥, p is the component of the electric field perpendicular to the
hole current. All other parameters in the model are material properties (note that δn
and δp have units of V/s). The material properties for silicon are also obtained from
this reference Ref. 19.

THEORY FOR THE CAUGHEY-THOMAS MOBILITY MODEL (E)

The Caughey-Thomas Mobility Model (E) adds high field velocity scattering to an
existing mobility model (or to a constant input mobility). It accepts input mobilities
of type L, LI, C, or S as well as user defined input mobilities. The model is based on
Ref. 20. The electron (μn,lo) and hole (μp,lo) mobilities are determined by the
following equations:

172 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 μ in ,n μ in ,p
μ n ,ct = -------------------------------------------------------- μ p ,ct = -------------------------------------------------------
-
μ in ,n F n n
 1 +  -------------------
α 1 ⁄ αn μ in ,p F p p
 1 +  -------------------
α 1 ⁄ αp
-
  v sat, n     v sat, p  
T β n, 1 T β p, 1
α n = α n, 0  ---------- α p = α p, 0  ----------
T ref T ref
T β n, 2 T β p, 2
v sat, n = v n, 0  ---------- v sat, p = v p, 0  ----------
 T ref  T ref

where T is the lattice temperature, μin,n and μin,p are the electron and hole input
mobilities and Fn and Fp are the driving forces for electrons and holes (currently
Fn=E||,n and Fp=E||,p where E||,n is the component of the electric field parallel to
the electron current and E||,p is the component of the electric field parallel to the
electron current). All other parameters in the model are material properties (note that
vn,0 and vp,0 are the saturation velocities for electrons and holes and have units of m/
s). The material properties for silicon are also obtained from Ref. 20.

THEORY FOR THE USER-DEFINED MOBILITY MODEL

The User-Defined Mobility Model can be used to create electron and hole mobilities
with user-defined expressions for the electron and hole mobilities. These mobility
models can be combined with other user-defined or predefined mobility models in
arbitrary combinations. These mobility models can take other defined mobilities as
inputs. By default the output mobility is set to the input mobility for both electrons
and holes. However, it is possible to change the expression for the output mobility so
that it is any function of the input mobility (or indeed a function that does not depend
on the input mobility). The default value is set in this way to make it straightforward
to access the variable in which the input mobility is stored.

In the finite volume method, the dependent variables are constant within
each mesh element and the gradients cannot be defined using the
differentiation operator in COMSOL. A detailed understanding of the
finite volume method is required in order to set up mobility models that
depend on the gradients of the dependent variables in COMSOL (∇n,
∇p, and ∇V). These variables include the currents jn and jp, the electric
field E, and the electric displacement field D, as well as the gradients of
the quasi-Fermi levels ∇Efn and ∇Efp. In the finite element method these
limitations do not apply and the gradients of the dependent variables (and
quantities which depend on them) can be used in expressions.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 173

 Defining the Carrier Mobility

Traps
Many semiconductor devices contain a distribution of traps within the band gap. These
traps are associated with donor and acceptor atoms, with other impurity atoms in the
bulk of the semiconductor, or with “dangling bonds” that occur at defects or exterior
surfaces and grain boundaries. It is useful to consider the processes that can occur
when trap states at a given energy, Et, exchange electrons or holes with states in the
valence or conduction bands at energy E. This situation is depicted in Figure 3-9.
There are four processes that occur, corresponding to the emission and capture of both
electrons and holes between states in the bands

E E

Et Et Et Et

E E

Figure 3-9: The four processes that contribute to SRH recombination. Left: An electron in
the conduction band with energy E is captured by a trap with energy Et (ec). Center left:
An electron in a trap at energy Et is emitted to an empty state with energy E in the
conduction band (ee). Center right: An electron in a trap at energy Et moves to an empty
state in the valence band at energy E. Equivalently a hole in the valence band is trapped
(hc). Right: An electron from an occupied state in the valence band at energy E is excited
into a trap with energy Et. Equivalently a hole in the trap is emitted (he).

In practice you may want to consider a set of discrete trap levels or a continuum of trap
states (with a density of states gt(Et)). Both of these approaches can be accommodated
by defining Nt, the density of traps per unit volume at a particular trap energy, Et. For
a continuum of trap states, Nt is given by:

N t = g t ( E t )dE t

174 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

For the trap energy Et, carriers in the conduction band or valence band with energy E
make the following contributions to the total recombination or generation rate per
unit volume for each process:

dr ec = f ( E )g c ( E ) ( 1 – f t )N t c ec ( E )dE
dg ee = ( 1 – f ( E ) )g c ( E )f t N t c ee ( E )dE
(3-80)
dr hc = ( 1 – f ( E ) )g v ( E )f t N t c hc ( E )dE
dg he = f ( E )g v ( E ) ( 1 – f t )N t c he ( E )dE

where Nt is the number of traps per unit volume, ft is the trap electron occupancy
(between 0 and 1), gc(E) is the conduction band density of states, gv(E) is the valence
band density of states, and cec(E), cee(E), chc(E), and che(E) are rate constants. The
net rate of capture for electrons and holes in the energy interval dE can be written as:

dr e = [ f ( E ) ( 1 – f t )c ec ( E ) – ( 1 – f ( E ) )f t c ee ( E ) ]g c ( E )N t dE
(3-81)
dr h = [ ( 1 – f ( E ) )c hc ( E )f t – f ( E ) ( 1 – f t )c he ( E ) ]g v ( E )N t dE

In thermal equilibrium the principle of detailed balance implies the reversibility of each
microscopic process that leads to equilibrium. Consequently, at equilibrium, the
expressions in the square brackets must be equal to zero. This leads to the following
relationships between the rate constants:

f ( E ) 1 – ft
c ee ( E ) = -------------------- ------------- c ec ( E )
1 – f ( E ) ft

1 – f ( E ) ft
c he ( E ) = -------------------- ------------- c hc ( E )
f ( E ) 1 – ft

In thermal equilibrium the occupancy of the electron traps ft is determined by

Fermi-Dirac statistics, fte=1/(1+exp[(Et-Ef)/(kBT)]/gD) (where gD is the
degeneracy factor). The above equations can be simplified to yield:

E – Et
c ee ( E ) = ------- exp  – ---------------- c ec ( E )
1
gD  kB T 
(3-82)
Et – E
c he ( E ) = g D exp  – ---------------- c hc ( E )
 kB T 

Equation 3-82 applies even away from equilibrium. Substituting this equation back
into Equation 3-81, rearranging and integrating, gives the total rate of electron (re) or
hole (rh) capture to traps at the specified energy, Et:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 175

 E fn – E t ∞
r e = N t 1 – f t  1 + ------- exp  – --------------------- 
1
 gD  kB T   E f ( E )cec ( E )gc ( E )dE
c

E t – E fp Ev
rh = N t f t – ( 1 – f t )g D exp  – --------------------- –∞ ( 1 – f ( E ) )chc ( E )gv ( E )dE
 kB T 

where the quasi-Fermi levels have been introduced.

Introducing the constants Cn and Cp, which represent the average capture probability
of an electron over the band:

∞
E f ( E )cec ( E )gc ( E )dE
C n = ------------------------------------------------------------
c

∞
-

E f ( E )gc ( E )dE
c

Ev

–∞
 ( 1 – f ( E ) )c hc ( E )g v ( E )dE
C p = -----------------------------------------------------------------------------
E v

–∞ ( 1 – f ( E ) )gv ( E )dE

and noting that:

∞ E c – E fn  E fn – E i
n = E f ( E )gc ( E )dE = N c γ n exp  – --------------------- = γ n n i ,eff exp  ---------------------
c
kB T  kb T 
Ev E fp – E v  E i – E fp
p =  ( 1 – f ( E ) )g v ( E )dE = N v γ p exp  – ---------------------- = γ p n i ,eff exp  ---------------------
–∞
 kB T   kb T 

the following equations are obtained:

E fn – E t n1
r e = nC n N t 1 – f t  1 + ------- exp  – --------------------- 
1
= C n N t n – nf t – ------- f t
 gD  kB T   gD
(3-83)
E t – E fp
rh = pC p N t f t – ( 1 – f t )g D exp  – --------------------- = C p N t [ pf t + g D ( p 1 f t – p 1 ) ]
 kB T 

where:

ΔE t ⁄ ( k B T ) – ΔE t ⁄ ( k B T )
n 1 = γ n n i ,eff e p 1 = γ p n i ,eff e
(3-84)
ΔE t = – E i + E t

176 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Equation 3-83 and Equation 3-84 define the electron and hole recombination rates
associated with the trap energy level.

For the density-gradient formulation, the following equivalent formulas are used:

E ft – E fn
r e = nC n N t ( 1 – f t ) 1 – exp  -----------------------
 kB T 
(3-85)
E fp – E ft
r h = pC p N t f t 1 – exp  -----------------------
 kB T 

E t – E fn
n 1 = n exp  ---------------------
 kB T 
(3-86)
E fp – E t
p 1 = p exp  ---------------------
 kB T 

The total rate of change of the number of trapped electrons is given by:

∂f t n1
N t ------- = r e – r h = C n N t n – nf t – ------- f t – C p N t [ pf t + g D ( p 1 f t – p 1 ) ] (3-87)
∂t gD

Equation 3-87 determines the occupancy of the traps at the level Et. For a continuous
distribution of traps the equivalent expression is:

∂f t n1
g t ( E t ) ------- = g t ( E t )C n n – nf t – ------- f t – g t ( E t )C p [ pf t + g D ( p 1 f t – p 1 ) ]
∂t gD

The total recombination rate for electrons and holes is given by integrating
Equation 3-83 over all the distributed traps and summing over the distinct discrete
traps (denoted by the superscript i), giving the following result:

Ec j i
j n1 j i n1 i
 
j j i i
Re = g t ( E t )C n n – nf t – ------
- f dE t +
j t
C n N t n – nf t – ------
-f
i t
j
Ev gD i
gD
(3-88)
Ec
 E 
j j j j j j j i i i i i i i
Rh = g t ( E t )C p [ pf t + gD ( p1 ft – p 1 ) ] dE t + C p N t [ pf t + gD ( p1 ft – p1 ) ]
v
j i

THEORY FOR THE SEMICONDUCTOR INTERFACE | 177

 Finally, the charge resulting from the occupied traps must also be computed. In
general the charge on a trap site depends on the nature of the trap. Table 3-1
summarizes the different trap types currently included in the Semiconductor Module.
TABLE 3-1: CHARGE DUE TO DIFFERENT TYPES OF TRAP

TRAP TYPE SPECIES CHARGE CHARGE DENSITY OF NUMBER

TRAPPED OCCUPIED UNOCCUPIED STATES DENSITY
D D
Donor trap Electron 0 + gt ( Et ) Nt
A A
Acceptor trap Hole 0 - gt ( Et ) Nt
e e
Neutral electron trap Electron - 0 gt ( Et ) Nt
h h
Neutral hole trap Hole + 0 gt ( Et ) Nt

The total density of traps is given by:

D A e h
gt ( Et ) = gt ( Et ) + gt ( Et ) + gt ( Et ) + gt ( Et )

For the discrete states the total number of traps at the discrete level Ei is:

i D, i A ,i e,i h ,i
Nt = Nt + Nt + Nt + Nt

The total charge density, Q, that results from the traps is given by:

Ev
D, j  k A, k l e, l h, m
E  ( 1 – ft )gt  ft gt  ft gt  ( 1 – ft )gt
j m
Q = q – + + dE t
c  
j k l m
(3-89)

 (1 –    (1 –
i D,i n A ,n o e ,o p h ,p
+q f t )N t – q ft Nt – q ft Nt +q f t )N t
i n o p

Equation 3-89 can be rewritten in the form:

Ec Ec
E [ gt E ft gt dEt + q  [ Nt  ft Nt
D h D,i h ,i i i
Q = q + g t ] dE t – q + Nt ] – q
v v
i i
Ec Ec E0 E0
E [ gt E ft gt dEt + q E ( 1 – ft )gt dEt – q E gt dEt
D h
Q = q + g t ] dE t – q
v 0 v v

   
D ,i h ,i i i i i i
+q [ Nt + Nt ] –q ft Nt +q (1 – f t )N t –q Nt
i Ei ≥ E0 Ei < E0 Ei < E0

where E0 is an energy within the band gap referred to as the neutral level. E0 is chosen
such that:

178 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 Ec E0
E [ gt  [ Nt E gt dEt + 
D h D ,i h ,i i
+ g t ] dE t + + Nt ] = Nt
v
i
v
Ei < E0

So if a neutral energy is employed the following equation applies for the charge on the
traps:

Ec E0
E ft gt dEt + q E ( 1 – ft )gt dEt – q  
i i i i
Q = –q ft Nt + q ( 1 – f t )N t (3-90)
Ei ≥ E0 Ei < E0
0 v

A neutral level is often a convenient way to characterize a set of traps at a boundary —

since the details of the types of trapping sites are frequently not known, but it is
possible to assign a neutral level to the boundary using experimental techniques such
as capacitance measurements.

Recombination and Generation

Generation and recombination processes produce a source or a sink for electrons or
holes in the current continuity equations (Equation 3-77). These terms usually serve
to restore the device to equilibrium.

From Equation 3-49 the np product at equilibrium is given by:

Ec – Ev
np = N c N v γ n γ p exp  – ------------------- = γ n γ p n i , eff ≡ n i, mod
2 2
(3-91)
 kB T 

where the effective intrinsic concentration is introduced:

Eg ΔE g
exp  – --------------- = n i exp  ---------------
1⁄2
n i ,eff = ( N c N v )
2k B T 2k B T
E g0
exp  – ---------------
1⁄2
ni = ( Nc Nv )
 2k B T

For the density-gradient formulation, instead of Equation 3-49, one similar to

Equation 3-53 is used, and the following development is similar but does not result in
exactly the same formulas (Ref. 46).

Here Eg=Eg0-ΔEg=Ec-Ev is the effective band gap that includes the material band gap
Eg0 and changes in the band gap (ΔEg), which result from effects such as band gap
narrowing. Equation 3-91 is known as the mass action law. Note that in an undoped
semiconductor, where charge conservation at equilibrium implies n = p, the carrier
concentrations are equal to the intrinsic concentration, ni, since band gap narrowing

THEORY FOR THE SEMICONDUCTOR INTERFACE | 179

 only occurs in heavily doped semiconductors. Equation 3-91 is frequently used in
recombination models to define the equilibrium carrier concentration.

In considering recombination processes, it is useful to note that photon momentum is

usually negligible in comparison to the carrier momentum, and that the energy of
phonons (lattice vibrations) is typically small compared to the band gap. Thus photon
mediated transitions are typically vertical in an E-k diagram and phonon mediated
transitions are frequently horizontal. Direct transitions involving both a photon and a
phonon are usually so unlikely that they do not contribute significantly to
recombination — in indirect semiconductors recombination is usually mediated by
traps (impurities with energies close to the midgap). The following mechanisms are
common in practical materials:

• Direct recombination: Important for direct band gap semiconductors such as

gallium arsenide. An electron-hole pair recombines with the emission of a photon
(of energy close to the band gap).
• Trap-assisted, or Shockley-Read-Hall recombination: Important in indirect band
gap semiconductors, such as silicon and germanium. A defect (usually with an
energy close to the midgap) is involved in the recombination process. An electron
or hole is first trapped by the defect and then emitted into the valence/conduction
band, resulting in a reduction in the number of available carriers. The carrier energy
is typically converted to heat.
• Auger recombination: In Auger recombination three carriers are involved. A
collision between two like carriers (for example, electrons) results in the
recombination of one of the electrons with a hole. The energy released by the
transition is transferred to the surviving electron. The resulting highly energetic
electron subsequently loses energy as it undergoes collisions.
• Impact ionization: Impact ionization occurs when, for example, an energetic
electron undergoes a collision in which it loses sufficient energy to promote an
electron in the valence band to the conduction band, resulting in an additional
electron-hole pair being produced. It is the mechanism responsible for avalanche
breakdown.

The Semiconductor interface has features to add Auger Recombination, Direct

Recombination, and Trap-Assisted Recombination to a semiconducting domain.
Impact Ionization Generation is also available. User-Defined Recombination or
User-Defined Generation (use a negative recombination rate for generation) can also
be added. Note that the recombination and generation features are additive, so it is
possible to model several processes simultaneously.

180 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

THEORY FOR THE DIRECT RECOMBINATION FEATURE
Direct recombination is usually the dominant recombination mechanism in direct
band gap semiconductors. The recombination rate can be derived phenomenologically
from the two process that contribute to the net recombination rate: recombination of
a conduction band electron with a hole in the valence band (caused by the emission of
a photon by the electron) and generation of an electron hole pair by a valence band
electron adsorbing a photon and moving into the conduction band. The
corresponding capture (c) and emission (e) processes, are shown in Figure 3-10.

Figure 3-10: The two processes involved in direct recombination. Left: A conduction band
electron is captured by an empty state (a hole) in the valence band. A photon is emitted.
Right: A valence band electron is emitted into the conduction band. A photon is absorbed.

The recombination process is easiest to treat phenomenologically. Consider electrons

in the conduction with energy E. A certain fraction of these electrons decay to states
in the valence band with energy E’. This process contributes an amount drc to the total
recombination process:

dr c = f ( E )g c ( E ) ( 1 – f ( E' ) )g v ( E' )c c ( E ,E' ) dE dE'

where f(E) is the Fermi-Dirac function, gc(E) is the density of states in the valence
band, gv(E) is the density of states in the conduction band and cc(E,E’) is the rate
constant for decay between states E and E’. If it is assumed that the rate constant does
not vary significantly in the vicinity of the band edges then cc(E,E’)~Cc and the
expression can be directly integrated to yield:

∞ Ev
Rc = Cc E f ( E )gc ( E ) dE –∞ ( 1 – f ( E' ) )gv ( E' ) dE' =
c
C c np

The generation process is more difficult to treat phenomenologically since it involves

photon mediated transitions that must occur vertically in the band structure with a
fixed energy difference (corresponding to the wavelength of the photons). However
for a given band structure and frequency, provided that the semiconductor is

THEORY FOR THE SEMICONDUCTOR INTERFACE | 181

 nondegenerate, the photon transitions occur at a rate approximately independent of
the carrier concentrations for a given illumination intensity. Thus a constant generation
rate is a reasonable approximation under these circumstances:

Ge = Ce

Where Ce is a rate constant (dependent on the wavelength and the intensity of the
incident light). These approximations do not always apply, and in circumstances where
a more detailed model is appropriate, the Semiconductor interface includes the Optical
Transitions feature.

In thermal equilibrium these rates must be equal so:

eq eq 2
R c = G e = C c γ n γ p n i ,eff = C e

So

2
C e = C c γ n γ p n i ,eff

The net rate of direct recombination is therefore given by:

d d d d 2
R n = R p = R c – G e = C ( np – γ n γ p n i ,eff )

where C=Cc is a material constant (SI unit: m3/s).

For the density-gradient formulation, the second term on the right hand side is
replaced with the product of equilibrium concentrations given by one similar to
Equation 3-53.

For indirect band gap semiconductors such as silicon and germanium, C is effectively
zero. In GaAs (a widely used direct band gap semiconductor) C is approximately
1·10-10 cm3/s.

THEORY FOR TRAP-ASSISTED RECOMBINATION: SHOCKLEY-READ-HALL

RECOMBINATION
In an indirect band gap semiconductor at low fields, trap-assisted recombination is
usually the dominant contributor to Un and Up. This recombination mechanism
involves the trapping of an electron or hole followed by re-emission into the valence
or conduction band (see Ref. 21 and Ref. 22). The details of this process are described
in the Traps section, and COMSOL Multiphysics provides features to model the traps
explicitly, solving for the occupancy of the traps. For less detailed modeling it is
common to use the original model of Shockley, Read, and Hall, in which steady-state

182 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

conditions are assumed for traps located at a single energy level. Equation 3-87
determines the occupancy factor for the state, ft. In the steady state the time derivative
is zero and the following occupancy factor is obtained by solving the equation:

Cn n + Cp p1
f t = -------------------------------------------------------------------------
n1
C n  n + ------- + C p ( p + g D p 1 )
gD

Consequently:

2 2
np – γ n γ p n i ,eff np – γ n γ p n i ,eff
R e = R h = N t ---------------------------------------------------------- = ---------------------------------------------------------------------
n
 n + ------1- n1
τ p  n + ------- + τ n ( p + g D p 1 )
 gD  ( p + gD p1 )  g D
---------------------- + ----------------------------
Cp Cn

where:

1
τ n = --------------
Nt Cn
1
τ p = --------------
Nt Cp

For the density-gradient formulation, the second term in the numerator on the right
hand side is replaced with the product of equilibrium concentrations given by one
similar to Equation 3-53.

Finally note that Cn and Cp can be written in terms of the thermal velocity of the
electrons and holes, respectively (vn,th/vp,th), as well as their average capture cross
sections (<σn>/<σp>):

C n = v n, th  σ n
C p = v p, th  σ p

THEORY FOR AUGER RECOMBINATION

Auger recombination becomes important at high nonequilibrium carrier densities,
since the process requires three carriers. For example, when two electrons collide the
collision can result in the recombination of one of the electrons with a hole. The
energy released by the transition is transferred to the surviving electron, which
subsequently returns to equilibrium as it undergoes collisions with the lattice. The
recombination rate is given by:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 183

 Au Au 2
Rn = Rp = ( C n n + C p p ) ( np – γ n γ p n i ,eff )

where Cn and Cp are material constants (SI unit: m6/s).

For the density-gradient formulation, the second term in the second factor on the right
hand side is replaced with the product of equilibrium concentrations given by one
similar to Equation 3-53.

For silicon Cp≈2.8·10-31 cm6/s and Cp≈9.9·10-32 cm6/s. In practice these

coefficients are weakly dependent on temperature and doping level.

THEORY FOR IMPACT IONIZATION

Impact ionization becomes important at high electric fields. When the carriers are
accelerated by the electric field in between collisions to velocities where their energies
are greater than the gap energy, they can dissipate enough energy during collisions that
additional electron hole pairs can be generated. Impact ionization is responsible for the
phenomenon of avalanche breakdown. The carrier generation rate due to impact
ionization is given by:

II II αn αp
R n = R p = – ------ J n – ------ J p
q q

For the values of αn and αp, the Semiconductor interface allows user-defined
expressions or using the model of Okuto and Crowell (Ref. 23):

b n ( 1 + d n ( T – T ref ) ) 2
α n = a n ( 1 + c n ( T – T ref ) )E || ,n exp  –  ------------------------------------------------------ 
 E || ,n 
b p ( 1 + d p ( T – T ref ) ) 2
α p = a p ( 1 + c p ( T – T ref ) )E || ,p exp  –  ----------------------------------------------------- 
E || ,p

Where E||,n and E||,p are the components of the electric field parallel to the electron
and hole currents, respectively, and Tref, an, ap, bn, bp, dn, and dp are material
properties (see Ref. 23 for values of these properties for silicon, germanium, gallium
arsenide, and gallium phosphate).

USER-DEFINED RECOMBINATION
The user-defined recombination feature allows the recombination rate to be specified
by the user for both electrons and holes.

184 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

USER-DEFINED GENERATION
The User-Defined Generation feature allows the carrier generation rate to be defined
for both the electrons and the holes.

In the finite volume method, the dependent variables are constant within
each mesh element and their gradients cannot be computed using the
differentiation operator in COMSOL Multiphysics. However, COMSOL
Multiphysics provides special variables that can be differentiated — see
Discretization and Formulation Options in the Modeling Guidelines
chapter.

Recombination and generation terms that are dependent on the gradients

of the dependent variables frequently introduce significant nonlinearity
into the equation system (whether the finite element or finite volume
methods are used). Consequently, it is often necessary to add additional
dependent variables to the equation system to assist the solver. Such
generation models are only recommended for advanced users.

Incomplete Ionization
Away from equilibrium the ionization of donors and acceptors is still given by
Equation 3-41 and Equation 3-42, except that the relevant quasi-Fermi level should
be used in the equation system:

+
Nd 1
------- = -------------------------------------------------------
Nd E fn – E d
1 + g d exp  ----------------------
kB T
(3-92)
–
Na 1
- = ------------------------------------------------------
---------
Na E a – E fp
1 + g a exp  ----------------------
kB T

It is often more convenient to express Equation 3-92 in terms of the carrier

concentrations rather than the relevant quasi-Fermi levels. From Equation 3-49:

E fn = k B T ln  ------------- + E c
n
 γ n N c
(3-93)
E fp = – k B T ln  ------------- + E v
n
 γ p N v

THEORY FOR THE SEMICONDUCTOR INTERFACE | 185

 Substituting Equation 3-92 into Equation 3-93 gives:

+
Nd 1
------- = --------------------------------------------------
-
Nd gd n ΔE d
1 + ------------- exp  -----------
γn Nc  k B T
(3-94)
–
Na 1
- = --------------------------------------------------
--------- -
Na ga p  ΔE a
1 + ------------- exp -----------
γp Nv  k B T

where ΔEd=Ec-Ed and ΔEa=Ea-Ev. Equation 3-94 is used in COMSOL Multiphysics

when incomplete ionization with the standard ionization model is chosen in the
Dopant Ionization section of the Semiconductor Material Model node. It is also
possible to select a user defined ionization model, in which case the ionization ratios
are specified directly by user-defined expressions. If complete ionization is chosen (the
default) then Nd+=Nd and Na–=Na.

Band Gap Narrowing

In heavily doped materials the swollen orbitals associated with the impurity atoms
begin to overlap and as a result the discrete energy level associated with the impurities
broadens to form a band of finite width. Potential fluctuations due to the random
distribution of the impurities also lead to a broadening of the impurity band. The
ionization level of the impurities is consequently reduced. Eventually the impurity
band overlaps the conduction band or valence band, effectively narrowing the band
gap. In this situation the density of states no longer have the same form as that derived
in Equation 3-21 or Equation 3-22. Nonetheless it is common to model band gap
narrowing using these equations, but assuming a band gap that varies as a function of
doping level.

Several options are available to specify the band gap narrowing.

USER DEFINED
The Semiconductor interface allows for models of band gap narrowing to be defined
using an arbitrary expression. To add band gap narrowing to a material, select User
defined in the Band Gap Narrowing section of the Semiconductor Material Model. Then
enter user-defined expressions for ΔEg (the amount of band gap narrowing) and for α
(the fraction of the band gap narrowing taken up by the conduction band). For
convenience the energy level is entered in units of volts and is converted to an energy

186 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

behind the scenes by multiplying by the electron charge, q. When band gap narrowing
is active the following equations apply:

E g = E g0 – ΔE g
E c = E ci – αΔE g
E v = E vi + ( 1 – α )ΔE g

where Eg is the band gap with band gap narrowing, Eg0 is the material band gap, Ec
is the conduction band edge with band gap narrowing, Eci is the conduction band
edge in the absence of band gap narrowing, Ev is the valence band edge with band gap
narrowing, and Evi is the valence band edge in the absence of band gap narrowing.

SLOTBOOM MODEL
The Slotboom model is frequently used to model band gap narrowing in silicon. It is
an empirical model that calculates the narrowing as a function of the total doping
concentration (Ref. 27). This empirical model combines all of the physical effects
(random potential fluctuations, electron-electron, carrier-impurity, and electron-hole
interactions) into one energy narrowing, and consequently the calculated narrowing is
the same in neutral and depleted parts of the device. The Slotboom model computes
the band gap narrowing according to the equation:

 NI NI 2
1
ΔE g = E ref  ln  ----------- + ln  ----------- + --- (3-95)
 N ref N ref 2

where NI = Nd + Na and the other parameters are material properties (Eref has the
same units as the band gap itself and Nref has SI units of 1/m3). For silicon the
Semiconductor Module material library uses the updated material properties due to
Klassen et. al. (Ref. 28) rather than the original properties given in Ref. 27. Note that
the fraction of the band gap narrowing taken up by the conduction band is also treated
as a material property (0.5 for silicon).

JAIN-ROULSTON MODEL
The model developed by Jain and Roulston (Ref. 29) is a physics-based model in
which the only empirical parameter is the fraction of the band gap narrowing taken up
by the conduction band (once again this is defined as a material property, and a default
of 0.5 is usually employed). The model can be rewritten in a simple form with three
coefficients. Coefficients for the model are available for a wide range of III-V materials,
as well as for silicon and germanium.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 187

  1
---
1
---
1
---
  Nd  3  Nd  4  Nd  2
 A n  ---------- + B n  ---------- + C n  ---------- Nd ≥ Na
 N ref N ref N ref
ΔE g =  (3-96)
 1
---
1
---
1
---
  Na  3 N 4 N
 ---------- + C  ----------
a a
2
 p N 
A ---------- + B p N  p N  Nd < Na
 ref ref ref

where An, Ap, Bn, Bp, Cn, and Cp are material properties (with the same units as the
band gap), and Nref is a reference doping level (with SI units of 1/m3). The fraction
of the band gap narrowing taken up by the conduction band is also treated as a material
property (0.5 for silicon). Material properties for a number of materials are available in
the material library, including the original materials described in Ref. 29 and several
materials whose properties were obtained from Ref. 30.

Optical Transitions
The Optical Transitions feature is designed to model optical absorption and stimulated
and spontaneous emission within the semiconductor. Stimulated emission or
absorption occurs when a transition takes place between two quantum states in the
presence of an oscillating electric field, typically produced by a propagating
electromagnetic wave. For a semiconductor the process of stimulated absorption
occurs when an electron in the valence band adsorbs a photon and transitions into the
conduction band (resulting in coherent absorption of light). Stimulated emission
occurs when an electron in the conduction band is stimulated by the field to transition
down to the valence band (resulting in coherent emission of light). Spontaneous
emission occurs when transitions from a high energy to a lower energy quantum state
occur, with the emission of light. It can be regarded as a process by which a system
returns to equilibrium, and correspondingly can be linked to the stimulated emission
in equilibrium by thermodynamic arguments. Spontaneous emission does not occur in
phase with any propagating waves in the system, and indeed can occur in the absence
of such waves. Currently the optical transitions feature is dedicated to treating both
stimulated and spontaneous emission in direct band gap semiconductors.
Consequently the theory in the subsequent sections assumes a direct band gap at
several points.

HAMILTONIAN IN THE PRESENCE OF AN ELECTROMAGNETIC WAVE

Stimulated emission is a quantum mechanical process that results from the change in
the electron Hamiltonian due to the electromagnetic wave. The Hamiltonian

188 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

appropriate for the motion of an electron in an electromagnetic field in the classical
limit (Ref. 31, §16) is given by:

1 2
H = ----------- ( p – qA ) + V
2m 0

where A is the magnetic vector potential and m0 is the electron mass (not the effective
mass).

The corresponding quantum mechanical operator is given by:

1 _ 2
H = ----------- ( – i h ∇ – qA ) + V
2m 0
_2
h 2 1 _ _ 2
= – ----------- ∇ + V + ----------- ( i h q ∇ ⋅ A + i h q A ⋅ ∇ + q ( A ⋅ A ) )
2m 0 2m 0

For most practical optical applications |qA| « |p| (Ref. 32) so the q2A⋅A term can be
neglected. When modeling electromagnetic fields in the frequency domain, COMSOL
employs a gauge that ensures that the time varying electric potential is zero and the
magnetic potential satisfies the equation ∇⋅A=0 (see Magnetic and Electric Potentials
in the COMSOL Multiphysics Reference Manual for further details on gauge fixing
for electromagnetic waves). With this gauge for A the ihq∇ ⋅ A term is zero.
Consequently the quantum mechanical Hamiltonian operator takes the form:

_2 _
h 2 i hq 1 2 e
H = – ----------- ∇ + V + ------------- A ⋅ ∇ = ----------- p + V + ----------- A ⋅ p
2m 0 2m 0 2m 0 2m 0

Since |qA| « |p| the time-dependent term that includes the vector potential can be
treated as a small perturbation to the original stationary Hamiltonian H0 such that:

H =H 0 + H'

where:

1 2
H 0 = ----------- p + V (3-97)
2m 0

e
H' = ----------- A ⋅ p (3-98)
2m 0

Equation 3-97 is the Hamiltonian for the semiconductor in the absence of the field,
which has eigenfunctions of the form given in Equation 3-14.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 189

 FERMI’S GOLDEN RULE AND CRYSTAL MOMENTUM CONSERVATION
Since the change to the electron Hamiltonian is small, time-dependent perturbation
theory can be used to solve the problem (see Ref. 8 for an introduction to
time-dependent perturbation theory that derives the main results used below). For
small, oscillatory perturbations to a stationary Hamiltonian, a key result of time
dependent perturbation theory (known as Fermi’s golden rule) predicts that the
probability of a transition from an occupied state (1) to an unoccupied state (2, in this
case at a higher energy than state 1) per unit time, W1→2 is given by:

2π- 2 _
W 1 → 2 = ------
_ H ' δ ( E 2 – E 1 – hω 0 ) (3-99)
h 12

where E1 and E2 are the corresponding eigenvalues of the states, ω0 is the angular
frequency of the oscillation and H12' is the matrix element corresponding to the
oscillatory perturbation (H') to the original Hamiltonian:

 Ψ2∗ H'Ψ1 d x
3
H 12 =

Equation 3-99 gives the transition probability between two discrete states. The delta
function indicates that the states must be separated by an energy equal to the photon
energy, that is the transition must conserve energy. An additional requirement on the
transition is that crystal momentum is conserved:

k 2 = k 1 + k opt (3-100)

where kopt is the photon wave vector. This requirement for crystal momentum
conservation can be derived using a periodic expansion of the Bloch functions for the
two wave functions in the matrix element H12' with a plane wave excitation (see, for
example, Ref. 32 and Ref. 33). However, the relation is in practice much more general
since it results from fundamental symmetries of the system Hamiltonian (see Appendix
M of Ref. 1). Typical optical wavelengths (400 to 700 nm in free space, dropping by
a factor of order 10 inside the semiconductor) are significantly larger than the unit cell
size of a semiconductor (typically less than 1 nm), which in turn determines the order
of the Bloch function wavelengths. Consequently kopt is usually neglected in
Equation 3-97 leading to k2 = k1.

The crystal momentum conservation requirement can therefore be rephrased as a

requirement that optical transitions take place vertically in the E-k diagram.

Equation 3-99 describes the transition rate from an unoccupied initial state to an
occupied final state. To obtain the net rate of stimulated emission (including both

190 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

upward and downward transitions) it is necessary to sum over all transitions from
occupied to unoccupied states. Figure 3-11 shows the transitions of interest in a simple
two-band model. Due to the spherical symmetry of the constant energy surfaces in
k-space (in a direct band-gap semiconductor), upward transitions occur between states
of fixed energy in the valence band (E1v) and in the conduction band (E2c).

Conduction band Energy, E Energy, E

E2c

Efn
Photon Ec
energy Eg
ħω0
Ev
Efp E1v k0

Valence band fc(E) fv(E)

Electron Occupancy

Figure 3-11: Diagram illustrating allowed optical transitions between the conduction
and valence bands in a two band model of a direct band-gap semiconductor. The electron
occupancy in the valence (fc(E)) and conduction (fv(E)) bands is also shown.

Assuming that the bands have a parabolic dispersion relation in the vicinity of k = 0,
E1v and E2c can be written as a function of the photon energy, hω 0 . The equations:

_ 2
h k0
E 1v = E v – ----------------
2m h
_ 2
h k0
E 2c = ---------------- + E c
2m e
_
h ω 0 = E 2c – E 1v

can be rearranged into the form of a single quadratic equation for k0. The equation
has a single physically significant root (for positive k):

THEORY FOR THE SEMICONDUCTOR INTERFACE | 191

 2m r _
k0 = ----------
_ 2- ( h ω 0 – E g )
h
mr _
E 1v = E v – -------- ( h ω 0 – E g )
mh
(3-101)
mr _
E 2c = E c + ------- ( h ω 0 – E g )
me
1- 1 1
------ = ------- + --------
mr me mh
av
For states with a given magnitude of k we can define an average matrix element H 12
which represents the averaged matrix element over all directions for a given excitation.
For unpolarized light the random orientation of the polarization can also be accounted
av
for in H 12 (note that usually in the literature unpolarized light is assumed, and matrix
elements quoted should be adjusted if polarized light is under consideration). Note
also that the matrix element is in general a function of the magnitude of k, but for small
values of k the assumption is generally made that it is independent of k.

Fermi’s golden rule now takes the form:

2π- av 2 _
W 1 → 2 = ------
_ H δ ( E 2 – E 1 – hω 0 ) (3-102)
h 12

STIMULATED ABSORPTION
Equation 3-102 is the equation for the transition rate between a single pair of states.
In a semiconductor, there are many states that satisfy Equation 3-101, and
correspondingly it is necessary to integrate over pairs of states for which the transition
can occur. Clearly an upward transition must take place from a full state into an empty
state, so the occupancy of the states must also be accounted for. One of the
assumptions underlying the drift diffusion equations is that the scattering within the
band occurs on a much smaller time scale than equilibrating processes that occur
between the bands. Consequently stimulated electrons can be considered to
equilibrate rapidly within the conduction band so that their distribution is described
by the Fermi-Dirac function with the appropriate electron quasi-Fermi level.
Correspondingly, the rate of stimulated absorption (the total number of upward
transitions per unit volume per unit time, SI unit: m-3s-1) is given by:

1→2 2π av 2 _
   ------
3
G stim = _- H g ( k )f v ( 1 – f c )δ ( E 2c – E 1v – hω 0 ) d k
h 12
(3-103)

192 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

where g(k) is the density of states in k-space (1/4π3 from Equation 3-19) and the
electron occupancy factors for the conduction and valence band, fc and fv, are given by:

1
f c = --------------------------------------------------
E 2c – E fn
1 + exp  ------------------------
 kB T 
(3-104)
1
f v = ---------------------------------------------------
E 1v – E fp
1 + exp  -------------------------
kB T

E2c and E1v can be written in terms of k, leading to the equation:

 h _2 2
1→2 1 av 2 k- _  3
G stim =  ----------------
2π h
2 _
- H 12 f v ( k ) ( 1 – f c ( k ) )δ  ---------------
 2m r
+ E g – h ω 0 d k


Transforming the integral to spherical polar coordinates and changing the integration
variable gives:

 h _2 2
1→2 1 - av 2 k- _ 

2
G stim = ---------------- H f ( k ) ( 1 – f ( k ) )δ  --------------- + E g – h ω 0 4πk dk
2 _ 12 v c 2m
2π h  r 

 h _2 2 _
1→2 2 - av 2 k _  m r k  h 2 k 2
G stim =  ---------
πh
_ H 12 f v ( k ) ( 1 – f c ( k ) )δ  ---------------- + E g – h ω 0 ----------
 2m r
_ - d
 h 2  2m r 
 ---------------
-

1→2 2m r av 2
G stim = ------------
_ 3- H 12 f c ( k 0 ) ( 1 – f v ( k 0 ) )k 0
πh

1→2 2π- av 2 1 2m r- ----------2m r _

G stim = ------
_ H 12 f v (ω 0 ) ( 1 – f c (ω 0 ) ) --------- ----------
_ _ 2- ( h ω 0 – E g )
h 2π h
2 2
h

Note that Equation 3-101 was used in the third and fourth steps. It is common to
rewrite this expression in the form:

1→2 2π- av 2 _
G stim = ------
_ H f ( 1 – f c )g red ( h ω 0 ) (3-105)
h 12 v

where the reduced density of states:

3
---
1  2m r 2
g red ( E ) = ---------2  ----------
_ 2- ( E – E g )
2π  h 

THEORY FOR THE SEMICONDUCTOR INTERFACE | 193

 is defined by analogy with Equation 3-21 and where the functional dependence of fc
and fv has been dropped.

PHYSICAL/THERMODYNAMIC CONSIDERATIONS, STIMULATED AND

SPONTANEOUS EMISSION
The rate of stimulated absorption is directly related to the rate of stimulated and
spontaneous emission as a result of physical and thermodynamic arguments originally
due to Albert Einstein (Ref. 34). Historically Einstein argued for the existence of
stimulated emission by means of an argument similar to the one presented here. In the
argument that follows the existence of stimulated emission is assumed from the outset.

Consider the physics of a semiconductor interacting with a radiation field whose

energy is distributed uniformly in frequency in the vicinity of the transition energy
(that is the spectral width of the radiation field is large compared to the transition line
width). In this case, a suitable phenomenological model for the rate of stimulated
emission and absorption (at a particular value of the photon energy E = hω , and in
a range of photon energies, dE) is:

1→2 _ _
g stim dE = B 12 n ( h ω )f v ( 1 – f c )g red ( h ω )dE
(3-106)
2→1 _ _
r stim dE = B 21 n ( h ω )f c ( 1 – f v )g red ( h ω )dE

where B12 and B21 represent constants related to the semiconductor material and
n ( hω ) is the mean number of photons per unit volume per unit photon energy
(SI unit: J−1m−3).

The rate of spontaneous emission should not be directly dependent on the radiation
field, so the following phenomenological model is adopted for this quantity:

2→1 _
r spon dE = A 21 f c ( 1 – f v )g red ( h ω )dE (3-107)

where A21 is another constant. When the semiconductor is in equilibrium with the
radiation field the two quasi-Fermi levels are equal and the emission and absorption
rates must balance. Consequently, for the equilibrium case:

1→2 2→1 2→1

g stim = r stim + r spon
_ _
B 12 n ( h ω )f v ( 1 – f c ) = B 21 n ( h ω )f c ( 1 – f v ) + A 21 f c ( 1 – f v )

After a little algebra (using Equation 3-104 with Efn = Efp = Ef) the result is:

194 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 _ _
B 12 n ( h ω )
- = exp  – ----------
hω-
---------------------------------------------
_ (3-108)
A 21 + n ( h ω )B 21  k B T

It is instructive to consider the high temperature limit of this equation. In that limit
the exponential term tends to unity and it is expected that stimulated emission
dominates over spontaneous emission, since n ( hω ) becomes large whilst A21 is a
constant. Consequently this limit implies that:

B 12 = B 21 (3-109)

Next, substituting Equation 3-109 into Equation 3-108 and rearranging gives:

 
_ A 21  1 
n ( h ω ) = ---------  ------------------------------------
_ - (3-110)
ω
exp  ----------- – 1
B 12  h
 kB T 

The form of Equation 3-110 is similar to that of the Planck blackbody radiation
spectrum, which is observed to hold experimentally for many materials (for example,
see Ref. 10 §63 or Ref. 32 section 9.2.1):

 
n E  ------------------------------------ 
3 2
1
n ( E ) = ---------------------
2 _ 3 3
- -
π h c  exp  ----------- – 1
E
  k B T 

Note that this expression is modified slightly from the standard form usually seen in
text books since the formula above gives the number of photons per unit photon energy
rather than the photon energy per unit frequency.

Consequently a relationship between B12 and A21 is implied:

3 2
A 21 n ω0
--------
- = ------------------
- (3-111)
B 12 _ 2 3
hπ c

where the relation E = hω is used. Equation 3-109 and Equation 3-111 are derived
assuming an exciting field with a distributed spectrum in thermal equilibrium.
However the phenomenological models described by Equation 3-106 and
Equation 3-107 are expected to apply away from equilibrium in the same form, so the
relationships between A21, B12, and B21 are general. Equation 3-111, which describes
spontaneous emission relates to a process that is entirely independent of the exciting

THEORY FOR THE SEMICONDUCTOR INTERFACE | 195

 radiation. Correspondingly, A21 = 1/τspon is written, which allows the
phenomenological equations to be written in the following form:

_ 2 3
1→2 hπ c - _ _
g stim = --------------------------
3 2
n ( h ω )f v ( 1 – f c )g red ( h ω )
n ω 0 τ spon
_ 2 3
2→1 hπ c - _ _ (3-112)
r stim = --------------------------
3 2
n ( h ω )f c ( 1 – f v )g red ( h ω )
n ω 0 τ spon
2→1 1 _
r spon = ------------ f c ( 1 – f v )g red ( h ω )
τ spon

These relationships apply for a distributed spectrum of incident radiation. In order to

relate the present analysis to that in the preceding section, consider how to relate the
case of a monochromatic field to that of a distributed spectrum, with a spectral width
that is large compared to the transition width. One way this can be done is to consider
1→2
the process of building up the distributed absorption rate g stim ( E ) from the
absorption rates due to a series of monochromatic fields, with appropriately scaled
energy densities (Ref. 35). For some coefficient C, a single, monochromatic line
spectrum at frequency ω0 leads to the following frequency domain contribution to the
rate of stimulated generation:

1→2 _
G stim ( E ) = Cn E δ ( E – h ω 0 )

where nE (SI unit: m-3) is the photon density at the photon energy E = hω 0 . To
build up the corresponding distributed spectrum requires summing over a series of
these line spectra, each with a different value for the driving photon energy E’ and
with, nE' = n(E’)d(E’). Correspondingly the sum can be transformed to an integral in
the following way:

1→2
g stim ( E ) =  n ( E' )Cδ ( E – E' ) d( E' ) = Cn ( E ) (3-113)

Comparing Equation 3-113 with Equation 3-112 shows that:

_ 2 3
hπ c - _
C = --------------------------
3 2
f v ( 1 – f c )g red ( h ω )
n ω 0 τ spon

The same argument can be applied to the rate of stimulated emission. Integrating the
monochromatic rate over all frequencies gives the total generation/recombination rate
due to the line source in both of these cases:

196 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 _ 2 3
1→2 h π c nE _
G stim = --------------------------
3 2
- f v ( 1 – f c )g red ( h ω 0 )
n ω 0 τ spon
(3-114)
_ 2 3
2→1 h π c nE _
R stim = --------------------------
3 2
- f c ( 1 – f v )g red ( h ω 0 )
n ω 0 τ spon

With a monochromatic line source the stimulated emission must occur at the
frequency of the source. In contrast, the spontaneous emission can occur at any
frequency allowed by the band structure. The rate of spontaneous recombination is
therefore computed by summing contributions from all possible transitions within the
band, for the electron distribution. The radiation emitted is not in phase with any
sources producing stimulated emission, and has a broader spectrum since transitions
can occur between states separated by a range of energies.

The total recombination due to spontaneous emission is given by:

∞
1
R spon = E ------------
τ spon c
g
f ( 1 – f v )g red ( E ) dE (3-115)

Note that fc and fv are both functions of E through Equation 3-104 and
Equation 3-101.

Equation 3-114 and Equation 3-115 determine the rates of stimulated and
spontaneous emission in terms of the spontaneous lifetime. The spontaneous lifetime
can be related to the matrix element for the transition, discussed in the previous
section. The energy density of the electromagnetic field for a traveling wave in a
dielectric medium is ε0εr′E02/2 where E0 is the electric field norm and εr′ is the real
part of the medium permittivity. Correspondingly the number density of the photons
is given by:

2
ε 0 ε r'E 0
n E = ------------------
_ -
2 h ω0

The following equation therefore applies for the rate of stimulated emission:

2 3 2
1→2 π c ε 0 ε r'E 0 _
G stim = ------------------------------
3 3
- f v ( 1 – f c )g red ( h ω 0 )
2n ω 0 τ spon

Comparing this equation with Equation 3-105 shows that the following relationship
exists between the matrix element and the spontaneous emission lifetime:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 197

 _ 3 2
h πc ε 0 ε r'E 0
τ spon = -----------------------------------
- (3-116)
3 3 av 2
4n ω 0 H 12

As a result of Equation 3-116 the rates of spontaneous and stimulated emission can
also be written in the form:

1→2 2π- av 2 _
G stim = ------
_ H f ( 1 – f c )g red ( h ω 0 )
h 12 v
2→1 2π- av 2 _
R stim = ------
_ H f ( 1 – f v )g red ( h ω 0 )
h 12 c
∞ 3 av 2
H 12
4n - ---------------

3
R spon = ------------------------
_4 3 - E f c ( 1 – f v )g red ( E ) dE
2
E g h πc ε
0 E0

Note that it is also common to define a net generation rate due to stimulated emission,
given by:

1→2 2→1 2π av 2 _
G stim = G stim – R stim = ------
_- H ( f v – f c )g red ( h ω 0 ) (3-117)
h 12

THE MATRIX ELEMENT

The preceding section derived the rates of stimulated and spontaneous emission in
terms of the perturbation to the Hamiltonian and the spontaneous emission lifetime.

Equation 3-98 showed the perturbation to the electric field could be written in the
form:

e
H' = ----------- A ⋅ p
2m 0

The matrix element in this form depends on the magnitude of the magnetic vector
potential. It is often more convenient to express the matrix element in terms of the
amplitude of the electric field oscillation (E0) and a unit vector in the direction of the
electric field (e). With the frequency domain gauge for the oscillatory field in
COMSOL Multiphysics, the electric field is related to the magnetic field by the
equation:

∂A
E = – = – i ωA
∂t

Writing the local electric field can be written in the form:

198 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 E = E0 e

gives:

iE 0
A = --------- e
ω

Consequently Equation 3-98 can be written in the form:

ieE 0
H' = ---------------- e ⋅ p
2m 0 ω

and:

eE 0 2
=  ---------------- e ⋅ p
2 2
H' (3-118)
2m 0 ω

The momentum matrix element between to states 1 and 2 (<2|e⋅p|1> using the
bracket notation) is frequently referred to in the literature as M12 — a convention
adopted by COMSOL Multiphysics.

An alternative form for the matrix element is sometimes used. As shown in section
9.1.3 of Ref. 32, the Hamiltonian can also be expressed as:

H' = E 0 e ⋅ μ

where μ is the dipole matrix element (μ=qr where r is the position vector) and
correspondingly:

2 2 2
H' = E0 e ⋅ μ (3-119)

The matrix element <2|e⋅μ|1> is written in the form μ12 in the user interface.

The two forms of the matrix element given in Equation 3-118 and Equation 3-119
can be used to specify the matrix element in COMSOL Multiphysics. Note that the
band averaged matrix element for the particular electric field orientation should be
av av
used ( M 12 or μ 12 ). As discussed in Ref. 36 and Ref. 38, the band-averaged matrix
elements can be derived using a k⋅p perturbation method. For a simple 4-band model
(Ref. 36), assuming unpolarized incident light, the matrix element takes the form
(Ref. 37):

THEORY FOR THE SEMICONDUCTOR INTERFACE | 199

 m0 m0 Eg ( Eg + Δ )
= --------  -------- – 1 -----------------------------
av 2
M 12
6 m∗   2
E g + --- Δ
3

where Eg is the gap energy and Δ is the spin orbital splitting energy of the valence band.
m* is the effective mass of electrons in the conduction band, but typically an
experimentally determined value should be used rather than the theoretical value (see
discussion in Ref. 37; for the case of GaAs, Ref. 38 provides values for other materials).

OPTICAL PROPERTIES
Poynting’s theorem for electromagnetic fields can be written in the form (Ref. 33):

∂ ε0 μ0 ∂P ∂M
 E × H ⋅ n dS =   E ⋅ j + ∂ t  ----2- E ⋅ E + -----2- H ⋅ H + E ⋅ ∂ t + μ0 H ⋅
∂t 
dV
S V
(3-120)

Here E is the electric field, H is the magnetic field strength, P is the material
polarization, M is its magnetization, and V is an enclosed volume within the material,
with surface S.

Each of the terms in Equation 3-120 has a physical interpretation that can be
associated with the flow of energy through the volume. The E×H term on the left hand
side of this equation describes the flow of energy out of the enclosed volume (E×H is
known as Poynting’s vector). The E⋅j term represents the energy expended on moving
charges within the volume. The next term represents the rate of change of the stored
electromagnetic energy in the vacuum and the final two terms represent the power per
unit volume expended on the electric and magnetic dipoles present in the material.

For optical frequencies, the E⋅j term is of order ρE2 for a material resistivity ρ (for
heavily doped silicon ρ is of order 10−5 Ω/m). Most practical semiconductors are
nonmagnetic, so the final term is usually not significant. In the frequency domain, the
term involving the polarization is of order ωε0E2 (ω is of order 1015 rad/s for optical
light, ε0 is 8.85·10-12 F/m so ωε0 is of order 104 Ω/m). Consequently, when
considering the interaction of semiconductors with propagating electromagnetic
waves it is only necessary to consider the material polarization. In the frequency
domain, power per unit volume lost to the material is given by the term:

P v = iωE ⋅ P

The polarization P is given by:

200 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 P = ε 0 χE = ε 0 ( χ' – iχ'' )E

where χ is the (complex) material susceptibility, with real part χ′ and imaginary part
χ″. In the frequency domain E(r,t) and P(r,t) take the form:

iω 0 t
E ( r ,t ) = E ( r )e
iω 0 t iω 0 t
P ( r ,t ) = P ( r )e = ε 0 ( χ' – iχ'' )E ( r )e

Consequently the time averaged dissipated power is given by:

1 1
 P v = --- Re [ iω 0 E∗ ( r ) ⋅ P ( r ) ] = --- ω 0 ε 0 Re [ iE∗ ( r ) ⋅ χE ( r ) ]
2 2

For an isotropic material χ becomes a scalar value:

1 1 2 1 2
 P v = --- Re [ iω 0 ε 0 ( χ' – iχ'' )E∗ ( r ) ⋅ E ( r ) ] = --- ω 0 ε 0 χ'' E ( r ) = --- ω 0 ε 0 χ''E 0
2 2 2
(3-121)

Equation 3-121 shows how the imaginary part of the susceptibility is related to the
power absorption by the material. Since each photon carries an energy hω the total
power adsorbed is directly related to the net rate of stimulated emission (given by
Equation 3-117).

_ 1 2
 P v = h ω 0 G stim = --- ω 0 ε 0 χ''E 0
2

Correspondingly:

_
2 h G stim 4π av 2 0 0 0
χ'' ( ω 0 ) = ------------------------
2
- = ------------
2
H 12 ( f v – f c )g red (3-122)
ε0 E0 ε0 E0

where the superscript 0 has been added to fv, fc, and gred to indicate that these
quantities are evaluated at the energy hω 0 which corresponds to the excitation
frequency.

Equation 3-122 shows how the imaginary part of the susceptibility is related to the
rate of stimulated emission and correspondingly to the transition matrix element.
Associated with the change in the imaginary part of the susceptibility is a small change
in the real part. If the complete frequency spectrum of the susceptibility is known, it is
possible to relate the imaginary part of the susceptibility to the real part using the
Kramers-Kronig relations. The Kramers-Kronig relations result from the constraint

THEORY FOR THE SEMICONDUCTOR INTERFACE | 201

 that in the time domain, the equations for the susceptibility at time t can only depend
on the electric field at times less than t (Ref. 39); that is, the equations are derived from
considering the system to be causal. Alternatively, the equations can be derived by
considering the analytic nature of the complex function χ for physical systems (see
Ref. 32 or Ref. 33 for derivations of this form). The Kramers-Kronig relation for the
susceptibility can be written in the form:

∞
2 ωχ'' ( ω )
χ' ( ω 0 ) = --- P
π 0 ω-------------------
2
–ω
2
dω (3-123)
0

Where the P indicates that the principle value of the integral is required.
Equation 3-123 gives the real part of the susceptibility in terms of the imaginary part,
provided that the entire frequency spectrum of the imaginary part is known. In
practice it is not possible to determine the entire frequency spectrum of the imaginary
part of the susceptibility, so instead a change in the imaginary part of the susceptibility
is considered due to the rearrangement of the carriers in the band. Taking as a
reference configuration for the semiconductor the case of an undoped semiconductor
at equilibrium (at temperatures such that the occupancy of the conduction band is
negligible) we define:

4π av 2 0
χ ref'' ( ω 0 ) = ------------2 H 12 g red (3-124)
ε0 E0

Equation 3-124 corresponds to a reference configuration in which the semiconductor

has an empty conduction band and a full valence band. The real part of the
susceptibility for the reference material can then be written as:

∞ ωχ
2 ref'' ( ω )
χ ref' ( ω 0 ) = --- P
π 0 -------------------------
ω –ω
2 2
dω
0

For a doped semiconductor at arbitrary temperature there is a change in the real and
imaginary parts of the susceptibility:

χ' ( ω ) = χ' ref ( ω ) + Δ ref χ' ( ω )

(3-125)
χ'' ( ω ) = χ'' ref ( ω ) + Δ ref χ'' ( ω )

where the subscript ref indicates a change from the reference configuration and where:

4π av 2
Δ ref χ'' ( ω ) = ------------2 H 12 ( f v – f c – 1 )g red (3-126)
ε0 E0

202 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Substituting Equation 3-125 into Equation 3-123 gives:

∞ ωχ
ref'' ( ω )
∞ ωΔ
2 2 ref χ'' ( ω )
χ ref' ( ω 0 ) + Δ ref χ' ( ω 0 ) = --- P
π 0 -------------------------
ω –ω
2 2
dω + --- P  -----------------------------
π 0 ω2 – ω2
dω
0 0

The χ0 terms can be eliminated using Equation 3-125 giving:

∞ ωΔ
2 ref χ'' ( ω )
Δ ref χ' ( ω 0 ) = --- P
π 0 -----------------------------
ω –ω
2 2
dω (3-127)
0

Equation 3-127 gives the change in the real part of the susceptibility from the
reference material susceptibility, for a given carrier concentration. This makes the
integral easier to perform in practice, since the changes in the susceptibility due to the
carriers is small at energies far from the band gap.

Equation 3-127 can be used to compute the change in the real part of the
susceptibility, given a knowledge of the susceptibility of the reference material. In
COMSOL Multiphysics, the real part of the susceptibility of the reference material is
treated as a user-specified material property (it is determined from the relative
permittivity or refractive index). The change in the refractive index due to the
excitation of carriers into the valence band (either as a result of doping or due to carrier
injection) is then accounted for by means of Equation 3-127. Note that χref′′(ω) is not
known over all frequencies and the real part of the susceptibility is likely to have
significant contributions from other processes that occur at frequencies away from
those corresponding to the band gap. However, since it is possible to define Δrefχ′′(ω)
from Equation 3-126 we can compute the change in the real part of the susceptibility
from Equation 3-127.

At the excitation frequency, ω0, the imaginary part of the susceptibility due to optical
transitions can be computed directly from Equation 3-124. Since the imaginary part
of the reference material susceptibility is not conveniently available as a material
property, COMSOL Multiphysics calculates this part of the susceptibility directly. It is
possible to include additional contributions to the imaginary part of the susceptibility
from mechanisms other than the optical transitions, as discussed in the next section.

MATERIAL PROPERTY INPUTS AND OUTPUTS FOR THE OPTICAL

TRANSITIONS FEATURE
The theory presented in the preceding section is focused on the determination of the
complex susceptibility of the material. Practically speaking, values of the complex

THEORY FOR THE SEMICONDUCTOR INTERFACE | 203

 refractive index (n-ik) or the complex relative permittivity (εr′-iεr″) are more often
available for a given material.

The permittivity is straightforwardly related to the susceptibility by the equations:

ε r' = 1 + χ'
ε r'' = χ''

The optical transitions feature computes changes in the material properties that result
from the transitions occurring in the semiconductor. The real part of the material
permittivity changes from an initial value εr,i′ to a modified value εr,i′ given by:

ε r, f' = ε r, i' + Δε r'

where in this case:

ε r, i' = 1 + χ' ref ( ω 0 )

Δε r' = Δ ref χ' ( ω 0 )

The results from the preceding section have been used in deriving the above equations.
Note that for the real part of the permittivity, the initial permittivity value is the real
part of the permittivity for the reference material at the excitation frequency
(χ′ref(ω0)), and the change in the permittivity is the same as Δrefχ′(ω0). However, this
is not the case for the imaginary part of the susceptibility, because the reference
material contribution to the imaginary part of the permittivity is not generally available
as a a material property. The change in the imaginary part of the permittivity is given
by:

ε r, f'' = ε r, i'' + Δε r''

Typically the initial value of the imaginary part of the permittivity (εr,i′′) is zero.
COMSOL Multiphysics allows for nonzero values of εr,i′′ for cases in which there are
other loss mechanisms in the material besides the optical transitions, so it is possible to
enter nonzero values for this parameter in the feature. These mechanisms contribute
in an additive manner to the material losses (or gain). The change in the imaginary part
of the permittivity is given by:

Δε r'' = χ'' ( ω 0 )

where χ′′(ω0) is given by Equation 3-122.

204 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

In optical applications it is often more common to specify the real and imaginary parts
of the refractive index (n-ik) than to specify the complex relative permittivity. Simple
relationships exist relating these quantities, allowing the material properties to be
expressed in either form. In general, the complex permittivity is equal to the square of
the complex refractive index (see Ref. 32, Ref. 33, or Ref. 35), so that:

( ( n i + Δn ) – i ( k i + Δk ) ) 2 = ε r, i' + Δε r' – i ( ε r, i'' + Δε r'' )

The relationships between these quantities can be obtained by expanding the left hand
side and equating the real and imaginary parts of the equations. For the initial values
of the permittivity and refractive index the following relationships hold:

2
ε r, i' = n i2 – k i
ε r, i'' = 2n i k i

Correspondingly, the remaining quantities are related by the equations:

2 2
Δε r' = 2n i Δn + [ Δn – 2k i Δk – Δk ]
Δε r'' = 2n i Δk + [ 2k i Δn + 2ΔkΔn ]

In practical semiconducting materials, ki, Δk, and Δn are small in comparison to ni, so
the terms in the square brackets are small. This is assumed when generating the
corresponding postprocessing variables.

Finally a further quantity, related to the imaginary part of the complex susceptibility,
is important because it can be measured experimentally. The absorption coefficient (α)
is given by:

ω
α = ------ χ''
cn

The absorption coefficient is available as a postprocessing variable. In laser applications

the gain (γ=−α) is also frequently encountered.

Boundary Conditions
The Semiconductor interface includes boundary conditions appropriate for the
Semiconductor Material Model feature. Additionally, if insulating regions are included
within the model using the Charge Conservation feature, separate boundary conditions
are available for these regions.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 205

 The theory for the following boundary conditions is discussed in the subsequent
sections:

• Metal Contacts
• Thin Insulating Gates
• Continuity/Heterojunction
• Boundary Conditions for Charge Conservation
• Boundary Conditions for the Density-Gradient Formulation

Metal Contacts
The Metal Contact boundary condition is used for modeling different types of
metal-semiconductor junctions. The Ideal Ohmic and Ideal Schottky types of contact
can be modeled with this feature.

IDEAL OHMIC
The ohmic contact option assumes local thermodynamic equilibrium at the contact.
Note that in practice it is often used in nonequilibrium situations where the boundary
condition imposed is no longer physical (for example in a forward biased p-n junction),
which is reasonable provided that the junction is located some distance from the region
of interest. Since equilibrium is assumed, both the hole and electron quasi-Fermi levels
are equal at the boundary. Charge neutrality at the boundary is also assumed so there
is no band bending and the band diagram takes the form shown in Figure 3-12.

206 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 V0=0, T=T0 V0=0, T=T V0>0, T=T

E0 qV0
E0 E0
qχ(T0) qV qχ(T)
qV qχ(T) qV0 qV
n Ec
Ec Ec Ef
0 Ef metal 0
metal Ef metal Eg(T)
Eg(T)
Eg(T0) ΔEf Ev Ev
Ev qV0
E0
E0 E0
qχ(T0) qV qχ
qV qV qχ(T) (T)
p
Ev qV0 Ec
Ec
Eg(T) Eg(T) Ef
Eg(T0)
Ef Ef metal Ev
0 metal metal 0
Ev Ev
ΔEf

Figure 3-12: Energy band diagram for an ohmic contact with n-type (top) and p-type
(bottom) semiconductors under different biasing and temperature conditions (left to
right). On the left the semiconductor is in the reference configuration at the equilibrium
temperature T 0 and at zero bias V0=0. In the center, the temperature of the semiconductor
is raised, changing the gap Eg and the electron affinity χ. Correspondingly, the energy of
the conduction band edge Ec and of the valence band edge Ev, is shifted as is the Fermi
level Ef. In this configuration the vacuum energy E0 also changes as a result of differences
in the space charge distribution in the device. On the right, in addition to the temperature
change, a bias V0 is applied to the metal, which shifts the entire band structure (as well as
the vacuum level) up in energy by qV0 from the configuration in the center.

Since equilibrium is assumed, Equation 3-49 is used for the carrier concentrations but
it is useful to write it in the alternative forms:

 E c – E f  E f – E i
n eq = γ n N c exp  – ------------------ = γ n n i ,eff exp  ------------------
 k b T   kb T 
(3-128)
 E f – E v  E i – E f
p eq = γ p N v exp  – ------------------- = γ p n i ,eff exp  ------------------
 k b T   kb T 

where Equation 3-77 is used to derive the expressions for n and p in terms of the
effective intrinsic carrier concentration. The charge neutrality condition states:

– +
n eq – p eq + N a – N d = 0 (3-129)

THEORY FOR THE SEMICONDUCTOR INTERFACE | 207

 Solve Equation 3-128 and Equation 3-129 in a form convenient for numerical analysis
by using the result neqpeq=γnγpni,eff2 (Equation 3-91). Using this result with
Equation 3-129 gives:

+ – 2
( p eq + ( N d – N a ) )p eq – γ n γ p n i ,eff = 0

This quadratic equation can be solved for peq giving:

1 + – 1 + – 2 2
p eq = – --- ( N d – N a ) ± --- ( N d – N a ) + 4γ n γ p n i ,eff
2 2

peq must be positive therefore:

1 + – 1 + – 2 2
p eq = – --- ( N d – N a ) + --- ( N d – N a ) + 4γ n γ p n i ,eff
2 2
(3-130)
1 + – 1 + – 2 2
n eq = --- ( N d – N a ) + --- ( N d – N a ) + 4γ n γ p n i ,eff
2 2

From Equation 3-128 the difference between the intrinsic level and the Fermi level can
be determined in the following manner:

n eq p eq
E f – E i = k B T ln  ------------------ = – k B T ln  ------------------
 γ n n i ,eff  γ p n i ,eff

This result applies at arbitrary temperatures. Using Equation 3-91 and the above
result, the conduction band energy level can be related to the intrinsic level and
correspondingly to the equilibrium Fermi level in the following manner:

Nv n eq Nv
E c = E i + --- E g – ------- ln  ------- = E f – k B T ln  ------------------ + --- E g – ------- ln  -------
1 kT 1 kT
2 2 Nc γ n n i ,eff 2 2 Nc

where the result Ev=Ec−Eg is used. The vacuum potential, V, is therefore:

kB T n eq Nv
V = –  --- E c + χ = -----------  ln  ------------------ + --- ln  -------  – ---  E f + --- E g – χ
1 1 1 1
(3-131)
q q γ n n i ,eff 2 Nc q 2

Equation 3-131 gives the vacuum potential relative to the Fermi level at an arbitrary
temperature. However, it does not fix the vacuum potential on an absolute scale (the
value of Ef is not known on this scale). As discussed in The Semiconductor Equations,
COMSOL Multiphysics references the vacuum potential to the Fermi level in the
equilibrium reference configuration (shown on the left of Figure 3-12). Since the
vacuum level adjusts to accommodate the space charge layers that are created in the
device, it does not remain constant when the temperature is changed away from

208 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

equilibrium. For an intrinsic semiconductor, however, the vacuum level remains
constant when the temperature is changed. This is shown in Figure 3-13. Ef on the
absolute scale must be set in a such a way that the limiting case of an intrinsic
semiconductor is correct.

T=T0 T=T

E0 E0
qV(T0)
qV(T)=qV(T0)
qχ(T0) qχ(T)

Ec Ec
Ei=Ef
Ei=Ef=0 0
Eg(T0) Eg(T)
Ev Ev

Figure 3-13: Diagram showing the band structure of an intrinsic semiconductor at the
equilibrium reference temperature and at the local temperature (with no applied
voltages). In this case the vacuum energy level does not change with a change in the local
temperature.

For the particular case of the intrinsic semiconductor:

V ( T ) = V ( T0 )

Using the above equation and Equation 3-131, and the fact that the intrinsic
semiconductor has no band gap narrowing,

k B T  N v ( T ) 1 
- ln ---------------- – --- E i ( T ) + --- E g ( T ) – χ ( T )
1 0 0
----------
2q  N c ( T ) q  2 
kB T0 Nv ( T0 ) 1
= -------------- ln  ------------------- – ---  E i ( T 0 ) + --- E g ( T 0 ) – χ ( T 0 )
1 0 0
2q Nc ( T 0 ) q 2

where χ0 is the electron affinity in the absence of band gap narrowing and Eg0 is the
energy gap in the absence of band gap narrowing. By definition Ei(T0) = Ef(T0) = 0
therefore:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 209

 kB Nv ( T ) Nv ( T0 )
E i ( T ) = E f ( T ) = -------  T ln  ---------------- – T 0 ln  -------------------  + χ ( T 0 ) – χ ( T )
0 0
2q   Nc ( T )   N c ( T 0 ) 
(3-132)
1 0 0
+ ------- ( E g ( T 0 ) – E g ( T ) )
2q

Equation 3-132 sets the value of the Fermi level relative to the zero in potential at
temperature T for an intrinsic semiconductor. Ef(T) is the offset in the energy scale
relative to the equilibrium temperature for the particular case of an intrinsic
semiconductor. Next consider a semiconductor with a nonuniform doping distribution
but with a region of intrinsic semiconductor. The entire semiconductor is heated
uniformly to raise its temperature from T0 to T1 (with no applied biases). The Fermi
level in the intrinsic semiconductor changes according to Equation 3-132. Since the
whole sample is still at equilibrium, the Fermi level everywhere else in the sample must
also change by the same amount. Equation 3-132 gives the value of the Fermi level at
equilibrium for any level of doping (in the absence of applied biases). The
temperature-dependent (and doping-independent) offset is defined in the Fermi level:

kB Nv ( T ) Nv ( T0 )
ΔE f ( T ) = -------  T ln  ---------------- – T 0 ln  -------------------  + χ ( T 0 ) – χ ( T )
0 0
2q Nc ( T ) Nc ( T0 )
(3-133)
1 0 0
+ ------- ( E g ( T 0 ) – E g ( T ) ) ≡ V eq, adj
2q

Equation 3-131 can be written as:

kB T n eq Nv
V = -----------  ln  ------------------ + --- ln  -------  – ---  ΔE f + --- E g – χ ≡ V eq
1 1 1
q   γ n n i ,eff 2  N c  q  2 

So with an applied bias V0, the boundary condition on V should be:

kB T n eq Nv
V = -----------  ln  ------------------ + --- ln  -------  – ---  ΔE f + --- E g – χ + V 0 = V eq + V 0(3-134)
1 1 1
q   γ n n i ,eff 2  N c  q  2 

The ohmic contact boundary condition imposes Equation 3-130 on the carrier
concentrations and Equation 3-134 on the potential at the boundary. For the
Semiconductor Equilibrium study step, only the constraint on the electric potential is
applied. Note that all metal contacts attached to the same Semiconductor Material Model
feature are biased at a common voltage, as dictated by the equilibrium condition. See
Semiconductor Equilibrium Study Settings on how to set the common bias voltage.

210 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

IDEAL SCHOTTKY
The metal-semiconductor interface is one of the most challenging problems in the field
of device physics. A variety of physical phenomena such as the influence of interface
states, the effect of interfacial layers, scattering, potential fluctuations, interface
roughness, image forces, and trap-assisted tunneling make practical detailed modeling
of these devices complex. In practice, boundary conditions used in simulations must
neglect much of the complexity of these devices.

The ideal Schottky option adopts a simplified model for Schottky contacts, which is
based largely on the approach introduced by Crowell and Sze (Ref. 24). The
semiconductor is assumed to be nondegenerate, since metal-degenerate
semiconductor contacts are usually best represented by the Ideal Ohmic option.

The contact acts as a source or sink for carriers and consequently it can be treated as a
surface recombination mechanism:

J n ⋅ n = – qv n ( n – n 0 )
J p ⋅ n = qv p ( p – p 0 )

Here, n is the outward normal of the semiconducting domain, vn and vp are the
recombination velocities for holes and electrons, respectively, and n0 and p0 are the
quasi-equilibrium carrier densities — that is, the carrier densities that would be
obtained if it were possible to reach equilibrium at the contact without altering the
local band structure. n0 and p0 are correspondingly defined as though the Fermi level
of the semiconductor at the boundary is equal to that of the metal. From Figure 3-14
n0 and p0 are given by:

 E c – E fm ΦB
n 0 = N c exp  – ----------------------- = N c exp  – -----------
 kB T   k B T
(3-135)
 E fm – E v Eg – ΦB
p 0 = N v exp  – ----------------------- = N v exp  – ---------------------
 kB T   kB T 

where:

ΦB = Φm – χ (3-136)

Here, Φm is the metal work function, Efm is the metal Fermi level, and ΦB is the
emission barrier height for electrons (from the metal). The observed barrier heights of
Schottky junctions frequently do not conform to Equation 3-136, largely as a result of
the complexities discussed previously. From a practical perspective, the value of ΦB can

THEORY FOR THE SEMICONDUCTOR INTERFACE | 211

 be determined experimentally and, consequently, can be considered as an input into
the model (COMSOL Multiphysics makes it possible to directly enter the barrier
height or to compute the ideal barrier height from the material properties).

V0=0, T=T0 V0=0, T=T V0<0, T=T

qV0
E0 E0
E0
qΦm qV qχ
qV qΦm (T) qV qχ(T)
qχ(T0) qΦm
n Ec
Ec Ec
metal Ef Ef
0 metal Eg(T) Eg(T) 0
Ef metal
Eg(T0) ΔEf
qV0 Ev
Ev Ev

V0=0, T=T0 V0=0, T=T V0>0, T=T

E0 qV0
E0
qχ(T0) E0
qV qχ(T) qV qχ(T)
qV
p qΦ qΦm
m
Ec qΦm Ec
Ec
Eg(T) Eg(T)
Eg(T0) metal Ef
Ef metal Ef Ev
0 metal Ev 0
Ev ΔEf qV0

Figure 3-14: Energy band diagram for a Schottky contact with n-type (top) and p-type
(bottom) semiconductors under different biasing and temperature conditions (left to
right). On the left, the semiconductor is in the reference configuration at the equilibrium
temperature T 0 and at zero bias V 0=0. In the center, the temperature of the semiconductor
is raised, changing the gap Eg and the electron affinity χ. Correspondingly, the energy of
the conduction band edge Ec and of the valence band edge Ev is shifted, as is the Fermi
level Ef, in equilibrium regions of the device (indicated to the right of the band diagram).
In this configuration the vacuum energy E0 also changes as a result of differences in the
space charge distribution in the device. On the right, in addition to the temperature
change, a forward bias V0 is applied to the metal.

The recombination velocities vn and vp are determined by assuming that the dominant
source of current across the junction is thermionic emission. Ref. 3 provides a full
derivation of this case, which shows that, if thermionic emission is dominant:

212 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 2
A n∗ T
v n = -----------------
qN c
(3-137)
2
A p∗ T
v p = -----------------
qN v

Here, An* and Ap* are the effective Richardson’s constants for electrons and holes,
respectively (these are essentially material properties related to the thermionic emission
— Ref. 3 has details). The Schottky contact boundary condition allows the
recombination velocities to be user defined or to be determined from Equation 3-137.
In addition, an extra current contribution can be specified with either a user-defined
scaling factor or the scaling factor computed using the WKB Tunneling Model.

Equation 3-135, Equation 3-136, and Equation 3-137 specify the boundary
condition on the currents imposed at a Schottky contact. The boundary condition on
the voltage can be determined from Figure 3-14:

ΔE f
V = – ( Φ B + χ ) – ---------- + V 0 (3-138)
q

Equation 3-138 is used by COMSOL Multiphysics to constrain the potential at the

boundary. For the Semiconductor Equilibrium study step, only the constraint on the
electric potential is applied. Note that all metal contacts attached to the same
Semiconductor Material Model feature are biased at a common voltage, as dictated by
the equilibrium condition. See Semiconductor Equilibrium Study Settings on how to
set the common bias voltage.

Thin Insulating Gates

The band diagram for a thin insulating gate is shown in Figure 3-15. As for the ohmic
contact, a change in the temperature of the device away from the equilibrium reference
results in a shift of the metal and semiconductor Fermi levels (ΔEf) on the absolute
energy scale, given by Equation 3-133. An applied potential further shifts the metal
Fermi level with respect to the semiconductor Fermi level (which can be defined in
regions away from the contact where little current flows).

THEORY FOR THE SEMICONDUCTOR INTERFACE | 213

 V0=0, T=T0 V0=0, T=T V0>0, T=T

qΦm
E0 E0 E0
qΦm qχ(T0) qΦm
n qV qχ(T) qV0 qV qχ(T)
qV
insulator

insulator

insulator
Ec metal Ec
Ec metal Ef Ef
0 metal Eg(T)
Eg(T0) Ef Eg(T) 0
ΔEf
Ev Ev Ev
dins
V0=0, T=T0 V0=0, T=T V0<0, T=T
E0 E0
E0
qχ(T0) qχ(T) qχ(T)
qV qV qV
qΦm
p Ec Ec
Ec
insulator
insulator

insulator
qΦm Eg(T) qΦm Eg(T)
Eg(T0) Ef
Ef metal Ef
0 Ev 0
metal Ev Ev
ΔEf metal
dins qV0

Figure 3-15: Energy band diagram for a thin insulating gate with n-type (top) and p-type
(bottom) semiconductors under different biasing and temperature conditions (left to
right). On the left, the semiconductor is in the reference configuration at the equilibrium
temperature T0 and at zero bias V0=0. In the center, the temperature of the semiconductor
is raised, changing the gap Eg and the electron affinity χ. Correspondingly, the energy of
the conduction band edge Ec and of the valence band edge Ev is shifted, as is the Fermi
level Ef, in equilibrium regions of the device (indicated to the right of the band diagram).
In this configuration the vacuum energy E0 also changes as a result of differences in the
space charge distribution in the device. On the right, in addition to the temperature
change, a bias V0 is applied to the metal, leading to an inversion layer under the gate.

When the Thin Insulator Gate boundary condition is used, the thin insulating layer is
not included in the COMSOL Multiphysics model but its effect is included in the
formulation of the boundary condition. The insulator is assumed to be so thin that the
electric field (which must be perpendicular to the metal surface) is to a good
approximation also perpendicular to the insulator-semiconductor boundary. When this
is the case, the normal electric displacement field (D) at the insulator-semiconductor
interface can be written as:

Vg – V
n ⋅ D = – ε ins ε 0 -----------------
d ins

214 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

where Vg is the potential on the gate, V is the potential at the insulator-semiconductor
interface, dins is the thickness of the insulator, εins is the relative dielectric permittivity
of the insulator, ε0 is the permittivity of free space, and n is the outward normal of the
semiconductor domain. From Figure 3-15, the gate voltage with applied potential V0
is given by:

ΔE f
V g = – Φ m + V 0 – ----------
q

where Φm is the metal work function. ΔEf is given by Equation 3-133. Therefore:

ΔE f
V + Φ m – V 0 + ----------
q
n ⋅ D = ε ins ε 0 ------------------------------------------------- (3-139)
d ins

The electron and hole currents flowing into the insulator must be zero therefore:

n ⋅ Jn = 0
(3-140)
n ⋅ Jp = 0

The thin insulating gate applies the Neumann boundary conditions given by
Equation 3-139 and Equation 3-140.

TRAPS ON THE SEMICONDUCTOR/INSULATOR BOUNDARY

When traps are present on the gate the traps accumulate charge as described in the
Traps section. As a result of the charge on the boundary the electric displacement field
is discontinuous across the interface and consequently:

Vg – V
n ⋅ D = – ε ins ε 0 ----------------- – Q
d ins

where Q is the surface charge density due to the traps, given by Equation 3-89 or
Equation 3-90. Consequently the form of Equation 3-139 is modified so that:

ΔE f
V + Φ m – V 0 + ----------
q
n ⋅ D = ε ins ε 0 ------------------------------------------------- – Q
d ins

Similarly, in a transient study, the traps can provide a sink for electrons or holes and
Equation 3-140 is changed such that:

THEORY FOR THE SEMICONDUCTOR INTERFACE | 215

 n ⋅ Jn = Re
n ⋅ Jp = Rh

where Re and Rh are given by Equation 3-88.

Continuity/Heterojunction
Space charge regions develop at the interface between two different semiconductors,
as a result of the change in the band structure that occurs at the junction. Such
interfaces are referred to as heterostructures. COMSOL Multiphysics handles all
interior boundaries in a model using the Continuity/Heterojunction feature, which
automatically detects whether a change in the band structure occurs at the interface.
For interior boundaries within a single material, the electron and hole densities are
continuous and the boundary has no effect on the solution. When a heterostructure is
present, one of the following Continuity model options should be selected from the
Heterojunction settings:

• Continuous quasi-Fermi levels

• Thermionic emission

The boundary conditions for these two options are described in more detail below.

CONTINUOUS QUASI-FERMI LEVELS

This boundary condition enforces continuity of the quasi-Fermi levels at the junction:

E fn1 = E fn2 E fp1 = E fp2

where Efn1 and Efp1 are the electron and hole quasi-Fermi levels in material 1 and Efn2
and Efp2 are the electron and hole quasi-Fermi levels in material 2.

This is an approximate boundary condition, and applies when the junction has
negligible resistivity.

Fermi-Dirac statistics are currently not supported when the thermionic

emission boundary condition is used.

THERMIONIC EMISSION
The thermionic emission boundary condition is based on Ref. 25.

In the limit of no tunneling, the normal currents across the junction are given by:

216 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

  – ( E c2 – E c1 ) ⁄ ( k B T )
 – q ( v n1 n 1 e – v n2 n 2 ) E c2 ≥ E c1
n 1 ⋅ J n1 = 
– ( E c1 – E c2 ) ⁄ ( k B T )
 q(v n e – v n1 n 1 ) E c2 < E c1
 n2 2

 – ( E v1 – E v2 ) ⁄ ( k B T ) (3-141)
 q ( v p1 p 1 e – v p2 p 2 ) E v1 ≥ E v2
n 1 ⋅ J p1 = 
 – q ( v p e –( E v2 – E v1 ) ⁄ ( k B T ) – v p ) E v1 < E v2
 p2 2 p1 1

n 2 ⋅ J n2 = – n 1 ⋅ J n1 n 2 ⋅ J p2 = – n 1 ⋅ J p1

where n1·Jn1 is the outward normal electron current from material 1, n2·Jn2 is the
normal electron current leaving material 2, n1·Jp1 is the outward normal hole current
from material 1, n2·Jp2 is the normal hole current leaving material 2. vn1, vn2, vp1,
and vp2 are the electron and hole recombination velocities for each material at the
boundary. n1, n2, p1, and p2 are the electron and hole concentrations on each side of
the boundary. Other quantities are defined in Figure 3-16. The recombination
velocities are given by:

2 2
A n1∗ T A n2∗ T
v n1 = -------------------- v n2 = --------------------
qN c1 qN c2
2 2
A p1∗ T A p2∗ T
v p1 = -------------------- v p2 = --------------------
qN v1 qN v2

where the effective Richardson’s coefficients An1*, An2*, Ap1*, and Ap2* are given by:

2 2
4πm e1∗ k B 4πm e2∗ k B
A n1∗ = ------------------------------
3
-q A n2∗ = ------------------------------
3
-q
h h
2 2
4πm h1∗ k B 4πm h2∗ k B
A p1∗ = -------------------------------
3
q A p2∗ = -------------------------------
3
q
h h

The effective masses in the above equations are assigned the smaller value of the
density of states effective mass between the 2 domains according to Ref. 25. The
effective mass can be altered in the equation view if required.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 217

 E0

E0
χ1
χ2

Ec1 Ec2

Efn1 Efn2
Eg1 Eg2
Efp1 Efp2
Ev1
Ev2
Semiconductor 1 n2 n1 Semiconductor 2

Figure 3-16: Heterostructure band diagram showing the vacuum energy level, E0, and
the conduction (Ec1, Ec2) and valence band (Ev1, Ev2) energies for the two materials.
The electron (Efn1, Efn2) and hole (Efp1, Efp2) quasi-Fermi levels are also shown, as are
the material band gaps (Eg1, Eg2) and affinities (χ1, χ2).

WKB Tunneling Model

The effect of tunneling can be accounted for by including an extra proportional factor
δ in the thermionic current density, using the formulas derived in Ref. 25 (based on
the WKB approximation).

The electron and hole normal current densities given in Equation 3-141 are multiplied
by a factor of (1+δnJn) and (1+δnJp), respectively. The extra current factors δnJn and
δnJp are given by the double integration along the electrical field line (dl) and along
the energy axis (dVx):


2
qV max qV – 4π
- V max – ---------x- ----------
-------------- max(0, 2m ( E b – qV x )) dl
q kB T h
V
kB T
δ = ----------- e e e 1
dV x (3-142)
kB T min

where h is the Planck constant, the limits of the energy axis integration are given by

V min = max(E b1, E b2) ⁄ q

V max = max(E b) ⁄ q

where the max function of the second equation is taken within the domain selection
of the potential barrier, and the potential barrier variable Eb is given by

218 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

  Ec electrons
Eb = 
 –Ev holes

and Eb1 and Eb2 are the values of Eb at the two opposite boundaries 1 and 2 across
the potential barrier and connected by an electric field line, evaluated on the side of
each boundary that is (immediately) outside of the potential barrier domain selection.

2
The integral … dl is the line integral along the electric field line between the two
1
opposite boundaries 1 and 2 across the potential barrier domain selection.

DENSITY-GRADIENT FORMULATION
For the density-gradient formulation, the Continuous Quasi-Fermi levels option is
automatically satisfied with the Lagrange shape functions used for the quasi-Fermi
level and Slotboom variables. This mimics the continuous nature of the quantum
mechanical wave function, although it should be treated as phenomenology at the best
(Ref. 45).

The Thermionic Emission option assumes the thermionic emission process dominates
and allows the quasi-Fermi level and the Slotboom variable to be discontinuous across
the heterojunction. The same formula as the drift-diffusion theory is used for the
thermionic current density, yielding similar results.

Boundary Conditions for Charge Conservation

When a charge conservation domain is added to a model all the boundary conditions
available for electrostatics can be applied to exterior boundaries of the domain
(boundaries adjacent to the semiconductor and interior boundaries automatically
obtain the correct boundary conditions). These boundary conditions include a work
function offset for the electric potential where appropriate. However, ΔEf is not added
automatically to the model. For this reason, it is recommended that for isothermal
models the equilibrium reference temperature (set this in the Semiconductor interface
Settings window under Reference Temperature) is set to the same temperature as the
model temperature. For nonisothermal models, ΔEf can be included in the work
function in the following manner:

Φm=Φm0+intop1(semi.V_eq_adj)

where is Φm0 the metal work function and an integration coupling operator (intop1)
has been added on a point within the semiconducting domain adjacent to the insulator
(if more than one Semiconductor interface is added to the domain, the prefix semi in
the variable within the integration operator might need to be incremented).

THEORY FOR THE SEMICONDUCTOR INTERFACE | 219

 Boundary Conditions for the Density-Gradient Formulation
For the density-gradient formulation, the additional dependent variables φn and φp
(Equation 3-78) usually obey simple natural boundary conditions (Ref. 47):

n ⋅ ( bn ∇ n ) = 0
(3-143)
n ⋅ ( bp ∇ p ) = 0

In the case of a boundary representing an abrupt potential barrier, for example, a

silicon-oxide interface, Jin et. al. in Ref. 48 suggested using the WKB approximation
to obtain the boundary condition:

– b n, ox
n ⋅ ( b n ∇ n ) = ----------------- n (3-144)
dn

where bn,ox is the density-gradient coefficient in the oxide (SI unit: V m^2), and dn is
the penetration depth into the oxide (m), given by

_2
h
b n, ox = -----------------------------
*
-
12 q m n, ox
_ (3-145)
h
d n = ----------------------------------------------
2 q m n, ox Φ n, ox

where m*n,ox and mn,ox are the effective masses in the oxide (SI unit: kg) and Φn,ox is
the potential barrier height (SI unit: V). For silicon-silicon dioxide interfaces, Jin et.
al. suggested the values of 0.22 m0, 0.5 m0, and 3.15 V, where m0 is the electron
mass.

Tunneling
Equation 3-37 in the section Electrons in a Perturbed Periodic Potential describes the
transport of both electrons and holes in a semiconductor with a potential that varies in
space. In many circumstances the potential varies slowly in space and wave packets
move according to the classical Hamiltonian, so correspondingly their dynamics is
determined by the semiclassical model, from which both drift-diffusion equations
follow. However, since Equation 3-37 is essentially a Schrödinger equation for
electrons and holes, it allows for quantum mechanical phenomena such as tunneling,
which are outside of the scope of the semiclassical model.

220 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

Tunneling occurs when an electron or hole wave packet is incident on a narrow
potential barrier with a barrier height greater than the energy associated with the group
velocity of the wave packet. A classical particle would be reflected at the boundary, but
in the solutions of the Schrödinger equation the envelope function Ψm(r) associated
with the wave packet decays exponentially inside the barrier. For thin barriers, Ψm(r)
is significant on the opposite side of the barrier, and correspondingly there is a
transmission probability associated with transport through the potential barrier. Since
the tunneling probability must be computed for wave packets with a range of energies
or k-vectors, it is usually only possible to solve the tunneling problem in 1D.
Consequently, the approaches adopted are necessarily phenomenological in nature.
The nature of the approximations required to arrive at the theory of tunneling is
discussed in detail in section 4 of Ref. 40. Since much of the underlying theory is
approximate in nature, it is sensible to favor numerically lighter solutions over more
detailed approaches.

Tunneling commonly occurs in semiconductor devices at a variety of locations,

including:

• Tunneling through the barrier formed at a Schottky diode when the semiconductor
is heavily doped underneath the metal contact (this is one way to produce an ohmic
contact).
• Tunneling through the potential barrier formed by the depletion layer associated
with a heterojunction.
• Direct tunneling through thin layers of oxide in very thin gates.
• Tunneling into the conduction band of an oxide material when a high bias is applied
to a thicker gate oxide (Fowler-Nordheim tunneling).
Note that for tunneling through direct band gap oxide barriers (such as silicon oxide)
in indirect band gap semiconductors (such as silicon), an additional complication is
encountered due to the band structure mismatch that occurs as a result of the change
in the material at the interface. This mismatch is not usually observed in practice — for
reasons that are currently not fully understood (see the discussion in chapter 5 of
Ref. 43). In this case the phenomenological perspective adopted in Ref. 40 is usually
employed, and tunneling through the barrier is treated without accounting for the
band structure mismatch, since this approach best describes the experimental results.

In COMSOL Multiphysics, features enabling the modeling of Fowler-Nordheim

tunneling are available.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 221

 FOWLER-NORDHEIM TUNNELING
When large fields are applied across an oxide layer, tunneling can occur directly into
the conduction band of the oxide, as shown in Figure 3-17 below.

E0

qV qχ
qΦm
tunneling Ec
Eg Ef E0
tunneling
Ev metal
qΦm
qV qχ
qV0
qV0

metal Ec
Ef
Eg
Ev
(a) (b)

Figure 3-17: Band diagrams showing Fowler-Nordheim electron tunneling into the
conduction band of an oxide. Eg is the band gap of the semiconductor, χ. is the electron
affinity and Φm is the work function of the metal. The energy of the conduction band edge
in the semiconductor is Ec, that of the valence band edge is Ev, the vacuum level is E0 and
the Fermi level is Ef. In both (a) and (b) the temperature of the device is the same as the
equilibrium reference temperature (temperature changes would cause an additional shift
in the Fermi-level which are not shown in the figure - see for example, Figure 3-15). In
(a) a large potential V0 is applied to the semiconductor. The band structure of the
insulator is such that electrons can tunnel through the triangular barrier formed by the
conduction band edge in the insulator, into the insulator conduction band. In (b) the
potential V0 is applied to the metal, and consequently electrons from the metal can tunnel
into the insulator conduction band. Once the electrons tunnel into the conduction band of
the insulator they are transported via drift diffusion into the metal (case (a)) or
semiconductor (case (b)). Note that the tunnel barrier presented to the electrons is
triangular in shape in both cases.

Fowler and Nordheim originally treated tunneling through triangular barriers

(Ref. 41) in the context of the Sommerfeld model. Their analysis was applied to
tunneling in oxides within a semiconductor device context by Lenzlinger and Snow
(Ref. 42). It is worth noting that Fowler and Nordheim did not make a WKB
approximation in their original analysis, but instead approximated the final form of the
analytic solution (a WKB approximation was used in Ref. 42 but for this reason is not
strictly necessary). The original result due to Fowler and Nordheim showed that the
n p
tunnel current through a triangular barrier for electrons ( J FN ) and/or holes ( J FN )
takes the form:

222 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 n
n n 2  B FN
J FN = A FN E ins exp  – -----------
 E ins
(3-146)
p
p p 2  B FN
J FN = A FN E ins exp  – -----------
 E ins
n p n p
where Eins is the electric field in the insulator and A FN , A FN , B FN , and B FN are
constants related to the material properties of the insulator and semiconductor. For the
commonly encountered case of the silicon/silicon oxide system, only electron
tunneling needs to be considered because the barrier for hole tunneling is significantly
higher than that for electrons.

Ref. 42 provides a detailed derivation of Equation 3-146, for the case of a

semiconductor-oxide system at finite temperature. The effect of a finite temperature
n n
was shown to be equivalent to a change in the values of the constants A FN and B FN .
Similarly image force lowering effects at the barrier also caused an effective change in
these constants. Since using the derivative material properties does not lead to a fully
consistent description of the tunneling current as a function of both electric field and
n n
temperature, it is customary to simply treat A FN and B FN as inputs to a model of the
tunneling process at a particular temperature. These values can be extracted from a plot
n
of log( J FN /Eins2) versus 1/Eins.

References for the Semiconductor Interface

1. N.W. Ashcroft and D.N. Mermin, Solid State Physics, Harcourt Publishing, 1976.

2. R.G. Chambers, Electrons in Metals and Semiconductors, Chapman and Hall,

1990.

3. M. Shur, Physics of Semiconductor Devices, Prentice Hall, 1990.

4. S.M. Sze and K.K. Ng, Physics of Semiconductor Devices, Wiley, 2007.

5. S.M. Sze and M.K. Lee, Semiconductor Devices Physics and Technology, John Wiley
& Sons, 2012.

6. S. Selberherr, Analysis and Simulation of Semiconductor Devices, Springer Verlag,

1984.

7. A.I.M. Rae, Quantum Mechanics, Taylor & Francis, 2007.

8. L.D. Landau and E.M. Lifshitz, Quantum Mechanics (Non-relativistic theory),

Course of Theoretical Physics (vol. 3), Butterworth-Heinmann, 2003.

THEORY FOR THE SEMICONDUCTOR INTERFACE | 223

 9. J. C. Slater, “Electrons in Perturbed Periodic Lattices,” Physical Review, vol. 76,
no. 11, pp. 1592–1601, 1949.

10. L.D. Landau and E.M. Lifshitz, Statistical Physics Part 1, Course of Theoretical
Physics (vol. 5), Butterworth-Heinmann, 1980.

11. R.C. Jaeger and F.H. Gaensslen, “Simulation of Impurity Freezeout Through
Numerical Solution of Poisson’s Equation with Application to MOS Device
Behavior,” IEEE Transactions on Electron Devices, vol. 27, no. 5, pp. 914–920,
1980.

12. U. Lindefelt, “Current-density Relations for Nonisothermal Modelling of

Degenerate Heterostructure Devices,” J. Applied Physics, vol. 75, no. 2, pp. 958–
966, 1994.

13. A.H. Marshak and C.M. Vliet, “Electrical Current and Carrier Density in
Degenerate Material with Nonuniform Band Structure,” Proceedings of the IEEE,
vol. 72, no. 2, pp. 148–164, 1964.

14. J.M. Dorkel, “On Electrical Transport in Non-Isothermal Semiconductors,”

Solid-State Electronics, vol. 26, no. 8. pp. 819–821, 1983.

15. R. Kim and M. Lundstrom, “Notes on Fermi-Dirac Integrals,” arXiv:0811.0116

[cond-mat.mes-hall], 2008. http://arxiv.org/abs/0811.0116.

16. N.D. Arora, J.R. Hauser, and D.J. Roulston, “Electron and Hole Mobilities in
Silicon as a Function of Concentration and Temperature,” IEEE Transactions on
Electron Devices, vol. 29, no. 2, pp. 292–295, 1982.

17. N.H. Fletcher, “The High Current Limit for Semiconductor Junction Devices,”
Proceedings of the IRE, vol. 45, no. 6, pp. 862–872, 1957.

18. J.M. Dorkel and Ph. Leturcq, “Carrier Mobilities in Silicon Semi-empirically
Related to Temperature, Doping and Injection Level,” Solid-State Electronics, vol.
24, no. 9, pp. 821–825, 1981.

19. C. Lombardi, S. Manzini, A. Saporito, and M. Vanzi, “A Physically Based Mobility

Model for Numerical Simulation of Nonplanar Devices,” IEEE Transactions on
Computer-Aided Design, vol. 7, no. 11, 1988.

20. C. Canali, G. Majni, R. Minder, and G. Ottaviani, “Electron and Hole Drift
Velocity Measurements in Silicon and Their Empirical Relation to Electric Field and
Temperature,” IEEE Transactions on Electron Devices, vol. 22, no. 11, pp. 1045–
1047, 1975. Note the correction in: G. Ottaviani, “Correction to ‘Electron and hole

224 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

drift velocity measurements in silicon and their empirical relation to electric field and
temperatures’,” IEEE Transactions on Electron Devices, vol. 23, no. 9, p. 1113,
1976.

21. R.N. Hall, “Electron-Hole Recombination in Silicon,” Physical Review, vol. 87,
no. 2, p. 387, 1952.

22. W. Shockley and W.T. Read, “Statistics of the Recombinations of Electrons and
Holes,” Physical Review, vol. 87, no. 5, pp. 835–842, 1952.

23. Y. Okuto and C. R. Crowell, “Threshold Energy Effect on Avalanche Breakdown

in Semiconductor Junctions,” Solid-State Electronics, vol. 18, pp. 161–168, 1975.

24. C.R. Crowell and S.M. Sze, “Current Transport in Metal-Semiconductor

Barriers,” Solid State Electronics, vol. 9, pp. 1035–1048, 1966.

25. K. Yang, J. R. East, and G. I. Haddad, “Numerical Modeling of Abrupt

Heterojunctions using a Thermionic-Field Emission Boundary Condition,” Solid
State Electronics, vol. 36, no. 3, pp. 321–330, 1993.

26. U. Lindefelt, “Heat Generation in Semiconductor Devices,” J. Applied Physics,

vol. 75, no. 2, pp. 942–957, 1994.

27. J. Slotboom and H. de Graaff, “Measurements of Bandgap Narrowing in Si

Bipolar Transistors,” Solid State Electron. vol. 19, no. 10, pp. 857–862, 1976.

28. D. B. M. Klaassen, J. W. Slotboom, and H. C. de Graaff, “Unified apparent

bandgap narrowing in n- and p-type silicon,” Solid State Electron, vol. 35, no. 2,
pp. 125–129, 1992.

29. S.C. Jain and D.J. Roulston, “A simple expression for band gap narrowing (BGN)
in heavily doped Si, Ge, GaAs and GexSi1•x strained layers,” Solid-State Electronics,
vol. 34, no. 5, pp. 453–465, 1991.

30. M. Levinshtein, S. Rumyantsev, and M. Shur, Handbook Series on Semiconductor

Parameters, vol. 1, World Scientific, 1996.

31. L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields, Course of
Theoretical Physics (vol. 2), Butterworth-Heinmann, 1975.

32. S.L. Chuang, Physics of Photonic Devices, John Wiley and Sons Inc., 2009.

33. A. Yariv, Quantum Electronics, John Wiley and Sons, 1989.

34. A. Einstein, “Zur Quantentheorie der Strahlung,” Physikalische Zeitschrift,

vol. 18, pp. 121–128, 1917 (in German).

THEORY FOR THE SEMICONDUCTOR INTERFACE | 225

 35. A. Yariv, Optical Electronics in Modern Communication, Oxford University
Press, 1997.

36. E. O. Kane, “Band structure of indium antimonide,” J. Physics and Chemistry of

Solids, vol. 1, pp. 249–262, 1957.

37. R.H. Yan, S. W. Corzine, L.A. Coldren, and I. Suemune, “Corrections to the
expression for gain in GaAs,” IEEE Journal of Quantum Electronics, vol. 26, no 2,
pp. 213–216, 1990.

38. C.Hermann and C. Weisbuch, “k⋅p perturbation theory in III-IV compounds and
alloys: a reexamination,” Physical Review B, vol. 15 no. 2, pp. 823–833, 1977.

39. B. Y. K. Hu, “Kramers–Kronig in two lines,” American Journal of Physics,

vol. 57, no. 9, p. 821, 1989.

40. C.B. Duke, “Tunneling in Solids,” Solid State Physics, supplement 10, Academic
Press, 1969.

41. R. H. Fowler and L. Nordheim, “Electron Emission in Intense Electric Fields,”

Proceedings of the Royal Society of London A, vol. 119, pp. 173–181, 1928.

42. M. Lenzlinger and E. H. Snow, “Fowler-Nordheim Tunneling into Thermally

Grown SiO2,” J. Applied Physics, vol. 40, no. 1, pp. 278–283,1969.

43. A. Schenk, “Advanced Physical Models for Silicon Device Simulation,”

Computational Microelectronics, ed. S. Selberherr, Springer Verlag, New York, 1998.

44. M. A. Green and M. J. Keevers, “Optical Properties of Intrinsic Silicon at 300 K,”
Progress in Photovoltaics: Research and Applications, vol. 3, pp. 189–192, 1995.

45. M.G. Ancona, “Density-gradient theory: a macroscopic approach to quantum

confinement and tunneling in semiconductor devices,” J. Comput. Electron., vol. 10,
p. 65, 2011.

46. M.G. Ancona, Z. Yu, R. W. Dutton, P.J. Vande Voorde, M. Cao, and D. Vook,
“Density-Gradient Analysis of MOS Tunneling,” IEEE Transactions On Electron
Devices, vol. 47, no. 12, p. 2310, 2000.

47. M.G. Ancona, D. Yergeau , Z. Yu and B.A. Biegel, “On Ohmic Boundary
Conditions for Density-Gradient Theory,” J. Comput. Electron., vol. 1, pp. 103–107,
2002.

48. S. Jin, Y.J. Park, and H.S. Min, “Simulation of Quantum Effects in the Nano-scale
Semiconductor Device,” J. Semicond. Tech. Sci., vol. 4, no. 1, p. 32, 2004.

226 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

The Semiconductor Optoelectronics,
Beam Envelopes Interface
This physics interface requires a Wave Optics Module license. For theory
and physics interface feature descriptions relating to the Beam Envelopes
interface, see the Wave Optics Module User’s Guide.

The Semiconductor Optoelectronics, Beam Envelopes interface ( ) is used to model the

interaction of electromagnetic waves with semiconductors. This multiphysics interface
adds Electromagnetic Waves, Beam Envelopes and Semiconductor interfaces. By
default an Optical Transitions node is added to the Semiconductor interface. The
multiphysics couplings add the stimulated and/or spontaneous generation rates to the
semiconductor and account for the electromagnetic gain or loss due to absorption or
emission of light by the material.

Combinations of frequency-domain modeling for the Electromagnetic Waves, Beam

Envelopes interface and stationary modeling for the Semiconductor interface, called
frequency-stationary and, similarly, frequency-transient modeling, are supported in 2D
and 3D.

When a predefined Semiconductor Optoelectronics, Beam Envelopes interface is added

from the Semiconductor branch ( ) of the Model Wizard or Add Physics windows,
Semiconductor and Electromagnetic Waves, Beam Envelopes interfaces are added to the
Model Builder.

In addition, a Multiphysics Couplings node is added, which automatically includes the

multiphysics coupling feature Semiconductor-Electromagnetic Waves Coupling.

On the Constituent Physics Interfaces

The Semiconductor interface solves Poisson’s equation for the electric potential and
the drift-diffusion equations for electrons and holes in a semiconductor material.

The Electromagnetic Waves, Beam Envelopes interface computes electric and

magnetic field distributions for systems and devices where the field amplitude varies
slowly on a wavelength scale. The physics interface can be used efficiently for
unidirectional and bidirectional propagation of electromagnetic beams. In this physics
interface the electric field is factored into a product of a slowly varying envelope
function (slowly on the scale of a wavelength) and a rapidly varying phase function.

THE SEMICONDUCTOR OPTOELECTRONICS, BEAM ENVELOPES INTERFACE | 227

 The phase function is a priori prescribed, so the physics interface solves the
time-harmonic wave equation for the slowly varying envelope function.

COUPLING FEATURES
If a multiphysics interface is added from the Model Wizard or using the Add Physics
window, then the appropriate coupling feature is automatically added to the model
builder tree structure. Alternatively, the physics interfaces can be added one at a time,
followed by the coupling features.

For example, if single Semiconductor and Electromagnetic Waves, Beam Envelopes

interfaces are added, then COMSOL Multiphysics adds an empty Multiphysics
Couplings node. You can then choose the coupling feature:
Semiconductor-Electromagnetic Waves Coupling.

Coupling features are available from the context menu (right-click the
Multiphysics Couplings node) or from the Physics toolbar, Multiphysics
menu.

PHYSICS INTERFACES AND COUPLING FEATURES

Use the online help in COMSOL Multiphysics to locate and search all the
documentation. All these links also work directly in COMSOL
Multiphysics when using the Help system.

Coupling Features
The Semiconductor-Electromagnetic Waves Coupling node is described in the section
Semiconductor-Electromagnetic Waves Coupling below.

Physics Interface Features

Physics nodes are available from the Physics ribbon toolbar (Windows users), Physics
context menu (Mac or Linux users), or right-click to access the context menu (all
users).

In general, to add a node, go to the Physics toolbar, no matter what

operating system you are using. Subnodes are available by clicking the
parent node and selecting it from the Attributes menu.

228 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

The available physics features for The Semiconductor Interface are listed in the section
Domain, Edge, Boundary, Pair, and Point Nodes for the Semiconductor Interface.

The available physics features for The Electromagnetic Waves, Beam

Envelopes Interface are listed in the section Domain, Boundary, Edge,
and Point Nodes for the Electromagnetic Waves, Beam Envelopes
Interface. As this physics interface requires the Wave Optics Module these
links do not work in the PDF but they do work in the on line help.

Semiconductor-Electromagnetic Waves Coupling

The Semiconductor-Electromagnetic Waves Coupling ( ) node passes the optical field
intensity from an electromagnetic waves interface to the Optical Transitions feature in
the Semiconductor interface. The change in the complex refractive index or permittivity
that results from the absorption in the semiconductor is then passed back to the
electromagnetic waves interface.

SETTINGS
The Label is the default multiphysics coupling feature name.

The Name is used primarily as a scope prefix for variables defined by the coupling node.
The software refer to variables in expressions using the pattern
<name>.<variable_name>. In order to distinguish between variables belonging to
different coupling nodes or physics interfaces, the name string must be unique. Only
letters, numbers, and underscores (_) are permitted in the Name field. The first
character must be a letter.

The default Name (for the first multiphysics coupling feature in the model) is semc1.

DOMAIN SELECTION
The default setting is All domains which couples all domains in the Semiconductor
interface that have an active Optical Transitions feature with the corresponding domains
in the electromagnetic waves interface.

COUPLED INTERFACES
This section defines the physics involved in the coupling. By default, the software
selects an appropriate physics interface for you from the Semiconductor and
Electromagnetic waves lists. These selections determine the two interfaces that are
coupled by the feature.

THE SEMICONDUCTOR OPTOELECTRONICS, BEAM ENVELOPES INTERFACE | 229

 The Semiconductor Optoelectronics,
Frequency Domain Interface
This physics interface requires a Wave Optics Module license. For theory
and physics interface feature descriptions relating to the Electromagnetic
Waves, Frequency Domain interface, see the Wave Optics Module User’s
Guide.

The Semiconductor Optoelectronics, Frequency Domain interface ( ) is used to model

the interaction of electromagnetic waves with semiconductors. This multiphysics
interface adds Electromagnetic Waves, Frequency Domain and Semiconductor
interfaces. By default an Optical Transitions node is added to the Semiconductor
interface. The multiphysics couplings add the stimulated and/or spontaneous
generation rates to the semiconductor and account for the electromagnetic gain or
loss, due to absorption or emission of light by the material.

Combinations of frequency-domain modeling for the Electromagnetic Waves,

Frequency Domain interface and stationary modeling for the Semiconductor interface,
called frequency-stationary and, similarly, frequency-transient modeling, are supported
in 2D and 3D.

When a predefined Semiconductor Optoelectronics, Frequency Domain interface is added

from the Semiconductor branch ( ) of the Model Wizard or Add Physics windows,
Semiconductor and Electromagnetic Waves, Frequency Domain interfaces are added to the
Model Builder.

In addition, a Multiphysics Couplings node is added, which automatically includes the

multiphysics coupling feature Semiconductor-Electromagnetic Waves Coupling.

On the Constituent Physics Interfaces

The Semiconductor interface solves Poisson’s equation for the electric potential and
the drift-diffusion equations for electrons and holes in a semiconductor material.

The Electromagnetic Waves, Frequency Domain interface is used to solve for

time-harmonic electromagnetic field distributions. For this physics interface, the
maximum mesh element size should be limited to a fraction of the wavelength. The
domain size that can be simulated thus scales with the amount of available computer

230 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

memory and the wavelength. This physics interface solves the time-harmonic wave
equation for the electric field.

COUPLING FEATURES
If a multiphysics interface is added from the Model Wizard or using the Add Physics
window, then the appropriate coupling feature is automatically added to the model
builder tree structure. Alternatively, the physics interfaces can be added one at a time,
followed by the coupling features.

For example, if single Semiconductor and Electromagnetic Waves, Frequency Domain

interfaces are added, then COMSOL Multiphysics adds an empty Multiphysics
Couplings node. You can then choose the coupling feature:
Semiconductor-Electromagnetic Waves Coupling.

Coupling features are available from the context menu (right-click the
Multiphysics Couplings node) or from the Physics toolbar, Multiphysics
menu.

PHYSICS INTERFACES AND COUPLING FEATURES

Use the online help in COMSOL Multiphysics to locate and search all the
documentation. All these links also work directly in COMSOL
Multiphysics when using the Help system.

Coupling Features
The Semiconductor-Electromagnetic Waves Coupling node is described in The
Semiconductor Optoelectronics, Beam Envelopes Interface section.

Physics Interface Features

Physics nodes are available from the Physics ribbon toolbar (Windows users), Physics
context menu (Mac or Linux users), or right-click to access the context menu (all
users).

In general, to add a node, go to the Physics toolbar, no matter what

operating system you are using. Subnodes are available by clicking the
parent node and selecting it from the Attributes menu.

THE SEMICONDUCTOR OPTOELECTRONICS, FREQUENCY DOMAIN INTERFACE | 231

 The available physics features for The Semiconductor Interface are listed in the section
Domain, Edge, Boundary, Pair, and Point Nodes for the Semiconductor Interface.

The available physics features for The Electromagnetic Waves, Frequency

Domain Interface are listed in the section Domain, Boundary, Edge,
Point, and Pair Nodes for the Electromagnetic Waves, Frequency
Domain Interface. As this physics interface requires the Wave Optics
Module these links do not work in the PDF but they do work in the on
line help.

232 | CHAPTER 3: SEMICONDUCTOR BRANCH INTERFACE

 4

The AC/DC Branch

This chapter describes the theory for the physics interfaces found under the
AC/DC branch ( ).

In this chapter:

• The Electrostatics Interface

• The Electrical Circuit Interface
• Theory for the Electrostatics Interface
• Theory for the Electrical Circuit Interface

See The Electromagnetics Interfaces in the COMSOL Multiphysics Reference

Manual for other AC/DC interface and feature node settings.

233
 The Electrostatics Interface
The Electrostatics (es) interface ( ), found under the AC/DC>Electric Fields and
Currents branch when adding a physics interface, is used to compute the electric field,
electric displacement field, and potential distributions in dielectrics under conditions
where the electric charge distribution is explicitly prescribed. The formulation is
stationary except for use together with other physics interfaces. Eigenfrequency,
frequency-domain, small-signal analysis, and time-domain modeling are supported in
all space dimensions.

The physics interface solves Gauss’ Law for the electric field using the scalar electric
potential as the dependent variable.

Charge Conservation is the main node, which adds the equation for the electric
potential and has a Settings window for defining the constitutive relation for the
electric displacement field and its associated properties such as the relative permittivity.

When this physics interface is added, these default nodes are also added to the Model
Builder — Charge Conservation, Zero Charge (the default boundary condition), and
Initial Values. Then, from the Physics toolbar, add other nodes that implement, for
example, boundary conditions and space charges. You can also right-click Electrostatics
to select physics features from the context menu.

Physics-Controlled Mesh
The physics-controlled mesh is controlled from the Mesh node’s Settings window (if the
Sequence type is Physics-controlled mesh). There, in the table in the Physics-Controlled
Mesh section, find the physics interface in the Contributor column and select or clear
the check box in the Use column on the same table row for enabling (the default) or
disabling contributions from the physics interface to the physics-controlled mesh.

Information from the physics, such as the presence of an infinite elements domain or
periodic condition, will be used to automatically set up an appropriate mesh sequence.

In the COMSOL Multiphysics Reference Manual see the

Physics-Controlled Mesh section for more information about how to
define the physics-controlled mesh.

SETTINGS
The Label is the default physics interface name.

234 | CHAPTER 4: THE AC/DC BRANCH

The Name is used primarily as a scope prefix for variables defined by the physics
interface. Refer to such physics interface variables in expressions using the pattern
<name>.<variable_name>. In order to distinguish between variables belonging to
different physics interfaces, the name string must be unique. Only letters, numbers, and
underscores (_) are permitted in the Name field. The first character must be a letter.

The default Name (for the first physics interface in the model) is es.

CROSS-SECTION AREA (1D COMPONENTS)

For 1D components, enter a default value for the Cross-section area A (SI unit: m2).
The default value of 1 is typically not representative for a thin domain. Instead it
describes a unit thickness that makes the 1D equation identical to the equation used
for 3D components. See also Change Cross Section.

THICKNESS (2D COMPONENTS)

For 2D components, enter a default value for the Out-of-plane thickness d (SI unit: m).
The default value of 1 is typically not representative for a thin dielectric medium, for
example. Instead it describes a unit thickness that makes the 2D equation identical to
the equation used for 3D components. See also Change Thickness (Out-of-Plane).

DEPENDENT VARIABLES
The dependent variable is the Electric potential V. You can change its name, which
changes both the field name and the variable name. If the new name coincides with the
name of another electric potential field in the model, the physics interfaces shares
degrees of freedom. The new name must not coincide with the name of a field of
another type or with a component name belonging to some other field.

DISCRETIZATION

In the COMSOL Multiphysics Reference Manual, see Table 2-4 for links
to common sections and Table 2-5 to common feature nodes. You can
also search for information: press F1 to open the Help window or Ctrl+F1
to open the Documentation window.

Electric Sensor: Application Library path

COMSOL_Multiphysics/Electromagnetics/electric_sensor

THE ELECTROSTATICS INTERFACE | 235

 Domain, Boundary, Edge, Point, and Pair Nodes for the
Electrostatics Interface
The Electrostatics interface has these domain, boundary, edge, point, and pair nodes
available.

ABOUT THE BOUNDARY CONDITIONS

The relevant physics interface condition at interfaces between different media is

n2 ⋅ ( D1 – D2 ) = ρs

In the absence of surface charges, this condition is fulfilled by the natural boundary
condition

n ⋅ [ ( ε 0 ∇V – P ) 1 – ( ε 0 ∇V – P ) 2 ] = – n ⋅ ( D 1 – D 2 ) = 0

AVAILABLE NODES
These nodes, listed in alphabetical order, are available from the Physics ribbon toolbar
(Windows users), Physics context menu (Mac or Linux users), or right-click to access
the context menu (all users). Also see Table 4-1 for a list of interior and exterior
boundary conditions, including edge, point, and pair availability.

In general, to add a node, go to the Physics toolbar no matter what

operating system you are using. Subnodes are available by clicking the
parent node and selecting it from the Attributes menu.

236 | CHAPTER 4: THE AC/DC BRANCH

 • Change Cross Section • Initial Values
• Change Thickness (Out-of-Plane) • Line Charge
• Charge Conservation • Line Charge (on Axis)
1
• Charge Conservation, Piezoelectric • Line Charge (Out-of-Plane)
• Dielectric Shielding • Periodic Condition
• Distributed Capacitance • Point Charge
• Electric Displacement Field • Point Charge (on Axis)
• Electric Potential • Space Charge Density
• Electrostatic Point Dipole • Surface Charge Density
• External Surface Charge • Terminal
Accumulation • Thin Low Permittivity Gap
• Floating Potential • Zero Charge (the default boundary
• Ground condition)
1 This feature is available with the Piezoelectricity interface, which requires the

Acoustics Module, MEMS Module, or the Structural Mechanics Module.

Table 4-1 lists the interior and exterior boundary conditions available with this physics
interface. It also includes edge, point, and pair availability.
TABLE 4-1: INTERIOR AND EXTERIOR BOUNDARY CONDITIONS (INCLUDING EDGE, POINT, AND PAIR
AVAILABILITY) FOR THE ELECTROSTATICS INTERFACE

NODE INTERIOR EXTERIOR ALSO AVAILABLE FOR

Change Cross Section x x pairs

Change Thickness (Out-of-Plane) x x pairs
Dielectric Shielding x x pairs
Distributed Capacitance x x pairs
Electric Displacement Field x x pairs
Electric Potential x x edges, points, and pairs
External Surface Charge x pairs
Accumulation
Floating Potential x x pairs
Ground x x edges, points, and pairs
Periodic Condition x not applicable
Surface Charge Density x x pairs

THE ELECTROSTATICS INTERFACE | 237

 TABLE 4-1: INTERIOR AND EXTERIOR BOUNDARY CONDITIONS (INCLUDING EDGE, POINT, AND PAIR
AVAILABILITY) FOR THE ELECTROSTATICS INTERFACE

NODE INTERIOR EXTERIOR ALSO AVAILABLE FOR

Terminal x x domains
Thin Low Permittivity Gap x not applicable
Zero Charge (the default) x x pairs

For axisymmetric components, COMSOL Multiphysics takes the axial

symmetry boundaries (at r = 0) into account and automatically adds an
Axial Symmetry node to the model that is valid on the axial symmetry
boundaries only. There are also Line Charge (on Axis) and Point Charge (on
Axis) available.

In the COMSOL Multiphysics Reference Manual, see Table 2-4 for links
to common sections and Table 2-5 to common feature nodes. You can
also search for information: press F1 to open the Help window or Ctrl+F1
to open the Documentation window.

Charge Conservation
The Charge Conservation node adds the equations for charge conservation according to
Gauss’ law for the electric displacement field. It provides an interface for defining the
constitutive relation and its associated properties such as the relative permittivity.

MATERIAL TYPE
The Material type setting decides how materials behave and how material properties are
interpreted when the mesh is deformed. Select Solid for materials whose properties
change as functions of material strain, material orientation, and other variables
evaluated in a material reference configuration (material frame). Select Nonsolid for
materials whose properties are defined only as functions of the current local state at
each point in the spatial frame, and for which no unique material reference
configuration can be defined. Select From material to pick up the corresponding setting
from the domain material on each domain.

238 | CHAPTER 4: THE AC/DC BRANCH

CONSTITUTIVE RELATION D-E
Select a Dielectric model to describe the macroscopic properties of the medium
(relating the electric displacement D with the electric field E) and the applicable
material properties, such as the relative permittivity. Select:

• Relative permittivity (the default) to use the constitutive relation D = ε0εrE. Then
the default is to take the Relative permittivity εr (dimensionless) values From material.
For User defined, select Isotropic, Diagonal, Symmetric, or Full and enter values or
expressions in the field or matrix. The default is 1.
• Polarization to use the constitutive relation D = ε0E + P. Then enter the components
based on space dimension for the Polarization vector P (SI unit: C/m2). The
defaults are 0 C/m2.
• Remanent electric displacement to use constitutive relation D = ε0εrE + Dr, where Dr
is the remanent displacement (the displacement when no electric field is present).
Then the default is to take the Relative permittivity εr (dimensionless) values From
material. For User defined, select Isotropic, Diagonal, Symmetric, or Full and enter
values or expressions in the field or matrix. Then enter the components based on
space dimension for the Remanent electric displacement Dr (SI unit: C/m2). The
defaults are 0 C/m2.

Conduction Loss (Time-Harmonic)

This feature requires the Acoustics Module, MEMS Module, or the

Structural Mechanics Module. See the individual documentation for
information.

The Conduction Loss (Time-Harmonic) subnode allows you to model possible conductive
losses in a dielectric material. The effect is only active in a Eigenfrequency or Frequency
response study.

THE ELECTROSTATICS INTERFACE | 239

 CONDUCTION CURRENT
By default, the Electrical conductivity σ for the media is defined From material. You can
also select User defined or Linearized resistivity.

• For User defined select Isotropic, Diagonal, Symmetric, or Full depending on the
characteristics of the electrical conductivity, and then enter values or expressions for
the Electrical conductivity σ in the field or matrix.
• For Linearized resistivity the default Reference temperature Tref, and Resistivity
temperature coefficient α, and Reference resistivity ρ0 are taken From material, which
means that the values are taken from the domain (or boundary) material. T is the
current temperature, which can be a value that is specified as a model input or the
temperature from a heat transfer interface. The definition of the temperature field
appears in the Model Inputs section.

LOCATION IN USER INTERFACE

Context Menus
Electrostatics>Charge Conservation>Conduction Loss (Time-Harmonic)

Ribbon
Physics tab with Charge Conservation node selected in the model tree:

Attributes>Conduction Loss (Time-Harmonic)

Initial Values
The Initial Values node adds an initial value for the electric potential V that can serve
as an initial condition for a transient simulation or as an initial guess for a nonlinear
solver.

INITIAL VALUES
Enter a value or expression for the initial value of the Electric potential V (SI unit: V).
The default value is 0 V.

Space Charge Density

The Space Charge Density node adds a space charge density ρ, which appears on the
right-hand side of the equation that the physics interface defines.

240 | CHAPTER 4: THE AC/DC BRANCH

SPACE CHARGE DENSITY
Enter a value or expression for the Space charge density ρv (SI unit: C/m3). The default
is 0 C/m3.

Zero Charge
The Zero Charge node adds the condition that there is zero charge on the boundary so
that n ⋅ D = 0. This boundary condition is also applicable at symmetry boundaries
where the potential is known to be symmetric with respect to the boundary. This is the
default boundary condition at exterior boundaries. At interior boundaries, it means
that no displacement field can penetrate the boundary and that the electric potential is
discontinuous across the boundary.

Ground
The Ground node implements ground (zero potential) as the boundary condition
V = 0.

Ground means that there is a zero potential on the boundary. This boundary condition
is also applicable at symmetry boundaries where the potential is known to be
antisymmetric with respect to the boundary.

For some physics interfaces, also select additional Ground nodes from the Edges (3D
components) or Points (2D and 3D components) submenus. For 2D axisymmetric
components, it can be applied on the Symmetry axis.

BOUNDARY, EDGE, OR POINT SELECTION

Beware that constraining the potential on edges or points in 3D or on

points in 2D usually yields a current outflow that is mesh dependent.

GROUND
Enter a value for the Contact work function Φc(SI unit: V). The default is 4 V.

CONSTRAINT SETTINGS
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options in the Show More Options dialog box.

THE ELECTROSTATICS INTERFACE | 241

 Electric Potential
The Electric Potential node provides an electric potential V0 as the boundary condition
V = V0.

Because the electric potential is being solved for in the physics interface, the value of
the potential is typically defined at some part of the geometry. For some physics
interfaces, also select additional Electric Potential nodes from the Edges (3D
components) or Points (2D and 3D components) submenus. For 2D axisymmetric
components, it can be applied on the symmetry axis.

BOUNDARY, EDGE, OR POINT SELECTION

Beware that constraining the potential on edges or points in 3D or on

points in 2D usually yields a current outflow that is mesh dependent.

ELECTRIC POTENTIAL
Enter the value or expression for the Electric potential V0 (SI unit: V). The default is
0 V.

Enter a value for the Contact work function Φc(SI unit: V). The default is 4 V.

CONSTRAINT SETTINGS
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options in the Show More Options dialog box.

Surface Charge Density

The Surface Charge Density node provides the following surface-charge boundary
condition for exterior boundaries (left) and interior boundaries (right):

–n ⋅ D = ρs , n ⋅ ( D1 – D2 ) = ρs

Specify the surface charge density ρs at an outer boundary or at an interior boundary

between two nonconducting media.

Add a contribution as a Harmonic Perturbation by right-clicking the parent node or

click Harmonic Perturbation on the Physics toolbar. For more information, see
Harmonic Perturbation — Exclusive and Contributing Nodes in the COMSOL
Multiphysics Reference Manual.

242 | CHAPTER 4: THE AC/DC BRANCH

SURFACE CHARGE DENSITY
Enter the value or expression for the Surface charge density ρs (SI unit: C/m2).

External Surface Charge Accumulation

The External Surface Charge Accumulation node implements the boundary condition

–n ⋅ D = ρs

where ρs is the solution of the following distributed ODE on the boundary:

dρ s
= n ⋅ Ji + n ⋅ Je
dt

where n·Ji is the normal component of the total ion current density on the wall and
n·Je is the normal component of the total electron current density on the wall, which
are feature inputs.

MATERIAL TYPE
The Material type setting decides how materials behave and how material properties are
interpreted when the mesh is deformed. Select Solid for materials whose properties
change as functions of material strain, material orientation, and other variables
evaluated in a material reference configuration (material frame). Select Nonsolid for
materials whose properties are defined only as functions of the current local state at
each point in the spatial frame, and for which no unique material reference
configuration can be defined. Select From material to pick up the corresponding setting
from the domain material on each domain.

EXTERNAL SURFACE CHARGE ACCUMULATION

Enter values or expressions for the Normal ion current density n·Ji (SI unit: A/m2)
and the Normal electron current density n·Je (SI unit: A/m2).

Electric Displacement Field

The Electric Displacement Field node adds the following electric-displacement
boundary condition:

n ⋅ D = n ⋅ D0

It specifies the normal component of the electric displacement field at a boundary.

THE ELECTROSTATICS INTERFACE | 243

 ELECTRIC DISPLACEMENT FIELD
Enter the coordinates of the Boundary electric displacement field D0 (SI unit: C/m2).

Periodic Condition
The Periodic Condition node defines periodicity or antiperiodicity between two
boundaries. If required, activate periodic conditions on more than two boundaries, in
which case the Periodic Condition tries to identify two separate surfaces that can each
consist of several connected boundaries. For more complex geometries, it might be
necessary to use the Destination Selection subnode. With this subnode the boundaries
which constitute the source and destination surfaces can be manually specified. The
Destination Selection subnode is available from the context menu (right-click the parent
node) as well as from the Physics toolbar, Attributes menu.

When this feature is used in conjunction with a Sector Symmetry feature

on connected boundaries, wherever the sector symmetry boundaries
connect with the periodic boundaries, the same periodic condition feature
cannot be used on both sides. At least two periodic condition features are
required for the model to compute correctly.

BOUNDARY SELECTION
When using nonconforming meshes on the source and destination of a periodic
boundary pair, for numerical stability, a finer mesh should be applied on the
destination side. Use conforming meshes if possible.

PERIODIC CONDITION
Select a Type of periodicity — Continuity (the default), Antiperiodicity, or Floquet
periodicity. Select:

• Continuity to make the electric potential periodic (equal on the source and
destination).
• Antiperiodicity to make it antiperiodic.
• Floquet periodicity (only available with products supporting piezoelectric modeling).
Specify the components of the k-vector for Floquet periodicity kF (SI unit: rad/m).

CONSTRAINT SETTINGS
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options in the Show More Options dialog box.

244 | CHAPTER 4: THE AC/DC BRANCH

For information about the Orientation of Source section, see Orientation of Source
and Destination in the COMSOL Multiphysics Reference Manual.

Thin Low Permittivity Gap

Use the Thin Low Permittivity Gap node

ε0 εr
n ⋅ D 1 = ---------- ( V 1 – V 2 )
d
ε0 εr
n ⋅ D 2 = ---------- ( V 2 – V 1 )
d

to model a thin gap of a material with a small permittivity compared to the adjacent
domains. The layer has the thickness d and the relative permittivity εr. The indices 1
and 2 refer to the two sides of the boundary.

THIN LOW PERMITTIVITY GAP

The default is to take the Relative permittivity εr (dimensionless) values From material.
For User defined, enter a different value or expression. Enter a Thickness d (SI unit: m).
The default is 5 mm.

Dielectric Shielding
The Dielectric Shielding node adds dielectric shielding as a boundary condition. It
describes a thin layer with thickness ds and a bulk relative permittivity εr that shields
the electric field:

n ⋅ D = – ∇ t ⋅ ε 0 ε rs d s ∇tV

Use this boundary condition when approximating a thin domain with a boundary to
reduce the number of mesh elements.

MATERIAL TYPE
The Material type setting decides how materials behave and how material properties are
interpreted when the mesh is deformed. Select Solid for materials whose properties
change as functions of material strain, material orientation and other variables
evaluated in a material reference configuration (material frame). Select Nonsolid for
materials whose properties are defined only as functions of the current local state at
each point in the spatial frame, and for which no unique material reference

THE ELECTROSTATICS INTERFACE | 245

 configuration can be defined. Select From material to pick up the corresponding setting
from the domain material on each domain.

ELECTRIC SHIELDING
The default is to take the Relative permittivity εr (dimensionless) values From material.
It takes it from the adjacent domains if not explicitly defined. For User defined select
Isotropic, Diagonal, Symmetric, or Full and enter values or expressions in the field or
matrix. Enter a Surface thickness ds (SI unit: m) of the shielding. The default is 1 m.

Terminal
The Terminal node provides a boundary or domain condition for connection to
external circuits, to transmission lines, or with a specified voltage or charge. By
specifying zero charge, a floating potential condition is obtained.

The Terminal node also enables the computation of the lumped parameters of the
system, such as capacitance

TERMINAL
Specify the terminal properties. Enter a unique Terminal name that will be used to
identify the global variables (such as the terminal voltage or current) that are defined
by the terminal. The Terminal name can contain alphanumeric characters, but it should
be numeric when performing terminal sweeps.

Enter a value for the Contact work function Φc(SI unit: V). The default is 4 V.

Select a Terminal type — Charge (the default), Voltage, Circuit, or Terminated. Select:

• Charge to prescribe the total charge deposited on the terminal boundaries. Enter the
total charge Q0. The default is 0 C, which corresponds to a uncharged floating
electrode.
• Voltage to specify a fixed electric potential at the terminal. Enter an electric potential
V0. The default is 1 V.

246 | CHAPTER 4: THE AC/DC BRANCH

• Circuit to specify a terminal connected to an Electrical Circuit physics interface that
models an external circuit. The Terminal node provides a current-voltage
characteristic to the circuit.

The Electrical Circuit interface requires a current from the Terminal, so

the Circuit terminal type can only be used in Time Dependent or
Frequency Domain studies, in which the current can be computed as the
time derivative of the charge.

The Circuit type cannot be used together with a terminal sweep.

• Terminated to connect the terminal to an impedance that might represent a load or

a transmission line. When Terminated is selected, the scattering parameters
(S-parameters) are computed. The Terminated setting can only be used in
frequency-domain studies.
- Enter a Terminal power P0 to specify the input power at this terminal from the
transmission line. This excitation can be toggled using a port sweep.
- Select an option from the Characteristic impedance from list to define the value of
the impedance — Physics interface or User defined. For Physics interface the
Reference impedance Zref defined on the Settings window for the physics under
Sweep Settings is used. For User defined enter a specific impedance Zref for this
terminal. The default is 50 Ω.

Initial values
The initial values subsection is used to provide initial values for the internal state
variables in the Terminal. These settings can be useful to provide an initial guess to the
nonlinear solver or the initial value in a time-dependent study.

The Initial value for voltage is available for all Terminal types except Voltage. Enter an
Initial value for voltage Vinit.

The Initial value for charge is available for the Circuit excitation. It specifies the initial
charge on the terminal when the circuit is connected. Enter an Initial value for charge
Qinit.

THE ELECTROSTATICS INTERFACE | 247

 CONSTRAINT SETTINGS
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options in the Show More Options dialog box.

Studies and Solvers in the COMSOL Multiphysics Reference Manual

Floating Potential
The Floating Potential node is used when modeling a metallic electrode at floating
potential. It applies a constant voltage V0 on the boundary (for domain features, this
is the boundary enclosing the selected domain), such that the total normal electric
displacement field D equals a specific charge Q0:

V = V0

 ( D ⋅ n ) dS = Q0 ,
∂Ω

where Ω represents the boundary and n refers to the surface normal. The constant
boundary voltage implies the tangential electric field equals zero; the electric field will
be perpendicular to the boundary:

–∇t V = Et = 0
n × E = 0.

In case of Q0 = 0 (the default case), the boundary will behave as an unconnected,

neutrally charged, good conductor under electrostatic conditions (a floating
equipotential). Although locally the displacement field may vary, the total electric
displacement field entering or leaving the boundary equals zero.

If the floating potential touches a point, boundary or domain feature that is not
floating (a Terminal or Ground feature), the floating potential will acquire that feature's
potential. If the floating potential is set to a certain charge, or connected to a circuit,
it behaves like a terminal.

FLOATING POTENTIAL
The Floating potential identifier text area shows the unique identifier for the floating
potential feature. It is used to identify the global variables created by the node, such as
the voltage. The Floating potential group check box on the Settings window for Floating

248 | CHAPTER 4: THE AC/DC BRANCH

Potential controls how potentials are assigned to boundary segments in the feature’s
boundary selection. If this check box is not selected, a single potential is applied to all
boundaries in the selection. If the check box is selected, each group of contiguous
boundaries in the selection is given a unique potential. This simplifies the setup of
models with many floating electrodes. The values of the potential at each group of
boundaries are then made available in postprocessing, collected in a vector variable.

The following options are not available if the Floating potential group check box is
selected. Select a specification for the Electric charge — choose User defined to specify
a total Charge Q0 deposited on the surface. Select Circuit to connect the floating
potential to an Electrical Circuit. Enter a value for the Contact work function Φc (SI
unit: V). The default is 4 V.

The Electrical Circuit interface requires a current from the Floating

Potential, so the Circuit terminal type can only be used in Time
Dependent or Frequency Domain studies, in which the current can be
computed as the time derivative of the charge.

Initial Values
The initial values subsection is used to provide initial values for the internal state
variables in the Floating potential. These settings can be useful to provide an initial
guess to the nonlinear solver or the initial value in a time-dependent study.

Enter an Initial value for voltage Vinit.

The Initial value for charge is available for the Circuit excitation. It specifies the initial
charge on the floating potential when the circuit is connected. Enter an Initial value for
charge Qinit.

CONSTRAINT SETTINGS
To display this section, click the Show More Options button ( ) and select Advanced
Physics Options in the Show More Options dialog box.

Settings and Properties Windows for Feature Nodes in the COMSOL

Multiphysics Reference Manual

THE ELECTROSTATICS INTERFACE | 249

 Distributed Capacitance
The Distributed Capacitance node adds a distributed capacitance boundary condition
according to the following equations for exterior boundaries (left) and interior
boundaries (right):

V ref – V V ref – V
– n ⋅ D = ε 0 ε r -------------------- n ⋅ ( D 1 – D 2 ) = ε 0 ε r --------------------
ds ds

Use this boundary condition to model a thin sheet or film of a dielectric material. The
sheet has the relative permittivity εr and the surface thickness ds, and it is connected
to the reference potential Vref.

DISTRIBUTED CAPACITANCE
Enter the values or expressions for Relative permittivity εr (dimensionless), Surface
thickness ds (SI unit: m), and Reference potential Vref. The default value for the surface
thickness is 10−3 m (1 mm) and 0 V for the reference potential.

Enter a value for the Contact work function Φc(SI unit: V). The default is 4 V.

Line Charge
For 3D components, use the Line Charge node to specify line charges along the edges
of a geometry. Add a contribution as a Harmonic Perturbation by right-clicking the
parent node or clicking Harmonic Perturbation on the Physics toolbar. For more
information see Harmonic Perturbation — Exclusive and Contributing Nodes in the
COMSOL Multiphysics Reference Manual.

EDGE SELECTION

Beware that constraining the potential on edges usually yields a current

outflow that is mesh dependent.

250 | CHAPTER 4: THE AC/DC BRANCH

LINE CHARGE
Enter a value or expression to apply a Line charge QL (SI unit: C/m). This source
represents electric charge per unit length and the default is 0 C/m.

Line Charge (on Axis) and Line Charge (Out-of-Plane)

Line Charge (on Axis)

For 2D axisymmetric components, use the Line Charge (on Axis) node to specify line
charges along the symmetry axis.

Add a contribution as a Harmonic Perturbation by right-clicking the parent node or

click Harmonic Perturbation on the Physics toolbar. For more information see
Harmonic Perturbation — Exclusive and Contributing Nodes in the COMSOL
Multiphysics Reference Manual.

LINE CHARGE (ON AXIS)

Enter a value or expression to apply a Line charge QL (SI unit: C/m). This source
represents electric charge per unit length and the default is 0 C/m.

Line Charge and Line Charge (Out-of-Plane)

Line Charge (Out-of-Plane)

For 2D and 2D axisymmetric components, points are selected and this is the same as
a line out-of-plane.

Add a contribution as a Harmonic Perturbation by right-clicking the parent node or

clicking Harmonic Perturbation on the Physics toolbar. For more information see
Harmonic Perturbation — Exclusive and Contributing Nodes in the COMSOL
Multiphysics Reference Manual.

Use the Line Charge (Out-of-Plane) node to specify line charges along the points of a
geometry for 2D and 2D axisymmetric components.

THE ELECTROSTATICS INTERFACE | 251

 POINT SELECTION

Beware that constraining the potential on points usually yields a current

outflow that is mesh dependent.

LINE CHARGE (OUT-OF-PLANE)

Enter a value or expression to apply a Line charge QL (SI unit: C/m). This source
represents electric charge per unit length and the default is 0 C/m.

Line Charge and Line Charge (on Axis)

Point Charge
The Point Charge node adds a point source to 3D components. The point charge
represents an electric displacement field flowing out of the point.

Add a contribution as a Harmonic Perturbation by right-clicking the parent node or

clicking Harmonic Perturbation on the Physics toolbar. For more information see
Harmonic Perturbation — Exclusive and Contributing Nodes in the COMSOL
Multiphysics Reference Manual.

POINT SELECTION

Beware that constraining the potential on points usually yields a current

outflow that is mesh dependent.

POINT CHARGE
Enter a value or expression to apply a Point charge QP (SI unit: C) to points. This
source represents an electric displacement field flowing out of the point. The default is
0 C.

Point Charge (on Axis) and Line Charge (Out-of-Plane)

252 | CHAPTER 4: THE AC/DC BRANCH

Point Charge (on Axis)
The Point Charge (on Axis) node adds a point source to 2D axisymmetric components.
The point charge represents an electric displacement field flowing out of the point.

Add a contribution as a Harmonic Perturbation by right-clicking the parent node or

clicking Harmonic Perturbation on the Physics toolbar. For more information see
Harmonic Perturbation — Exclusive and Contributing Nodes in the COMSOL
Multiphysics Reference Manual.

POINT SELECTION

Beware that constraining the potential on points usually yields a current

outflow that is mesh dependent.

POINT CHARGE (ON AXIS)

Enter a value or expression to apply a Point charge QP (SI unit: C) to points on an axis.
This source represents an electric displacement field flowing out of the point. The
default is 0 C.

Point Charge and Line Charge (Out-of-Plane)

Change Cross Section

This node is available with 1D components. This setting overrides the global
Cross-Section Area setting made in any physics interface that uses this feature. For 2D
components, see Change Thickness (Out-of-Plane).

Use the Change Cross Section node to set the cross-section area for specific geometric
entities.

CHANGE CROSS SECTION

Enter a value or expression for the Cross-section area A. The default value of 1 unit
length is typically not representative for a thin domain. Instead it describes a unit
thickness that makes the 1D equation identical to the equation used for 3D
components.

THE ELECTROSTATICS INTERFACE | 253

 Change Thickness (Out-of-Plane)
This node is available for 2D components. This setting overrides the global Thickness
setting made in any physics interface that uses this node. For 1D components, see
Change Cross Section.

Use the Change Thickness (Out-of-Plane) node to set the out-of-plane thickness for
specific geometric entities.

CHANGE THICKNESS (OUT-OF-PLANE)

Enter a value or expression for the Out-of-plane thickness d (SI unit: m). The default
value is, in most cases, 1 unit length, which is typically not representative for a thin
domain. Instead it describes a unit thickness that makes the 2D equation identical to
the equation used for 3D components.

Charge Conservation, Piezoelectric

This feature is available with the Piezoelectricity interface, which requires

the Acoustics Module, MEMS Module, or the Structural Mechanics
Module. See the individual documentation for information.

The Charge Conservation, Piezoelectric node is normally used together with a

Piezoelectric Effect multiphysics coupling node and a corresponding Piezoelectric
Material node in the Solid Mechanics interface. The node is added by default to the
Electrostatics interface when adding a Piezoelectricity interface. It is also available from
the context menu (right-click the Electrostatics interface parent node) or from the
Physics toolbar.

When the Charge Conservation, Piezoelectric node is added to the

electrostatics interface in the absence of an active Piezoelectric Effect
multiphysics coupling node, the material behaves similarly to a Charge
Conservation node, with electric properties corresponding to the relative
permittivity entered (see below). The piezoelectric effect is not included
in the corresponding equation system.

ELECTRIC DISPLACEMENT
If the node is used together with an active Piezoelectric Effect multiphysics coupling
node, then these settings are locked. Note that if they are unlocked, then the material

254 | CHAPTER 4: THE AC/DC BRANCH

behaves like a dielectric and not a piezoelectric. In this case, the default is to take the
Relative permittivity εrS (dimensionless) values From material. For User defined, select
Isotropic, Diagonal, Symmetric, or Full and enter values or expressions in the field or
matrix.

Electrostatic Point Dipole

Add an Electrostatic Point Dipole node to 3D and 2D components. Add a contribution
as a Harmonic Perturbation by right-clicking the parent node or click Harmonic
Perturbation on the Physics toolbar. For more information see Harmonic Perturbation
— Exclusive and Contributing Nodes in the COMSOL Multiphysics Reference
Manual.

The Electrostatic Point Dipole represents the limiting case of zero separation distance
between two equally strong point sources of opposing signs while maintaining the
product between separation distance and source strength at a fixed value (p). The
dipole moment is a vector entity with positive direction from the negative charge to
the positive one.

ELECTROSTATIC POINT DIPOLE

Select a Dipole specification — Magnitude and direction (the default) or Electric dipole
moment.

• For Magnitude and direction enter coordinates for the Electric dipole moment
direction np (dimensionless) and the Electric dipole moment, magnitude p.
• For Electric dipole moment enter coordinates for the Electric dipole moment p.

THE ELECTROSTATICS INTERFACE | 255

 The Electrical Circuit Interface
The Electrical Circuit (cir) interface ( ), found under the AC/DC branch ( ) when
adding a physics interface, is used to model currents and voltages in circuits including
voltage and current sources, resistors, capacitors, inductors, and semiconductor
devices. Models created with the Electrical Circuit interface can include connections
to distributed field models. The physics interface supports stationary,
frequency-domain and time-domain modeling and solves Kirchhoff’s conservation
laws for the voltages, currents and charges associated with the circuit elements.

When this physics interface is added, it adds a default Ground Node feature and
associates that with node zero in the electrical circuit.

Circuit nodes are nodes in the electrical circuit (electrical nodes) and
should not be confused with nodes in the Model Builder tree of the
COMSOL Multiphysics software. Circuit node names are not restricted
to numerical values but can contain alphanumeric characters.

DEVICE NAMES
Each circuit component has an associated Device name, which is constructed from a
prefix identifying the type of the device and a string. The string can be specified in the
feature’s Settings window. The Device name is used to identify variables defined by the
component, and for the SPICE Import and Export functionality.

SETTINGS
The Label is the default physics interface name.

The Name is used primarily as a scope prefix for variables defined by the physics
interface. Refer to such physics interface variables in expressions using the pattern
<name>.<variable_name>. In order to distinguish between variables belonging to
different physics interfaces, the name string must be unique. Only letters, numbers, and
underscores (_) are permitted in the Name field. The first character must be a letter.

The default Name (for the first physics interface in the model) is cir.

256 | CHAPTER 4: THE AC/DC BRANCH

RESISTANCE IN PARALLEL TO PN JUNCTIONS
For numerical stability, a large resistance is added automatically in parallel to the pn
junctions in diodes and BJT devices. Enter a default value for the Resistance in parallel
to pn junctions Rj (SI unit: Ω). The default value is 1·1012 Ω.

CREATE UNIQUE NODES FOR NEW DEVICES

When this setting is selected (the default), newly added devices will be assigned unused
node names. The devices will be disconnected from the rest of the circuit and the
nodes should be updated to reflect the actual circuit connections. When this setting is
deselected, new devices will be connected to the lowest-numbered nodes starting from
0.

ELECTRICAL CIRCUIT TOOLBAR

The following nodes are available from the Electrical Circuit ribbon toolbar (Windows
users), Electrical Circuit context menu (Mac or Linux users), or right-click to access the
context menu (all users):

For step-by-step instructions and general documentation

descriptions, this is the Electrical Circuit toolbar.

THE ELECTRICAL CIRCUIT INTERFACE | 257

 • Ground Node • Subcircuit Definition
• Voltmeter • Subcircuit Instance
• Ampère Meter • Mutual Inductance
• Resistor • Transformer
• Capacitor • NPN BJT and PNP BJT2
• Inductor • n-Channel MOSFET and
• Voltage Source p-Channel MOSFET2

• Current Source • External I vs. U3

• Diode • External U vs. I3

• Voltage-Controlled Voltage Source1 • External I-Terminal3

• Voltage-Controlled Current Source1 • SPICE Circuit Import

1 • SPICE Circuit Export
• Current-Controlled Voltage Source
1
• Current-Controlled Current Source
1 Selected from the Dependent Sources submenu when you right-click main node.
2
Selected from the Transistors submenu when you right-click main node.
3
Selected from the External Couplings submenu when you right-click main node.

Ground Node
The Ground Node ( ) feature adds a ground node with the default node number zero
to the electrical circuit. This is the default node in the Electrical Circuit interface. More
ground nodes can be added but those must have unique node numbers and are by
default given higher node numbers.

GROUND CONNECTION
Set the Node name for the ground node in the circuit. The convention is to use 0 (zero)
for the ground node. If adding more ground nodes, each must have a unique node
name (number).

Voltmeter
The Voltmeter ( ) feature connects a voltmeter (voltage measurement device)
between two nodes in the electrical circuit. A voltmeter behaves electrically as an open
circuit. The voltmeter node adds a Probe sampling the voltage across it.

258 | CHAPTER 4: THE AC/DC BRANCH

DEVICE NAME
Enter a Device name for the voltmeter. The name is also used in the probe.

NODE CONNECTIONS
Set the two Node names for the connecting nodes for the resistor.

Ampère Meter
The Ammeter ( ) feature connects an ammeter (current measurement device)
between two nodes in the electrical circuit. An ammeter behaves electrically as a short
circuit. The ammeter node adds a Probe sampling the current through it.

DEVICE NAME
Enter a Device name for the ammeter. The name is also used in the probe.

NODE CONNECTIONS
Set the two Node names for the connecting nodes for the resistor.

DEVICE PARAMETERS
Enter the Resistance of the resistor.

Resistor
The Resistor ( ) feature connects a resistor between two nodes in the electrical
circuit.

DEVICE NAME
Enter a Device name for the resistor. The prefix is R.

NODE CONNECTIONS
Set the two Node names for the connecting nodes for the resistor.

DEVICE PARAMETERS
Enter the Resistance of the resistor.

Capacitor
The Capacitor ( ) feature connects a capacitor between two nodes in the electrical
circuit.

THE ELECTRICAL CIRCUIT INTERFACE | 259

 DEVICE NAME
Enter a Device name for the capacitor. The prefix is C.

NODE CONNECTIONS
Set the two Node names for the connecting nodes for the capacitor.

DEVICE PARAMETERS
Enter the Capacitance of the capacitor.

Inductor
The Inductor ( ) feature connects an inductor between two nodes in the electrical
circuit.

DEVICE NAME
Enter a Device name for the inductor. The prefix is L.

NODE CONNECTIONS
Set the two Node names for the connecting nodes for the inductor.

DEVICE PARAMETERS
Enter the Inductance of the inductor.

Voltage Source
The Voltage Source ( ) feature connects a voltage source between two nodes in the
electrical circuit.

DEVICE NAME
Enter a Device name for the voltage source. The prefix is V.

NODE CONNECTIONS
Set the two Node names for the connecting nodes for the voltage source. The first node
represents the positive reference terminal.

260 | CHAPTER 4: THE AC/DC BRANCH

DEVICE PARAMETERS
Enter the Source type that should be adapted to the selected study type. It can be
DC-source, AC-source, or a time-dependent Sine source or Pulse source. Depending on
the choice of source, also specify the following parameters:

• For a DC-source, the Voltage Vsrc (default value: 1 V). DC-sources are active in
Stationary and Time-Dependent studies.
• For an AC-source: the Voltage Vsrc (default value: 1 V) and the Phase Θ (default
value: 0 rad). AC-sources are active in Frequency Domain studies only.
• For a sine source: the Voltage Vsrc (default value: 1 V), the Offset Voff (default value:
0 V), the Frequency (default value: 1 kHz), and the Phase Θ (default value: 0 rad).
The sine sources are active in Time-Dependent studies and also in Stationary
studies, providing that a value for t has been provided as a model parameter or
global variable.
• For a pulse source: the Voltage Vsrc (default value: 1 V), the Offset Voff (default value:
0 V), the Delay td (default value: 0s), the Rise time tr and Fall time tf (default values:
0 s), the Pulse width pw (default value: 1 μs), and the Period Tper (default value: 2 μs).
The pulse sources are active in Time-Dependent studies and also in Stationary
studies, providing that a value for t has been provided as a model parameter or
global variable.

All values are peak values rather than RMS.

For the AC source, the frequency is a global input set by the solver. AC
sources should be used in Frequency-domain studies only. Do not use the
Sine source unless the model is time-dependent.

Current Source
The Current Source ( ) feature connects a current source between two nodes in the
electrical circuit.

DEVICE NAME
Enter a Device name for the current source. The prefix is I.

THE ELECTRICAL CIRCUIT INTERFACE | 261

 NODE CONNECTIONS
Set the two Node names for the connecting nodes for the current source. The first node
represents the positive reference terminal from where the current flows through the
source to the second node.

DEVICE PARAMETERS
Enter the Source type that should be adapted to the selected study type. It can be
DC-source, AC-source, or a time-dependent Sine source or Pulse source. Depending on
the choice of source, also specify the following parameters:

• For a DC-source, the Current isrc (default value: 1 A). DC-sources are active in
Stationary and Time-Dependent studies.
• For an AC-source: the Current isrc (default value: 1 A) and the Phase Θ (default
value: 0 rad). AC-sources are active in Frequency Domain studies only.
• For a sine source: the Current isrc (default value: 1 A), the Offset ioff (default value:
0 A), the Frequency (default value: 1 kHz), and the Phase Θ (default value: 0 rad).
The sine sources are active in Time-Dependent studies and also in Stationary
studies, providing that a value for t has been provided as a model parameter or
global variable.
• For a pulse source: the Current isrc (default value: 1 A), the Offset ioff (default value:
0 A), the Delay td (default value: 0 s), the Rise time tr and Fall time tf (default values:
0 s), the Pulse width pw (default value: 1 μs), and the Period Tper (default value: 2 μs).
The pulse sources are active in Time-Dependent studies and also in Stationary
studies, providing that a value for t has been provided as a model parameter or
global variable.

All values are peak values rather than RMS.

For the AC source, the frequency is a global input set by the solver. AC
sources should be used in frequency-domain studies only. Do not use the
Sine source unless the model is time-dependent.

Voltage-Controlled Voltage Source

The Voltage-Controlled Voltage Source ( ) feature connects a voltage-controlled
voltage source between two nodes in the electrical circuit. A second pair of nodes
define the input control voltage.

262 | CHAPTER 4: THE AC/DC BRANCH

DEVICE NAME
Enter a Device name for the voltage-controlled voltage source. The prefix is E.

NODE CONNECTIONS
Specify four Node names: the first pair for the connection nodes for the voltage source
and the second pair defining the input control voltage. The first node in a pair
represents the positive reference terminal.

DEVICE PARAMETERS
Enter the voltage Gain. The resulting voltage is this number multiplied by the control
voltage.

Voltage-Controlled Current Source

The Voltage-Controlled Current Source ( ) feature connects a voltage-controlled
current source between two nodes in the electrical circuit. A second pair of nodes
define the input control voltage.

DEVICE NAME
Enter a Device name for the voltage-controlled current source. The prefix is G.

NODE CONNECTIONS
Specify four Node names: the first pair for the connection nodes for the current source
and the second pair defining the input control voltage. The first node in a pair
represents the positive voltage reference terminal or the one from where the current
flows through the source to the second node.

DEVICE PARAMETERS
Enter the source Gain (SI units: S). The resulting current is this number multiplied by
the control voltage. It represents the transconductance of the source.

Current-Controlled Voltage Source

The Current-Controlled Voltage Source ( ) feature connects a current-controlled
voltage source between two nodes in the electrical circuit. The input control current
is the one flowing through a two-pin device.

DEVICE NAME
Enter a Device name for the current-controlled voltage source. The prefix is H.

THE ELECTRICAL CIRCUIT INTERFACE | 263

 NODE CONNECTIONS
Set two Node names for the connection nodes for the voltage source. The first node in
a pair represents the positive reference terminal.

DEVICE PARAMETERS
Enter the voltage Gain and select the Device whose current is taken as the control
current. The resulting voltage is this number multiplied by the control current through
the named Device (any two-pin device). Thus it formally has the unit of resistance.

Current-Controlled Current Source

The Current-Controlled Current Source ( ) feature connects a current-controlled
current source between two nodes in the electrical circuit. The input control current
is the one flowing through a named device that must be a two-pin device.

DEVICE NAME
Enter a Device name for the current-controlled current source. The prefix is F.

NODE CONNECTIONS
Specify two Node names for the connection nodes for the current source. The first node
in a pair represents the positive reference terminal from where the current flows
through the source to the second node.

DEVICE PARAMETERS
Enter the current Gain and select the Device whose current is taken as the control
current. The resulting current is this number multiplied by the control current
through the Device.

Subcircuit Definition
The Subcircuit Definition ( ) feature is used to define subcircuits, which can be
inserted as devices into the main circuit using Subcircuit Instance nodes. Create the
subcircuit by adding subnodes to the Subcircuit Definition node, either by using the
Physics toolbar, or by right-clicking the Subcircuit Definition.

SUBCIRCUIT PINS
Define the Pin names at which the subcircuit connects to the main circuit or to other
subcircuits when referenced by a Subcircuit Instance node. The Pin names refer to
circuit nodes in the subcircuit. The order in which the Pin names are defined is the

264 | CHAPTER 4: THE AC/DC BRANCH

order in which they are referenced by a Subcircuit Instance node. The devices
constituting the subcircuit should be connected only to the subcircuit’s pins and to
themselves.

Subcircuit Instance
The Subcircuit Instance ( ) feature represents an instance of a subcircuits defined by
a Subcircuit Definition feature.

DEVICE NAME
Enter a Device name for the subcircuit instance. The prefix is X.

NODE CONNECTIONS
Select the Name of subcircuit link from the list of defined subcircuits in the circuit model
and the circuit Node names at which the subcircuit instance connects to the main circuit
or to another subcircuit if used therein.

NPN BJT and PNP BJT

The NPN BJT ( ) and the PNP BJT ( ) device models are large-signal models for
bipolar junction transistors (BJT). It is an advanced device model and no thorough
description and motivation of the many input parameters are attempted here. Many
device manufacturers provide model input parameters for this BJT model. For any
particular make of BJT, the device manufacturer should be the primary source of
information.

DEVICE NAME
Enter a Device name for the BJT. The prefix is Q.

NODE CONNECTIONS
Specify three Node names for the connection nodes for the BJT device. These represent
the collector, base, and emitter nodes for the NPN transistor, and the emitter, base, and
collector nodes for the PNP transistor.

THE ELECTRICAL CIRCUIT INTERFACE | 265

 MODEL PARAMETERS
Specify the Model Parameters. Reasonable defaults are provided but for any particular
BJT, the device manufacturer should be the primary source of information.

The interested reader is referred to Ref. 1 for more details on

semiconductor modeling within circuits.

For an explanation of the Model Parameters see Bipolar Transistors.

n-Channel MOSFET and p-Channel MOSFET

The n-Channel MOSFET ( ) and the p-Channel MOSFET ( ) device models are
large-signal models for, respectively, an n-Channel MOS field-effect transistor
(MOSFET) and p-Channel MOSFET. These are advanced device models and no
thorough description and motivation of the many input parameters are attempted
here. Many device manufacturers provide model parameters for the MOSFET models.
For any particular make of MOSFET, the device manufacturer should be the primary
source of information.

DEVICE NAME
Enter a Device name for the MOSFET. The prefix is M.

NODE CONNECTIONS
Specify four Node names for the connection nodes for the n-Channel MOSFET or
p-Channel MOSFET device. These represent the drain, gate, source, and bulk nodes,
respectively.

MODEL PARAMETERS
Specify the Model Parameters. Reasonable defaults are provided but for any particular
MOSFET, the device manufacturer should be the primary source of information.

The interested reader is referred to Ref. 1 for more details on

semiconductor modeling within circuits.

For an explanation of the Model Parameters see MOSFET Transistors.

266 | CHAPTER 4: THE AC/DC BRANCH

Mutual Inductance
The Mutual Inductance allows specifying a coupling between two existing Inductor
features in the circuit. The mutual inductance of the coupling is

M = k L1 L2

where k is the coupling factor and L1 and L2 are the inductances of the inductors.

DEVICE PARAMETERS
Enter values or expressions for the:

• Coupling factor k (dimensionless). The value must be between 0 and 1, and the
default is 0.98.
• First inductance L1 (SI unit: H) and Second inductance L2 (SI unit: H). These must
be set to two different Inductor features in the circuit.

Transformer
The Transformer feature represents either a combination of two Inductor and a Mutual
Inductance features, or an ideal transformer.

NODE CONNECTIONS
Enter or edit the table in the Node names column for the primary and secondary node
connections.

DEVICE PARAMETERS
Choose a Transformer model — Specify inductors (the default) or Ideal transformer.

For Specify inductors enter values or expressions for the:

• Coupling factor k (dimensionless). The default is 0.98.

• First inductance L1 (SI unit: H). The default is 1 mH.
• Second inductance L2 (SI unit: H). The default is 1 mH.

For Ideal transformer enter values or expressions for the Winding ratio N1/N2
(dimensionless). The default is 10.

THE ELECTRICAL CIRCUIT INTERFACE | 267

 Diode
The Diode device model ( ) is a large-signal model for a diode. It is an advanced
device model and no thorough description and motivation of the many input
parameters are attempted here. The interested reader is referred to Ref. 1 for more
details on semiconductor modeling within circuits. Many device manufacturers
provide model parameters for this diode model. For any particular make of diode, the
device manufacturer should be the primary source of information.

DEVICE NAME
Enter a Device name for the diode. The prefix is D.

NODE CONNECTIONS
Specify two Node names for the positive and negative nodes for the Diode device.

MODEL PARAMETERS
Specify the Model Parameters. Reasonable defaults are provided but for any particular
diode, the device manufacturer should be the primary source of information.

For an explanation of the Model Parameters see Diode.

External I vs. U
The External I vs. U ( ) feature connects an arbitrary voltage measurement (for
example, a circuit terminal or circuit port boundary or a coil domain from another
physics interface) as a voltage source between two nodes in the electrical circuit. The
resulting circuit current from the first node to the second node is typically coupled
back as a prescribed current source in the context of the voltage measurement.

DEVICE NAME
Enter a Device name for the External I vs. U node.

NODE CONNECTIONS
Specify the two Node names for the connecting nodes for the voltage source. The first
node represents the positive reference terminal.

268 | CHAPTER 4: THE AC/DC BRANCH

EXTERNAL DEVICE
Enter the source of the Voltage. If circuit or current excited terminals or circuit ports
are defined on boundaries or domains or a multiturn coil domains is defined in other
physics interfaces, these display as options in the Voltage list. Also select the User defined
option and enter your own voltage variable, for example, using a suitable coupling
operator. For inductive or electromagnetic wave propagation models, the voltage
measurement must be performed as an integral of the electric field because the electric
potential only does not capture induced EMF. Also the integration must be performed
over a distance that is short compared to the local wavelength.

Except when coupling to a circuit terminal, circuit port, or coil, the

current flow variable must be manually coupled back in the electrical
circuit to the context of the voltage measurement. This applies also when
coupling to a current excited terminal. The name of this current variable
follows the convention cirn.IvsUm_i, where cirn is the tag of the
Electrical Circuit interface node and IvsUm is the tag of the External I vs.
U node. The tags are typically displayed within curly brackets {} in the
Model Builder.

Nonlocal Couplings and Coupling Operators in the COMSOL

Multiphysics Reference Manual

External U vs. I
The External U vs. I ( ) feature connects an arbitrary current measurement (for
example, a coil domain from another physics interface) as a current source between
two nodes in the electrical circuit. The resulting circuit voltage between the first node
and the second node is typically coupled back as a prescribed voltage source in the
context of the current measurement.

DEVICE NAME
Enter a Device name for the External U vs. I node.

NODE CONNECTIONS
Specify the two Node names for the connecting nodes for the current source. The
current flows from the first node to the second node.

THE ELECTRICAL CIRCUIT INTERFACE | 269

 EXTERNAL DEVICE
Enter the source of the Current. Voltage excited terminals or lumped ports defined on
boundaries in other physics interfaces are natural candidates but do not appear as
options in the Voltage list because those do not have an accurate built-in current
measurement variable. A User defined option must be selected and a current variable
entered, for example, using a suitable coupling operator.

The voltage variable must be manually coupled back in the electrical

circuit to the context of the current measurement. This applies also when
coupling to a voltage excited terminal or lumped port. The name of this
voltage variable follows the convention cirn.UvsIm_v, where cirn is the
tag of the Electrical Circuit interface node and UvsIm is the tag of the
External U vs. I node. The tags are typically displayed within curly
brackets {} in the Model Builder.

Nonlocal Couplings and Coupling Operators in the COMSOL

Multiphysics Reference Manual

External I-Terminal
The External I-Terminal ( ) feature connects an arbitrary voltage-to-ground
measurement (for example, a circuit terminal from another physics interface) as a
voltage-to-ground assignment to a node in the electrical circuit. The resulting circuit
current from the node is typically coupled back as a prescribed current source in the
context of the voltage measurement. This node does not apply when coupling to
inductive or electromagnetic wave propagation models because then voltage must be
defined as a line integral between two points rather than a single point measurement
of electric potential. For such couplings, use the External I vs. U node instead.

DEVICE NAME
Enter a Device name for the External I-terminal.

NODE CONNECTIONS
Set the Node name for the connecting node for the voltage assignment.

270 | CHAPTER 4: THE AC/DC BRANCH

EXTERNAL TERMINAL
Enter the source of the Voltage. If circuit- or current-excited terminals are defined on
boundaries in other physics interfaces, these display as options in the Voltage list. Also
select the User defined option and enter a voltage variable, for example, using a suitable
coupling operator.

Except when coupling to a circuit terminal, the current flow variable must
be manually coupled back in the electrical circuit to the context of the
voltage measurement. This applies also when coupling to a current
excited terminal. The name of this current variable follows the convention
cirn.termIm_i, where cirn is the tag of the Electrical Circuit interface
node and termIm is the tag of the External I-Terminal node. The tags are
typically displayed within curly brackets {} in the Model Builder.

Nonlocal Couplings and Coupling Operators in the COMSOL

Multiphysics Reference Manual.

SPICE Circuit Import

Right-click the Electrical Circuit ( ) feature node to import an existing SPICE netlist
(select Import Spice Netlist). A window opens — enter a file location or browse your
directories to find one. The default file extension for a SPICE netlist is .cir. The
SPICE circuit import translates the imported netlist into Electrical Circuit interface
nodes so these define the subset of SPICE features that can be imported.

See SPICE Import and Export about the supported SPICE commands.

SPICE Circuit Export

Right-click the Electrical Circuit ( ) feature node to export the current circuit to the
SPICE netlist file format (select Export Spice Netlist ). A window opens — enter a

THE ELECTRICAL CIRCUIT INTERFACE | 271

 file location or browse your directories to find one. The default file extension for a
SPICE netlist is .cir. The compatible circuit nodes are exported as SPICE devices

See SPICE Export for more details on the supported SPICE commands.

272 | CHAPTER 4: THE AC/DC BRANCH

Theory for the Electrostatics
Interface
The Electrostatics Interface is available for 3D, 2D in-plane, and 2D axisymmetric
components. Applications with electrostatic equations include high-voltage apparatus,
electronic devices, and capacitors. The term “statics” is not to be interpreted literally
— it is the observation time, or time scale at which the applied excitation changes, that
is short compared to the charge relaxation time; also, the electromagnetic wavelength
and skin depth are very large compared to the size of the domain of interest.

If you do not know whether to use the Electric Currents or the Electrostatics interface,
which both solve for the scalar electric potential V, consider using an explicit charge
transport model. See Charge Relaxation Theory.

Charge Relaxation Theory

COMSOL Multiphysics includes physics interfaces for the modeling of static electric
fields and currents. Deciding what specific physics interface and study type to select for
a particular modeling situation requires a basic understanding of the charge dynamics
in conductors.

The different physics interfaces involving only the scalar electric potential can be
interpreted in terms of the charge relaxation process. The fundamental equations
involved are Ohm’s law for the conduction current density

J c = σE

the equation of continuity

∂-----
ρ-
+ ∇ ⋅ Jc = 0
∂t

and Gauss’ law

∇ ⋅ ( εE ) = ρ

By combining these, one can deduce the following differential equation for the space
charge density in a homogeneous medium

THEORY FOR THE ELECTROSTATICS INTERFACE | 273

 ∂-----
ρ- σ
+ --- ρ = 0
∂t ε

This equation has the solution

–t ⁄ τ
ρ ( t ) = ρ0 e

where

ε
τ = ---
σ

is called the charge relaxation time. For a good conductor like copper, τ is of the order
of 10−19 s, whereas for a good insulator like silica glass, it is of the order of 103 s. For
a pure insulator, it becomes infinite.

When modeling real-world devices, there is not only the intrinsic time scale of the
charge relaxation time but also an external time scale t at which a device is energized
or the observation time. It is the relation between the external time scale and the
charge relaxation time that determines what physics interface and study type to use.
The results are summarized in Table 4-2 below,
TABLE 4-2: SUITABLE PHYSICS INTERFACE AND STUDY TYPE FOR DIFFERENT TIME-SCALE REGIMES.

CASE PHYSICS INTERFACE STUDY TYPE

τ>>t Electrostatics Stationary

τ<<t Electric Currents Stationary
τ~t Electric Currents Time Dependent orFrequency Domain

FIRST CASE: τ >> T

If the external time scale is short compared to the charge relaxation time, the charges
do not have time to redistribute to any significant degree. Thus the charge distribution
can be considered as a given model input. The best approach is to solve the
Electrostatics formulation using the electric potential V.

By combining the definition of the potential with Gauss’ law, you can derive the
classical Poisson’s equation. Under static conditions, the electric potential V is defined
by the equivalence E = −∇V. Using this together with the constitutive relation D = ε0E
+ P between D and E, you can rewrite Gauss’ law as a variant of Poisson’s equation

– ∇ ⋅ ( ε 0 ∇V – P ) = ρ

274 | CHAPTER 4: THE AC/DC BRANCH

This equation is used in the Electrostatics interface. It is worth noting that Gauss’ law
does not require the charge distribution to be static. Thus, provided dynamics are slow
enough that induced electric fields can be neglected and hence a scalar electric
potential is justified, the formulation can be used also in the Time Dependent study
type. That typically involves either prescribing the charge dynamics or coupling a
separate formulation for this.

Such separate charge transport formulations can be found in the Plasma

Module, the Semiconductor Module, and the Chemical Reaction
Engineering Module.

SECOND CASE: τ <<T

If the external time scale is long compared to the charge relaxation time, the stationary
solution to the equation of continuity has been reached. In a stationary coordinate
system, a slightly more general form of Ohm’s law than above states that

J c = σE + J e

where Je is an externally generated current density. The static form of the equation of
continuity then reads

∇ ⋅ J c = – ∇ ⋅ ( σ ∇V – J e ) = 0

To handle current sources, the equation can be generalized to

– ∇ ⋅ ( σ ∇V – J e ) = Q j

This equation is used in the static study type for the Electric Currents interface.

GENERAL CASE: CHARGE DYNAMICS

If the charge relaxation time is comparable to the external time scale, the Time
Dependent or Frequency Domain study types for the Electric Currents interface must
be used.

Combining the time-harmonic equation of continuity

∇ ⋅ Jc = ∇ ⋅ ( σE + J e ) = –jωρ

with the equation (∇⋅ D = ρ) yields the following equation for the frequency domain
study type:

THEORY FOR THE ELECTROSTATICS INTERFACE | 275

 – ∇ ⋅ ( ( σ + jωε 0 )∇V – ( J e + jωP ) ) = 0

For the Time Dependent study type, use the transient equation of continuity

∂ρ
∇ ⋅ J c = ∇ ⋅ ( σE + J e ) = –
∂t

and the resulting equation becomes

– ∇ ⋅ ∂ ( ε 0 ∇V + P ) – ∇ ⋅ ( σ∇V – J e ) = 0
∂t

These dynamic formulations are valid as long as induced electric fields can be ignored
and hence the electric field is essentially curl free. This condition is fulfilled provided
that skin effect and wave propagation effects can be ignored. The skin depth must be
much larger than the geometrical dimensions of the modeled device and so must the
wavelength. Note also that these formulations can be used to model dielectric regions
of capacitive/resistive devices even though the interior of electrodes sometimes does
not meet the large skin depth condition. In that case, the electrodes must only be
represented as boundary conditions (fixed or floating potential). The interior, metallic
domains are not included in the analysis. Obviously, this is only a valid approach for
devices where metallic electrodes do not entirely bypass (short circuit) the
capacitive/resistive layers. If metallic electrodes short circuit the capacitive/resistive
layers, the time evolution of the current is determined by inductive and resistive effects
with very little influence from the capacitive layers. Then the Magnetic Fields interface
is the appropriate modeling tool.

Electrostatics Equations
Under static conditions, the electric potential, V, is defined by the relationship:

E = – ∇V

Combining this equation with the constitutive relationship D = ε0E + P between the
electric displacement D and the electric field E, it is possible to represent Gauss’ law
as the following equation:

– ∇ ⋅ ( ε 0 ∇V – P ) = ρ

In this equation, the physical constant, ε0 (SI unit: F/m) is the permittivity of vacuum,
P (SI unit: C/m2) is the electric polarization vector, and ρ (SI unit: C/m3) is a space
charge density. This equation describes the electrostatic field in dielectric materials.

276 | CHAPTER 4: THE AC/DC BRANCH

For in-plane 2D modeling, the Electrostatics interface assumes a symmetry where the
electric potential varies only in the x and y directions and is constant in the z direction.
This implies that the electric field, E, is tangential to the xy-plane. With this symmetry,
the same equation is solved as in the 3D case. The physics interface solves the following
equation where d is the thickness in the z direction:

– ∇ ⋅ d ( ε 0 ∇V – P ) = ρ

The axisymmetric version of the physics interface considers the situation where the
fields and geometry are axially symmetric. In this case, the electric potential is constant
in the ϕ direction, which implies that the electric field is tangential to the rz-plane.

The Electrostatics Interface in Time Dependent or Frequency Domain

Studies
The Electrostatics Interface can also be solved in a dynamic study (Time Dependent
or Frequency Domain). The equation system solved, however, is typically always the
one presented in the previous section for the stationary case, in which no transient
electromagnetic effects are taken into account. The difference is that the sources of the
problem (charge densities, electric potential) are assumed to be time-varying (in a
Time Dependent study) or time-harmonic (in a Frequency Domain study).

The support for dynamic studies simplifies the coupling of the Electrostatics interface
with other physics interfaces. Using the physics interface in a dynamic study is a valid
approximation only if the time-scale (or the frequency) of the study is so slow that
transient electromagnetic effects can be neglected; for example, in acoustic or
structural problems.

Most of the features in the Electrostatics interface have the same behavior in all study
types, with the important exception of the Terminal node.

The Electrostatics interface also supports the small-signal analysis study sequence,
which can be used when a time-harmonic perturbation is superposed on a static bias
charge or voltage.

THEORY FOR THE ELECTROSTATICS INTERFACE | 277

 Theory for the Electrical Circuit
Interface
The Electrical Circuit Interface theory is discussed in this section:

• Electric Circuit Modeling and the Semiconductor Device Models

• Bipolar Transistors
• MOSFET Transistors
• Diode
• Reference for the Electrical Circuit Interface

Connecting to Electrical Circuits

Electric Circuit Modeling and the Semiconductor Device Models

Electrical circuit modeling capabilities are useful when simulating all sorts of electrical
and electromechanical devices ranging from heaters and motors to advanced plasma
reactors in the semiconductor industry. There are two fundamental ways that an
electrical circuit model relates to a physical field model.

• The field model is used to get a better, more accurate description of a single device
in the electrical circuit model.
• The electrical circuit is used to drive or terminate the device in the field model in
such a way that it makes more sense to simulate both as a tightly coupled system.

The Electrical Circuit interface makes it possible to add nodes representing circuit
elements directly to the Model Builder tree in a COMSOL Multiphysics model. The
circuit variables can then be connected to a physical device model to perform
co-simulations of circuits and multiphysics. The model acts as a device connected to
the circuit so that its behavior is analyzed in larger systems.

The fundamental equations solved by the Electrical Circuit interface are Kirchhoff’s
circuit laws, which in turn can be deduced from Maxwell’s equations. The supported
study types are Stationary, Frequency Domain, and Time Dependent.

278 | CHAPTER 4: THE AC/DC BRANCH

There are three more advanced large-signal semiconductor device features available in
the Electrical Circuit interface. The equivalent circuits and the equations defining their
non-ideal circuit elements are described in this section. For a more complete treatise
on semiconductor device modeling see Ref. 1.

Bipolar Transistors
Figure 4-1 illustrates the equivalent circuit for the npn bipolar junction transistor.

Figure 4-1: A circuit for the bipolar transistor.

The pnp transistor model is similar in all regards to the npn transistor, with the
difference that the polarities of the currents and voltages involved are reversed. The
following equations are used to compute the relations between currents and voltages
in the circuit.

THEORY FOR THE ELECTRICAL CIRCUIT INTERFACE | 279

 R B – R BM
v rb = ----  R BM – -------------------------- i b
1
A f bq 

 v be
 --------------
v bc
--------------  
  – 1
NF VT NR VT
= -----------------------------------------------  1 + 1 + 4I S  ----------------------- + ------------------------ 
1 e –1 e
f bq
v bc v be  I KF A I KR A 
2  1 – ----------- – -----------    
 V AF V AR 
v be v be
 I S  -------------
N V
-   --------------
N V 
i be = A  -------  e F T – 1 + I SE  e E T – 1 
 BF    
v bc v bc
 I S  --------------
N V   --------------
N V 
i bc = A  --------  e R T – 1 + I SC  e C T – 1 
B
 R   
v be v bc
 I S  -------------
N V
- --------------
N V 
i ce = A  -------  e F T + e C T 
f
 bq  
k B T NOM
V T = ------------------------
q

There are also two capacitances that use the same formula as the junction capacitance
of the diode model. In the parameter names below, replace x with C for the
base-collector capacitance and E for the base-emitter capacitance.

 v bx  – MJx
 1 – ---------
 -
 V Jx v bx < F C V Jx
C jbx 
= AC Jx × 
v bx  v bx ≥ F C V Jx
 ( 1 – F ) – 1 – M Jx  1 – F ( 1 + M ) + M ---------
-
 C  C Jx Jx V 
Jx

The model parameters are listed in the table below.

TABLE 4-1: BIPOLAR TRANSISTOR MODEL PARAMETERS

PARAMETER DEFAULT DESCRIPTION

BF 100 Ideal forward current gain

BR 1 Ideal reverse current gain
2
CJC 0 F/m Base-collector zero-bias depletion capacitance
CJE 0 F/m2 Base-emitter zero-bias depletion capacitance
FC 0.5 Breakdown current
2
IKF Inf (A/m ) Corner for forward high-current roll-off

280 | CHAPTER 4: THE AC/DC BRANCH

TABLE 4-1: BIPOLAR TRANSISTOR MODEL PARAMETERS

PARAMETER DEFAULT DESCRIPTION

IKR Inf (A/m2) Corner for reverse high-current roll-off

IS 1e-15 A/m2 Saturation current
ISC 0 A/m2 Base-collector leakage saturation current
ISE 0 A/m2 Base-emitter leakage saturation current
MJC 1/3 Base-collector grading coefficient
MJE 1/3 Base-emitter grading coefficient
NC 2 Base-collector ideality factor
NE 1.4 Base-emitter ideality factor
NF 1 Forward ideality factor
NR 1 Reverse ideality factor
RB 0 Ωm2 Base resistance
RBM 0 Ωm2 Minimum base resistance
RC 0 Ωm2 Collector resistance
RE 0 Ωm2 Emitter resistance
TNOM 298.15 K Device temperature
VAF Inf (V) Forward Early voltage
VAR Inf (V) Reverse Early voltage
VJC 0.71 V Base-collector built-in potential
VJE 0.71 V Base-emitter built-in potential

THEORY FOR THE ELECTRICAL CIRCUIT INTERFACE | 281

 MOSFET Transistors
Figure 4-2 illustrates an equivalent circuit for the n-channel MOSFET transistor. The
p-channel MOSFET transistor is treated similarly, but the polarities of the involved
voltages are reversed.

Figure 4-2: A circuit for the MOSFET transistor.

The following equations are used to compute the relations between currents and
voltages in the circuit.

282 | CHAPTER 4: THE AC/DC BRANCH

  KP
 W - -------
 ----
L 2
( 1 + Λv ds )v ds ( 2v th – v ds ) v ds < v th

i ds =  W KP 2
 ----- ------- ( 1 + Λv ds )v th v ds ≥ v th
 L 2
 0 v ds < v th ≤ 0

v th = v gs – ( V TO + Γ ( Φ – v bs – Φ ) )
v bd
 -----------
NV
- 
i bd = I S  e T – 1
 
v bs
 -----------
NV
- 
i bs = I S  e T – 1
 
k B T NOM
V T = ------------------------
q

There are also several capacitances between the terminals

C gd = C gd0 W
C gs = C gs0 W

1 – v
–M J
 bd
--------
-
  PB  v bx < F C P B
C jbd = C BD × 
v bx v bx ≥ F C P B
 ( 1 – F ) – 1 – M J  1 – F ( 1 + M ) + M -------
-
 C  C J J PB 

The model parameters are as follows:

TABLE 4-2: MOSFET TRANSISTOR MODEL PARAMETERS

PARAMETER DEFAULT DESCRIPTION

CBD 0 F/m Bulk-drain zero-bias capacitance

CGDO 0 F/m Gate-drain overlap capacitance
CGSO 0 F/m Gate-source overlap capacitance
FC 0.5 Capacitance factor
IS 1e-13 A Bulk junction saturation current
KP 2e-5 A/V2 Transconductance parameter
L 50e-6 m Gate length
MJ 0.5 Bulk junction grading coefficient

THEORY FOR THE ELECTRICAL CIRCUIT INTERFACE | 283

 TABLE 4-2: MOSFET TRANSISTOR MODEL PARAMETERS

PARAMETER DEFAULT DESCRIPTION

N 1 Bulk junction ideality factor

PB 0.75 V Bulk junction potential
RB 0Ω Bulk resistance
RD 0Ω Drain resistance
RDS Inf (Ω) Drain-source resistance
RG 0Ω Gate resistance
RS 0Ω Source resistance
TNOM 298.15 K Device temperature
VTO 0V Zero-bias threshold voltage
W 50e-6 m Gate width
Γ (GAMMA) 1 V0.5 Bulk threshold parameter
Φ (PHI) 0.5 V Surface potential
Λ (LAMBDA) 0 1/V Channel-length modulation

284 | CHAPTER 4: THE AC/DC BRANCH

Diode
Figure 4-3 illustrates equivalent circuit for the diode.

Figure 4-3: A circuit for the diode.

The following equations are used to compute the relations between currents and
voltages in the circuit.

THEORY FOR THE ELECTRICAL CIRCUIT INTERFACE | 285

 i d = i dhl + i drec + i db + i c
vd
 -----------
NV
-
 1
i dhl = I S  e T – 1 ---------------------------------------------------
  vd
I S  -----------
NV T
-

1 + ---------  e – 1
I KF  
vd
 --------------
N V 
i drec = I SR  e R T – 1
 
vd + BV
– -----------------
-
N BV V T
i db = I BV e

 vd  –M
  1 – ------
- vd < FC VJ
  V J
C j = C J0 × 
 – 1 – M vd
 ( 1 – FC )  1 – F C ( 1 + M ) + M ------- v d ≥ F C V J
 VJ

k B T NOM
V T = ------------------------
q

where the following model parameters are required

TABLE 4-3: DIODE TRANSISTOR MODEL PARAMETERS

PARAMETER DEFAULT DESCRIPTION

BV Inf (V) Reverse breakdown voltage

CJ0 0F Zero-bias junction capacitance
FC 0.5 Forward-bias capacitance coefficient
IBV 1e-09 A Current at breakdown voltage
IKF Inf (A) Corner for high-current roll-off
IS 1e-13 A Saturation current
M 0.5 Grading coefficient
N 1 Ideality factor
NBV 1 Breakdown ideality factor
NR 2 Recombination ideality factor
RS 0Ω Series resistance
TNOM 298.15 K Device temperature
VJ 1.0 V Junction potential

286 | CHAPTER 4: THE AC/DC BRANCH

Reference for the Electrical Circuit Interface
1. P. Antognetti and G. Massobrio, Semiconductor Device Modeling with Spice, 2nd
ed., McGraw-Hill, 1993.

THEORY FOR THE ELECTRICAL CIRCUIT INTERFACE | 287

288 | CHAPTER 4: THE AC/DC BRANCH
 5

Schrödinger Equation Interfaces

This chapter describes the Schrödinger Equation interface ( ) and the

Schrödinger-Poisson Equation multiphysics interface ( ).

In this chapter:

• The Schrödinger Equation Interface

• Theory for the Schrödinger Equation Interface
• The Schrödinger-Poisson Equation Multiphysics Interface

289
 The Schrödinger Equation Interface
The Schrödinger Equation (schr) interface ( ), found under the Semiconductor branch
( ) when adding a physics interface, solves the Schrödinger equation for the wave
function of a single particle in an external potential. By default, the Schrödinger
equation is added to all the domain selections of the interface.

When this physics interface is added, these default nodes are added to the Model
Builder: Effective Mass, Electron Potential Energy, Zero Flux, and Initial Values. Then,
from the Physics toolbar, add other nodes that implement, for example, boundary
conditions or additional contributions to the potential energy. You can also right-click
Schrödinger Equation to select physics features from the context menu.

SETTINGS
The Label is the default physics interface name.

The Name is used primarily as a scope prefix for variables and coupling operators
defined by the physics interface. Refer to such physics interface variables in expressions
using the pattern <name>.<variable_name>. In order to distinguish between
variables belonging to different physics interfaces, the name string must be unique.
Only letters, numbers, and underscores (_) are permitted in the Name field. The first
character must be a letter.

The default Name (for the first physics interface in the model) is schr.

MODEL PROPERTIES
Use Model properties to set the particle type and study-specific variables in the model.

Particle type
Select an option from the Particle type list: Electrons (the default) or Holes.

• For electron-like particles, the direction of lower energy is in the negative direction
of the electron potential energy. Pictorially, this type of particles tend to fall down
to the bottom of an electron potential energy well.
• For hole-like particles, the direction of lower energy is in the positive direction of
the electron potential energy. Pictorially, this type of particles tend to float up to the
top of an electron potential energy peak.

290 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

Azimuthal quantum number
For axisymmetric models, enter an integer for the azimuthal quantum number m
(unitless). The default is 0.

Eigenvalue scale
• Enter a constant value for the Eigenvalue scale λscale (J) for eigenvalue studies. The
eigenenergy is defined as the product of the eigenvalue scale and the eigenvalue
(dimensionless). The default is 1 eV, so that the eigenvalue will be the numerical
value of the eigenenergy in unit of eV.

Energy
Enter a constant value or a global parameter for the Energy E (J) for stationary studies.
This specifies the total energy of the stationary Schrödinger equation.

DISCRETIZATION
Use this section to change the discretization of the Schrödinger equation.

DEPENDENT VARIABLES
The dependent variable (field variable) is for the particle Wave function psi. The name
can be changed but the names of fields and dependent variables must be unique within
a model.

• Domain and Boundary Nodes for the Schrödinger Equation Interface

• Theory for the Schrödinger Equation Interface

Domain and Boundary Nodes for the Schrödinger Equation Interface

The Schrödinger Equation Interface has these domain and boundary nodes available
from the Physics ribbon toolbar (Windows users), Physics context menu (Mac or Linux
users), or by right-clicking the main physics interface node to access the context menu
(all users).

In general, to add a node, go to the Physics toolbar, no matter what

operating system you are using. Subnodes are available by clicking the
parent node and selecting it from the Attributes menu.

THE SCHRÖDINGER EQUATION INTERFACE | 291

 In the COMSOL Multiphysics Reference Manual see Table 2-4 for links
to common sections and Table 2-5 to common feature nodes. You can
also search for information: press F1 to open the Help window or Ctrl+F1
to open the Documentation window.

• The Schrödinger Equation Interface

• Theory for the Schrödinger Equation Interface

These nodes and subnodes are described in this section and listed in alphabetical order:

• Effective Mass • Periodic Condition

• Electron Potential Energy • Zero Flux
• Initial Values • Zero Probability
• Open Boundary

Effective Mass
Use the Effective Mass node to set the effective mass of the particle. The default is
0.067*me_const for electron-like particles and 0.45*me_const for hole-like
particles.

It is common for electrons or holes to have different effective masses in different

regions of a heterostructure. This can be easily specified by using multiple Effective
Mass nodes. Later nodes overrides earlier nodes in the Model Builder tree.

Electron Potential Energy

The Electron Potential Energy feature adds contribution to the total potential energy
experienced by the particle. The contribution from all Electron Potential Energy nodes
are summed up to give the total potential energy. This makes it easy to construct
complicated potential profiles by employing a number of nodes.

ELECTRON POTENTIAL ENERGY

Select a profile option for the Electron Potential Energy Ve (for electrons) or Vh (for
holes): Simple harmonic potential (the default), Quantum well, or User defined.

292 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

Simple harmonic potential
For this profile, enter the

• Angular frequency ω0 (rad/s). The default is 1[eV]/hbar_const.

• Center of potential r0 (m). The default is the origin of the geometry. For
axisymmetric models, the center of potential is fixed at the origin (not editable).
• Constant offset V0 (J). The default is 0[eV].
The contribution to the total electron potential energy is

1 2 2
V 0 ± --- m eff ω 0 r – r 0
2

where the plus sign is for electrons and minus sign for holes.

Quantum well
For this profile, enter the

• Width of quantum well d0 (m). The default is 10[nm].

• Depth of quantum well Vw (J). The default is 100[meV].
• Center of potential r0 (m). The default is the origin of the geometry. For
axisymmetric models, the center of potential is fixed at the origin (not editable).
• Constant offset V0 (J). The default is 0[eV].
The contribution to the total electron potential energy is

V 0 ± V w  r – r 0 > --- d 0
1
 2 

where the plus sign is for electrons and minus sign for holes.

If the effective mass of the particle varies from the well region to the barrier region, as
is very often the case, then it will be easier to not use this option. Instead, define
different domains for the well and barrier regions and use the User defined option to
specify the potential and effective mass in each domain.

User defined
For this option, enter the contribution to the electron potential energy in the input
field. The default is 0.5*schr.meff*((1[eV]/hbar_const)^2)*(x^2) for electrons
and -0.5*schr.meff*((1[eV]/hbar_const)^2)*(x^2) for holes.

THE SCHRÖDINGER EQUATION INTERFACE | 293

 Initial Values
The Initial Values node provides initial values for the wave function ψ. The default value
is eps, a tiny nonzero value.

Zero Flux
The Zero Flux node is the default boundary condition for external boundaries where
the normal component of the probability current density is set to zero by the nature
of the equation system.

Zero Probability
Use the Zero Probability node to constrain the value of the wave function to zero at a
boundary.

Periodic Condition
The Periodic Condition allows modeling just one unit cell of a periodic structure, such
as superlattices.

PERIODICITY SETTINGS
Select an option from the Type of periodicity list: Continuity (the default) or
Floquet-Bloch periodicity.

For Floquet-Bloch periodicity, enter the k-vector for Floquet periodicity kF (rad/m). The
default is the null vector, reducing the condition to be equivalent to Continuity. For
time dependent studies, the Floquet-Bloch periodicity option does not apply. If
erroneously selected, it will be reduced to the continuity condition.

DESTINATION SELECTION (SUBNODE)

If desired, use this subnode to manually select the destination boundary for the
periodic condition.

Open Boundary
The Open Boundary condition allows outgoing waves leave the modeling domain
without being reflected back from the external boundary. It allows an optional
incoming wave to be specified by the user. An advanced physics option allows the
reversal of the open boundary condition: an unspecified incoming wave comes into the

294 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

modeling domain through the boundary and no outgoing wave leaves the domain
through the boundary.

For stationary studies, an alternative for modeling nonreflecting outgoing

waves is to use an exterior domain specified as a Perfectly Matched Layer to
absorb the outgoing waves.

WAVE TYPE
This setting specifies the outgoing wave type. Depending on the spatial dimension of
the geometry, one or more of the following outgoing wave types may be supported:
Plane wave, Cylindrical wave and Spherical wave. If only one wave type is supported by
the geometry, the drop down menu is unavailable and displays the supported wave
type. If more than one is supported, the drop down menu is active and allows the user
to make a selection.

The “no reflection” condition works well only if the outgoing wave propagating
direction is normal to the boundary, otherwise some reflection will occur. Therefore
for the Plane wave type, the wave vector is assumed to be in the direction of the
outgoing normal of the boundary.

For the Cylindrical wave type, enter the Axial direction of cylindrical wave and one Point
on center axis of cylindrical wave.

For the Spherical wave type, enter the Center of spherical wave.

INCOMING WAVE
This setting specifies an optional incoming wave if the check box is checked. The
following settings may appear only if the check box is checked.

Incoming wave type

If more than one wave type is supported by the spatial dimension of the geometry, then
the Incoming wave type drop down menu appears and allow the user to make a
selection.

For any incoming wave type, enter the Incoming wave amplitude ψ0. The default is 1.

For incoming Plane wave, enter the Incoming wave direction.

For incoming Cylindrical wave, enter the Axial direction of incoming cylindrical wave and
one Point on center axis of incoming cylindrical wave.

For incoming Spherical wave, enter the Center of incoming spherical wave.

THE SCHRÖDINGER EQUATION INTERFACE | 295

 OPEN BOUNDARY TYPE
This setting is available if the Advanced Physics Options is enabled by clicking the Show
More Options button ( ) and selecting Advanced Physics Options in the Show More
Options dialog box. Select either Outgoing or Incoming.

Outgoing
This is the default. The open boundary allows outgoing waves leave the modeling
domain without being reflected back from the boundary.

Incoming
This reverses the usual open boundary condition: an unspecified incoming wave comes
into the modeling domain through the boundary and no outgoing wave leaves the
domain through the boundary. This particular option is mainly for analyzing resonant
tunneling conditions, see the Double Barrier 1D tutorial in the Application Libraries.

Coupling Operators
For convenience, an integration operator schr.int and a maximum operator
schr.max are pre-defined over the selection of the physics interface. See the Settings
section for the physics interface to see more explanation about the scoping Name for
the physics interface (default is schr).

296 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

Theory for the Schrödinger Equation
Interface
For an introduction to the Schrödinger equation and its application in quantum
confined semiconductor systems, the user is encouraged to consult textbooks such as
Ref. 1 and Ref. 2. This section is not meant to be a replacement for general textbooks
on the subject. Only those parts of the theory pertaining to the usage of the physics
interface will be covered.

In this section:

• Sign Convention and Energy Units • Sign Convention of the Electron

• Units of Wave Functions Potential Energy
• References for the Schrödinger
Equation Interface

Sign Convention and Energy Units

The single-particle Schrödinger Equation implemented in the physics interface reads

_2 ∇ψ ( r , t ) _ ∂
– h ∇• ----------------------- + V ( r, t )ψ ( r, t ) = – i h ψ ( r, t ) (5-1)
2m eff ( r ) ∂t

Note that the energy operator on the right-hand side of the equation takes an opposite
sign convention as the one adopted in most textbooks of quantum mechanics. This is
because COMSOL Multiphysics takes the engineering convention of exp(+iωt) for
time-harmonic solutions, as opposed to the physics convention of exp(−iωt). The
engineering convention is adopted for the Schrödinger Equation interface so that the
sign convention is consistent within the COMSOL product family. Under this unusual
sign convention, the momentum operator also acquires an opposite sign, since a plane
wave is now exp(−ikx+iωt), not exp(+ikx−iωt) as in most textbooks.

The stationary Schrödinger Equation reads

_2 ∇ψ ( r )
– h ∇• ----------------------- + V ( r )ψ ( r ) = Eψ ( r ) (5-2)
 2m eff ( r )

THEORY FOR THE SCHRÖDINGER EQUATION INTERFACE | 297

 For stationary studies, where the total energy E is a known quantity, enter a constant
value or a global parameter for the Energy E (J) in the physics interface settings window
(see section Energy).

For eigenvalue studies, where the eigenenergy E is unknown and to be solved for,
enter a constant value for the Eigenvalue scale λscale (J) in the physics interface settings
window (see section Eigenvalue scale). The eigenenergy E is given by the product of
the eigenvalue scale λscale and the eigenvalue λ (dimensionless). The default value for
λscale is 1 eV, so that the eigenvalue λ takes on the numerical value of the eigenenergy
E in units of eV.

Units of Wave Functions

Since the COMSOL Multiphysics software does not allow fractional units, the wave
functions in all spatial dimensions are set to unitless. The built-in variable for the
probability density (schr.Pr by default) has the unit of 1/m in 1D, 1/m2 in 2D or
1D axisymmetry, and 1/m3 in 3D or 2D axisymmetry.

Sign Convention of the Electron Potential Energy

The sign convention for the Electron Potential Energy domain condition follows the
one adopted by the Semiconductor Interface:

• For electron-like particles, the direction of lower energy is in the negative direction
of the electron potential energy. Pictorially, this type of particles tend to fall down
to the bottom of an electron potential energy well.
• For hole-like particles, the direction of lower energy is in the positive direction of
the electron potential energy. Pictorially, this type of particles tend to float up to the
top of an electron potential energy peak.

References for the Schrödinger Equation Interface

1. L. I. Schiff, Quantum Mechanics, 3rd edition (1968), McGraw-Hill.

2. P. Harrison, Quantum Wells, Wires and Dots, 3rd edition (2009), Wiley.

298 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

The Schrödinger-Poisson Equation
Multiphysics Interface
The Schrödinger-Poisson Equation multiphysics interface creates a bidirectional
coupling between an Electrostatics physics interface and a Schrödinger Equation physics
interface, to model charge carriers in quantum-confined systems. The electric potential
from the Electrostatics contributes to the potential energy term in the Schrödinger
Equation. A statistically weighted sum of the probability densities from the eigenstates
of the Schrödinger Equation contributes to the space charge density in the
Electrostatics. All spatial dimensions (1D, 1D axial symmetry, 2D, 2D axial symmetry,
and 3D) are supported. A dedicated Schrödinger-Poisson study type is available for
automated generation of self-consistent iterations in the solver sequence.

When a predefined Schrödinger-Poisson Equation multiphysics interface is added from

the Semiconductor branch ( ) of the Model Wizard or Add Physics windows,
Schrödinger Equation and Electrostatics interfaces are added to the Model Builder.

In addition, a Multiphysics Couplings node is added, which automatically includes the

multiphysics coupling feature Schrödinger-Poisson Coupling.

On the constituent physics interfaces

The Schrödinger Equation interface solves the Schrödinger equation for general
quantum mechanical problems, as well as for the electron and hole wave functions in
quantum-confined semiconductors under the assumption of the envelope function
approximation.

The Electrostatics interface is used to compute the electric field, the electric
displacement field and potential distributions in dielectrics under conditions where the
electric charge distribution is explicitly prescribed. The formulation is stationary but
for use together with other physics, also eigenfrequency, frequency-domain,
small-signal analysis and time-domain modeling are supported in all space dimensions.
The physics interface solves Gauss' Law for the electric field using the scalar electric
potential as the dependent variable.

COUPLING FEATURES
If a multiphysics interface is added from the Model Wizard or using the Add Physics
window, then the appropriate coupling feature is automatically added to the model

THE SCHRÖDINGER-POISSON EQUATION MULTIPHYSICS INTERFACE | 299

 builder tree structure. Alternatively, the physics interfaces can be added one at a time,
followed by the coupling features.

For example, if single Electrostatics and Schrödinger Equation interfaces are added, then
COMSOL Multiphysics adds an empty Multiphysics Couplings node. You can then
choose the coupling feature: Schrödinger-Poisson Coupling.

Coupling features are available from the context menu (right-click the
Multiphysics Couplings node) or from the Physics toolbar, Multiphysics
menu.

PHYSICS INTERFACES AND COUPLING FEATURES

Use the online help in COMSOL Multiphysics to locate and search all the
documentation. All these links also work directly in COMSOL
Multiphysics when using the Help system.

Coupling Features
The Schrödinger-Poisson Coupling node is described in the section
Schrödinger-Poisson Coupling below.

Physics Interface Features

Physics nodes are available from the Physics ribbon toolbar (Windows users), Physics
context menu (Mac or Linux users), or right-click to access the context menu (all
users).

In general, to add a node, go to the Physics toolbar, no matter what

operating system you are using. Subnodes are available by clicking the
parent node and selecting it from the Attributes menu.

The available physics features for The Schrödinger Equation Interface are listed in the
section Domain and Boundary Nodes for the Schrödinger Equation Interface.

For theory and physics interface feature descriptions relating to the Electrostatics
interface, see the COMSOL Multiphysics Reference Manual.

300 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

Schrödinger-Poisson Coupling
The Schrödinger-Poisson Coupling ( ) node takes user inputs to generate the variables
and equations necessary for the bidirectional coupling between the Electrostatics
physics interface and the Schrödinger Equation physics interface, to model charged
particles in quantum-confined systems.

SETTINGS
The Label is the default multiphysics coupling feature name.

The Name is used primarily as a scope prefix for variables defined by the coupling node.
The software refer to variables in expressions using the pattern
<name>.<variable_name>. In order to distinguish between variables belonging to
different coupling nodes or physics interfaces, the name string must be unique. Only
letters, numbers, and underscores (_) are permitted in the Name field. The first
character must be a letter.

The default Name (for the first multiphysics coupling feature in the model) is schrp1.

DOMAIN SELECTION
The default setting is All domains which couples all domains in the selected Schrödinger
Equation physics interface that overlap with the domain selection of the selected
Electrostatics interface.

COUPLED INTERFACES
This section defines the physics involved in the coupling. By default, the software
selects an appropriate physics interface for you from the Schrödinger Equation and
Electrostatics lists. These selections determine the two physics interfaces that are
coupled by this feature.

MODEL INPUT
Enter the temperature T (SI unit: K) of the system to be modeled. The default value
is 293.15[K].

PARTICLE DENSITY COMPUTATION

From the Particle density computation drop-down menu, select Fermi-Dirac statistics,
parabolic band (the default) or User defined.

Fermi-Dirac statistics, parabolic band

This option computes the statistically weighted effective density of states for each
eigenstate, assuming Fermi-Dirac statistics and parabolic band approximation. See the

THE SCHRÖDINGER-POISSON EQUATION MULTIPHYSICS INTERFACE | 301

 Equation section of the user interface for the formula used, which varies according to
the spatial dimension.

Enter the Fermi energy level Ef (SI unit: J), density-of-state effective mass md (SI unit:
kg), and the degeneracy factor gi (dimensionless). The default values are 0[eV],
0.067*me_const, and 1, respectively.

User defined
Directly enter the formula for the statistically weighted effective density of states Ni
(1/m2 in 1D, 1/m in 2D and 1D axial symmetry, dimensionless in 3D and 2D axial
symmetry). The default expression is 0.

CHARGE DENSITY COMPUTATION

Enter the charge number zq (dimensionless) for the particle, for example, -1 for
electrons and 1 for positrons. The default value is -1.

From the Charge density computation list, select Modified Gummel iteration (the default),
Damped iteration, or User defined.

If the Minimization of global variable option is used in the Schrödinger-Poisson study

step settings, enter the formula for the Global error variable (dimensionless). The
default expression is (schrp1.max(abs(V-schrp1.V_old)))/1[V], which
computes the max difference between the electric potential profiles from the two most
recent iterations, scaled by 1 V.

Modified Gummel iteration

This option updates the space charge density contribution from the sum of the
statistically weighted probability density profile for each eigenstate, by making a
prediction based on the high temperature limit (Maxwell-Boltzmann statistics). See
the Equation section of the user interface for the formula used.

An additional tuning parameter α (dimensionless) can be entered to either accelerate

the convergence at low temperatures using positive values, or provide extra damping
using negative values. The default value is 0.

Damped iteration
This option updates the space charge density contribution from the sum of the
statistically weighted probability density profile for each eigenstate, by adding a
fractional contribution to the space charge density from the previous iteration (with
the same fraction removed from the previous charge density). See the Equation section
of the user interface for the formula used.

302 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

The fraction is specified by the damping factor α (dimensionless). The default value is
0.1.

User defined
Directly enter the formula for the space charge density ρv (SI unit: C/m3). The default
expression is 0.

Schrödinger-Poisson Study Step

The dedicated Schrödinger-Poisson study step automatically generates the iterations in
the solver sequence for the self-consistent solution of the fully coupled
Schrödinger-Poisson equation.

To take advantage of the default settings detailed below, use the Model Wizard or the
Add Study button in the ribbon/toolbar of the graphical user interface. A study step
added by right-clicking a Study node does not include any default settings suggested
by the physics.

STUDY SETTINGS
The default option for the Eigenfrequency (Eigenvalue) search method list is Manual,
because for a completely new problem, it is often necessary to use this option to find
the range of the eigenenergies and a rough estimate of the number of eigenstates.
Make sure that the Unit list is left as blank (default).

Once the range and number are found, switch to the Region search option with
appropriate settings for the range and number of eigenvalues, in order to ensure that
all significant eigenstates are found by the solver.

Search region (Region search option)

Make sure the Unit list is left as blank (default).

The real and imaginary parts of the input fields refer to the real and imaginary parts of
the eigenvalue, respectively (not the real and imaginary part of the eigenfrequency).
So, to look for the eigenenergies of bound states, set the input fields for the real parts
to the expected energy ranges, and set the input fields for the imaginary parts to a small
range around zero to capture numerical noise or slightly leaky quasi-bound states.

PHYSICS AND VARIABLES SELECTION

If there are extra domain or boundary conditions used to obtain the initial value for
the fully coupled problem, remember to disable them here. For example, see the
schrodinger_poisson_nanowire tutorial model, where the Thomas-Fermi solution

THE SCHRÖDINGER-POISSON EQUATION MULTIPHYSICS INTERFACE | 303

 is used as the initial condition, and the space charge density contribution from the
Thomas-Fermi approximation is disabled here.

ITERATIONS
The default option for the Termination method drop-down menu is Minimization of
global variable, which updates a table displaying the history of a global error variable
after each iteration during the solution process. This provides a good indication of the
solution process to monitor whether the iteration is converging.

The default expression for the Global variable input field uses the built-in global error
variable schrp1.global_err, which computes the max difference between the
electric potential fields from the two most recent iterations, in the unit of V, as
discussed in the section Charge Density Computation for the Schrödinger-Poisson
Coupling multiphysics node earlier. Note that the prefix for the variable, in this case
schrp1, should match the Name input field of the Schrödinger-Poisson Coupling
multiphysics node. (See section Settings) Setting the Absolute tolerance to 1e-6 thus
means the iteration ends after the max difference is less than 1 uV.

VALUES OF DEPENDENT VARIABLES

Use this section to configure the initial conditions for the study step. Consult the
COMSOL Multiphysics Reference Manual for details.

STUDY EXTENSIONS
Use the Auxiliary sweep option to solve for a set of parameters, for example, in the
schrodinger_poisson_nanowire tutorial model, it is used to solve for a set of
azimuthal quantum numbers. The total particle density will be computed using the
statistically weighted sum of all the eigenstates for all parameters.

304 | CHAPTER 5: SCHRÖDINGER EQUATION INTERFACES

 6

Glossary

This Glossary of Terms contains modeling terms relevant to Semiconductor

modeling. For mathematical terms as well as geometry and CAD terms specific to
the COMSOL Multiphysics® software and documentation, see the glossary in the
COMSOL Multiphysics Reference Manual. For references to more information
about a term, see the index.

305
 Glossary of Terms
acceptor Dopant atoms that, when added to the semiconductor, have the effect of
removing free electrons from the system at sufficiently high temperatures.

Auger recombination Auger recombination becomes important at high

nonequilibrium carrier densities since the process requires three carriers. For example,
when two electrons collide the collision can result in the recombination of one of the
electrons with a hole. The energy released by the transition is transferred to the
surviving electron, which subsequently returns to equilibrium as it undergoes
collisions with the lattice.

avalanche breakdown See impact ionization.

band A collection of states that can be occupied by electrons, spanning a discrete range
of energies.

band bending A shift in the energy levels associated with the band structure as a
function of position in space. Band bending is usually caused by electric fields
originating from space charge layers in the semiconductor.

band gap A range of energies in which no states are available for electrons to occupy.

band gap narrowing A reduction in the band gap of a heavily doped semiconductor
resulting from the merging of the impurity band with the conduction or valence band.

base (transistor) See bipolar transistor.

basis The specific pattern of atoms that is repeated periodically at each point in a
lattice.

bipolar transistor/bipolar junction transistor/BJT A transistor that uses two back to

back p-n junctions to achieve power amplification. These transistors can be of the type
PNP or NPN. There are three terminals in a bipolar transistor: the emitter, the base,
and the collector, each of which is connected to one of the differently doped regions
of the semiconductor (the base is connected to the region between the collector and
emitter). During operation, a small current flowing between the base and the emitter
usually controls a much larger current between the collector and emitter terminals,
either to amplify a signal or to implement a switch.

306 | CHAPTER 6: GLOSSARY

Bloch’s theorem A theorem that describes restrictions on the form of the electron
wave function in a periodic potential.

Bloch wave function/Bloch state A wave function that satisfies Bloch’s theorem.

Boltzmann equation A partial differential equation that, for semiconductors, describes

the evolution of the hole and electron distribution functions.

body (transistor) See field effect transistor.

collector (transistor) See bipolar transistor.

conduction band A band with a range of energies that lie just above the band gap in
which the Fermi level lies.

degenerate semiconductor A semiconductor where Fermi-Dirac statistics are required

to describe the carrier distribution functions.

depletion region A region of the semiconductor, usually at a p-n junction or metal

semiconductor contact, in which the carrier concentration is greatly suppressed at
equilibrium. There is usually a space charge and corresponding electric field associated
with the depletion region.

density of states The number of states per unit energy (or, in some cases, per unit
volume of k-space) that can be occupied by electrons.

diode A two terminal device with much lower resistance when current flows in one
direction (the forward direction) than when it flows in the other direction (the reverse
direction).

direct recombination This recombination process is possible in direct band gap

semiconductors and occurs when electron-hole pairs recombine with the emission of
photons (with an energy close to the band gap).

donors Dopant atoms which, when added to the semiconductor, have the effect of
adding additional free electrons to the system at sufficiently high temperatures.

dopants Impurities that are deliberately added to a semiconductor in order to modify

the conductivity of the material.

drain (transistor) See field effect transistor.

GLOSSARY OF TERMS | 307

 drift-diffusion equations Equations describing the migration and diffusion of electrons
and holes in a semiconductor under the influence of applied fields, temperature
gradients, or concentration gradients.

electron distribution function A function that determines the fraction of available

states that are full as a function of energy.

electron quasi-Fermi level See quasi-Fermi level.

electrons An elementary particle with a negative elementary electric charge.

emitter (transistor) See bipolar transistor.

Fermi-Dirac distribution The general form of the electron (or hole) distribution
function in equilibrium.

Fermi-Dirac statistics Electron or hole statistics that are described by the Fermi-Dirac
distribution.

Fermi level The energy associated with states which, at a specified temperature, have
a 50% probability of being occupied. The Fermi level is the chemical potential for
electrons.

Fermi surface The surface in k-space that corresponds to the Fermi level.

field effect transistor A transistor that uses an electric field to alter the geometry of a
conducting channel adjacent to the surface of a semiconductor. Usually the field is
applied to the surface of a semiconductor in an oppositely doped region (sometimes
connected to the body terminal) between two doped-like regions of semiconductor
(connected to the source and the drain terminals). The field results from a voltage
applied to a terminal referred to as the gate. In a MOSFET, the field produces band
bending near the surface of the semiconductor, resulting in an inversion layer, which
produces a conducting channel between the source and the drain. The extent of the
channel is controlled by the applied field allowing the channel resistance to be
modulated. For a JFET or MESFET, the field is applied to a p-n junction or a Schottky
contact, altering the size of the depletion layer and consequently modifying the
channel thickness.

Gaussian doping A distribution of dopants where the concentration decays from a

peak value following a Gaussian distribution.

308 | CHAPTER 6: GLOSSARY

Generation A process by which electron hole pairs are created.

gate (transistor) See field effect transistor.

heterojunction An interface between two different semiconductors.

hole distribution function A function which determines fraction of available states

which are empty as a function of energy.

hole quasi-Fermi level See quasi-Fermi level.

holes Particles evoked to simplify the computation of current from bands that are
almost completely full of electrons.

impact ionization A carrier generation mechanism that occurs when the carriers are
accelerated by the electric field in between collisions to velocities where the energies
are greater than the gap energy, these can dissipate enough energy during collisions
that additional electron hole pairs can be generated. This process is referred to as
impact ionization and is responsible for the phenomenon of avalanche breakdown.

incomplete ionization For semiconductors with a wide band gap or for conventional
semiconductors at low temperatures, the thermal energy is not sufficient to completely
ionize the dopant atoms and only a fraction of the dopants are ionized. This situation
is referred to as incomplete ionization.

insulator A solid that has no observable conductivity at temperatures below its melting
point. The band structure of insulators is such that there are only completely full or
completely empty bands, and a band gap of sufficient size exists between the full and
empty bands that there is negligible thermal excitation.

inversion layer A layer of semiconductor in which band bending leads to a change of

the material from n-type to p-type or vice versa.

JFET (Junction Field Effect Transistor) A field effect transistor in which the electric
field is applied to the semiconductor by a biased p-n junction.

k-space The space in which the wave vectors of the electron wave function (usually
denoted k), are defined.

GLOSSARY OF TERMS | 309

 lattice A set of points in space that are all identical due to the periodicity of the
structure represented.

Maxwell-Boltzmann distribution The limit of the Fermi-Dirac distribution when the

Fermi energy is several times the thermal energy within the band gap.

Maxwell-Boltzmann statistics Electron or hole statistics that are described by the

Maxwell-Boltzmann distribution.

metal A solid with good electrical conductivity. A metal has one or more partially filled
bands.

MESFET (Metal Semiconductor Field Effect Transistor) A field effect transistor in

which the electric field is applied to the semiconductor by a Schottky contact.

mobility A quantity (strictly a tensor) relating the velocity of electrons or holes to the
applied electric field.

MOSFET (Metal Oxide Semiconductor Field Effect Transistor) A field effect transistor
in which the electric field is applied to the semiconductor by a metal contact (or in
some cases a heavily doped polysilicon) separated from the semiconductor by an
insulator (often silicon oxide).

nearly free electron model A quantum mechanical model for the behavior of electrons
moving in a weakly periodic potential.

nondegenerate semiconductor A semiconductor in which it is sufficient to use

Maxwell-Boltzmann statistics to describe the carrier distribution functions.

n-type A region of semiconductor where the number of donors exceeds the number
of acceptors.

ohmic contact A contact between a metal and a semiconductor that has a linear IV
characteristic.

p-type A region of semiconductor where the number of acceptors exceeds the number
of donors.

p-n junction A junction between p- and n- type regions of a semiconductor.

310 | CHAPTER 6: GLOSSARY

Pauli exclusion principle The Pauli exclusion principle states that two Fermions (such
as electrons) cannot occupy the same quantum mechanical state. A consequence of the
principle is that each state in a crystal can be occupied only by two electrons, one spin
up and another spin down.

quasi-Fermi level Net current flows and/or thermal gradients in a semiconductor

represent nonequilibrium conditions and consequently the concept of the Fermi level
(derived from equilibrium thermodynamics) does not apply to these processes.
However, in many devices it is possible to assume a collective (and different)
quasi-equilibrium for electrons and holes, since the relaxation time for scattering
within the band is much less than that for recombination processes. The electron and
hole quasi-Fermi levels are the Fermi levels associated with these two quasi-equilibrium
populations.

recombination A process by which electrons and holes recombine, resulting in a

return of the system to equilibrium.

rectifier A device with much lower resistance when current flows in one direction (the
forward direction) than when it flows in the other direction (the reverse direction).

reciprocal lattice The lattice obtained by taking the Fourier transform of the lattice in
real space. The reciprocal lattice is a set of vectors in k-space.

Schottky contact A metal semiconductor contact where the barrier height and width
is sufficient to result in rectifying behavior.

Schrödinger equation A partial differential equation that describes the evolution of the
wave function.

semiclassical model A model that describes the dynamics of electrons moving in

periodic solids in between collisions.

semiconductor A solid that is an insulator at zero temperature but that has observable
conductivity at temperatures below its melting point. The conductivity results from a
significant number of electrons that are thermally excited across the band gap;
consequently the gap energy must be of the same order as the thermal energy (most
important semiconductors have a gap energy of less than 2 eV).

Shockley-Read-Hall recombination This recombination mechanism is important in

indirect band gap semiconductors, such as silicon and germanium (for these materials

GLOSSARY OF TERMS | 311

 direct recombination is not possible). A defect (usually with an energy close to the
mid-gap) is involved in the process. An electron or hole is first trapped by the defect
and then emitted into the valence/conduction band, resulting in a reduction in the
number of available carriers. The carrier energy is typically converted to heat.

source (transistor) See field effect transistor.

spin An intrinsic form of angular momentum carried by elementary particles. The

electron spin can take two values, +1/2 and -1/2, frequently referred to as up and
down, respectively.

transistor A device used to amplify or switch electrical signals.

valence band A band with a range of energies that lie just below the band gap in which
the Fermi-level lies.

wave function A complex-valued function that describes the quantum state of a

particle. The modulus squared of the wave function represents the probability density
of finding a given particle in a given region of space at a particular time.

312 | CHAPTER 6: GLOSSARY

I n d e x
A acceptors 144 172
acoustic phonon scattering 163 change cross-section (node) 253
adiabatic approximation 114 change thickness (out-of-plane) (node)
ammeter (node) 259 254
analytic doping model (node) 88 charge conservation (node) 238
analytic trap density (node) 93 charge conservation, piezoelectric
Application Libraries window 19 (node) 254
application library examples charge relaxation theory 273
electrical circuits 42 circuit import, SPICE 271
electrostatics 235 collector node 265
Lombardi surface mobility 86 common settings 15
Arora mobility model (LI) (node) 87 conduction band 130
theory 170 conduction loss (time-harmonic) (node)
Auger recombination (node) 102 239
Auger recombination theory 180 continuity/heterojunction (node) 76
auxiliary sweep settings 54 continuous energy level (node)

B band gaps 126 boundaries 111

base node 265 domains 100

bipolar junction transistor 265 heterointerfaces 112

bipolar transistor 12 coupling, to the electrical circuits inter-

Bloch wave function 121 face 43

Bloch’s theorem 121 crystal momentum 138

boundary nodes current sources, theory 275

electrostatics 236 current-controlled current source

boundary selection for doping profile (node) 264

(node) 93 current-controlled voltage source

boundary selection for trap density pro- (node) 263

file (node) 96 D degenerate semiconductors 25

Brillouin zone 123, 125 density-gradient formulation 27, 30, 58,
bulk node 266 62–63, 66, 76, 110, 113–114, 153–154,

C capacitor (node) 259 163, 167, 177, 179, 182–184, 219–220

carrier mobility 31 depletion region 155

Caughey-Thomas mobility model (E) device models, electrical circuits 279

(node) 83 dielectric shielding (node) 245

Caughey-Thomas mobility model, theory diode (node) 268

diode transistor model 285

INDEX| 313
 Dirac comb 118 extended band scheme 126
direct recombination (node) 103 external I vs. U (node) 268
direct recombination, theory 180 external I-terminal (node) 270
discrete energy level (node) 99 external surface charge accumulation
boundaries 110 (node) 243
heterointerfaces 111 external U vs. I (node) 269
discretization 27
F Fermi surface 117
distributed capacitance (node) 250
Fermi-Dirac function 164
documentation 18
Fermi-Dirac integral 152
domain nodes
Fermi-Dirac statistics 25
electrostatics 236
Fermi-energy 130
donors 144
finite element 27
doping distribution 33
finite volume 27
drain node 266
Fletcher mobility model (C) (node) 81
drift-diffusion equations 13, 165
Fletcher mobility model, theory 171
E edge nodes floating gate (node) 112
electrostatics 236 floating potential (node)
effective mass (node) 292 electrostatics 248
E-k relationship 125 Fowler-Nordheim tunneling 221
electric displacement field (node) 243 frequency domain study
electric fields theory 273 theory, electric currents 275
electric potential (node) 242 frequency domain, perturbation study
electrical circuit interface 256 step 50
theory 278
G gate node 266
electrical circuits
Gauss’ law and charge relaxation theory
modeling techniques 42
273
electron mobility models 31
Gauss’ law equation 276
electron potential energy (node) 292
geometric doping model (node) 92
electrostatic point dipole (node) 255
geometric trap density (node) 95
electrostatics interface 234
grid independence 52
theory 273
ground (node) 241
emailing COMSOL 20
ground node (node) 258
emitter node 265
H harmonic perturbation, modeling 50
energy bands 126
Heisenberg uncertainty principle 120,
equation of continuity 273
139
error message, electrical circuits 42
hole mobility models 31
exporting
SPICE netlists 49 I impact ionization 180

314 | I N D E X
 impact ionization generation (node) 104 metal contact (node) 106
importing minority carrier concentration 25
SPICE netlists 48, 271 MPH-files 19
indirect optical transitions (node) 74 multiphysics 227, 230
inductor (node) 260 mutual inductance (node) 267
initial conditions 54
N n-Channel MOS transistor 266, 282
initial values (node)
n-Channel MOSFET (node) 266
electrostatics 240
netlists, SPICE 48, 271
Schrodinger equation interface 294
nodes, common settings 15
semiconductor interface 79
nondegenerate limit 153
insulation (node) 75
nondegenerate semiconductors 24
insulator interface (node) 113
NPN bipolar junction transistor 279
internet resources 18
NPN BJT (node) 265
ionized impurity scattering 163
O Ohm’s law and charge relaxation theory
K Kirchhoff’s circuit laws 278
273
knowledge base, COMSOL 21
one electron wave function 114
k-space 120
open boundary (node) 294
L lattice temperature 24 optical transitions (node) 70
lattice vectors 118
P pair nodes
line charge (node) 250
electrostatics 236
line charge (on axis) (node) 251
Pauli exclusion principle 116
line charge (out-of-plane) (node) 251
Pauli repulsion effect 129
Liouville’s theorem 143
p-Channel MOS transistor 266
Lombardi surface mobility model (S)
p-Channel MOSFET (node) 266
(node) 85
periodic condition (node) 244
Lombardi surface mobility model, theory
Schrodinger equation interface 294
172
phonons 31
M majority carrier devices 25 physics interfaces, common settings 15
many body effects 129 PNP BJT (node) 265
many body Schrodinger equation 114 point charge (node) 252
many electron wavefunction 114 point charge (on axis) (node) 253
mass action law 179 point nodes
Matthiessen’s rule 169 electrostatics 236
Maxwell’s equations power law mobility model (L) (node) 80
electrical circuits and 278 theory 170
Maxwell-Boltzmann distribution 24 predefined couplings, electrical circuits
meshing strategies 52 43

INDEX| 315
 Q quasi-Fermi level formulation 27, 30, 62– subcircuit definition (node) 264
63, 79, 154, 163, 167 subcircuit instance (node) 265
quasi-Fermi levels 159 surface charge density (node) 242

R reciprocal lattice 119 T Taylor series 130

reciprocal space 120 technical support, COMSOL 20
resistor (node) 259 terminal (node)

S Schottky contacts 211 electrostatics 246

Schrödinger equation 115, 134, 297 theory

time-independent 121 electric fields 273

Schrödinger equation interface 290 electrical circuit interface 278

Schrödinger-Poisson Equation multi- electrostatics interface 273

physics interface 299 semiconductor interface 114

semiconductor device models 279 thin insulator gate (node) 108

Semiconductor Equilibrium study 55, 66, thin low permittivity gap (node) 245

153, 210, 213 time dependent study 275

semiconductor interface 60 time-dependent models 55

theory 114 time-independent Schrödinger equation

semiconductor material model (node) 65 121

semiconductor optoelectronics, beam transformer (node) 267

envelopes interface 227 transport properties 142

semiconductor optoelectronics, fre- trap assisted recombination (node) 97

quency domain interface 230 Trap-Assisted Heterointerface Recom-

semiconductor-electromagnetic waves bination 111

coupling (node) 229, 301 Trap-Assisted Surface Recombination

Shockley-Read-Hall recombination 180 110

silicon band structure 132 triangular mesh 52

small-signal analysis, frequency domain U user-defined generation (node) 105

study 50 user-defined mobility model (node) 79
solver tolerance 56 user-defined mobility model, theory 173
Sommerfeld model 115, 117 user-defined recombination (node) 105
source node 266
V valence band 130
space charge density (node) 240
voltage source (node) 260
SPICE
voltage-controlled current source
exporting 49
(node) 263
SPICE netlists 48, 271
voltage-controlled voltage source (node)
standard settings 15
262
stationary models 54
voltmeter (node) 258
stationary, study step 50

316 | I N D E X
W Wannier function 131
websites, COMSOL 20

Z zero charge (node) 241

zero flux (node) 294
zero probability (node) 294

INDEX| 317
318 | I N D E X