Particle Tracing Module

User’s Guide
Particle Tracing Module User’s Guide
© 1998–2017 COMSOL
Protected by U.S. 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; and 9,454,625. 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, Capture the Concept, COMSOL Desktop,
LiveLink, and COMSOL Server 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 5.3

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: CM022701

C o n t e n t s

Chapter 1: Introduction

About the Particle Tracing Module 10

What Can the Particle Tracing Module Do? . . . . . . . . . . . . . 10
Particle Tracing Plots, Data Sets, Derived Values, and Studies . . . . . . 11
The env and bndenv Operators . . . . . . . . . . . . . . . . . 12
The Particle Tracing Module Physics Interface Guide . . . . . . . . . 13
Common Physics Interface and Feature Settings and Nodes. . . . . . . 14
Where Do I Access the Documentation and Application Libraries? . . . . 15

Overview of the User’s Guide 18

C h a p t e r 2 : Pa r t i c l e Tr a c i n g M o d e l i n g

Particle Tracing 20
Introduction to Particle Tracing Modeling . . . . . . . . . . . . . 20
Charged Particle Tracing . . . . . . . . . . . . . . . . . . . . 21
Particle Tracing for Fluid Flow . . . . . . . . . . . . . . . . . . 23
Mathematical Particle Tracing . . . . . . . . . . . . . . . . . . 30
Multiphysics Interfaces . . . . . . . . . . . . . . . . . . . . . 30

Modeling Tools 32
Choosing a Formulation . . . . . . . . . . . . . . . . . . . . 32
Geometry and Mesh Settings . . . . . . . . . . . . . . . . . . 36
Special Variables . . . . . . . . . . . . . . . . . . . . . . . 38
Component Couplings. . . . . . . . . . . . . . . . . . . . . 42
Monte Carlo Modeling. . . . . . . . . . . . . . . . . . . . . 44
Study Setup . . . . . . . . . . . . . . . . . . . . . . . . . 46
Auxiliary Dependent Variables and Residence Time . . . . . . . . . . 48
Initialization of Auxiliary Dependent Variables . . . . . . . . . . . . 50
Accumulators . . . . . . . . . . . . . . . . . . . . . . . . 51
Overriding Particle Properties . . . . . . . . . . . . . . . . . . 52

CONTENTS |3
 Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Particle Counters . . . . . . . . . . . . . . . . . . . . . . 54
Improving Plot Quality . . . . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . 57

C h a p t e r 3 : M a t h e m a t i c a l Pa r t i c l e Tr a c i n g

The Mathematical Particle Tracing Interface 60

Domain, Boundary, Pair, and Global Nodes for the Mathematical
Particle Tracing Interface . . . . . . . . . . . . . . . . . . 65
Wall. . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Periodic Condition . . . . . . . . . . . . . . . . . . . . . . 70
Accumulator (Boundary) . . . . . . . . . . . . . . . . . . . . 71
Particle Counter . . . . . . . . . . . . . . . . . . . . . . . 73
Secondary Emission . . . . . . . . . . . . . . . . . . . . . . 73
Particle Properties . . . . . . . . . . . . . . . . . . . . . . 77
Force . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Rotating Frame . . . . . . . . . . . . . . . . . . . . . . . 80
Velocity Reinitialization . . . . . . . . . . . . . . . . . . . . 81
Accumulator (Domain) . . . . . . . . . . . . . . . . . . . . 82
Particle-Particle Interaction . . . . . . . . . . . . . . . . . . . 84
Release . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Axial Symmetry . . . . . . . . . . . . . . . . . . . . . . . 93
Inlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Nonlocal Accumulator. . . . . . . . . . . . . . . . . . . . . 98
Outlet . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Particle Continuity . . . . . . . . . . . . . . . . . . . . . 100
Auxiliary Dependent Variable . . . . . . . . . . . . . . . . . 100
Release from Edge . . . . . . . . . . . . . . . . . . . . . 101
Release from Point . . . . . . . . . . . . . . . . . . . . . 101
Release from Grid . . . . . . . . . . . . . . . . . . . . . 102
Override Properties . . . . . . . . . . . . . . . . . . . . 103
Release from Data File. . . . . . . . . . . . . . . . . . . . 103

Theory for the Mathematical Particle Tracing Interface 106

Newtonian Formulation . . . . . . . . . . . . . . . . . . . 106

4 | CONTENTS
 Hamiltonian Formulation . . . . . . . . . . . . . . . . . . . 108
Lagrangian Formulation . . . . . . . . . . . . . . . . . . . 108
Massless Formulation . . . . . . . . . . . . . . . . . . . . 109
Particle Tracing in Rotating Frames . . . . . . . . . . . . . . . 109
Initial Conditions — Position . . . . . . . . . . . . . . . . . 110
Initial Conditions — Velocity . . . . . . . . . . . . . . . . . 111
Particle-Particle Interactions . . . . . . . . . . . . . . . . . 113
Auxiliary Dependent Variables . . . . . . . . . . . . . . . . . 116
About the Boundary Conditions for the Particle Tracing Interfaces . . . 116
Accumulator Theory: Domains . . . . . . . . . . . . . . . . 119
Accumulator Theory: Boundaries . . . . . . . . . . . . . . . 120
References for the Mathematical Particle Tracing Interface . . . . . . 122

C h a p t e r 4 : C h a rg e d Pa r t i c l e Tr a c i n g

The Charged Particle Tracing Interface 124

Domain, Boundary, Pair, and Global Nodes for the Charged Particle
Tracing Interface. . . . . . . . . . . . . . . . . . . . . 127
Collisions. . . . . . . . . . . . . . . . . . . . . . . . . 128
Elastic . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Excitation . . . . . . . . . . . . . . . . . . . . . . . . 132
Attachment . . . . . . . . . . . . . . . . . . . . . . . . 132
Ionization. . . . . . . . . . . . . . . . . . . . . . . . . 133
Resonant Charge Exchange . . . . . . . . . . . . . . . . . . 134
Nonresonant Charge Exchange . . . . . . . . . . . . . . . . 135
User Defined . . . . . . . . . . . . . . . . . . . . . . . 136
Friction Force . . . . . . . . . . . . . . . . . . . . . . . 137
Particle-Matter Interactions . . . . . . . . . . . . . . . . . . 138
Ionization Loss. . . . . . . . . . . . . . . . . . . . . . . 139
Nuclear Stopping . . . . . . . . . . . . . . . . . . . . . . 140
Particle Beam . . . . . . . . . . . . . . . . . . . . . . . 141
Thermionic Emission . . . . . . . . . . . . . . . . . . . . 151
Particle Properties . . . . . . . . . . . . . . . . . . . . . 153
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 154
Electric Force . . . . . . . . . . . . . . . . . . . . . . . 154
Magnetic Force . . . . . . . . . . . . . . . . . . . . . . 156

CONTENTS |5
 Space Charge Density Calculation . . . . . . . . . . . . . . . 158
Surface Charge Density . . . . . . . . . . . . . . . . . . . 159
Current Density . . . . . . . . . . . . . . . . . . . . . . 159
Heat Source. . . . . . . . . . . . . . . . . . . . . . . . 160
Etch . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Theory for the Charged Particle Tracing Interface 162

Introduction to the Charged Particle Tracing Interface Theory . . . . 162
Electric Force Theory . . . . . . . . . . . . . . . . . . . . 164
Magnetic Force Theory . . . . . . . . . . . . . . . . . . . 164
Collisional Force Theory . . . . . . . . . . . . . . . . . . . 165
Particle-Matter Interaction Theory . . . . . . . . . . . . . . . 169
Particle Beam Theory . . . . . . . . . . . . . . . . . . . . 172
Thermionic Emission Theory . . . . . . . . . . . . . . . . . 178
Surface Charge Density Theory . . . . . . . . . . . . . . . . 183
Current Density Theory . . . . . . . . . . . . . . . . . . . 183
Heat Source Theory . . . . . . . . . . . . . . . . . . . . 184
Etch Theory. . . . . . . . . . . . . . . . . . . . . . . . 185
References for the Charged Particle Tracing Interface . . . . . . . . 186

C h a p t e r 5 : Pa r t i c l e Tr a c i n g f o r F l u i d F l ow

The Particle Tracing for Fluid Flow Interface 188

Domain, Boundary, Pair, and Global Nodes for the Particle Tracing
for Fluid Flow Interface . . . . . . . . . . . . . . . . . . 192
Particle Properties . . . . . . . . . . . . . . . . . . . . . 194
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 195
Drag Force . . . . . . . . . . . . . . . . . . . . . . . . 195
Lift Force . . . . . . . . . . . . . . . . . . . . . . . . . 199
Brownian Force . . . . . . . . . . . . . . . . . . . . . . 200
Gravity Force . . . . . . . . . . . . . . . . . . . . . . . 202
Acoustophoretic Force . . . . . . . . . . . . . . . . . . . 203
Electric Force . . . . . . . . . . . . . . . . . . . . . . . 204
Magnetic Force . . . . . . . . . . . . . . . . . . . . . . 205
Dielectrophoretic Force . . . . . . . . . . . . . . . . . . . 205
Shell. . . . . . . . . . . . . . . . . . . . . . . . . . . 207

6 | CONTENTS
 Magnetophoretic Force . . . . . . . . . . . . . . . . . . . 208
Thermophoretic Force . . . . . . . . . . . . . . . . . . . 209
Erosion . . . . . . . . . . . . . . . . . . . . . . . . . 211
Mass Deposition . . . . . . . . . . . . . . . . . . . . . . 212
Boundary Load . . . . . . . . . . . . . . . . . . . . . . 213
Mass Flux. . . . . . . . . . . . . . . . . . . . . . . . . 213
Volume Force Calculation . . . . . . . . . . . . . . . . . . 213
Convective Heat Losses . . . . . . . . . . . . . . . . . . . 215
Radiative Heat Losses . . . . . . . . . . . . . . . . . . . . 215
Heat Source. . . . . . . . . . . . . . . . . . . . . . . . 216
Droplet Breakup . . . . . . . . . . . . . . . . . . . . . . 216
Kelvin-Helmholtz Breakup Model. . . . . . . . . . . . . . . . 217
Rayleigh-Taylor Breakup Model . . . . . . . . . . . . . . . . 218
Nozzle. . . . . . . . . . . . . . . . . . . . . . . . . . 218
Nozzle Domain . . . . . . . . . . . . . . . . . . . . . . 221

The Droplet Sprays in Fluid Flow Interface 222

Theory for the Particle Tracing for Fluid Flow Interface 223
Particle Motion in a Fluid . . . . . . . . . . . . . . . . . . . 223
Particle Motion in a Turbulent Flow . . . . . . . . . . . . . . . 230
Particle Motion in a Shear Flow . . . . . . . . . . . . . . . . 236
Additional Forces. . . . . . . . . . . . . . . . . . . . . . 240
Erosion Theory . . . . . . . . . . . . . . . . . . . . . . 249
Droplet Breakup Theory . . . . . . . . . . . . . . . . . . . 251
Nozzle Theory . . . . . . . . . . . . . . . . . . . . . . 255
Computing Particle Temperature . . . . . . . . . . . . . . . . 257
Computing Particle Mass . . . . . . . . . . . . . . . . . . . 259
References for the Particle Tracing for Fluid Flow Interface . . . . . . 259

Chapter 6: Multiphysics Interfaces

The Particle Field Interaction, Non-Relativistic Interface 264

Electric Particle Field Interaction . . . . . . . . . . . . . . . . 266
Space Charge Limited Emission . . . . . . . . . . . . . . . . 268

CONTENTS |7
 Theory for the Particle Field Interaction, Non-Relativistic
Interface 270
Electrostatics and Charged Particle Tracing Equations . . . . . . . . 270
Space Charge Density Calculation . . . . . . . . . . . . . . . 271
Stabilization of the Space Charge Density Calculation . . . . . . . . 275
Theory for the Space Charge Limited Emission Node . . . . . . . . 276

The Particle Field Interaction, Relativistic Interface 280

Magnetic Particle Field Interaction . . . . . . . . . . . . . . . 283

Theory for the Magnetic Particle Field Interaction, Relativistic

Interface 285
Electrostatics and Charged Particle Tracing Equations . . . . . . . . 285
Current Density Calculation . . . . . . . . . . . . . . . . . 287

The Fluid-Particle Interaction Interface 291

Fluid-Particle Interaction . . . . . . . . . . . . . . . . . . . 293

Theory for the Fluid-Particle Interaction Interface 295

Laminar Flow and Particle Tracing Equations . . . . . . . . . . . 295
Volume Force Calculation . . . . . . . . . . . . . . . . . . 296

Chapter 7: Glossary

Glossary of Terms 302

8 | CONTENTS
 1

Introduction

This guide describes the Particle Tracing Module, an optional add-on package for
COMSOL Multiphysics® designed to compute particle trajectories. The particles
can interact with boundaries and their motion can be affected by external fields,
which may be user-defined or solved for by other physics interfaces.

This chapter introduces you to the capabilities of this module. A summary of the
physics interfaces and where you can find documentation and model examples is
also included. The last section is a brief overview with links to each chapter in this
guide.

In this chapter:

• About the Particle Tracing Module

• Overview of the User’s Guide

9
 About the Particle Tracing Module
These topics are included in this section:

• What Can the Particle Tracing Module Do?

• Particle Tracing Plots, Data Sets, Derived Values, and Studies
• The env and bndenv Operators
• The Particle Tracing Module Physics Interface Guide
• Where Do I Access the Documentation and Application Libraries?

The Physics Interfaces and Building a COMSOL Multiphysics Model in

the COMSOL Multiphysics Reference Manual

What Can the Particle Tracing Module Do?

The Particle Tracing Module is a general purpose, flexible tool that is used to compute
the trajectories of particles. The particles can be affected by various forces,
particle-boundary interactions, and particle-domain interactions. One-way and
two-way couplings between particles and external fields are also supported.

A custom physics interface is available for modeling the motion of electrons and ions
in electromagnetic fields. This makes it possible to model devices such as magnetic
lenses, electron guns, and mass spectrometers. There are a number of tools available to
extract typical quantities of interest and to make plots of the particle trajectories,
including phase portraits and Poincaré maps. Built-in multiphysics interfaces are
available to model the interaction of particles with electric and magnetic fields.
Dedicated features are available to perform Monte Carlo simulations of the collisions
between ions or electrons and the molecules in a rarefied gas.

There is also a dedicated physics interface for computing particle trajectories in a fluid
system. There is a wide variety of predefined forces, including drag, gravity, electric,
and Brownian force. Fluid-based particle tracing allows for detailed investigation of
mixing, separation, and filtering devices. Furthermore, built-in features are available
for modeling complicated physical processes such as thermophoresis,
dielectrophoresis, magnetophoresis, and acoustophoresis.

10 | CHAPTER 1: INTRODUCTION
A mathematical particle tracing interface provides complete freedom and flexibility to
define the equations of motion governing particle trajectories. In addition to built-in
models to describe Newton’s law of motion, there are dedicated formulations available
to specify the Lagrangian or Hamiltonian that dictates the motion of the particles.

All of the physics interfaces provide many different ways of releasing particles. Particles
can be released from geometric entities of any level, including domains, boundaries,
edges, and points. They can also be released by specifying the initial coordinates
directly or by importing sets of initial coordinates from a file. Particles can also be
released at multiple different times. There are many options available for describing
how the particles interact with boundaries, including diffuse and specular reflection,
sticking probabilities, and secondary particle emission.

The physics interfaces can all be used with COMSOL Multiphysics, but when tracing
charged particles the AC/DC Module is often beneficial because it can be used to
model complex alternating or direct current systems and includes more options for
computing magnetic fields. If the particles move through a rarefied gas, as is often the
case for ion and electron beams in vacuum systems, the Molecular Flow Module is
useful because it provides built-in tools to compute the density of extremely rarefied
gases. Similarly, when tracing particles in a fluid system the CFD Module or
Microfluidics Module is often useful due to the multitude of advanced fluid flow
features available in those modules.

Particle Tracing Plots, Data Sets, Derived Values, and Studies

The Particle Tracing Module uses some operators, studies, and postprocessing features
that are only available with this module. The descriptions for these features are,
however, described with the generic plot, data set, and study types. Go to the links
below for more information.

In the COMSOL Multiphysics Reference Manual:

• Particle Trajectories (study)

• Bidirectionally Coupled Particle Tracing (study)
• Particle (Data Set) (data set)
• Particle Trajectories (plot) and Filter for Particle Trajectories
• Phase Portrait (plot)
• Poincaré Map (plot)
• Particle Evaluation

A B O U T T H E P A R T I C L E TR A C I N G M O D U L E | 11
 The links to the nodes described in the COMSOL Multiphysics Reference
Manual do not work in the PDF, only from the on line help in COMSOL
Multiphysics.

For an example of:

• Particle Trajectories see Ion Cyclotron Motion: Application Library

path: Particle_Tracing_Module/Charge_Particle_Tracing/
ion_cyclotron_motion.
• Poincaré Map see Rössler Attractor: Application Library path
Particle_Tracing_Module/Tutorials/rossler_attractor.
• Phase Portrait see Ideal Cloak: Application Library path
Particle_Tracing_Module/Ray_Tracing/ideal_cloak.

The env and bndenv Operators

The Particle Tracing Module has these operators that are only available
with this module. See Built-In Operators in the COMSOL Multiphysics
Reference Manual for additional information.

• Evaluating env(expr) on a particle, evaluates the expression expr at the point in the
domain where the particle is. When evaluating a variable var on a particle, if the
variable is not defined on the particle it is automatically replaced by env(var).
Therefore the env operator can often be omitted. If env(var) is evaluated when a
particle is on a boundary, and var is defined on two domains adjacent to the
boundary, the operator returns the arithmetic mean of var on both sides of the
boundary.
• Evaluating bndenv(expr) on a particle, evaluates expr at the point on the boundary
where the particle is. If the particle is not on a boundary, the evaluation fails. Use
this operator instead of env when evaluating expressions that are only defined on
boundaries.
• The operators env_in(expr) and env_out(expr) can only be used when a particle
is in contact with a boundary. These operators evaluate the expression expr in the
domain the particle reaches the boundary from and the domain on the opposite side

12 | CHAPTER 1: INTRODUCTION
 of the boundary, respectively. These operators are often used to evaluate an
expression that changes discontinuously at the boundary.
• The operators env_rel(expr) and bndenv_rel(expr) are analogous to
env(expr) and bndenv(expr), respectively, except that they always evaluate
expressions at the initial particle positions.

The Particle Tracing Module Physics Interface Guide

The Particle Tracing 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 Particle Tracing Module are listed in the table. The functionality of
the COMSOL Multiphysics base package is given in the COMSOL Multiphysics
Reference Manual.

In the COMSOL Multiphysics Reference Manual:

• Introduction to Solvers and Studies

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

PHYSICS INTERFACE ICON TAG SPACE AVAILABLE PRESET STUDY

DIMENSION TYPE

AC/DC

Particle Tracing

Particle Field Interaction, — 3D, 2D, 2D bidirectionally coupled

Non-Relativistic axisymmetric particle tracing; time
dependent
Particle Field Interaction, — 3D, 2D, 2D bidirectionally coupled
Relativistic (requires the axisymmetric particle tracing; time
AC/DC Module and dependent
Particle Tracing Module)

A B O U T T H E P A R T I C L E TR A C I N G M O D U L E | 13
 PHYSICS INTERFACE ICON TAG SPACE AVAILABLE PRESET STUDY
DIMENSION TYPE

Charged Particle Tracing cpt 3D, 2D, 2D bidirectionally coupled

axisymmetric particle tracing; time
dependent
Fluid Flow

Particle Tracing

Particle Tracing for Fluid fpt 3D, 2D, 2D bidirectionally coupled

Flow axisymmetric particle tracing; time
dependent
Fluid-Particle Interaction — 3D, 2D, 2D bidirectionally coupled
axisymmetric particle tracing; time
dependent
Droplet Sprays in Fluid — 3D, 2D, 2D bidirectionally coupled
Flow axisymmetric particle tracing; time
dependent
Mathematics

Mathematical Particle pt 3D, 2D, 2D bidirectionally coupled

Tracing axisymmetric particle tracing; time
dependent

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.

In each module’s documentation, 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 2-3 for

links to common sections and Table 2-4 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.

14 | CHAPTER 1: INTRODUCTION
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 (as long as you have a license), 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 many of the features on the GUI. 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.

A B O U T T H E P A R T I C L E TR A C I N G M O D U L E | 15
 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 application includes documentation with the theoretical background and
step-by-step instructions to create a model application. The applications are available
in COMSOL as MPH-files that you can open for further investigation. You can use the
step-by-step instructions and the actual applications as a template for your own
modeling and applications. In most models, SI units are used to describe the relevant
properties, parameters, and dimensions in most examples, 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 application and its
properties, including options to open it or a PDF document.

The Application Libraries Window in the COMSOL Multiphysics

Reference Manual.

16 | CHAPTER 1: INTRODUCTION
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 info@comsol.com.

To receive technical support from COMSOL for the COMSOL products, please
contact your local COMSOL representative or send your questions to
support@comsol.com. An automatic notification and a case number are sent to you by
email.

COMSOL ONLINE RESOURCES

COMSOL website www.comsol.com

Contact COMSOL www.comsol.com/contact
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

A B O U T T H E P A R T I C L E TR A C I N G M O D U L E | 17
 Overview of the User’s Guide
The Particle Tracing Module User’s Guide gets you started with modeling using
COMSOL Multiphysics. The information in this guide is specific to this module.
Instructions how to use COMSOL in general are included with 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.

TA B L E O F C O N T E N T S , G L O S S A R Y, A N D I N D E X
To help you navigate through this guide, see the Contents, Glossary of Terms, and
Index.

P A R T I C L E TR A C I N G M O D E L I N G
In Particle Tracing Modeling, the different particle tracing interfaces are discussed
(Charged Particle Tracing, Particle Tracing in Fluid Flow, and Mathematical Particle
Tracing). In the Modeling Tools section, special variables, Monte Carlo modeling,
filters, and auxiliary dependent variables and residence time are discussed.

M A T H E M A T I C A L PA R T I C L E TR A C I N G
The Mathematical Particle Tracing Interface and its underlying theory are described in
this chapter.

C H A R G E D P A R T I C L E TR A C I N G
The Charged Particle Tracing Interface and its underlying theory are described in this
chapter.

P A R T I C L E TR A C I N G F O R F L U I D F L OW
The Particle Tracing for Fluid Flow Interface, The Droplet Sprays in Fluid Flow
Interface, and their underlying theory are described in this chapter.

MULTIPHYSICS INTERFACES
The built-in multiphysics interfaces, including The Particle Field Interaction,
Non-Relativistic Interface, The Particle Field Interaction, Relativistic Interface, and
the The Fluid-Particle Interaction Interface, are described in this chapter.

18 | CHAPTER 1: INTRODUCTION
 2

Particle Tracing Modeling

This chapter gives an overview of the physics interfaces available for modeling the
trajectories of particles and provides guidance on choosing the appropriate physics
interface for a specific problem.

In this chapter:

• Particle Tracing
• Modeling Tools

19
 Particle Tracing
In this section:

• Introduction to Particle Tracing Modeling

• Charged Particle Tracing
• Particle Tracing for Fluid Flow
• Mathematical Particle Tracing

Introduction to Particle Tracing Modeling

Particle tracing provides a Lagrangian description of a problem by solving ordinary
differential equations using Newton’s law of motion. The trajectories of individual
particles are always solved for in the time domain. Newton’s law of motion requires
specification of the particle mass and all forces acting on the particle. The forces can be
divided into two categories, those due to external fields and due to interactions
between particles. Forces due to external fields are typically computed from a finite
element model, using the physics interfaces available in COMSOL Multiphysics and its
other optional add-on modules.

For each particle, a second-order ordinary differential equation is solved for each
component of the position vector. This means that three ordinary differential
equations are solved for each particle in 3D and two in 2D. A first-order formulation
of Newton’s law of motion is also available, in which coupled first-order ordinary
differential equations are solved for the components of the particle position and
velocity. The Mathematical Particle Tracing interface also includes a Hamiltonian
formulation that solves coupled first-order ordinary differential equations for the
particle position and generalized momentum.

At each time step taken by the solver, the forces acting on each particle are queried
from the external fields at the current particle position. If particle-particle interaction
forces are included in the model then they are added to the total force. The particle
position is then updated and the process repeats until the specified end time for the
simulation is reached. During each time step, the particles may interact with
boundaries in the geometry, or they may be subjected to other phenomena that can
discontinuously change the particle velocity.

Because the Particle Tracing Module uses a very general formulation for computing
particle trajectories, the particle tracing interfaces can be used to model charged

20 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
particle motion in electromagnetic fields, large scale planetary and galactic movement,
and particle motion in laminar, turbulent, and multiphase fluid systems.

Charged Particle Tracing

The Charged Particle Tracing Interface is designed to model the motion of electrons,
individual ions, or small ion clusters in electric and magnetic fields.

SPACE CHARGE EFFECTS

A computationally efficient way to approach charged particle tracing modeling is to
first solve for any external electric or magnetic fields, using for example the
Electrostatics interface or the Magnetic Fields interface, and then use these fields to
exert electric and magnetic forces on the ions or electrons via a unidirectional, or
one-way, coupling. In this approach, the effect of the particles on the fields is
neglected. The external fields can be solved for using a Stationary, Frequency Domain,
or Time Dependent study. The particle trajectories are then solved for in a separate
Time Dependent study. The fact that the particle trajectories are computed in their
own study allows efficient and computationally inexpensive iterative solvers to be used.

If the number density of the charged particles is sufficiently large, the particles may
begin to cause significant perturbations in the external fields. There is no specific
magnitude of the number density at which the space charge effects become significant;
rather, as a general rule they should be included if their contributions to the external
electric or magnetic fields is of a comparable order of magnitude to the sources or
boundary conditions included in the other physics, such as surfaces maintained at
specified potentials or external current sources. The bidirectional, or two-way,
coupling between particles and fields can be included in the model by using The
Particle Field Interaction, Non-Relativistic Interface, which automatically adds the
dedicated Electric Particle Field Interaction Multiphysics node to account for space
charge effects.

If, in addition, the particles move at relativistic speeds, the current density due to
particle motion may become significant. The Particle Field Interaction, Relativistic
Interface automatically adds the Electric Particle Field Interaction and Magnetic
Particle Field Interaction Multiphysics nodes to account for the space charge density
and current density of particles, respectively. Magnetic particle-field interactions are
usually negligibly small compared to electric particle-field interactions at
non-relativistic speeds.

P A R T I C L E TR A C I N G | 21
 Electron Beam Divergence Due to Self Potential: Application Library
path Particle_Tracing_Module/Charged_Particle_Tracing/
electron_beam_divergence

Relativistic Paraxial Electron Beam: Application Library path

Particle_Tracing_Module/Charged_Particle_Tracing/
electron_beam_divergence_relativistic

The computational requirements for models that include particle-field interactions

increase significantly over those that neglect them. If the fields and particle trajectories
are directly coupled to each other, both must be computed in the same Time
Dependent study, and a fairly small time step must be taken by the solver to account
for the constantly changing electric potential. In addition, the space charge density and
current density are computed using variables that are constant over each mesh element,
so it may be necessary to refine the mesh or increase the number of model particles to
more accurately model particle-field interactions.

If the fields are stationary, as often occurs when beams of particles are released at
constant current, it is possible to significantly reduce the computational cost of the
model by using a Stationary solver to compute the fields and a Time Dependent solver
to compute the particle trajectories. It is also possible to create a solver loop that
alternates between the Stationary and Time Dependent solvers so that a bidirectional
coupling between the trajectories and fields can be established; a dedicated
Bidirectionally Coupled Particle Tracing study step is available for setting up such a
solver loop. The process of combining these solvers is described in the section Study
Setup.

COULOMB FORCES
If the density of charged particles is extremely high then it can be necessary to include
the Coulomb force that acts between the particles. This is done by adding a
Particle-Particle Interaction node to the model. When particle-particle interactions are
included in a model the computational requirements increase and scale as the number
of particles squared. When including the Coulomb force, it is often best to start with
a small number of particles, solve the model, and then assess whether or not the effect
is important to the model.

22 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
 Coulomb forces and bidirectionally coupled particle-field interactions
usually shouldn’t be included together in the same model because this
causes the force exerted by each charged particle on other charged
particles to be double-counted, once directly and once via the electric
field defined in the modeling domain.

Particle Tracing for Fluid Flow

The Particle Tracing for Fluid Flow Interface is designed for modeling microscopic
and macroscopic particles in a background fluid. There are two phases in the system:
a particle phase consisting of discrete bubbles, particles, droplets, and so forth, and a
continuous phase in which the particles are immersed. In order for the particle tracing
approach to be valid, the fluid system should be a dilute or dispersed flow. This means
that the volume fraction of the particles is much smaller than the volume fraction of
the continuous phase, generally less than 1%. When the volume fraction of the particles
is not small, the fluid system is categorized as a dense flow and a different modeling
approach is required.

It is important to realize that with the particle tracing approach, particles do not
displace the fluid they occupy. In addition, the finite size of the particle is not taken
into account when modeling particle-wall interactions. In other words, for purposes of
detecting particle-boundary interactions, the particles are treated as point masses. The
specification of particle diameter is mostly used for size-dependent forces, such as the
drag and dielectrophoretic forces.

SPARSE FLOW
In a sparse flow, the continuous phase affects the motion of the particles but not
vice-versa. This is often referred to as “one-way coupling.” When modeling such a
system in COMSOL, it is usually most efficient to solve for the continuous phase and
the dispersed phase in separate studies. The fluid usually affects the particle motion
through the Drag Force feature, which defines a drag force based on the fluid
properties, particle properties, and the particle velocity relative to the flow. Several
built-in drag laws can be used, and the optimal drag law usually depends on the size
and speed of the particles. Built-in options to apply random perturbations to account
for turbulence in the fluid are also available.

P A R T I C L E TR A C I N G | 23
 For example, in the following model example, the velocity field is first computed using
a Stationary study, then the particle trajectories are computed using a separate Time
Dependent study.

Particle Trajectories in a Laminar Static Mixer: Application Library

path Particle_Tracing_Module/Fluid_Flow/laminar_mixer_particle

Flow type Required features

Sparse flow Drag force

Increasing Dilute flow Drag force,

particle mass (two-way fluid-particle
and volume coupling) interaction
density

Dispersed Drag force,

flow (four-way fluid-particle
coupling) interaction,
particle-particle
interaction

Dense flow Cannot be solved

with the Particle
Tracing Module

Figure 2-1: The physics features required to model sparse, dilute, and dispersed flows.

DILUTE FLOW
In a dilute flow the continuous phase affects the motion of the particles and the particle
motion in turn disrupts the continuous phase. This is often referred to as a
bidirectional coupling or “two-way coupling.” The bidirectional coupling between
particles and fluids can be modeled using the Fluid-Particle Interaction Multiphysics
node. This node can be added manually if the necessary physics interfaces are already

24 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
present. Alternatively, The Fluid-Particle Interaction Interface can be used to
automatically add the necessary physics interfaces and Multiphysics nodes.

The body force exerted on the fluid by the particle is applied in an approximate way,
in that it is smeared out over a mesh element. This smearing effect makes the volume
force computed by the Fluid-Particle Interaction node somewhat mesh dependent.
When modeling fluid-particle interactions for which the mass flow rate of particles is
not constant, the continuous phase and dispersed phase must be computed
simultaneously in the same study. The computational demand is significantly higher
than in the Sparse Flow case.

If the fields are stationary, as often occurs when particles are released at constant mass
flow rate, it may be possible to compute the particle trajectories using a Time
Dependent solver while computing the fluid flow variables using a Stationary solver. It
is also possible to create a solver loop that alternates between the Stationary and Time
Dependent solvers so that a bidirectional coupling between the trajectories and fields
can be established; a dedicated Bidirectionally Coupled Particle Tracing study step is
available for setting up such a solver loop. The process of combining these solvers is
described in the section Study Setup.

DISPERSED FLOW
In addition to the effects mentioned above, particle-particle interactions also need to
be taken into account. This is often referred to as “four-way coupling.”
Particle-particle interactions can be included in a model by adding a Particle-Particle
Interaction node.

The following limitations apply:

• Hard sphere collisions are not supported. Forces must vary continuously with
respect to the distance between particles as in, for example, the Coulomb force
between charged particles.
• The computation time scales as the number of particles squared. This is because
every particle interacts with every other particle over all distances. Although it is
possible to apply a cut-off radius beyond which particles do not interact with each
other, the interaction force must still be evaluated for every pair of particles.
• If the particle-particle interaction law is highly nonlinear, it can be necessary to use
a very small time step. This is particularly true if the Lennard-Jones option is selected.
• The Particle-Particle Interaction feature produces a Jacobian matrix that is
completely full. For a large number of particles, this is very expensive to factorize.
By default, the Exclude Jacobian contribution for particle-particle interaction check

P A R T I C L E TR A C I N G | 25
 box is selected, which preserves the sparseness in the Jacobian. Clearing this check
box is likely to result in a dramatic increase in the amount of memory and time
needed to solve the problem.

Figure 2-2: The Settings window for Particle-Particle Interaction.

The Particle Tracing for Fluid Flow interface provides a convenient way of computing
the particle mass and temperature. Built-in auxiliary dependent variables for the mass
and temperature can be activated by selecting the Compute particle mass and Compute
particle temperature check boxes, respectively, in the physics interface Additional
Variables section. When the option to compute particle mass is activated, it is possible
to set the initial particle mass in particle release features, such as Release, Inlet, and
Release from Grid. It is also possible to select a distribution function for the initial
mass. This is important when modeling separation devices where the goal is to
understand the transmission probability of particles of various sizes.

26 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
Figure 2-3: Settings window for the Particle Tracing for Fluid Flow interface.

MODELING ADVECTION AND DIFFUSION

Continuum methods such as FEM have one major drawback when it comes to
modeling the advection and diffusion of a particulates in a fluid. The higher the Péclet
number, the more numerically unstable the method becomes. The Péclet number is
the ratio of the rate of advection to the rate of diffusion:

Lu
Pe = -----------
D

where L is the characteristic dimension, u is the advection velocity, and D is the

diffusion coefficient of the particles. In general, continuum methods cannot handle

P A R T I C L E TR A C I N G | 27
 systems where the Péclet number is greater than around 1000. The diffusion
coefficient D (SI unit: m2/s) for spherical particles is, following Einstein’s relation,

kB T
D = ----------------
6πμr p

where

• μ (SI unit: Pa·s) is the fluid dynamic viscosity,

• kB = 1.3806488 × 10-23 J/K is Boltzmann’s constant,
• T (SI unit: K) is the temperature, and
• rp (SI unit: m) is the radius of the particles.

For 100 nm diameter particles in water at room temperature, this results in a diffusion
coefficient of around 4 × 10-12 m2/s. For a microfluidic device with characteristic size
1 mm and velocity of 1 m/s, this results in a Péclet number of 2.5 × 108. Handling
such a large Péclet number with continuum methods is clearly not possible (Ref. 1).

Particle trajectories are computed in a Lagrangian reference frame, removing the

restriction on the Péclet number. The Péclet number can be anything from 0 to infinity
without introducing numerical instabilities. Advection is added to the particles via the
Drag Force. Diffusion is added to particles by adding the Brownian Force. If the
background velocity field is zero then particle motion is purely diffusive (zero Péclet
number). If the Brownian force is neglected and the background velocity is nonzero,
the motion is pure advection (infinite Péclet number).

PARTICLES IN A RAREFIED GAS

Many drag laws, such as the Stokes drag law, are based on the assumption of
continuum flow, in which the particle Knudsen number Kn (dimensionless) is very
small,

λ
Kn = ---- « 1
L

where λ (SI unit: m) is the mean free path of molecules in the surrounding fluid, and
L (SI unit: m) is a characteristic length of the particle, which is often the particle radius
or diameter. The exact definition of the characteristic length may vary depending on
the source being cited and should be considered with caution.

When the particles are extremely small or they are surrounded by a rarefied gas, the
assumption of continuum flow may not be valid. By selecting the Include rarefaction

28 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
effects check box in the physics interface Particle Release and Propagation section, it is
possible to apply correction factors to the Drag Force and Thermophoretic Force,
enabling accurate modeling of particle motion in the slip flow, transitional flow, and
free molecular flow regimes.

Continuum flow

Slip flow
Transitional
flow
Free molecular
flow

Figure 2-4: A plot showing the main fluid flow regimes for rarefied gas flows. Different
regimes are separated by lines of constant Knudsen numbers. The number density of the
gas is normalized to the number density of an ideal gas at a pressure of 1 atmosphere and
a temperature of 0° C (n0).

MODELING MOTION OF LIQUID DROPLETS

Several built-in features are available for modeling the formation, propagation, and
breakup of sprays of liquid droplets.

The Droplet Breakup feature and its two subnodes, Kelvin-Helmholtz Breakup Model
and Rayleigh-Taylor Breakup Model, can be used to model the breakup of liquid
droplets into successively smaller child droplets. The Nozzle feature can be used to
release a spray of droplets at a specified location.

The Droplet Sprays in Fluid Flow Interface functions like The Particle Tracing for
Fluid Flow Interface, with appropriate default features and settings for modeling the
breakup of liquid droplets.

P A R T I C L E TR A C I N G | 29
 Mathematical Particle Tracing
The Mathematical Particle Tracing Interface gives access to the underlying
mathematical formalism on which the Charged Particle Tracing and Particle Tracing
for Fluid Flow interfaces are built. The Mathematical Particle Tracing interface allows
for specification of particle motion in terms of either a Lagrangian or Hamiltonian.
Often it is easier to write down an expression for the Lagrangian or Hamiltonian for
particles rather than deriving the equations of motion. The Hamiltonian formulation
solves for both the particle position and the particle momentum, so the number of
degrees of freedom is doubled when the Hamiltonian formulation is activated.

Multiphysics Interfaces
The Charged Particle Tracing and Particle Tracing for Fluid Flow interfaces can be
used to create unidirectional, or one-way, couplings between particles and fields. For
example, when using the Charged Particle Tracing interface, it is possible to compute
the electric potential using a Stationary study and then use the computed potential to
exert an electric force on the particles.

It is also possible to create bidirectional couplings, in which the particles contribute to

fields in the surrounding domains, which may then exert forces on the particles. A
typical example is a diverging beam of charged particles.

Dedicated Multiphysics nodes are available for several of the most common examples
of bidirectional coupling between particles and fields. The following Multiphysics
nodes are available depending on the other physics interfaces that are present:

• Electric Particle Field Interaction

• Space Charge Limited Emission
• Magnetic Particle Field Interaction
• Fluid-Particle Interaction

In addition, it is possible to add all of the physics interfaces necessary for a specific
Multiphysics node by using the dedicated Multiphysics interfaces:

• The Particle Field Interaction, Non-Relativistic Interface

• The Particle Field Interaction, Relativistic Interface
• The Fluid-Particle Interaction Interface

For example, to model a non-relativistic beam of electrons that diverges due to

self-potential, add the Particle Field Interaction, Non-Relativistic interface. This

30 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
automatically adds instances of the Charged Particle Tracing and Electrostatics physics
interfaces and the Electric Particle Field Interaction Multiphysics node.

P A R T I C L E TR A C I N G | 31
 Modeling Tools
In this section:

• Choosing a Formulation
• Geometry and Mesh Settings
• Special Variables
• Component Couplings
• Monte Carlo Modeling
• Study Setup
• Auxiliary Dependent Variables and Residence Time
• Initialization of Auxiliary Dependent Variables
• Accumulators
• Overriding Particle Properties
• Filters
• Particle Counters
• Improving Plot Quality
• References

See Particle Tracing Plots, Data Sets, Derived Values, and Studies for links
to the COMSOL Multiphysics Reference Manual.

Choosing a Formulation
The formulation of the equations of motion is specified by selecting an option from
the Formulation list in the physics interface Particle Release and Propagation section.

NEWTONIAN
The Newtonian formulation is the default and the most common formulation. It
defines a set of second-order ordinary differential equations for the components of the
particle position based on Newton’s 2nd law of motion,

d  m -------
dq
= F
d t  p dt 

32 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
where q (SI unit: m) is the particle position, mp (SI unit: kg) is the particle mass, and
F (SI unit: N) is the total force on the particles.

NEWTONIAN, FIRST ORDER

The Newtonian, first order formulation is an alternative to the Newtonian formulation.
It defines a set of coupled first-order ordinary differential equations for the
components of the particle position and velocity.

The default time stepping method for the first-order Newtonian formulation is the
Dormand-Prince 5 Runge-Kutta method, an explicit time stepping method. By
comparison, the second-order Newtonian formulation uses the Generalized alpha
implicit method by default.

The explicit time stepping method is less suitable for stiff problems, meaning that the
Newtonian formulation is more robust when the particles are subjected to extremely
large, abrupt accelerations. The Newtonian formulation is also favorable for problems
involving ultrarelativistic particles. However, for some nonstiff problems, the explicit
method can give comparable or even better accuracy and performance, compared to
the implicit method.

Generally, it is most convenient to begin with the default Newtonian formulation, then
to consider switching to the Newtonian, first order formulation to optimize
performance if the problem is not overtly stiff.

Motion of Trapped Protons in Earth’s Magnetic Field: Application

Library path Particle_Tracing_Module/Charged_Particle_Tracing/
trapped_protons

MASSLESS
The Massless formulation defines a set of first-order ordinary differential equations for
the components of the particle position only. The particle velocity is directly specified,
either by an expression of by using a previously computed field.

The Massless formulation is useful for modeling the motion of extremely small particles
in a fluid. If Brownian diffusion and turbulent dispersion are neglected, the ratio of the
drag force to the mass of an extremely small particle is often so large that the particle
reaches equilibrium with the surrounding flow almost instantaneously, so that the
particles follow fluid streamlines. Such particles are sometimes referred to as tracer
particles, as opposed to inertial particles which are allowed to have some nonzero
velocity relative to the fluid.

M O D E L I N G TO O L S | 33
 LAGRANGIAN
The Lagrangian formulation (Ref. 2, Chapter 1) defines a set of second-order ordinary
differential equations for the components of the particle position. Compared to the
Newtonian formulation, the Lagrangian formulation has the same number of degrees of
freedom and offers greater flexibility in specifying the equations of motion, but it is not
possible to use the Force feature.

The equation of motion for a system with Lagrangian L (SI unit: J) is

d  ∂L = ∂L (2-1)
d t  ∂ v ∂q

where v (SI unit: m/s) is the particle velocity and q (SI unit: m) is the particle
position. Equation 2-1 is called the Euler-Lagrange equation.

The Lagrangian of a free, nonrelativistic particle is

1
L = T = --- m p ( v ⋅ v ) (2-2)
2

For example, for an instance of The Mathematical Particle Tracing Interface with tag
pt in 3D, the expression pt.mp*(pt.vx^2+pt.vy^2+pt.vz^2)/2 is the Lagrangian
for a free particle that is not subjected to any forces. Note that substitution into the
Euler-Lagrange equation yields

dv
m p ------- = 0
dt

If all forces can be expressed as the gradients of potentials, it is possible to specify

Newton’s law of motion in terms of a Lagrangian,

L = T–U

where T (SI unit: J) is the particle kinetic energy and U (SI unit: J) is the total
potential energy. For example, if U only depends on particle position, not velocity,
then substitution into the Euler-Lagrange equation yields

dv
m p ------- = – ∇U
dt

HAMILTONIAN
The Hamiltonian formulation defines a set of coupled first-order ordinary differential
equations for the components of the particle position and generalized momentum.

34 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
Following Chapter 7 in Ref. 2 the Hamiltonian H (SI unit: J) can be derived directly
from an expression for the Lagrangian L. The degrees of freedom are the position
vector components qi and the generalized momenta pi, defined as

∂L- · dq i
p i = -------
· q i = ---------
∂q i dt

For example, for a free, nonrelativistic particle, the generalized momenta are

∂ -1 · 2
·  --- m p
p i = -------
∂q i  2  qj 
j

Where the sum is over space dimensions in the model. This yields the simplified result

·
pi = mp qi (2-3)

for i from 1 to the total number of space dimensions. Thus, in this case the generalized
momentum is simply the particle momentum.

For some definitions of the Hamiltonian, the generalized momentum and

particle momentum may differ, so Equation 2-3 is not necessarily true for
any arbitrary Hamiltonian. However, certain features that accept
expressions for momentum components, like the General reflection
condition for the Wall feature, treat the specified expressions as values of
the particle momentum, not the generalized momentum. When using the
Hamiltonian formulation, always begin by checking whether the
Equation 2-3 holds, and use extra caution when entering user-defined
expressions for momentum components if it does not.

The Hamiltonian is then defined as

·
H =  pj qj – L
j

Using this Hamiltonian the following first-order equations are defined:

dq i ∂H- dp i ∂H-
--------- = ------- --------- = – ------- (2-4)
dt ∂p i dt ∂q i

These are known as Hamilton’s equations.

M O D E L I N G TO O L S | 35
 For example, for a free particle,

· p j2
H =  pj qj –  -----------
2m p
j j

Substitution with Equation 2-2 and Equation 2-3then yields

1
H = ----------- ( p ⋅ p )
2m p

For example, for an instance of The Mathematical Particle Tracing Interface with tag
pt in 3D, using default names for the generalized momentum components, the
expression (px^2+py^2+pz^2)/(2*pt.mp) is the Hamiltonian for a free particle that
is not subjected to any forces. Substitution into Hamilton’s Equations then yields

dq 1 dp
------- = -------- p -------- = 0
dt mp dt

As expected, the particle moves in a straight line and its momentum is conserved.

Geometry and Mesh Settings

When particles reach the boundaries of geometric entities in a model, they do not
interact with an exact parameterized representation of the geometry. Rather, they
propagate through the mesh elements that discretize the modeling domain and
interact with the boundary elements that cover the surfaces of the geometric entities.

REPRESENTATION OF CUR VED SURFACES

When the surfaces of the geometry are flat, the shape of the surface mesh is
indistinguishable from the shape of the geometric entities themselves. Therefore, the
fact that particles interact with the mesh instead of the geometry does not introduce
any discretization error, and it is possible to accurately compute particle trajectories
even when the mesh is extremely coarse.

Curved surfaces in the geometry, however, usually incur a significant amount of

discretization error when predicting how particles will interact with them. The time
and location at which the particle interacts with the boundary mesh element might be
slightly different from the time at which it would have interacted with an exact
representation of the surface. In addition, the tangential and normal directions on the
boundary mesh element may differ from the tangential and normal directions on the
surface, affecting the accuracy of boundary conditions that involve the tangential and

36 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
normal directions, such as the Bounce condition which causes particles to undergo
specular reflection.

The order of the curved mesh elements used to determine the geometry shape is
controlled by the Geometry shape order list in the Model Settings section of the Settings
window for the main Component node. If Automatic, the default, is selected, the curved
mesh elements are usually represented by quadratic curves; in some cases, linear
functions are used to prevent inverted mesh elements from being created.

The effect of the geometry shape order is most notable on a coarse mesh, as shown in
Figure 2-5. The mesh elements are shown as pale gray lines in the background and the
particle trajectories are represented as thick red arrows. The particles initially move
downward and are specularly reflected by a parabolic surface. If Linear is selected from
the Geometry shape order list, all particles that hit the same boundary element are
specularly reflected in the same direction, as shown on the left. Even though the
bottom surface is parabolic, the particles don’t all intersect at a single focus due to the
discretization error. If Quadratic or Automatic is selected, particles that hit the same
boundary element can still be reflected in different directions because the tangential
and normal directions can vary along the surface of the curved element. As a result,
the particles all intersect at a well-defined focal point.

Figure 2-5: Comparison of particles being specularly reflected at a curved boundary

represented using linear elements (left) and quadratic elements (right).

PA R T I C L E TR A C I N G I N A N I M P O R T E D M E S H
It is also possible to compute particle trajectories in an imported mesh. The mesh can
be imported from a COMSOL Multiphysics file (.mphbin for a binary file format

M O D E L I N G TO O L S | 37
 or .mphtxt for a text file format) or from a NASTRAN file (.nas, .bdf, .nastran,
or .dat).

If the mesh is imported from a COMSOL Multiphysics file, the imported mesh always
uses linear geometry shape order for the purpose of modeling particle-boundary
interactions, even if the model used to generate the mesh had a higher geometry shape
order.

If the mesh is imported from a NASTRAN file, the particle-boundary interactions may
be modeled using either linear or higher geometry shape order. If Export as linear
elements is selected when generating the NASTRAN file, or if Import as linear elements
is selected when importing the file, then linear geometry shape order will be used.

Special Variables
The particle tracing interfaces define a number of special variables, some of which can
only be used during results processing. These variables can be found in the Particle
statistics plot group during results processing, as shown in Figure 2-6. All of the
variables described in this section are preceded by its Name; see The Mathematical
Particle Tracing Interface for more information.

Figure 2-6: An example of variables available from the particle statistics menu and
available with the Mathematical Particle Tracing interface.

38 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
The following variables are defined:

• Particle index, pidx. Each particle is assigned a unique index starting from 1 up to
the total number of particles. This expression can be passed into a function, which
can create, for example, random forces that are unique for each particle. Suppose a
random function has already been defined with name rn1, which takes 2 input
arguments. Then a random force can be constructed with the expression
rn1(pidx,t).
• Particle release feature, prf. If there are multiple release features in a model, it is
useful to be able to visualize how the particles mix together based on their initial
release position. The Particle release feature variable takes a numeric value, starting
at 1, which is unique to each release feature.

When you also have the CFD Module, see Particle Tracing in a
Micromixer: Application Library path CFD_Module/Particle_Tracing/
micromixer_particle_tracing.

• Particle release time, prt. Since particles can be released at arbitrary points in time,
it is often useful to visualize at what time a specific particle was released. The Particle
release time is only available for primary particles, not those released due to
secondary emission.

If the Store particle status data check box is selected in the physics interface Additional
Variables section, then the following additional variables are created:

• The release time of a given particle (variable name rti). This works for secondary
particles and thus allows for extraction of the time at which a secondary particle was
released.
• The time at which a particle stopped at a boundary (variable name st).
• The final status of the particle (variable name fs). This indicates the status of a
particle at a given point in time. When used during postprocessing, the value always

M O D E L I N G TO O L S | 39
 indicates the status of the particle at the last time step. The value is an integer which
has one of the following values:
- 0 for unreleased particles.
- 1 for particles that are still in the modeling domain.
- 2 for frozen particles.
- 3 for stuck particles.
- 4 for particles that have disappeared.

• The status of the particle (variable name particlestatus, with no scope). This
indicates the current status of each particle. The value is an integer with the same
meaning as the final status of the particle. When the Store particle status data check
box is cleared, the variable particlestatus exists while computing particle
trajectories, so it can be used in weak expressions defined on the particles, but
cannot be used during postprocessing.

The following variables are global and can therefore be evaluated using
the Global Evaluation node under Derived Values. They do not have unique
values for each particle.

• Total number of particles, Nt. This total includes both primary and secondary
particles, and includes particles that have disappeared or have not been released.
• Total number of particles in selection, Nsel. If a selection has been applied to the
Particle data set, the number of particles in that selection can be evaluated.
• Transmission probability, alpha. Often the transmission probability is the main
quantity of interest in a particle tracing model. The transmission probability is
usually computed by dividing the total number of particles which reach an outlet by
the number of particles released at an inlet. The Transmission probability variable is
much more general, and can be used on domains, boundaries, or combinations of
both.

• Particle Trajectories in a Laminar Static Mixer: Application Library

path Particle_Tracing_Module/Fluid_Flow/laminar_mixer_particle
• Brownian Motion: Application Library path Particle_Tracing_Module/
Tutorials/brownian_motion

40 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
The following variables are found in the Release statistics plot group during results
processing, as shown in Figure 2-7.

Figure 2-7: An example of variables available from the release statistics menu and
available with the Charged Particle Tracing interface.

• Total number of particles released by feature, <tag>.Ntf, where <tag> is the tag of
a particle release feature, such as the Release, Inlet, or Release from Grid feature.
This global variable is uniquely defined for each release feature, and gives the total
number of particles that are released by that feature. This includes particles that have
not yet been released and particles that have disappeared or become frozen or stuck
due to collisions with walls. It does not include any secondary particles.
• Release frequency, <tag>.frel (SI unit: Hz), where <tag> is the tag of a particle
release feature. This variable appears when the Particle release specification in The
Charged Particle Tracing Interface or The Particle Tracing for Fluid Flow Interface
is set to Specify current or Specify mass flow rate respectively, and yields the number
of physical particles per model particle per second which must be released in order
to generate the user defined Release current magnitude or Mass flow rate, respectively.

M O D E L I N G TO O L S | 41
 • Release frequency, frel (SI unit: Hz). This variable also appears when the Particle
release specification is set to Specify current or Specify mass flow rate, and is defined
for every particle, yielding the release frequency of the feature which produced that
particle.
• Release current magnitude, <tag>.rc (SI unit: A). This variable is available when the
Particle release specification in The Charged Particle Tracing Interface is set to
Specify current, and yields the magnitude of the release current at the release feature
identified by <tag>.
• Mass flow rate, <tag>.mdot (SI unit: kg/s). This variable is available when the
Particle release specification in The Particle Tracing for Fluid Flow Interface is set to
Specify mass flow rate, and yields the mass flow rate at the release feature identified
by <tag>.

Component Couplings
The purpose of a model is often to compute the sum, average, maximum value, or
minimum value of a quantity over a group of particles, such as the average kinetic
energy or the maximum residence time. A particle tracing interface with Name <name>
creates the following operators:

• <name>.<name>op1(expr) evaluates the sum of the expression expr over the

particles. The sum includes all particles that are active, frozen, or stuck to
boundaries. It excludes particles that have not yet been released and those that have
disappeared.
• <name>.<name>op_all1(expr) evaluates the sum of the expression expr over all
particles, including particles those that are not yet released or have disappeared.
Since the coordinates of unreleased and disappeared particles are not-a-number
(NaN), the sum may return NaN if the model includes unreleased or disappeared
particles. An expression such as pt.ptop1(isnan(qx)) can be used to compute the
total number of unreleased and disappeared particles.
• <name>.<name>aveop1(expr) evaluates the average of the expression expr over
the active, frozen, and stuck particles. Unreleased and disappeared particles
contribute to neither the numerator nor the denominator of the arithmetic mean.
• <name>.<name>aveop_all1(expr) evaluates the average of the expression expr
over all particles. It is likely to return NaN if the model includes unreleased or
disappeared particles.
• <name>.<name>maxop1(expr) evaluates the maximum value of the expression
expr over all active, frozen, and stuck particles.

42 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
• <name>.<name>maxop_all1(expr) evaluates the maximum value of the
expression expr over all particles.
• <name>.<name>minop1(expr) evaluates the minimum value of the expression
expr over the active, frozen, and stuck particles.
• <name>.<name>minop_all1(expr) evaluates the minimum value of the expression
expr over all particles.

• <name>.<name>maxop1(expr, evalExpr) evaluates the expression evalExpr for

the particle that has the maximum value of the expression expr out of all active,
frozen, and stuck particles. For example, in a model that uses the Mathematical
Particle Tracing interface with name pt, the expression pt.ptmaxop1(pt.V, qx)
would evaluate the x-coordinate qx of the particle with the greatest velocity
magnitude pt.V.
• <name>.<name>maxop_all1(expr, evalExpr) evaluates the expression
evalExpr for the particle that has the maximum value of the expression expr for all
particles, including disappeared and unreleased particles.
• <name>.<name>minop1(expr, evalExpr) evaluates the expression evalExpr for
the particle that has the minimum value of the expression expr out of all active,
frozen, and stuck particles. For example, in a model that uses the Mathematical
Particle Tracing interface with name pt, the expression pt.ptminop1(pt.V, qx)
would evaluate the x-coordinate qx of the particle with the smallest velocity
magnitude pt.V.
• <name>.<name>minop_all1(expr, evalExpr) evaluates the expression
evalExpr for the particle that has the minimum value of the expression expr for all
particles, including disappeared and unreleased particles.

An instance of the Mathematical Particle Tracing interface with the default name pt
defines the built-in component couplings shown in Table 2-1.
TABLE 2-1: BUILT-IN COMPONENT COUPLINGS FOR THE MATHEMATICAL PARTICLE TRACING INTERFACE

NAME DESCRIPTION

pt.ptop1(expr) Sum over particles

pt.ptop_all1(expr) Sum over all particles
pt.ptaveop1(expr) Average over particles
pt.ptaveop_all1(expr) Average over all particles
pt.ptmaxop1(expr) Maximum over particles
pt.ptmaxop_all1(expr) Maximum over all particles
pt.ptminop1(expr) Minimum over particles

M O D E L I N G TO O L S | 43
 TABLE 2-1: BUILT-IN COMPONENT COUPLINGS FOR THE MATHEMATICAL PARTICLE TRACING INTERFACE

NAME DESCRIPTION

pt.ptminop_all1(expr) Minimum over all particles

pt.ptmaxop1(expr, evalExpr) Evaluate at maximum over particles
pt.ptmaxop_all1(expr, evalExpr) Evaluate at maximum over all particles
pt.ptmaxop1(expr, evalExpr) Evaluate at minimum over particles
pt.ptmaxop_all1(expr, evalExpr) Evaluate at minimum over all particles

Monte Carlo Modeling

Many particle tracing physics features define expressions that include random
numbers. In the Particle Tracing for Fluid Flow interface, the Drag Force and
Brownian Force physics features are forces that potentially include components which
are random in nature. In addition, some settings for particle-wall interaction and
emission of secondary particles depend on random numbers. When these features are
included in a model, it can be necessary to solve the problem multiple times and take
a statistical average of the results. The seeding of random numbers is controlled by the
Arguments for random number generation list in the physics interface Advanced Settings
section. The following options are available:

• Generate unique arguments — seed random numbers based on the position of each
physics feature in the Model Tree. This ensures that random numbers generated for
different physics features are independent of each other. Use this option to make the
results of a study reproducible when running the study multiple times in succession.
The random numbers may still differ when running the same study on different
architectures or in different versions of the software.
• Generate random arguments — seed random numbers using randomly generated
double-precision numbers. The seeds are regenerated whenever the study is run,
causing the results to differ when running the same study multiple times.
• User defined — When this option is selected, additional text fields appear in the
settings windows for all features that use random numbers. The number entered in
this text field is used as an additional argument for random number generation. A
set of distinct solutions can be obtained by running a Parametric Sweep over several
values of this additional argument.

For more information about the available options, see Particle Release and
Propagation in The Mathematical Particle Tracing Interface.

44 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
 Brownian Motion: Application Library path Particle_Tracing_Module/
Tutorials/brownian_motion

Figure 2-8: An example of the Settings window for Brownian Force.

All three particle tracing interfaces include a Velocity Reinitialization feature. This
allows for general purpose Monte Carlo modeling because the velocity vector can be
discretely changed at each time step according to some logical expression. The
Charged Particle Tracing interface also includes dedicated features for reinitializing

M O D E L I N G TO O L S | 45
 particle velocity to model the interaction of model particles with molecules in a rarefied
background gas.

The Collisions node supports a variety of subnodes, such as Elastic collisions and
Ionization reactions, which can be used to model the interactions that can occur when
particles collide with atoms or molecules in a rarefied background gas. If a collision
should occur, the charged particle has its velocity vector reinitialized by sampling the
velocity of the background gas particle at random from a drifting Maxwellian
distribution. This provides an accurate description of a charged particle interacting
with a background gas.

Study Setup
For all of the particle tracing physics interfaces, the particle trajectories must be
computed using a Time Dependent solver. For single-physics particle tracing, this can
be set up by using the Time Dependent study.

UNIDIRECTIONAL COUPLINGS
If a unidirectional, or one-way, coupling between particle trajectories and fields is set
up, it is often useful to first compute the fields in the surrounding domains and then
to compute the particle trajectories using a Time Dependent study. For example, it is
possible to model the motion of particles in a channel containing fluid by first setting
up the Laminar Flow interface and computing the fluid velocity and pressure using a
Stationary study. Then couple the flow profile to the Particle Tracing for Fluid Flow
interface using the Drag Force feature and solve for the particle trajectories in the time
domain.

The unidirectional coupling can be used when the particles do not significantly perturb
the surrounding fields. In the example of a distribution of particles in a fluid, this
means that the force exerted by particles on the fluid is not large enough to noticeably
redirect the flow. When modeling the motion of charged particles in an external
electric field, this means that the density of charged particles is not large enough to
significantly affect the electric potential.

BIDIRECTIONAL COUPLINGS
If the particles do significantly perturb the fields in the surrounding domains, it is
necessary to create a bidirectional, or two-way, coupling. The most straightforward
way to set up a bidirectional coupling is by using one of the Multiphysics Interfaces.
Alternatively, if the necessary physics interfaces are already present, it is possible to
manually add the contributions from the particles to the surrounding fields. For

46 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
example, if instances of the Charged Particle Tracing and Electrostatics interfaces are
present, the contribution of the charged particles to the space charge density can be
included by adding the Electric Particle Field Interaction Multiphysics node.

When setting up bidirectional couplings between physics interfaces, it is important to

determine whether the fields are stationary or not. For example, if a beam of charged
particles is released at constant current, then it is possible for the space charge density
at any point in the beam to remain constant over time.

Bidirectionally Coupled Time Domain Calculation

If the fields are not constant, it is necessary for the particle trajectories and fields to be
computed together in the time domain. For a charged particle beam, this might mean
that the beam is pulsed, and thus the electric potential is time-dependent. Full
time-domain calculation of the particles and fields is much more computationally
demanding than a unidirectional coupling between particles and fields, because it
requires particles to be released at a very large number of time steps. For fully
time-domain calculations, Specify release times should be selected from the Particle
release specification list in the physics interface Particle Release and Propagation section.

Using Iterative Solver Loops

If the fields are stationary, it is possible to significantly reduce the computational cost
of the model by using an iterative solver loop in which the particle trajectories are
computed using a Time Dependent solver and the fields are computed using a
Stationary solver. To obtain a self-consistent solution, it is necessary to ensure that the
result from each of these solvers is used to set the value of variables not solved for by
the other solver. The For and End For nodes, when added to a solver sequence, can be
used to set up an iterative loop that does the following:

1 Set all contributions from particles to external fields to zero.

2 Compute all field variables, using a Stationary solver, using the contributions from
particles to external fields computed in the previous step.
3 Compute the particle trajectories and their contributions to external fields, using the
field variables computed in the previous step.
4 Repeat steps 2 and 3 until a specified number of iterations has been reached or
another convergence criterion has been satisfied.

The Bidirectionally Coupled Particle Tracing study step automatically sets up an

iterative solver loop, and can be used in place of a Time Dependent study to facilitate
the calculation of bidirectionally coupled particle trajectories and fields.

M O D E L I N G TO O L S | 47
 If you have the AC/DC Module, see Electron Beam Divergence Due to
Self Potential, Application Library path ACDC_Module/Particle_Tracing/
electron_beam_divergence.

When solving for the fields using a Stationary study, it is possible to considerably
reduce the number of particles in the simulation by selecting Specify current or Specify
mass flow rate from the Particle release specification list in the physics interface Particle
Release and Propagation section. This causes each model particle to represent a number
of real particles per unit time, all moving along the same path, and therefore it is only
necessary to release particles during the first time step.

Auxiliary Dependent Variables and Residence Time

Auxiliary dependent variables are additional degrees of freedom that can be defined on
each particle and used to solve additional first-order ordinary differential equations
along each particle trajectory. They can be used to help keep track of things like
residence time, particle trajectory length, integrated shear rate, and so forth. When an
Auxiliary Dependent Variable feature is added to the physics interface an additional
ordinary differential equation is solved for each particle.

To compute the residence time of particles:

1 From the Physics toolbar, Global menu, select Auxiliary Dependent Variable.
2 In the Settings window for Auxiliary Dependent Variable, Auxiliary Dependent Variable
section, enter 1 in the R field.
For each particle, the differential equation
d ( rp ) = 1
dt

is solved, and so the variable rp represents the residence time.

To compute the length of the particle trajectories:
3 In the Auxiliary Dependent Variable section, from the Integrate list, select Along
particle trajectory.

The following differential equation is now solved for each particle:

48 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
 d ( rp ) = 1
ds

and so the variable rp is now the total length of the particle trajectory. The initial values
for the auxiliary dependent variables are set in the release features included in the
model.

By defining an appropriate source term, it is possible to evaluate the time integral or

path integral of any quantity that is known at the particle’s position. It is also possible
to change the value of an auxiliary dependent variable discontinuously when a particle
interacts with a boundary by entering an expression for the new value in the New Value
of Auxiliary Dependent Variables section of the settings window for the Wall node. The
new value of an auxiliary dependent variable may be defined recursively, for example,
to count collisions of a particle with the surrounding walls. In addition, the value of an
auxiliary dependent variable can be changed when the Velocity reinitialization condition
specified in a Velocity Reinitialization feature is satisfied, enabling the variable to be
reinitialized at any location within a domain.

Several built-in features are available for creating auxiliary dependent variables for
quantities that are frequently used. For example, the Compute particle mass and
Compute particle temperature check boxes in the settings window for The Particle
Tracing for Fluid Flow Interface create auxiliary dependent variables for the particle
mass and temperature, respectively. Several built-in features are available for applying
heat sources to the particles when the particle temperature is computed.

The residence time can also be computed in a different way by selecting the Store
particle status data check box. This creates variables for the particle release time and
the particle stop time (the time when a particle freezes or sticks to a boundary, or leaves
the modeling domain altogether). The residence time is then simply the difference
between the two.

It is also possible to integrate a quantity over time by using the timeint() operator.
The expression timeint(t1,t2,expr) evaluates the integral of the expression expr
from initial time t1 to final time t2. The accuracy of the integral depends on the total
number of time steps taken by the solver. The timeint() operator offers a convenient
way to evaluate the integrals of many different quantities without recomputing the
solution.

timeint and timeavg in the COMSOL Multiphysics Reference

Manual

M O D E L I N G TO O L S | 49
 Initialization of Auxiliary Dependent Variables
In addition to user-defined variables that can be inserted into a model using the
Auxiliary Dependent Variable feature, many physics features and physics interface
settings automatically define built-in auxiliary dependent variables for quantities such
as particle temperature, mass, and number of collisions.The particle release features
include options to control the way in which these auxiliary dependent variables are
initialized when particles are released into the model.

CO NT RO LL IN G TH E ORDER IN WH IC H VA RIA BLES A RE IN ITIA LIZED

When assigning initial values to particle degrees of freedom, the components of the
particle position vector are always assigned first. The initial values of the remaining
degrees of freedom, which may include velocity or momentum as well as auxiliary
dependent variables, may depend on the initial particle position. In addition, the initial
value of any variable may be defined in terms of other variables that are initialized
before it.

By default, user-defined auxiliary dependent variables are initialized after all other
particle variables. However, in the Initial Value of Auxiliary Dependent Variables section
of the Settings window for each release feature, the text field for the initial value of each
auxiliary dependent variable is accompanied by a check box, Initialize before particle
momentum. When this check box is selected, the corresponding variable is initialized
immediately after the particle position vector components and before any other
degrees of freedom. If multiple user-defined auxiliary dependent variables are present,
the order in which they are initialized is determined by their relative position in the
model tree; variables that appear earlier in the model tree are evaluated before variables
that appear below them.

The particle degrees of freedom are then initialized in the following order:

1 Position vector components.

2 User-defined auxiliary dependent variables for which the Initialize before particle
momentum check box has been selected. These variables are initialized in the order
in which the Auxiliary Dependent Variable nodes appear in the model tree.
3 Particle mass (Particle Tracing for Fluid Flow interface only).
4 Particle temperature (Particle Tracing for Fluid Flow interface only).
5 Particle velocity or momentum.
6 Other built-in auxiliary dependent variables, including the out-of-plane degrees of
freedom in a 2D and 2D axisymmetric geometry. This also includes collision

50 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
 counters created by the Collisions node and the various collision types that can be
added to it (Charged Particle Tracing interface only).
7 User-defined auxiliary dependent variables for which the Initialize before particle
momentum check box is cleared.

It is possible to express the initial values of auxiliary dependent variables in terms of

other particle degrees of freedom, including position, velocity, and other auxiliary
dependent variables. However, particle degrees of freedom can only be defined in
terms of other degrees of freedom that are initialized before them, that is, quantities
that appear earlier in the above list.

REL EASIN G D IS TRIBUT IO NS OF AUXILIAR Y DEPEN DEN T VARIABL ES

It is possible to assign a single initial value to an auxiliary dependent variable or to
sample multiple values from a distribution. The type of distribution can be selected
using the Distribution function setting in the Initial Value of Auxiliary Dependent Variables
section of all release features. Select None to enter a single initial value, Normal to create
a normal distribution function, Lognormal to create a log-normal distribution function,
Uniform to create a uniform distribution function, and List of values to enter an a list of
numerical values. For the Normal, Lognormal, and Uniform distributions, enter a
user-defined Mean (default 0) and Standard deviation (default 1). It is possible to create
a set of uniformly distributed numbers between 0 and 1 by selecting the Uniform
distribution and entering a Mean of 0.5 and a Standard deviation of sqrt(1/12).

Accumulators
Often the quantity of interest in particle tracing simulations involves the interaction of
particles with fields defined on a set of domains or boundaries. Examples of
particle-domain interactions include the accumulation of space charge density due to
ions and electrons, and the volume force exerted by moving particles on the
surrounding fluid. Examples of notable particle-boundary interactions include erosion
due to the impact of solid particles on a surface, sputtering of surface molecules due to
the impact of ions at a high velocity, and boundary loads generated by a stream of
particles impinging on a surface.

An Accumulator is a physics feature that allows dependent variables that are defined on
domain or boundary mesh elements to be affected by particles that interact with those
elements. When an Accumulator is added directly to a physics interface, it defines a
variable, called an accumulated variable, in each mesh element in a set of domains. The
value of the accumulated variable in a mesh element is affected by the presence of

M O D E L I N G TO O L S | 51
 particles within that element. Particles may either affect the accumulated variable
directly, by changing the value of the accumulated variable within the mesh element
occupied by the particle; or by changing the accumulated variable’s time derivative.

When an Accumulator is added as a subnode to a Wall, Outlet, or Axial Symmetry node,

it defines an accumulated variable on boundary elements in the selection list of its
parent node, defining one degree of freedom per element. These degrees of freedom
are incremented as particles hit the boundary. For more information on the
Accumulator features, see Accumulator (Boundary) and Accumulator (Domain).

The Accumulator features are available with all particle tracing interfaces. In addition,
the following features are available in The Charged Particle Tracing Interface to define
accumulated variables for more specialized applications: Etch, Surface Charge Density,
Current Density, and Heat Source.

The following features in The Particle Tracing for Fluid Flow Interface define
accumulated variables for specific applications: Erosion, Mass Deposition, Boundary
Load, and Mass Flux.

The dedicated Multiphysics nodes and interfaces described in the section Multiphysics
Interfaces also use special cases of accumulators on domains to model particle-field and
fluid-particle interactions.

Because each particle’s location is defined as a point whereas the accumulated variables
are defined on mesh elements of finite size, accumulators are inherently
mesh-dependent. If the accumulated variable is the density of a physical quantity, it
tends to become infinitely large in the limit as the mesh elements become
infinitesimally small. To obtain reasonable solutions, the number of particles moving
through a domain or interacting with the boundary should typically be at least an order
of magnitude larger than the total number of domain elements in a typical
cross-section of the domain or the total number of elements on the boundary,
respectively.

Overriding Particle Properties

It is often useful to compute the trajectories of multiple species of particle
simultaneously. By default, all features that release particles, such as the Inlet, Release
from Grid, and Secondary Emission features, create particles with properties defined
in the Particle Properties 1 node. These properties can include particle mass, charge
number, diameter, and density. The Override Properties node can be used to create
alternative sets of properties, which can then be assigned to groups of particles by

52 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
changing the Inherit properties setting for a release feature. For example, the Override
Properties node can be used to create one Release from Grid feature that produces ions
and a second Release from Grid feature that produces electrons.

Figure 2-9: The Inherit Properties section of the Settings window for Release from Grid.

Filters
Visualizing the trajectory of systems with a very large number of particles can consume
a lot of computer resources and often particles obscure one another. It is possible to
filter the type of particle and the number of particles which should be rendered. To do
this, right-click the Particle Trajectories plot type and select Filter.

Figure 2-10: An example of the Settings window for the Filter node.

M O D E L I N G TO O L S | 53
 The Particle type can be set to render primary particles, secondary particles, both, or
by a logical expression.

If particles are obscuring one another or the burden on the graphics card is very high,
the number of particles rendered can be reduced by changing the Particles to render
option. A fraction of the total number of particles to render or the total number of
particles to render can be set.

Particle Counters
A Particle Counter feature is a domain or boundary feature that provides information
about particles arriving on a set of selected domains or surfaces from a release feature.
Such quantities include the number of particles transmitted, the transmission
probability, transmitted current, mass flow rate etc. The feature provides convenient
expressions that can be used in the Filters node of the Particle Trajectories plot, which
allows only the particles which reach the particle counter selection to be visualized.
The following variables are provided by the Particle Counter feature, with the feature
tag <tag>:

• <tag>.Nfin is number of transmitted particles from the release feature to the

particle counter at the final time.
• <tag>.Nsel is number of transmitted particles from the release feature to the
particle counter.
• <tag>.alpha is the transmission probability from the release feature to the particle
counter.
• <tag>.rL is a logical expression for particle inclusion. This can be set in the Filter
node of the Particle Trajectories plot in order to visualize the particles which connect
the release feature to the counter.
• <tag>.It is the transmitted current from the release feature to the particle counter.
This variable is only available for the Charged Particle Tracing interface when Specify
current is selected from the Particle release specification list in the physics interface
Particle Release and Propagation section.
• <tag>.mdott is the transmitted mass flow rate from the release feature to the
particle counter. This variable is only available for the Particle Tracing for Fluid Flow
interface Specify mass flow rate is selected from the Particle release specification list
in the physics interface Particle Release and Propagation section.

54 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
If the release feature is a Particle Beam feature in the Charged Particle Tracing interface,
additional variables for the average position, velocity and energy of the transmitted
particles are available.

The Particle Counter feature only creates variables, which do not affect the
solution. Therefore, they can be added to a model without the need to
re-compute the solution, it just needs to be updated. To do this, right
click on the Study node and select Update Solution. The new variables
described above will be immediately available for results processing.

Improving Plot Quality

Particles can interact with boundaries at any time during a simulation. However, the
solution is only stored at a finite number of time steps. This can sometimes cause the
Particle Trajectories plot to give misleading results, since the exact instant at which a
particle-wall interaction occurs is often not included in the plot.

It is possible to improve the quality of the Particle Trajectories plot by selecting the
Store extra time steps for wall interactions check box in the physics interface Particle
Release and Propagation section. When this check box is selected, the exact times of
particle-wall interactions are stored in memory. When the solution is plotted, a number
of extra time steps are included, which are typically close to the times at which
particle-wall interactions occur.

In the following example, a particle is released at point (0.2, 0.5) in a unit square with
initial velocity (0.6, 11.2). The walls of the square are given the Bounce wall condition.
The Time-Dependent study is run until 1 s, with output every 0.05 s.

The particle trajectory is computed with the Store extra time steps for wall interactions
check box cleared and with the check box selected. The results are shown in
Figure 2-11.

Note that the solution at the stored times is equally valid in both cases; even though a
time step isn’t always stored at the exact instant a particle-wall collision occurs, the
collision is still handled at the correct time, using an extrapolation algorithm that is
determined by the Wall accuracy order setting in the physics interface Advanced Settings
section.

M O D E L I N G TO O L S | 55
 Figure 2-11: Comparison of the Particle Trajectories plot without extra time steps (left)
and with extra time steps (right).

Selecting the Store extra time steps for wall interactions check box does not guarantee
that all particle-wall interaction times will be shown in the Particle Trajectories plot.
Limitations on the number of extra time steps are imposed to avoid degrading
performance by rendering an overwhelming number of particle positions.

For example, consider a model in which 10,000 particles are traced and each particle
undergoes approximately 100 wall collisions. There could be as many as 1,000,000
distinct solution times at which a particle-wall collision occurs, each of which would
require all 10,000 particles to be rendered because every particle must be shown at the
same list of solution times.

To prevent this, the maximum number of extra time steps shown in the Particle
Trajectories plot is restricted by options in the plot settings. It is constrained either by
an absolute number of extra time steps or by a multiple of the number of time steps
stored in the solution.

The following plot illustrates this behavior. In this example, 50 model particles are
released with varying initial y-coordinates and initial x-coordinates of 0.2. The particles
are specularly reflected by the inclined wall so that they bounce upward.

In the image on the left, a total of 5 time steps are stored in the solution, so no more
than 5 additional particle-wall interaction times are plotted. In the middle image, a
total of 10 time steps are stored in the solution. In the right image, a total of 20 time
steps are stored in the solution. As the number of time steps increases, an increasing
number of model particles are shown to touch the wall because increasing the number

56 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
of solution times also increases the number of wall interaction times that can be
rendered.

Figure 2-12: Comparison of the stored extra time steps for three different numbers of stored
solution times.

Typically the Store extra time steps for wall interactions check box will have a very
pronounced effect on models with a small number of model particles, whereas for
models with a very large number of particles it may be necessary to specify a very large
maximum number of extra steps to generate higher-quality plots.

References
1. E.M. Lifshitz and L.P. Pitaevskii, Fluid Mechanics, Butterworth, 1987.

2. L.D. Landau and E.M. Lifshitz, Mechanics, 3rd ed., Elsevier, 1976.

M O D E L I N G TO O L S | 57
58 | C H A P T E R 2 : P A R T I C L E TR A C I N G M O D E L I N G
 3

Mathematical Particle Tracing

This chapter describes the Mathematical Particle Tracing interface found under
the Mathematics branch ( ) when adding a physics interface.

In this chapter:

• The Mathematical Particle Tracing Interface

• Theory for the Mathematical Particle Tracing Interface

59
 The Mathematical Particle Tracing
Interface
The Mathematical Particle Tracing (pt) interface ( ), found under the Mathematics
branch ( ) when adding a physics interface, computes the trajectories of particles
through a geometry. The particle motion is usually driven by external fields. The
particles can be massless or have their motion determined by Newton’s second law.
You can also specify a Hamiltonian or Lagrangian to dictate the motion of the particles.

When this physics interface is added, these default nodes are also added to the Model
Builder — Wall and Particle Properties. Then, from the Physics toolbar, add other nodes
that implement, for example, boundary conditions and particle release features. You
can also right-click Mathematical Particle Tracing 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 pt.

PARTICLE RELEASE AND PROPAGATION

Use the settings in this section to control how particles are released and how the
particle trajectories are stored in the solution.

Formulation
Select a Formulation— Newtonian (the default), Newtonian, first order, Lagrangian,
Massless, or Hamiltonian. This selection changes the equations of motion that are solved
for the particles. It affects the inputs for the Particle Properties node. For Newtonian
or Newtonian, first order a Force node can also be added to the model.

60 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 For a comparison of the first- and second-order Newtonian formulations,
see Choosing a Formulation in the Modeling Tools section.

Relativistic Correction
If Newtonian or Newtonian, first order is selected from the Formulation list, the
Relativistic correction check box is available but by default it is not selected. If you select
the check box, then the formulation is applicable for particles with very high speed,
meaning you can take relativistic effects on the particle mass into account. The particle
mass mp (SI unit: kg) is then computed as

mr
m p = ----------------------------------
2
1–v⋅v⁄c

where mr (SI unit: kg) is the rest mass, v (SI unit: m/s) is the particle velocity, and
c = 2.99792458 × 108 m/s is the speed of light in a vacuum.

Store Extra Time Steps for Wall Interactions

Typically, the exact times at which particle-wall interactions occur do not coincide with
time steps taken by the solver. This does not actually affect the accuracy of the solution,
but it can lead to misleading-looking plots when the boundaries reflect particles,
because the particle may appear to change direction at a finite distance from the wall,
instead of while hitting the wall.

Select the Store extra time steps for wall interactions check box to plot the solution at
extra time steps, in addition to the specified time steps in the output of the
time-dependent solver. Typically these additional time steps are close to times at which
the particles hit walls.

For an illustration of the effect of the Store extra time steps for wall
interactions check box, see Improving Plot Quality in the Modeling Tools
section.

Maximum Number of Secondary Particles

Some physics features, such as the Secondary Emission subnode for the Wall boundary
condition, release additional particles other than those that are specified by particle
release features. These particles are called secondary particles because they are created

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 61
 as a result of an existing particle interacting with a domain or boundary feature. The
memory for all secondary particles must be preallocated when beginning the study.

The Maximum number of secondary particles field ensures that sufficient memory for all
of the secondary particles is preallocated. It also prevents an inordinate number of
particles from being generated by capping them at the number supplied in the field.
The default value is 10,000.

If no sources of secondary particles such as Secondary Emission nodes are present, no

unreleased secondary particles are created. If at least one feature is capable of emitting
secondary particles, degrees of freedom are allocated for the total number of particles
entered in the Maximum number of secondary particles text field, even if some of these
particles are never released.

ADDIT IO N AL VA RIABLES
Use the settings in this section to determine what additional information is stored in
the solution while computing the particle trajectories, such as information about the
status of each particle and the entity that released it.

Store Particle Status Data

By default the Store particle status data check box is not selected. When selected it adds
new variables for quantities that cannot necessarily be recovered from the particle
trajectory data alone. This is especially true if automatic remeshing has been used in a
model. The following variables are created:

• The release time of a given particle (variable name rti).

• The time at which a particle stopped at a boundary (variable name st).
• The final status of the particle (variable name fs). This indicates the status of a
particle at a given point in time. When used during postprocessing, the value always
indicates the status of the particle at the last time step. The value is an integer which
has one of the following values:

- 0 for unreleased particles.

- 1 for particles which are still in the modeling domain.
- 2 for frozen particles.
- 3 for stuck particles.
- 4 for particles that have disappeared.

In addition, the particle status (variable name particlestatus, with no scope) can
be used during postprocessing. This variable is always defined while computing the

62 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
particle trajectories, but its value is only stored if the Store particle status data check
box is selected.

To summarize the total number of particles having each final status, the following
global variables are also defined.
TABLE 3-1: GLOBAL VARIABLES BASED ON PARTICLE STATUS

NAME DESCRIPTION

fac Fraction of active particles at final time step

fds Fraction of disappeared particles at final time step
ffr Fraction of frozen particles at final time step
fse Fraction of secondary particles released
fst Fraction of stuck particles at final time step

The global variable names in Table 3-1 all take the unreleased secondary particles into
account. For example, suppose an instance of the Mathematical Particle Tracing
interface includes 100 primary particles and 100 allocated secondary particles. At the
last time step, suppose that 80 of the primary particles have stuck to boundaries and
that 40 secondary particles have been emitted, all of which are still active. Then the
variable pt.fac, the fraction of active particles at the final time step, would have the
value (20 + 40)/(100 + 100) or 0.3.

Store Particle Release Statistics

By default the Store particle release statistics check box is not selected. When selected
it adds a new internal variable called releaseindex to identify the release feature that
creates each particle. Every node that is capable of releasing particles, including
secondary particles, is associated with a unique positive integer based on its position
relative to other nodes in the Model Builder.

When this check box is cleared, the variable <scope>.prf can still be used to identify
the release feature that creates each particle, but this variable does not account for
secondary particle emission.

ADVANCED SETTINGS
This section is only shown when Advanced Physics Options are enabled (click the Show
button on the Model Builder).

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 63
 The Wall accuracy order sets the accuracy order of the time stepping used for time steps
during which a particle-wall interaction happens:

• Order 1 means that a forward Euler step is used to compute the motion both before
and after the wall collision.
• Order 2 (the default) means that a second-order Taylor method is used to compute
the motion before the wall collision. After the collision a second-order Runge-Kutta
method is used.

The Wall accuracy order also controls the time stepping that is used when particles are
released during a time step taken by the solver.
Select an option from the Arguments for random number generation list: Generate unique
arguments, Generate random arguments, or User defined. This setting determines how
the additional argument to random functions is defined in features such as the Wall
boundary condition with the Diffuse scattering wall condition and the Brownian Force
in The Particle Tracing for Fluid Flow Interface. Typically the random numbers are
functions of the particle index, position, time, and another argument i, which is
defined as follows:

• For Generate unique arguments the additional argument is based on the position of
each node in the Model Builder. As a result, random numbers generated in different
nodes are created independently of each other, but the same result can be
reproduced by running the same study several times.
• For Generate random arguments the additional argument is randomly created,
causing the random functions to return different results each time the study is run.
For User defined the additional argument is defined by a user input in the Settings
window each feature. Independent solutions can be obtained by running a parametric
sweep for different values of i.

Enter a value for the Maximum number of wall interactions per time step. The default
value is 1000. If a particle undergoes more than the specified number of boundary
interactions or velocity reinitializations in a single time step taken by the solver, the
particle will disappear. This is included as a safeguard to prevent particles from getting
stuck in infinite loops if the time between successive particle-wall interactions becomes
infinitesimally small.

DEPENDENT VARIABLES
The dependent variables (field variables) are the Particle position, Particle position
components, Particle momentum, Particle momentum components, Particle velocity, and
Particle velocity components. Note that not all of these dependent variables are needed

64 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
for every formulation of the equations of motion; for example, the field variable names
for Particle momentum and Particle momentum components are only used if Hamiltonian
is selected from the Formulation list. The names can be changed but the names of fields
and dependent variables must be unique within a model.

• Theory for the Mathematical Particle Tracing Interface

• Domain, Boundary, Pair, and Global Nodes for the Mathematical
Particle Tracing Interface

• Rössler Attractor: Application Library path

Particle_Tracing_Module/Tutorials/rossler_attractor
• Ion Cyclotron Motion: Application Library path:
Particle_Tracing_Module/Charged_Particle_Tracing/ion_cyclotron_motion

Domain, Boundary, Pair, and Global Nodes for the Mathematical

Particle Tracing Interface
The Mathematical Particle Tracing Interface has these domain, boundary, pair, and
global nodes 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.

DOMAIN
When Newtonian or Newtonian, first order is selected as the Formulation, the following
are available:

• Force
• Rotating Frame
• Particle-Particle Interaction

When Newtonian, Newtonian, first order, Lagrangian, or Hamiltonian is selected as the

Formulation, the following are available:

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 65
 • Velocity Reinitialization
• Secondary Emission (subnode to Wall)
• Secondary Emission (subnode to Velocity Reinitialization)

For all Formulation choices the following are available:

• Release
• Accumulator (Domain)
• Particle Counter

BOUNDARY
• Axial Symmetry
• Inlet
• Nonlocal Accumulator (subnode to Inlet)
• Outlet
• Wall (default)
• Periodic Condition
• Accumulator (Boundary) (subnode to Wall, Outlet, and Axial Symmetry)
• Particle Counter

PAIRS
• Inlet
• Particle Continuity

EDGE
• Release from Edge

POINT
• Release from Point

GLOBAL
• Auxiliary Dependent Variable
• Particle Properties (default, singleton)
• Release from Grid
• Override Properties

66 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
• Release from Data File

In the COMSOL Multiphysics Reference Manual see Table 2-3 for links
to common sections and Table 2-4 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.

Wall
Use the Wall node to determine what happens to the particles when contact with a wall
is made. The Accumulator (Boundary) subnode is available from the context menu
(right-click the parent node) or from the Physics toolbar, Attributes menu. If
Newtonian, Newtonian, first order, Lagrangian, or Hamiltonian is selected from the
Formulation list in the physics interface Particle Release and Propagation section, the
Secondary Emission subnode is also available.

The Wall node can be applied to interior or exterior boundaries, but the instance of this
node that is created by default only applies to exterior boundaries.

WAL L CO N DIT IO N
Select a Wall condition — Freeze (the default), Bounce, Stick, Disappear, Pass through,
Diffuse scattering, Mixed diffuse and specular reflection, or General reflection.

If Massless is selected from the Formulation list in the physics interface Particle Release
and Propagation section, only the Freeze (default), Stick, and Disappear conditions are
available.
TABLE 3-2: WALL CONDITION OPTIONS

OPTIONS DESCRIPTION

Freeze Select to fix the particle position and velocity at the instant a wall is
struck. So, the particle position no longer changes after contact with
the wall and the particle velocity remains at the same value as when the
particle struck the wall. This boundary condition is typically used to
recover the velocity or energy distribution of charged particles at the
instant contact was made with the wall.
Bounce Select to specularly reflect from the wall such that the particle kinetic
energy is conserved. It is typically used when tracing microscopic
particles in a fluid.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 67
 TABLE 3-2: WALL CONDITION OPTIONS

OPTIONS DESCRIPTION

Stick Select to fix the particle position at the instant the wall is struck. The
particle velocity is set to zero. This can be used if the velocity or energy
of the particles striking a wall is not of interest.
Disappear This option means that the particle is not displayed once it has made
contact with the wall. Use this option if display of the particle location
after contact with the wall is not of interest.
Pass through When applied to an interior boundary, the Pass through option allows
particles to cross the boundary unimpeded. When applied to an
exterior boundary, this option behaves like the Disappear option.
Diffuse Select to bounce particles off a wall according to Knudsen’s cosine law.
scattering That is, the probability a particle bouncing off the surface in a given
direction within a solid angle dω is given by cos(θ)dω, where θ is the
angle between the direction of the reflected particle and the wall
normal. The total particle kinetic energy is conserved.
Mixed Select to bounce particles off a wall either specularly or according to
diffuse and Knudsen’s cosine law, based on a user-defined probability. By combining
specular this wall condition with a primary particle condition, it is possible to
reflection include up to three different types of particle-wall interactions at a
single boundary. Also enter the Probability of specular reflection γs.
General Select to allow an arbitrary velocity to be specified after a particle
reflection makes contact with the wall. The velocity components can be functions
of the incident particle velocity, energy, or any other quantity. Note
that the total momentum of the particle does not necessarily have to
be conserved with this option. See General Reflection Settings.

GENERAL REFLECTION SETTINGS

This section is available when General reflection (see Table 3-2) is selected as the Wall
condition.

Enter values for the Reflected particle velocity vp (SI unit: m/s) either in Cartesian
coordinates (x, y, z) (the default) or select the Specify tangential and normal velocity
components check box to enter coordinates in the tangent-normal coordinate system
(t1, t2, n). In this case the normal direction is selected so that an incident particle is
reflected back into the domain it previously occupied if the specified normal velocity
component is positive. The tangential directions are oriented so that they form a
right-handed coordinate system, together with the normal direction. As a result, the
tangential and normal directions may point in the opposite direction of the
corresponding vectors defined for the geometry (for example, root.nx) and for the
physics interface (for example, pt.nx).

68 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
PRIMARY PARTICLE CONDITION
Select a Primary particle condition — None (the default), Probability, or Expression.
When the default, None, is kept, it means that the Wall condition is always respected by
the incident particles.

Probability
If Probability is selected, the Wall condition is applied with a certain probability. Enter
a value for the Probability γ (dimensionless) of the particle behaving according to the
Wall condition. Otherwise the particle behaves according to the Otherwise setting.

The value of γ should always be between 0 and 1.

For example, if the Wall condition is set to:

• Freeze and γ is set to 0.1, then for every 10 particles that strike the wall, on average
one freezes and the remaining 9 particles behave according to the Otherwise
setting.
• Stick and γ is set to 0.5, then on average half of the particles stick to the wall and the
other half behave according to the Otherwise setting.

Expression
If Expression is selected, the Evaluation expression e (dimensionless) is evaluated
whenever the particle strikes the wall. The default value is 1. If the Evaluation expression
is nonzero, the particle behaves according to the Wall condition, otherwise the particle
behaves according to the Otherwise setting.

Otherwise
The options available for the Otherwise setting are the same as for the Wall Condition,
except that Mixed diffuse and specular reflection and General reflection are not available.
The Otherwise setting can be used to make particles interact with a wall differently
when a certain condition is satisfied. For example, to specify that particles with a
diameter greater than 1 micron stick to an interior boundary, and the rest pass
through:

• Select Stick as the Wall condition,

• set the Primary particle condition to pt.dp>1E-6, and
• select Pass through as the Otherwise option.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 69
 N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
This section is available if an Auxiliary Dependent Variable has been added to the
model.

When a particle crosses or touches a boundary, the values of its auxiliary dependent
variables can be changed. The new values can be functions of any combinations of
particle variables and variables defined on the boundary. A simple application is to use
this to count the number of times a particle strikes the wall.

Select the Assign new value to auxiliary variable check box or boxes based on the
number of auxiliary variables in the model. Then enter the new value or expression in
the field. For example, if there is an auxiliary variable, psi, then enter a value for psinew
in the field. So, to increment the value of psi by 1 when a particle touches or crosses
a boundary, enter psi+1 in the text field for psinew.

ADVANCED SETTINGS
If the Primary particle condition is set to Probability, or if the Diffuse scattering wall
condition is used, then the Wall feature generates random numbers. If, in addition, the
Arguments for random number generation setting is User defined, the Advanced Settings
section is available. Enter the Additional input argument to random number generator.
The default value is 1.

Periodic Condition
The Periodic Condition can be used to remove particles that make contact with a source
boundary and re-insert them at some other boundary, the destination. The boundaries
must be planar and cannot be curved. At least two boundaries must be selected.

TY P E O F PE R I O D I C I T Y
Select an option from the Type of periodicity list: Continuity or User defined.

If Continuity is selected, the particle velocity at the destination boundary is equal to the
velocity at the source boundary.

If the source and destination boundaries are parallel, the components of the particle
velocity vector are unchanged. If the source and destination boundaries are not
parallel, then the two boundaries are interpreted at the extents of a unit cell exhibiting
sector symmetry; thus the velocity vector will be rotated based on the orientation of
the destination with respect to the source. This is illustrated in Figure 3-1; the angle
between the incident particle velocity v1 and the reinitialized particle velocity v2 is
equal to the angle between the source and destination boundaries (shown in red).

70 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Figure 3-1: Transformation of particle velocity in a geometry exhibiting sector symmetry.

If User defined is selected, enter values or expressions for the components of the
Reinitialized particle velocity vr (SI unit: m/s). The default is zero. Expressions for the
reinitialized particle velocity are evaluated at the source boundary, not the destination;
for example, expressions in terms of tangential and normal vector components will use
the directions tangent or normal to the source boundary. If the velocity is such that
the particle will immediately leave the modeling domain through the destination
boundary, the particle will instead disappear, to prevent the particle from being
subjected to an arbitrarily large number of successive boundary interactions.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
This works in the same way as for the Wall feature. The expressions are evaluated at the
source boundary. See New Value of Auxiliary Dependent Variables.

Accumulator (Boundary)
The Accumulator subnode is available from the context menu (right-click the Wall,
Outlet, Periodic Condition or Axial Symmetry parent node) or from the Physics
toolbar, Attributes menu. Each Accumulator subnode defines a variable, called the
accumulated variable, on each boundary element in the selection of the parent node.
Whenever a particle hits a boundary element, the value of the accumulated variable in
that element is incremented based on the value of the user-defined Source term R for
the incident particle.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 71
 ACCUMULATOR SETTINGS
Select an option from the Accumulator type list: Density (default) or Count.

• For Density the accumulated variable is divided by the surface area (in 3D) or length
(in 2D) of the boundary element where it is defined.
• For Count the accumulated variable is the sum of the source terms of all particles that
hit the boundary element, and is unaffected by the boundary element size.

Select an option from the Accumulate over list: Particle-wall interactions (default) or
Particles in boundary elements.

• For Particle-wall interactions the accumulated variable is affected by all particles that
hit the boundary element, including those that pass through or are reflected by the
boundary.
• For Particles in boundary elements the accumulated variable is only affected by
particles that freeze or stick to the boundary element.

Enter the Accumulated variable name. The default is rpb. The accumulated variable is
defined as <scope>.<name>, where <scope> includes the name of the physics
interface node, parent boundary condition, and the Accumulator node; and <name> is
the accumulated variable name.

For example, if the Accumulator subnode is added to a Wall node in an instance of the
Mathematical Particle Tracing interface using the default variable name rpb, the
accumulated variable name might be pt.wall1.bacc1.rpb.

Enter a Source R. The unit of the source term depends on the settings in the Units
section. Whenever a particle collides with a boundary element in the selection of the
parent node, the accumulated variable in that element is incremented by the source
term. If the Accumulator type is set to Density, the source term is divided by the area of
the boundary element (in 3D) or the length of the boundary element (in 2D).

UNITS
Select a Dependent variable quantity from the list; the default is Dimensionless [1]. To
enter a unit, select None from the list and in the Unit field enter a value, for example, K,
m/s, or mol/m^3.

SMOOTHING
The accumulated variables are computed using discontinuous shape functions. Select
the Compute smoothed accumulated variable check box to compute a smoothed
accumulated variable by computing the average value of the variable within a sphere of

72 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
a user-defined radius. Then enter a Smoothing radius r (SI unit: m). The default is
0.1[m].

Accumulator Theory: Boundaries

Particle Counter
Use the Particle Counter feature to compute information about particles that are
located in a set of selected domains or on a set of selected boundaries. The counter may
detect all particles or only the particles released by a specified release feature.
Computed variables are the number of particles transmitted, the number of particles
transmitted at the final time, the transmission probability, and a logical expression
which can be used to filter the rendered particles during results processing.

When used with The Particle Tracing for Fluid Flow Interface, the Particle Counter also
computes the transmitted mass flow rate when the Particle release specification is set to
Specify mass flow rate.

When used with The Charged Particle Tracing Interface, the Particle Counter also
computes the transmitted current when the Particle release specification is set to Specify
current. If the release feature is a Particle Beam feature in the Charged Particle Tracing
interface, additional variables for the average position, velocity and energy of the
transmitted particles are available.

PARTICLE COUNTER
Select an option from the Release feature list. If All (the default) is selected, the Particle
Counter collects information about all particles in the selected domains or boundaries,
regardless of how they were released. If any particle release features have been added
to the model, they can also be selected from the list. Selecting any particle release
feature from the list causes particles from all other sources to be ignored by the
counter.

Secondary Emission
The Secondary Emission subnode is available from the context menu (right-click the
Wall, Axial Symmetry, or Velocity Reinitialization parent node) or from the Physics

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 73
 toolbar, Attributes menu. When Massless is selected as the Formulation, the Secondary
Emission subnode is not available.

If the Secondary Emission node is added to a boundary feature such as a

Wall or Axial Symmetry node, secondary particles are released when
particles hit the boundary. If the Secondary Emission node is added to a
Velocity Reinitialization node, secondary particles are released when the
particle velocity is reinitialized.

SECONDARY EMISSION CONDITION

For Secondary Emission subnodes applied to boundary nodes, select an option from the
Activate on wall conditions list: All, Primary only, or Otherwise only. These settings
determine which wall conditions in the parent node cause secondary particles to be
released.

• For All both the Wall condition and Otherwise condition cause secondary particles to
be released.
• For Primary only secondary emission only occurs if the incident particle behaves
according to the Wall condition setting.
• For Otherwise only secondary emission only occurs if the incident particle behaves
according to the Otherwise only setting.

Select a Secondary emission condition — None, Probability, or Expression. For Probability

enter the Probability γ (dimensionless) for secondary emission to occur. For Expression
enter an Evaluation expression e (dimensionless) that must be nonzero for secondary
emission to occur.

SECONDARY PARTICLES

Number of Secondary Particles

In the Number of secondary particles field Ns, specify the number of secondary particles
to release per incident particle. The default value is 1.

Initial Velocity
Select an Initial velocity. If the Secondary Emission feature is added to a Wall or Axial
Symmetry node, the available options are Isotropic hemisphere (the default), Reflection
of primary particle, Diffuse scattering, and User defined. If the Secondary Emission feature
is added to a domain level feature such as Velocity Reinitialization, the available options

74 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
are Constant speed, spherical and User defined (default). The following briefly explains
each option.

• Isotropic hemisphere: releases the secondary particles with a constant speed and
hemispherical velocity direction with the north pole directed in the normal direction
away from the wall. The speed of the secondary particles is equal to the speed of the
incident particle, divided by the number of secondary particles.
• Constant speed, spherical: releases the secondary particles with a constant speed and
spherical distribution of velocity directions. Enter a Speed with which all secondary
particles are released.
• Diffuse scattering: causes secondary particles to be released with a probability
distribution based on Knudsen’s cosine law, like the Diffuse scattering option for the
Wall node.
• Reflection of primary particle: releases the secondary particles in the direction the
primary particle would go if it were specularly reflected, like the Bounce option for
thee Wall node.
• User Defined: allows for an arbitrary velocity vector to be set for the secondary
particles. Enter values or expressions for the Initial particle velocity v0 (SI unit: m/s)
either in Cartesian coordinates (x, y, and z for 3D) (the default) or select the Specify
tangential and normal velocity components check box (available only when the parent
node is a Wall or Axial Symmetry node) to enter coordinates in the tangent-normal
coordinate system (t1, t2, n), a specification of the tangential components of the
velocity and a normal component (in 2D) and two tangential components and a
normal component (in 3D). In this case, the normal is directed away from the wall
on which the particle is incident. As a result, the tangential and normal directions
may point in the opposite direction of the corresponding vectors defined for the
geometry (for example, root.nx) and for the physics interface (for example,
pt.nx).

Scale Velocity by Number of Secondary Particles

For Secondary Emission nodes added to boundary features, select the Scale velocity by
number of secondary particles check box when the velocity of the secondary particles
depends on the velocity of the primary particle to divide the velocity of each secondary
particle by the number of secondary particles.

For example, suppose an incident particle with speed 2 m/s hits a Wall with a Secondary
Emission subnode. If the Secondary Emission node releases 4 particles in an Isotropic
hemisphere and the Scale velocity by number of secondary particles check box is selected,
each reflected particle will have a speed of 0.5 m/s. If the Scale velocity by number of

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 75
 secondary particles check box is cleared, each reflected particle will have a speed of
2 m/s.

Offset Initial Position

This section is only shown if the Secondary Emission node is a subnode of a
domain-level feature such as the Velocity Reinitialization node, not a boundary feature
such as Wall node.

Select the Offset initial position check box to offset the initial positions of the secondary
particles before releasing them. The secondary particles can then appear in the region
surrounding the parent particle, instead of its exact position.

Select an Offset method—Using initial velocity (the default) or Isotropic sphere.

• For Using initial velocity enter a Time interval Δt (SI unit: s). The default is 0. Each
secondary particle is then displaced by the product of its initial velocity and the
specified time interval before being released.
• For Isotropic sphere enter a Particle displacement magnitude Δr (SI unit: m). The
default is 0. All secondary particles are then moved a distance equal to the
displacement magnitude before they are released. If the offset would cause
secondary particles to be placed outside of the modeling domain, they are instead
released at the location of the primary particle.

ADVANCED SETTINGS
For any of the following combinations of settings, the Secondary Emission feature
generates random numbers:

• Probability is selected from the Secondary emission condition list.

• Isotropic sphere, Diffuse scattering, or Constant speed, spherical, is selected from the
Initial velocity list.
• The secondary particle positions are offset using the Isotropic sphere offset method
(shown for domain-level secondary emission only).

If, in addition, User Defined is selected from the Arguments for random number
generation list in the physics interface Advanced Settings section, the Advanced Settings
section is available. Enter the Additional input argument to random number generator.
The default value is 1.

INHERIT PROPERTIES
Select an option from the Inherit Properties list. The Particle Properties node (default)
is always available. If any Override Properties nodes have been added to the model,

76 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
they can also be selected. Use this setting to determine what set of properties is given
to the particles that are released by this feature.

IN IT IA L VA LUE OF AUXILIAR Y DEPEN DENT VARIABLES

This section is available if an Auxiliary Dependent Variable has been added to the
model. For each of the active Auxiliary Dependent Variable features in the model, enter
a value or expression for its initial value. For example, if the auxiliary dependent
variable has the default name rp, enter a value or expression for its initial value rp0.

About the Wall Boundary Conditions

Particle Properties
Use the Particle Properties node to set properties based on the Formulation selected
from the physics interface Particle Release and Propagation section. Set the particle
velocity for Massless formulations, or particle mass for Newtonian, Newtonian, first
order, Lagrangian, or Hamiltonian formulations.

For theory see:

• Newtonian Formulation
• Hamiltonian Formulation
• Lagrangian Formulation
• Massless Formulation

PARTICLE MASS
This section is shown when Newtonian, Newtonian, first order, Hamiltonian, or
Lagrangian is selected as the Formulation.

Enter a value or expression for the Particle mass mp (SI unit: kg). The default
expression is me_const, which is a predefined physical constant for the electron mass,
9.10938188 × 10−31 kg.

If the Relativistic correction check box is selected in the physics interface Advanced
Settings section, instead enter a value or expression for the Particle rest mass mr
(SI unit: kg). The default expression is me_const.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 77
 LAGRANGIAN
This section is shown for The Mathematical Particle Tracing Interface and when
Lagrangian is selected as the Formulation.

Enter a value or expression for Lagrangian L (SI unit: J). The default is 0. The
Lagrangian is usually a function of the particle kinetic energy and fields, which exert a
force on the particles. To include field forces in the Lagrangian, it should be possible
to express the field in terms of the gradient of a potential.

P A R T I C L E VE L O C I T Y
This section is shown when Massless is selected as the Formulation.

Enter a vector for the Particle velocity v (SI unit: m/s) based on space dimension. The
Massless formulation means that the particles follow streamlines of the particle velocity
expression.

HAMILTONIAN
This section is shown for The Mathematical Particle Tracing Interface and when
Hamiltonian is selected as the Formulation.

Select an option from the Specify Hamiltonian list — Directly (the default) or Manually.

• For Directly enter a value for Hamiltonian H (SI unit: J) The default is 0. The
Hamiltonian is typically a function of the particle kinetic energy and any fields which
would induce a force on the particles.
• For Manually enter coordinates for the Particle velocity vH (SI unit: m/s) and
Hamiltonian Force FH (SI unit: N) based on space dimension.

The Hamiltonian formulation solves a system of first order ordinary

differential equations for the particle coordinate and the particle
momentum. Thus, when using the Hamiltonian formulation, the number
of degrees of freedom is the same as the Newtonian, first order formulation
and twice as high as the Massless, Lagrangian, and Newtonian formulations.

78 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Force

The Force node is only available when Newtonian or Newtonian, first order
is selected as the Formulation in the physics interface Particle Release and
Propagation section.

Use the Force node to exert user-defined forces on particles to influence their motion.
All forces defined in the model are added together to compute the total force on the
particles. The force can be specified directly or using a susceptibility and field.

FORCE
Select an option from the Specify force list: Directly (the default) or Using susceptibility
and field. For Directly enter values or expressions for the Force F (SI unit: N) based on
space dimension.

For Using susceptibility and field:

• Select Isotropic, Diagonal, Symmetric, or Anisotropic and enter a value or expression

in the field or matrix (based on space dimension) for the Susceptibility χ.
• Enter values or expressions for the Field Γ.

The units of the susceptibility and field are controlled by the Units section.

AFFECTED PARTICLES
Select an option from the Particles to affect list: All (the default), Single species, or
Logical expression.

• If All is selected, the force is applied to every particle in the selected domains.
• For Single species select an option from the Affected particle properties list. The
Particle Properties node is always available and is the default. In addition, if any
Override Properties nodes have been added to the physics interface, they can be
selected. Use this option to apply a force to a specific type of particle.
• For Logical expression enter a Logical expression for inclusion e. The default is 1. The
force is applied to all particles in the selected domains for which the logical
expression is nonzero.

UNITS
This section is available if Using susceptibility and field is selected from the Specify force
list. Select a Susceptibility quantity from the list; the default is Dimensionless [1]. To

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 79
 enter a unit, select None from the list and in the Unit field enter a value, for example, K,
m/s, or mol/m^3. The unit of the Susceptibility is changed to the specified unit. The
unit of the Field is changed so that the product of the susceptibility and field has a unit
of force.

Rotating Frame
Use the Rotating Frame feature to exert fictitious forces on particles when computing
their trajectories in a rotating frame of reference, including the centrifugal, Coriolis,
and Euler forces.

ROTATING FRAME
In 2D, enter the coordinates of the Rotation axis base point rbp (SI unit: m) about
which the frame rotates. The default is the origin (0,0).

In 3D, select an option from the Axis of rotation list: x-axis, y-axis, z-axis (the default),
or User defined. For User defined, enter the coordinates of the Rotation axis base point
rbp (SI unit: m) about which the frame rotates. The default is the origin (0,0,0). Then
enter the components of the Rotation axis direction eax (dimensionless). The default
direction is parallel to the positive z-axis, (0,0,1). It is not necessary to enter a unit
vector for the axis direction because the expression is automatically normalized.

The remaining inputs are the same for any space dimension.

Select an option from the Rotational direction list: Counterclockwise (the default) or
Clockwise. The sense of rotation follows the standard right-hand rule; that is, if you curl
the fingers on your right hand in the direction of the rotation, your thumb points in
the direction of the axis of rotation (in 3D) or in the positive z-direction (in 2D).

Select an option from the Rotational frequency list: Angular velocity (the default) or
Revolutions per time. For Angular velocity enter the Angular velocity magnitude Ω
(SI unit: rad/s). The default is 0[rad/s]. For Revolutions per time enter the
Revolutions per time f (SI unit: Hz). The default is 0[rpm].

By default the Centrifugal force check box is selected. This means that the fictitious
centrifugal force is exerted on the particles.

By default the Coriolis force check box is selected. This means that the fictitious Coriolis
force, which arises when the particles have nonzero velocity in the rotating frame, is
exerted on the particles.

80 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
By default the Euler force check box is selected. This means that the fictitious Euler
force, which arises when the frame has nonzero angular acceleration, is exerted on the
particles.

AFFECTED PARTICLES
Use this section to exert the fictitious forces on specific particles. The available settings
are the same as for the Force node.

Particle Tracing in Rotating Frames

Velocity Reinitialization

The Velocity Reinitialization node is only available when Newtonian,

Newtonian, first order, Lagrangian, or Hamiltonian is selected as the
Formulation in the physics interface Particle Release and Propagation
section.

Use the Velocity Reinitialization node to reset the particle velocity to an arbitrary
expression if a specific condition is satisfied.

The Secondary Emission subnode is available from the context menu (right-click the
parent node) or from the Physics toolbar, Attributes menu.

VE L O C I T Y R E I N I T I A L I Z A T I O N
Enter an expression for the Velocity reinitialization condition e (dimensionless). If this
expression is nonzero then the effect of the Velocity Reinitialization node is applied to
the particle. This expression is evaluated at every time step taken by the solver.

Select the Specify reinitialization time check box to make the velocity reinitialization
occur at a specific time. If this check box is selected, enter a value or expression for the
Reinitialization time tr (SI unit: s). The default value is 0.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 81
 If the Specify reinitialization time check box is cleared, the reinitialization
condition is only evaluated at the beginning of each time step taken by the
solver, so each particle’s velocity can only be reinitialized once per time
step.

If the Specify reinitialization time check box is selected, the reinitialization

condition and time are reevaluated after every velocity reinitialization and
wall interaction, so it is possible for each particle’s velocity to be
reinitialized multiple times in a single time step taken by the solver.

Select an option from the Effect on primary particle list—None, Reinitialize (the
default), Freeze, Stick, or Disappear. If None is selected, the trajectory of the primary
particle is not affected, although it is still possible to release secondary particles at the
primary particle’s position and to reinitialize its auxiliary dependent variables.

For Reinitialize the velocity is reinitialized at each time step according to the expression
specified in the Reinitialized particle velocity vr (SI unit: m/s) field. Both the Velocity
reinitialization condition and the Reinitialized particle velocity can be functions of any of
the particle variables, that is, particle position, energy, and so forth.

The options Freeze, Stick, and Disappear are identical to the same options described for
the Wall node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
This section is available if an Auxiliary Dependent Variable has been added to the
model. The values of auxiliary variables can be changed whenever the Velocity
reinitialization condition is satisfied. See Wall for all settings.

Accumulator (Domain)
Use the Accumulator node to define additional degrees of freedom on a domain. Each
Accumulator defines a variable, called the accumulated variable, on each domain
element in the set of selected domains. The values of the accumulated variables are
determined by the properties of particles in each domain element.

82 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
ACCUMULATOR SETTINGS
Select an option from the Accumulator type list: Density (default) or Count.

• For Density the accumulated variable is divided by the volume of the mesh element
where it is defined.
• For Count the accumulated variable is unaffected by the element size.

Select an option from the Accumulate over list: Elements (default) or Elements and time.

• For Elements the value of the accumulated variable in an element is the sum of the
source terms of all particles in that element. If the Accumulator type is set to Density,
this sum is divided by the mesh element volume. At a later time, a particle has no
effect on the value of the accumulated variable in an element it passed through
previously.
• For Elements and time the time derivative of the accumulated variable in an element
is the sum of the source terms of all particles in that element. If the Accumulator type
is set to Density, this sum is divided by the mesh element volume. As each particle
moves through a series of mesh elements, it leaves behind a contribution to the
accumulated variable that remains even after the particle has moved on.

Enter the Accumulated variable name. The default is rpd. The accumulated variable is
defined as <name>.<varname>, where <name> is the physics interface name and
<varname> is the accumulated variable name. For example, in an instance of the
Mathematical Particle Tracing interface with default name pt and default accumulated
variable name rpd, the variable would be named pt.rpd.

Enter a Source R. The unit of the source depends on the settings in the Units section.
The source term is used to calculate the accumulated variable in a manner specified by
the Accumulate over and Accumulator type settings.

If Elements and time is selected from the Accumulate over list, select an option from the
Source interpolation list: Constant, Linear (the default), Quadratic, or Exponential. The
Source interpolation determines what functional form the Source is assumed to follow
during each time step taken by the solver. This information is used to compute the
accumulated variable in mesh elements that the particles pass through during each time
step.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 83
 UNITS
Select a Dependent variable quantity from the list; the default is Dimensionless [1]. To
enter a unit, select None from the list and in the Unit field enter a value, for example, K,
m/s, or mol/m^3.

Accumulator Theory: Domains

Particle-Particle Interaction

The Particle-Particle Interaction node is only available when Newtonian or

Newtonian, first order is selected as the Formulation in the physics interface
Particle Release and Propagation section.

This feature does not support hard sphere collisions; the interaction
forces must be finite and must be expressed as continuous functions of the
distance between particles.

Use the Particle-Particle Interaction node to define how the particles interact with each
other. There are predefined options available for the Coulomb, Lennard-Jones, and
linear elastic forces. It is also possible to define arbitrary expressions for the interaction
force.

FORCE
Select an option from the Interaction force list: Coulomb (the default), Linear elastic,
Lennard-Jones, or User defined. If the default, Coulomb, is kept, a Coulomb force
describes the interaction between charged particles. No user input is necessary.

Linear elastic
For Linear elastic enter the Spring constant ks (SI unit: N/m). The default is 1 N/m.
Enter the Equilibrium distance between particles r0 (SI unit: m). The default is 1 mm.

84 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Lennard-Jones
For Lennard-Jones it uses the Lennard-Jones potential to approximate the interaction
between neutral particles. The value of these parameters depend on the gas molecules
interacting and can usually be found from a literature search.

• Enter the Collision diameter σ (SI unit: m), typically in the order of a few angstroms
(the default value is 3.3 × 10−10 m or 3.3 Å).
• Enter the Interaction strength ε (SI unit: J), usually in the order of 10−21 Joules (the
default value is 1.6 × 10−21 J).

User Defined

There is specific syntax that you must use to specify user-defined

interaction forces.

For User defined enter a user-defined expression for the interaction Force Fu
(SI unit: N) based on space dimension.

The particle degrees of freedom are given the variable names qx, qy, and qz (in 3D),
but to access the position vector of neighboring particles use the expression dest(qx),
dest(qy), and dest(qz).

A predefined expression for the distance between particles is available because the
particle-particle interaction forces usually depend inversely on the distance between
the particles. This expression is accessed using <name>.r, where <name> is the physics
interface node name. So, if the name is pt, then this variable is accessed using pt.r.

As an example, the gravitational force on particle i depends on the position vector and
mass of all other particles:

N
( ri – rj )
 --------------------
2
Fi = –G m -
3
r –r i j
j=1

where ri is the position vector of the ith particle, G is the gravitational constant, and
m is the mass. To enter this as a user-defined force, enter (in 2D):

G*m^2*(qx-dest(qx))/sqrt((qx-dest(qx))^2+(qy-dest(qy))^2+tol)^3
G*m^2*(qy-dest(qy))/sqrt((qx-dest(qx))^2+(qy-dest(qy))^2+tol)^3

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 85
 where tol is a user-defined parameter to prevent divide by zero for the ith particle. In
practice it is quite difficult to choose the value of tol. It should in general be a small
fraction of the smallest distance you want to allow between particles.

USING THE DEST() OPERATOR

When reading or writing the expressions for a user-defined particle-particle interaction
force, the dest() operator is used to identify properties of the particle that is being
subjected to a force, while the omission of the dest() operator indicates some
property or a particle that is exerting the force. For example, when computing the
force on the ith particle, the expression qx-dest(qx) is the difference between the
x-coordinate of some other particle (say, the jth particle) and the x-coordinate of the
ith particle.

ADVANCED SETTINGS
The fact that all particles can interact with all other particles in the system means that
a full Jacobian matrix is generated at each time step when solving. Assembly and
factorization of such a matrix is very expensive in both time and memory.

By default, the Exclude Jacobian contribution for particle-particle interaction force check
box is selected. This means the contribution to the Jacobian matrix is ignored due to
the particle-particle interaction force. This also means that the problem solves much
faster and requires much less memory. The drawback of this is that the Jacobian is not
exact, and the solver therefore needs to take very small time steps when solving.

Rotating Galaxy: Application Library path

Particle_Tracing_Module/Tutorials/rotating_galaxy

By default, the Apply cutoff length check box is cleared. If this check box is selected,
enter a value or expression for the Cutoff length rc (SI unit: m). The default is 5 mm.
If the distance between particles is greater than the cutoff length, the particle-particle
interaction force between them is set to zero.

Release
Use the Release node to release particles into the model from a selected set of domains.
The release times, initial position, initial velocity, and initial value of any auxiliary
dependent variables can be specified. The sections available are based on the
Formulation in the physics interface Particle Release and Propagation section. If

86 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Newtonian, Newtonian, first order, Lagrangian, or Hamiltonian are selected, also enter an
initial velocity.

RELEASE TIMES
Enter Release times (SI unit: s) or click the Range button ( ) to select and define a
range of specific times. At each release time, particles are released with initial position
and velocity as defined in the following sections.

For models using The Charged Particle Tracing Interface the Release
Current Magnitude section is available instead of Release Times when the
Particle release specification is set to Specify current. Enter a value for the
Release current magnitude I (SI unit: A). The default is 1 nA. For models
using The Particle Tracing for Fluid Flow Interface the Mass Flow Rate
section is available instead of Release Times when the Particle release
specification is set to Specify mass flow rate. Enter a value for the Mass flow
·
rate m (SI unit kg/s). The default is 1 mg/s. In either case, all particles
are released at time t = 0. The expression for the Release current
magnitude or Mass flow rate should only depend on parameters and global
variables, not on particle variables such as the particle mass.

INITIAL POSITION
Select an Initial position—Mesh based (the default), Density, or Random.

Mesh Based
For Mesh based the particles are released from a set of positions determined by a
selection of geometric entities (of arbitrary dimension) in the mesh.

Select a Refinement factor between 1 and 5 (the default is 1), where the centers of the
refined mesh elements are used. Thus, the number of positions per mesh element is
2
refine^dim, except for pyramids, where it is (4*refine -1)*refine/3.

Density
For Density the particles are positioned in the selected domains by sampling from a
user-defined spatial distribution. Enter a value for the Number of particles per release N
(dimensionless). The default is 1. Enter a value or expression for the Density
proportional to ρ (dimensionless). The default is 1.

The field Density proportional to is an expression—the resulting particle distribution

approximately has a density that is proportional to this expression. The resulting

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 87
 distribution looks a bit random, and it depends on the order in which the mesh
elements are numbered.

The distribution is not necessarily the same in different COMSOL

versions, but the total number of particles released is always N.

Select a Release distribution accuracy order between 1 and 5 (the default is 5), which
determines the integration order that is used when computing the number of particles
to release within each mesh element. The higher the accuracy order, the more
accurately particles will be distributed among the mesh elements.

The Position refinement factor (default 0) must be a nonnegative integer. When the
refinement factor is 0, each particle is always assigned a unique position, but the density
is taken as a uniform value over each mesh element. If the refinement factor is a positive
integer, the distribution of particles within each mesh element is weighted according
to the density, but it is possible for some particles to occupy the same initial position.
Further increasing the Position refinement factor increases the number of evaluation
points within each mesh element to reduce the probability of multiple particles
occupying the same initial position.

If the model includes a Particle-Particle Interaction then there is a risk

that the force may become infinite if the initial particle positions are not
unique. When modeling particle-particle interactions the Position
refinement factor should be 0 to ensure that initial positions are unique.
However, it may then be necessary to refine the mesh to get an accurate
initial distribution of particle positions if the density is not uniform.

Random
For Random the particles are released at random positions within the selected entities.
If particles are released at multiple release times, the initial positions are uniquely
generated for each release time. By contrast, for Density the particle positions are the
same at each release time.

The particle positions are determined in the following way. First a random mesh
element is selected for each particle with probability proportional to the element size,
so that particles are more likely to be released from larger mesh elements than smaller
elements. After the mesh element is selected, random local coordinates are chosen

88 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
within the element, then converted to global coordinates. The particle is then placed
at this location.

The initial particle coordinates are pseudorandom; that is, they are not
derived from a natural entropy source. Therefore, if the same study is
re-run on the same architecture multiple times, the same initial particle
positions will be used each time.

The Number of particles per release is the number before velocity, release time and
auxiliary dependent variable multiplication, so the actual number of model particles
generated by the Release feature may be significantly larger.

I N I T I A L VE L O C I T Y
For The Mathematical Particle Tracing Interface this section is shown when Newtonian,
Newtonian, first order, Lagrangian, or Hamiltonian is selected as the Formulation.

For The Charged Particle Tracing Interface and The Particle Tracing for Fluid Flow
Interface this section is shown when Newtonian or Newtonian, first order is selected as
the Formulation.

Select an option from the Initial velocity list: Expression (the default), Kinetic energy and
direction, Constant speed, spherical, Constant speed, hemispherical, Constant speed, cone,
Constant speed, Lambertian (3D only), or Maxwellian.

• For Expression enter coordinates for the Initial particle velocity v0 (SI unit: m/s)
based on space dimension. The defaults are 0 m/s.
• For Kinetic energy and direction enter the Initial kinetic energy E0 (SI unit: J). The
default is 1 keV. Then enter coordinates for the Initial particle direction L0
(dimensionless) based on space dimension. It is not necessary to normalize the
components of the initial direction vector because this is done automatically when
computing the initial velocity.

The option Kinetic energy and direction is not supported for The Particle
Tracing for Fluid Flow Interface.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 89
 • For Constant speed, spherical a total of Nvel particles are released at each release
point, each with the same speed. The velocity distribution is spherical for 3D
components and circular for 2D. Enter the following:
- Speed v (SI unit: m/s). The default is 1 m/s.
- Number of particles in velocity space Nvel (dimensionless). The default is 200.

• For Constant speed, hemispherical a total of Nvel particles are released at each release
point, each with the same speed. The velocity distribution is hemispherical for 3D
components and a half circle for 2D. Enter the:
- Speed v (SI unit: m/s). The default is 1 m/s.
- Number of particles in velocity space Nvel (dimensionless). The default is 200.
- A direction vector for the Hemisphere axis r (dimensionless). This sets the
direction of the north pole of the hemisphere or half circle.

• For Constant speed, cone a total of Nvel particles are released at each release point,
each with the same speed. The velocity distribution is a cone of angle α for 3D
components and an arc of half-angle α for 2D. Enter the:
- Speed v (SI unit: m/s). The default is 1 m/s.
- Number of particles in velocity space Nvel (dimensionless). The default is 200.
- A direction vector for the Cone axis r (dimensionless). This sets the direction of
the cone axis for 3D components or the centerline of the arc for 2D.
- A Cone angle α (SI unit: rad).

• The Constant speed, Lambertian option is only available in 3D. It generates Nvel
particles for each release point, each with the same speed. The velocity distribution
is a hemisphere in which the distribution of polar angles follows the cosine law.
Enter the:
- Speed v (SI unit: m/s). The default is 1 m/s.
- Number of particles in velocity space Nvel (dimensionless). The default is 200.
- A direction vector for the Hemisphere axis r (dimensionless). This sets the
direction of the north pole of the hemisphere.

90 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
• For Maxwellian enter the:
- Number of particles per velocity space direction Nv. The default is 50.
- Temperature T0 (SI unit: K). The default is 293.15 K.

This option can generate an overwhelming number of particles because

for each release time and release position, this number is multiplied by
Nvsdim, where sdim is the number of space dimensions. Typically only
choose this option when releasing particles at a single release time at a
single point in space.

For Constant speed, spherical, Constant speed, hemispherical, Constant speed, cone,
Constant speed, Lambertian, and Maxwellian, select an option from the Sampling from
distribution list: Deterministic (the default) or Random. If Deterministic is selected, the
initial velocity is computed by sampling from the velocity distribution in a
deterministic and reproducible manner. If Random is selected, the particle velocity is
sampled from the distribution using pseudo-random numbers and may not always
generate the same results.

For Constant speed, spherical, Constant speed, hemispherical, Constant speed, cone, and
Constant speed, Lambertian, select the Allow nonuniform speeds in distribution check box
to allow particles from the same release point to have different initial speeds. This
check box has no effect on the initial velocity directions of the released particles, only
their velocity magnitudes. If this check box is selected, expressions for the initial speed
in terms of the particle index can return a unique value for each particle in the
distribution.

For example, if Constant speed, hemispherical is selected from the Initial

velocity list, the Allow nonuniform speeds in distribution check box is
selected, and the expression for the initial speed is
2[m/s]*(random(pt.pidx)+0.5), then particles will be released with
directions sampled from an isotropic hemisphere and with speeds sampled
pseudorandomly between 0 and 2 m/s.

INHERIT PROPERTIES
Select an option from the Inherit Properties list. The Particle Properties node (default)
is always available. If any Override Properties nodes have been added to the model,

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 91
 they can also be selected. Use this setting to determine what set of properties is given
to the particles that are released by this feature.

I N I T I A L P A R T I C L E TE M P E R A T U R E
This section is only available for The Particle Tracing for Fluid Flow Interface and is
only shown when the Compute particle temperature check box is selected in the physics
interface Additional Variables section. Enter a value or expression for the Initial particle
temperature Tp,0 (SI unit: K). The default value is 293.15[K].

INITIAL PARTICLE MASS

This section is only available for The Particle Tracing for Fluid Flow Interface and is
only shown when the Compute particle mass check box is selected in the physics
interface Additional Variables section.

The initial particle mass can be a single value or can be sampled from a distribution.
Select an option from the Distribution function list—None (the default), Normal,
Lognormal, Uniform, or List of values.

When None is selected, enter an initial value mp,0. The default value is 10-9 kg.

Select Normal to create a normal distribution function, Lognormal to create a

log-normal distribution function, or Uniform to create a uniform distribution function.
The Number of values N (default 1) sets the number of points in the distribution
function. Enter the Mean particle mass (default 10-9 kg) and Particle mass standard
deviation (default 10-10 kg). Select List of values to enter a list of numerical values for
the initial particle mass directly.

INITIAL MULTIPLICATION FACTOR

This section is only available for The Particle Tracing for Fluid Flow Interface and is
only shown when the Enable macroparticles check box is selected in the physics
interface Additional Variables section. Enter a value or expression for the Initial
multiplication factor nn,0 (dimensionless). The default value is 1.

I NI TI AL VAL UE O F AUX I L IA R Y D EP E ND EN T VAR IA BL E S

This section is available if an Auxiliary Dependent Variable has been added to the
model.

For each of the Auxiliary Dependent Variables added to the model, choose a
Distribution function for the initial value of the auxiliary dependent variables and
whether the initial value of the auxiliary dependent variables should be a scalar value
or a distribution function.

92 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
The number of particles simulated can increase substantially and the following options
are available for each of the Auxiliary Dependent Variables added to the model.

When None is selected, enter an initial value (rp0).

For the initial value of the auxiliary dependent variables, select Normal to sample from
a normal distribution function, Lognormal to sample from a log-normal distribution
function, or Uniform to sample from a uniform distribution function. For any selection,
the Number of values sets the number of points in the distribution function. Enter a
user-defined Mean (default 0) and Standard deviation (default 1). Select List of values to
enter a list of numerical values for the auxiliary dependent variable directly.

By default auxiliary dependent variables are initialized after all other degrees of
freedom. Select the Initialize before particle momentum check box to compute the
initial value of the auxiliary dependent variable immediately after computing the initial
particle position. By selecting this check box it is possible to define the initial particle
velocity as a function of the auxiliary dependent variables. The order of initialization
for the auxiliary dependent variables is described in more detail in Initialization of
Auxiliary Dependent Variables in the Particle Tracing Modeling chapter.

ADVANCED SETTINGS
This section is shown if Newtonian or Newtonian, first order is selected from the
Formulation list in the physics interface Particle Release and Propagation section.

By default the Subtract moving frame velocity from initial particle velocity check box is
cleared. Select this check box to offset the initial velocity of released particles if the
frame is moving. The frame velocity is taken from an active instance of the Rotating
Frame feature if it has been added to the model. Thus, if the particle trajectories are
computed in a rotating frame, clearing this check box means that the particles are
released with respect to the rotating frame; selecting the check box means that the
particle velocities are taken with respect to the non-rotating (inertial) frame.

Axial Symmetry
An Axial Symmetry node is automatically added to 2D axisymmetric components. The
options available for the feature are exactly the same as for Wall, except the Boundary
Selection is locked and is only applicable on the symmetry axis. The Accumulator
(Boundary) and/or Secondary Emission subnodes are available from the context
menu (right-click the parent node) or from the Physics toolbar, Attributes menu.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 93
 Inlet
Use the Inlet node to release particles into the modeling domain from selected
boundaries. The Nonlocal Accumulator subnode is available from the context menu
(right-click the parent node) or from the Physics toolbar, Attributes menu.

Go to Release for information about the following sections: Release Times, Release
Current Magnitude, Mass Flow Rate, Inherit Properties, Initial Particle Temperature, Initial
Particle Mass, Initial Multiplication Factor, Initial Value of Auxiliary Dependent Variables,
and Advanced Settings.

COORDINATE SYSTEM SELECTION

It is possible to specify the initial particle velocity in terms of the global coordinates or
in another coordinate system defined for the model Component. Select an option from
the Coordinate system list. By default Global coordinate system is selected. If other
coordinate systems are defined, they can also be selected from the list. When specifying
the initial particle velocity (see the Initial Velocity section), velocity or direction
components can be specified using the basis vectors of whichever coordinate system
has been selected from the list.

When a coordinate system other than Global coordinate system is selected from the
Coordinate system list, arrows will appear in the Graphics window to indicate the
orientation of the basis vectors of the coordinate system on the selected boundaries.

INITIAL PO SITION
Select an Initial position—Mesh based (the default), Uniform distribution (2D
components) Projected plane grid (3D components), Density, or Random. Mesh Based,
Density, and Random have the same settings as described for the Release node.

For 2D components, if Uniform distribution is selected, enter the Number

of particles per release N (dimensionless). The default is 1. The union of
the selected edges is divided into N segments of approximately equal
length, and a particle is placed in the middle of each segment.

94 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 For 3D components, if Projected plane grid is selected, enter the Number
of particles per release N (dimensionless). The default is 1. The particles
are distributed on a plane grid in planes that are approximately tangential
to the selected boundaries (for a plane boundary, you thus get a uniform
distribution).

For inlets defined on identity pairs in an assembly, select an option from the Release
particles from boundaries list: Source (the default), Destination, or Source and
Destination. This option determines which boundary mesh to use when determining
the initial particle positions. The mesh on either side of an inlet pair can be quite
different, as shown in Figure 3-2.

Figure 3-2: Mesh-based release of particles from a source boundary (left), destination
boundary (middle), or both source and destination (right). In each rectangle, the source
boundary is on the side of the lighter-colored mesh.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 95
 I N I T I A L VE L O C I T Y
Select an Initial velocity: Expression (the default), Kinetic energy and direction, Constant
speed, hemispherical, Constant speed, cone, or Constant speed, Lambertian (3D only).

• For Expression enter expressions for the Initial particle velocity v0 (SI unit: m/s)
based on space dimension. Select the Specify tangential and normal vector
components check box to enter expressions for the initial velocity in terms of the
tangential and normal directions at the surface.

When using The Particle Tracing for Fluid Flow Interface, enter a value
or expression for the Velocity field u. If another physics interface is present
which computes the velocity field then this can be selected from the list.

• For Kinetic energy and direction enter the Initial kinetic energy E0 (SI unit: J). The
default is 1 keV. Then enter coordinates for the Initial particle direction L0
(dimensionless) based on space dimension. It is not necessary to normalize the
components of the initial direction vector because this is done automatically when
computing the initial velocity. Select the Specify tangential and normal vector
components check box to enter expressions for the initial direction in terms of the
tangential and normal directions at the surface.

The option Kinetic energy and direction is not supported for The Particle
Tracing for Fluid Flow Interface.

• For Constant speed, hemispherical enter a Speed v0 (SI unit: m/s), the Number of
particles in velocity space Nvel (dimensionless), and the Hemisphere axis r. Select the
Specify tangential and normal vector components check box to enter expressions for
the hemisphere axis in terms of the tangential and normal directions at the surface.
• For Constant speed, cone enter a Speed v0 (SI unit: m/s), the Number of particles in
velocity space Nvel (dimensionless), the Cone axis r, and the Cone angle α (SI
unit: rad). Select the Specify tangential and normal vector components check box to
enter expressions for the cone axis in terms of the tangential and normal directions
at the surface.
• The Constant speed, Lambertian option is only available in 3D. Enter a Speed v0
(SI unit: m/s), the Number of particles in velocity space Nvel (dimensionless), and
the Hemisphere axis r. The velocity distribution is a hemisphere in which the
distribution of polar angles follows the cosine law, similar to the Diffuse scattering
option for the Wall feature. Select the Specify tangential and normal vector

96 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 components check box to enter expressions for the hemisphere axis in terms of the
tangential and normal directions at the surface.

When the Specify tangential and normal velocity components or Specify

tangential and normal vector components check box is selected, arrows
indicating the normal direction on the selected boundaries will appear in
the Graphics window.

Note that the normal direction may be opposite the built-in variable for
the boundary normal (for example, nx, ny, and nz) to ensure that a
positive value causes particles to be released into the simulation domain.
This often occurs when the Inlet is applied to exterior boundaries.

When the normal direction used by the Inlet feature is opposite the
normal vector defined by the geometry, the tangential directions are
similarly inverted to ensure that the boundary coordinate system is
right-handed.

For Constant speed, hemispherical, Constant speed, cone, and Constant speed,
Lambertian, select an option from the Sampling from distribution list: Deterministic (the
default) or Random. If Deterministic is selected, the initial velocity is computed by
sampling from the velocity distribution in a deterministic and reproducible manner. If
Random is selected, the particle velocity is sampled from the distribution using
pseudo-random numbers and may not always generate the same results.

For Constant speed, spherical, Constant speed, hemispherical, Constant speed, cone, and
Constant speed, Lambertian, select the Allow nonuniform speeds in distribution check box
to allow particles from the same release point to have different initial speeds. This
check box has no effect on the initial velocity directions of the released particles, only
their velocity magnitudes. If this check box is selected, expressions for the initial speed
in terms of the particle index can return a unique value for each particle in the
distribution.

For example, if Constant speed, hemispherical is selected from the Initial

velocity list, the Allow nonuniform speeds in distribution check box is
selected, and the expression for the initial speed is
2[m/s]*(random(pt.pidx)+0.5), then particles will be released with
directions sampled from an isotropic hemisphere and with speeds sampled
pseudorandomly between 0 and 2 m/s.

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 97
 About the Inlet Boundary Conditions

Nonlocal Accumulator
Use the Nonlocal Accumulator subnode to communicate information from a particle’s
current position to the surface from which it was released.

The subnode is available from the context menu (right-click the Inlet parent node) or
from the Physics toolbar, Attributes menu.

Each Nonlocal accumulator subnode defines a variable, called the accumulated variable,
that is computed using variables defined on particles released by the parent Inlet node
or on domains and boundaries encountered by such particles.

ACCUMULATOR SETTINGS
Select an Accumulator type: Density (default) or Count.

• For Density the accumulated variable is divided by the volume of the mesh element
where it is defined.
• For Count the accumulated variable is unaffected by the element size.

Select an option from the Accumulate over list: Elements (default) or Elements and
time.

• For Elements the accumulated variable is proportional to the instantaneous value of

the Source term R for all applicable particles.
• For Elements and time the time derivative of the accumulated variable is proportional
to the instantaneous value of the Source term R for all applicable particles, and thus
the accumulated variable considers the time history of particles in the modeling
domain instead of just their current values.

Enter the Accumulated variable name. The default is rpi.

Enter a Source R. The unit of the source depends on the settings in the Units section.
The source term is used to calculate the accumulated variable in a manner specified by
the Accumulate over and Accumulator type settings.

98 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Select a Source geometric entity level — Domains, Boundaries, or Domains and
boundaries.

• If Domains is selected, particles only contribute to the accumulated variable on their

releasing surface if they are still active; that is, they are still propagating through a
domain.
• If Boundaries is selected, the particles only contribute to the accumulated variable if
they have become stuck or frozen to a boundary somewhere in the model.
• If Domains and boundaries is selected, all of the active, stuck, and frozen particles
released by a feature can contribute to the accumulated variable.

UNITS
Select a Dependent variable quantity from the list; the default is Dimensionless [1]. To
enter a unit, select None from the list and in the Unit field enter a value, for example, K,
m/s, or mol/m^3.

SMOOTHING
The accumulated variable is piecewise discontinuous across different boundary
elements. Select the Compute smoothed accumulated variable check box to define a
spatially smoothed accumulated variable; at each point on the boundary, the smoothed
accumulated variable is the average of the accumulated variable over a circle. Enter a
Smoothing radius r (SI unit: m). The default is 0.1 m.

Outlet
Use the Outlet node to determine what should happen to particles when exiting the
modeling domain. The Accumulator (Boundary) subnode is available from the context
menu (right-click the parent node) or from the Physics toolbar, Attributes menu.

OUTLET
Select a Wall condition — Freeze (the default) or Disappear. When Freeze is selected, the
particle position and velocity remains frozen once the particle has made contact with
the outlet boundary. This allows information about the particle velocity, energy and so
forth to be recovered at the instant the particles left the modeling domain. For

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 99
 Disappear it means that the particles are not visualized once they strike the outlet
boundary.

• About the Outlet Boundary Conditions

• Theory for the Mathematical Particle Tracing Interface

Particle Continuity
Use the Particle Continuity node to specify that particles should cross a pair boundary
as if it were invisible. Pair boundaries appear when the geometry sequence ends in Form
Assembly instead of Form Union. Such boundaries require special handling because the
mesh elements on either side of the pair boundary aren’t required to match up exactly.

PAIR SELECTION
Select one or more identity pairs to allow particles to cross between the source and
destination boundaries of these pairs. Such identity pairs are typically created
automatically on interior boundaries when the geometry sequence ends in a Form
Assembly node instead of a Form Union node.

The Particle Continuity node does not cause the particle position
components to change discontinuously; the source and destination
boundaries for the identity pairs must be overlapping.

If the geometry sequence ends in a Form Union node, the Particle

Continuity node usually isn’t necessary because particles can freely cross
interior boundaries where no boundary condition has been applied.

Auxiliary Dependent Variable

Use the Auxiliary Dependent Variable node to add additional degrees of freedom for
each particle. These can be used to solve for the particle size, path length, residence
time, number of wall collisions, spin, and so forth.

• Auxiliary Dependent Variables and Residence Time

• Initialization of Auxiliary Dependent Variables

100 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
AUXILIAR Y DEPEN DENT VARIABLE
Enter a Field variable name. The default is rp and can be changed to anything provided
it does not conflict with the name of the variables for the position or momentum
degrees of freedom. The name should not conflict with other auxiliary dependent
variables.

Enter a Source R. The unit of the source depends on the settings in the Units section.

Under Integrate chose whether to integrate the equation you have defined With respect
to time or Along particle trajectory. For example, to compute the residence time of a
group of particles in a given system, set the Source to 1 and set Integrate to With respect
to time. To compute the length of the particle trajectory, set the Source to 1 and set
Integrate to Along particle trajectories.

UNITS
Select a Dependent variable quantity from the list; the default is Dimensionless [1]. To
enter a unit, select None from the list and in the Unit field enter a value, for example, K,
m/s, or mol/m^3.

Release from Edge

Use the Release from Edge node to release particles into the modeling domain from a
set of selected edges in a 3D geometry.

Go to Release for information about the following sections: Release Times, Release
Current Magnitude, Mass Flow Rate, Initial Velocity, Inherit Properties, Initial Particle
Temperature, Initial Particle Mass, Initial Multiplication Factor, Initial Value of Auxiliary
Dependent Variables, and Advanced Settings.

INITIAL POSITION
Select an Initial position—Mesh based (the default), Density, or Random. Mesh Based,
Density, and Random have the same settings as described for the Release node. If
Uniform distribution is selected, enter the Number of particles per release N
(dimensionless). The union of the selected edges is divided into N segments of
approximately equal length, and a particle is placed in the middle of each segment.

Release from Point

Use the Release from Point node to release particles into the modeling domain from
selected points in the geometry. By default, one particle is released at each selected

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 101
 point, although a larger number of particles will be released if the initial velocity or an
auxiliary dependent variable is sampled from a distribution of values.

Go to Release for information about the following sections: Release Times, Release
Current Magnitude, Mass Flow Rate, Initial Velocity, Inherit Properties, Initial Particle
Temperature, Initial Particle Mass, Initial Multiplication Factor, Initial Value of Auxiliary
Dependent Variables, and Advanced Settings.

Release from Grid

Use the Release from Grid node to release particles from a grid of initial positions with
user-defined coordinates.

Go to Release for information about the following sections: Release Times, Release
Current Magnitude, Mass Flow Rate, Initial Velocity, Inherit Properties, Initial Particle
Temperature, Initial Particle Mass, Initial Multiplication Factor, Initial Value of Auxiliary
Dependent Variables, and Advanced Settings.

INITIAL COORDINATES
Enter Initial coordinates based on space dimension (qx,0, qy,0, and qz,0 for 3D
components) for the particle positions or click the Range button ( ) to select and
define a range of specific coordinates.

Select an option from the Grid type list: All combinations (the default) or Specified
combinations. If Specified combinations is selected, the number of initial coordinates
entered for each space dimension must be equal, and the total number of particles
released is equal to the length of one of the lists of initial coordinates. If All
combinations is selected, the total number of particles released is equal to the product
of the lengths of each list of initial coordinates.

For example, suppose a 2D model component includes a Release from Grid node with
the following initial coordinates:

• qx,0 = range(0,1,3)
• qy,0 = range(2,2,8)

If All combinations is selected, a total of 16 particles will be released, including every

possible combination of the initial x- and y-coordinates. If Specified combinations is
selected, 4 particles will be released with initial positions (0,2), (1,4), (2,6), and (3,8).

The number of released particles may increase if the initial velocity or initial values of
auxiliary dependent variables involve sampling from a distribution, in which case a

102 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
large number of particles may be released at each initial position. The position for any
particles with initial coordinates outside the geometry are set to not-a-number (NaN),
so the particles do not appear when plotted during postprocessing.

Override Properties
By default, all particles are assigned properties such as mass, charge number, and
Lagrangian based on the settings entered for Particle Properties. Add an Override
Properties node to create alternative sets of properties that can be assigned to the
particles. The user inputs in the Settings window for Override Properties are identical to
those in the Settings window for Particle Properties. The properties in an Override
Properties node can then be applied to particles using the Inherit Properties section in
the Release, Inlet, Release from Grid, and Secondary Emission features.

Overriding Particle Properties

Release from Data File

Use a Release from Data File node to release particles by specifying the initial position,
velocity, and values of auxiliary dependent variables using data from a text file.

Go to Release for information about the following sections: Release Times, Release
Current Magnitude, Mass Flow Rate, Inherit Properties, Initial Particle Temperature, Initial
Multiplication Factor, and Advanced Settings.

Formatting Guidelines for Files Containing Particle Data

The imported data file should be a text file (*.txt) arranged in a spreadsheet format;
that is, each row corresponds to a distinct particle and should have the same number
of columns as all other rows. Columns can be separated by spaces, tabs, or a
combination of the two. Begin a line with the percent (“%”) character to include
comments or empty lines in the data file.

For example, a data file containing the following text would insert particles at the
positions (0.1, 0.2, 0.6) and (0.2, 0.4, 0.8) in a three-dimensional geometry:

% Initial particle positions

% qx0 qy0 qz0
0.1 0.2 0.6
0.2 0.4 0.8

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 103
 INITIAL PO SITION
Browse your computer to select a text file, then click Import to import the data. To
remove the imported data, click Discard. Enter the Index of first column containing
position data i to indicate which column represents the first coordinate of the particle
position vectors. The default value, 0, indicates the first column.

I N I T I A L VE L O C I T Y
Select an option from the Initial velocity list: Expression (the default), From file, Kinetic
energy and direction, Constant speed, spherical, Constant speed, hemispherical, Constant
speed, cone, Constant speed, Lambertian (3D only), or Maxwellian.

• For From file, enter the Index of first column containing velocity data i. The default is
3. The columns are zero-indexed; that is, an index of 0 corresponds to the first
column.
• For all other options, the settings are the same as for the Release node.

INITIAL PARTICLE MASS

This section is only available for The Particle Tracing for Fluid Flow Interface and is
only shown when the Compute particle mass check box is selected in the physics
interface Additional Variables section.

The initial particle mass can be a single value or can be sampled from a distribution.
Select an option from the Distribution function list—From file, None (the default),
Normal, Lognormal, Uniform, or List of values.

• When From file is selected, enter the Index of column containing mass data. The
default value is 3. The columns are zero-indexed; that is, an index of 0 corresponds
to the first column.
• For all other options, the settings are the same as for the Release node.

I NI TI AL VAL UE O F AUX I L IA R Y D EP E ND EN T VAR IA BL E S

This section is available if an Auxiliary Dependent Variable has been added to the
model.

104 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
For each of the active Auxiliary Dependent Variable nodes in the model, choose an
option from the Distribution function list: From file, None (the default), Normal,
Lognormal, Uniform, or List of Values.

• When From file is selected, enter the Index of column containing data. The default
value is 3. The columns are zero-indexed; that is, an index of 0 corresponds to the
first column.
• For all other options, the settings are the same as for the Release node..

T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 105
 Theory for the Mathematical Particle
Tracing Interface
The equation system solved by the Mathematical Particle Tracing interface depends on
the formulation selected. In the following, the particle has a position vector q, a
velocity vector v, and a momentum vector p.

The Mathematical Particle Tracing Interface theory is described in this section:

• Newtonian Formulation
• Hamiltonian Formulation
• Lagrangian Formulation
• Massless Formulation
• Particle Tracing in Rotating Frames
• Initial Conditions — Position
• Initial Conditions — Velocity
• Particle-Particle Interactions
• Auxiliary Dependent Variables
• About the Boundary Conditions for the Particle Tracing Interfaces
• Accumulator Theory: Domains
• Accumulator Theory: Boundaries
• References for the Mathematical Particle Tracing Interface

Newtonian Formulation
The Mathematical Particle Tracing interface makes it possible to solve ordinary
differential equations for the particle positions. When the Newtonian formulation is
selected, the particle position is computed using Newton’s second law:

d m
( v) = F
dt p

where mp is the particle mass (SI unit: kg), v is the particle velocity (SI unit: m/s), and
F is the force exerted on the particle (SI unit: N). The force is either specified directly
or using a susceptibility multiplied by a suitable field:

106 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 F = χ⋅Γ

where χ is the susceptibility tensor and Γ is a vector field. The particle velocity is
defined as:

dq
v =
dt

where q is the particle position vector (SI unit: m).

When using a relativistic correction of the particle mass for particles with very high
velocity, the particle mass mp is defined as

mr
m p = ----------------------------------
2
1–v⋅v⁄c

where mr is the rest mass (SI unit: kg) and c = 2.99792458 × 108 m/s is the speed of
light in a vacuum, which is a built-in physical constant.

If the default Newtonian formulation is selected, Newton’s second law is expressed as a

set of second-order ordinary differential equations for the particle position vector
components:

d  m dq
------- = F
d t  p dt 

If the first-order Newtonian formulation is selected, Newton’s second law is instead

expressed as a set of coupled first-order ordinary differential equations:

d m
( v) = F
dt p

dq
v =
dt

USING FIRST-ORDER AND SECOND-ORDER FORMULATIONS

The advantage of using the first-order Newtonian formulation is that it avoids mixing
first- and second-order equations when any Auxiliary Dependent Variables are also
solved for, since the equations for auxiliary dependent variables are always first-order.
This allows efficient, high-order explicit time stepping methods to be used for a wider
class of problems compared to the second-order Newtonian formulation.

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 107
 The explicit time stepping methods, such as Runge-Kutta methods, are most suitable
for nonstiff problems. In a particle tracing context this often means that the particle
position, velocity, or other dependent variables aren’t expected to change abruptly over
an extremely short time interval.

Hamiltonian Formulation
Sometimes it is more convenient to describe the particle motion using a Hamiltonian
formulation. In this case, the following system of first-order ODEs is solved:

dq
= v H ( q, t )
dt

dp
= F H ( q , v, t )
dt

The particle velocity vH (SI unit: m/s) and particle force FH (SI unit: N) can be
entered directly, or, more conveniently by specifying the Hamiltonian, H (SI unit: J).
The Hamiltonian is related to the particle velocity and force via:

∂H
vH =
∂p

and

∂H
FH = –
∂q

For a more in-depth example, see Choosing a Formulation in the Particle Tracing
Modeling chapter.

Lagrangian Formulation
Instead of a Hamiltonian, a Lagrangian formulation can be used, in which case
Lagrange’s equation is solved:

d  ∂L = ∂L
d t  ∂ v ∂q

Lagrange’s equation is rearranged before solving into the following form:

108 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 ∂  m ∂q = N – 1  ∂L –  ⋅ ∇∂L 
v
∂ t ∂t  ∂q  ∂ v 

where N is the Hessian of the Lagrangian with respect to the velocities:

2
∂ L
N ij =
∂ v i ∂v j

For a more in-depth example, see Choosing a Formulation in the Particle Tracing
Modeling chapter.

Massless Formulation
The pathline of a massless particle is given by:

dq
= v
dt

where v is a user-defined velocity vector.

When the Newtonian, Lagrangian, or Massless formulations are used there are degrees
of freedom for the particle position only. When the Newtonian, first order formulation
is used there are degrees of freedom for the particle position and velocity. When the
Hamiltonian formulation is used, there are degrees of freedom for the particle position
and momentum.

Particle Tracing in Rotating Frames

The Rotating Frame feature exerts fictitious forces on particles that are moving in a
rotating frame of reference. Optionally, it also applies an offset to the initial velocity of
released particles, depending on whether their velocity is initialized with respect to the
inertial (nonaccelerating) frame or the rotating frame.

Consider a particle of mass mp (SI unit: kg) moving in a noninertial frame of reference
with translational acceleration W (SI unit: m/s2) that is rotating at angular velocity Ω
(SI unit: rad/s) about a center of rotation with position rbp (SI unit: m). Note that Ω
is a vector quantity that also indicates the orientation of the axis of rotation and the
sense of rotation (clockwise or counterclockwise), according to the standard
right-hand rule; that is, if your right thumb points in the direction of the angular
velocity vector, your fingers sweep in the direction of rotation. The total fictitious force
Ffr (SI unit: N) exerted on a particle in this noninertial frame of reference is (Ref. 1)

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 109
 F fr = F cen + F cor + F eul + F acc

where Fcen is the centrifugal force,

F cen = m p Ω × ( r × Ω )

Fcor is the Coriolis force,

F cor = 2m p v × Ω

Feul is the Euler force,

·
F eul = m p r × Ω

and Facc is the force resulting from linear acceleration of the frame,

F acc = – m p W

The displacement r (SI unit: m) of the particle is defined with respect to the center of
rotation,

r = q – r bp

where q (SI unit: m) is the particle position. The Rotating Frame feature only accounts
for rotational motion of the frame, not translational motion, so the term Facc is
canceled.

When the Rotating Frame feature is active, it is possible to subtract the frame velocity
from the initial particle velocity when releasing particles. This is equivalent to
specifying the initial particle velocity with respect to the inertial frame. The relationship
between the initial velocity in the inertial frame vi,in and in the rotating frame vi,rot is

v i, in = v i, rot – v f

where vf is the frame velocity at the particle’s position,

vf = Ω × r

Initial Conditions — Position

Initial conditions for the particle position are either mesh based, from a user-specified
grid, uniformly distributed or based on an analytic expression.

110 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Initial Conditions — Velocity
There are several options available for specifying the initial velocity of particles:

EXPRESSION
The default is to specify an initial expression for each component of the velocity, v0. In
the case of the Newtonian, first-order Newtonian, and Lagrangian formulations the
following condition is implemented:

dq
= v0
dt

In this case the initial velocity of every particle is defined using the same expression.
The particles may have different initial velocity if v0 is a function of initial position.

KINETIC ENERGY AND DIRECTION

If the initial kinetic energy E0 (SI unit: J) and initial particle direction L0
(dimensionless) are specified, for the Newtonian, first-order Newtonian, and
Lagrangian formulations the following condition is implemented:

dq L 0 2E
= ---------- ----------0
dt L0 mp

where mp is the mass of the particle. If either Newtonian formulation is used and a
relativistic correction is enabled, instead the following condition is implemented:

dq cL 0 1
= ---------- 1 – ------------------------------2-
dt L0  E0 
 ------------2- + 1
 mr c 

where mr (SI unit: kg) is the rest mass of the particle and c = 2.99792458 × 108 m/s
is the speed of light in a vacuum.

MAXWELLIAN
When a Maxwellian initial velocity distribution is selected, an array of Nv initial velocity
values is created for each particle in each velocity direction. For a given initial
temperature, in each velocity direction, the distribution function is given by:

2
mp  mp vi 
f ( vi ) = - exp  – -----------------
--------------------
2πk B T 0  2k B T 0

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 111
 where

• vi (SI unit: m/s) is the initial velocity component in the ith direction,
• mp (SI unit: kg) is the particle mass,
• T0 (SI unit: K) is the temperature, and
• kB = 1.3806488 × 10-23 J/K is the Boltzmann constant.

So the total distribution function is:

nsdim

f(v) = ∏ f ( vi )
i=1

The probability a particle has an initial velocity vi is given by:

nsdim

g( v) = ∏ vi f ( vi )
i=1

This can clearly generate a large number of particles because the total number of initial
velocities are Nvnsdim. This option is particularly useful in plasma modeling, AC/DC
modeling, and in Monte Carlo simulations for molecular dynamics applications.

CONSTANT SPEED, SPHERICAL

Sometimes it is desirable to release Nvel particles with the same speed isotropically in
velocity space. Defining the speed as c, the following expressions generate such a
velocity distribution in 2D according to:

v x = c cos θ
v y = c sin θ

where θ goes from 0 to 2π in Nvel steps.

In 3D the velocity distribution is given by:

v x = c cos θ sin ϕ
v y = c sin θ sin ϕ
v z = c cos ϕ

The azimuthal angle θ is uniformly distributed from 0 to 2π. The polar angle ϕ is
sampled from the interval [0, π] with probability density proportional to sin ϕ. The

112 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
polar angle is arbitrarily chosen as the angle that the initial velocity makes with the
positive z-axis, but any direction could be chosen because the sphere is isotropic.

CONSTANT SPEED, HEMISPHERICAL

The Constant speed, hemispherical option is the same as the Constant speed, spherical
option, except that in 2D θ goes from 0 to π and in 3D ϕ goes from 0 to π/2. The angle
(θ in 2D or ϕ in 3D) is measured from the direction given by the Hemisphere axis
setting.

C O N S TA N T S P E E D, C O N E
The Constant speed, cone option is the same as the Constant speed, spherical option,
except that in 2D θ goes from 0 to α and in 3D ϕ goes from 0 to α. The angle (θ in 2D
or ϕ in 3D) is measured from the direction given by the Cone axis setting.

CONSTANT SPEED, LAMBERTIAN

The Constant speed, Lambertian option releases particles within a hemisphere in 3D
velocity space, but the probability distribution function is different from that of the
Constant speed, hemisphere option. Recall that for an isotropic hemispherical
distribution the polar angle ϕ has a probability density proportional to sin ϕ; for the
Lambertian distribution the probability density is instead proportional to sin ϕ cos ϕ.
Because of this extra cosine term, distributions following this probability density are
said to follow Lambert’s cosine law or (in molecular dynamics) Knudsen’s cosine law.

Particle-Particle Interactions

COULOMB FORCE
The Coulomb force on particle i in a system of N particles is defined as:

N
e
2 ( ri – rj )
F i = ------------
4πε0  Zi Zj --------------------
r –r
-
3
j=1 i j

where

• e = 1.602176565 × 10-19 C is the elementary charge,

• ε0 = 8.854187817 × 10-12 F/m is the permittivity of vacuum,
• ri is the position vector of the ith particle (SI unit: m),

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 113
 • rj is the position vector of the jth particle (SI unit: m), and
• Z (dimensionless) is the particle charge number.

The Coulomb force is repulsive for particles of the same charge and attractive for
particles with opposite charge.

LENNARD-JONES FORCE
The Lennard-Jones interaction is often used as the intermolecular potential function
to estimate the transport properties of a gas. The Lennard-Jones potential is given by:

σ 12 σ 6
U ( r ) = 4ε  --- –  ---
r r

where r is the distance between the particles (SI unit: m), ε is the interaction strength
(SI unit: N·m), and σ is the collision diameter (SI unit: m).

Figure 3-3: Plot of the Lennard-Jones potential normalized to the interaction strength
versus the nondimensional radial distance.

Using the fact that the force is computed as:

F = – ∇U

the expression for the force on the ith particle becomes:

114 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 N
σ 13 σ 7 ri – rj
2  ------------------ –  ------------------  ------------------
24ε
F i = ---------
σ  ri – rj ri – rj ri – rj
j=1

A typical interaction strength is around 1.6 × 10−21 J and a typical collision diameter
is around 3.3 × 10−10 m (3.3 Å). The first term in the square brackets is a repulsive
force due to internuclear repulsion, and the second term is an attractive force known
as the London dispersion force (LDF). The variation of the force with distance is
plotted in Figure 3-4.

Figure 3-4: Plot of the Lennard-Jones force versus nondimensionalized radial distance
between particles.

LINEAR ELASTIC FORCE

The linear elastic force Fi on particle i in a system of N particles is defined as

N
( ri – rj )
Fi = –ks  ( r i – r j – r 0 ) --------------------
ri – rj
j = 1 ;i ≠ j

where ks is the spring constant (SI unit: N/m) and r0 is the equilibrium distance
between particles (SI unit: m).

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 115
 Auxiliary Dependent Variables
In certain cases it is desirable to add additional degrees of freedom to compute particle
mass, temperature, size, spin, and so forth. The additional degree of freedom is solved
for each particle. For each degree of freedom, w, the following ODE is solved for each
particle:

dw
= R
dt

where R is a user-specified source term. When the Integrate option is set to Along
particle trajectory, the following ODE is instead solved for each particle:

dw
= R
ds

where s (SI unit: m) is the direction tangential to the motion of the particle.

About the Boundary Conditions for the Particle Tracing Interfaces

The equations of motion for the particle trajectories are supplemented with a variety
of options to describe how the particles behave when they contact surfaces in the
geometry.

ABO U T TH E WAL L BO UN DA R Y C ON DITIO N S

The following nomenclature is used with time as an example:

• Value of a variable precontact, w.

• Value of a variable at the moment of contact, wc.
• Value of a variable postcontact, w'.

When a particle comes in contact with a wall, at time tc, the following options are
available for what happens as a result of the particle-wall interaction.

Disappear
When the particle strikes the wall it disappears from view. Mathematically the particle
location is set to not-a-number (NaN) at all time steps after the particle makes contact
with the wall:

q′ = NaN

This condition means that the particle is not rendered during results processing.

116 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
Freeze
The particle position and velocity remain the same at all time steps t > tc:

q′ = q c

v′ = v c

Because the velocity is frozen at the time of impact, this boundary condition allows for
recovery of velocity and energy distribution functions.

Stick
The particle follows the motion of the wall after contact:

q′ = q w

v′ = v w

Bounce
The particle is reflected from the wall in the tangent plane. The incident angle and
reflected angle are the same with respect to the surface normal:

q′ = q

v′ = v – 2 ( n ⋅ v )n

This conserves energy if the Hamiltonian depends isotropically on momentum.

Diffuse Scattering
The particle is reflected from the wall with a velocity vector according to Knudsen’s
cosine law. Specifically, for the particle position:

q′ = q

and for the particle velocity in 3D:

v t1 = v c sin θ sin φ

v t2 = v c sin θ cos φ

v n = v c cos θ

where vc is the velocity vector when the particle strikes the wall vt1 and vt2 are the
outcoming tangential components of the velocity and vn is the normal component of

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 117
 the velocity. The angle φ is a uniformly distributed random number between 0 and 2π.
The angle θ is defined as:

θ = asin ( Γ )

where Γ is a uniformly distributed random number between 0 and 1. The probability

distribution of the normal velocity component is directly proportional to cos(θ).

In 2D, the tangential and normal velocity components are:

v t = v c sin θ

v n = v c cos θ

where the angle θ is defined as:

π
θ = acos ( Γ ) – ---
2

where Γ is now a uniformly distributed random number between -1 and 1.

Mixed Diffuse and Specular Reflection

The particle has probability γ to be reflected specularly, as if using the Bounce
condition. Otherwise the particle is reflected diffusely, as if using the Diffuse scattering
condition.

General Reflection
The general reflection option allows for arbitrary velocities to be specified for the
particles after they strike the wall:

q′ = q

and

v' = v p

where vp (SI unit: m/s) is the user-defined velocity vector. The velocity can be
specified either in Cartesian coordinates or in the normal-tangent coordinate system.

ABOUT THE INLET BOUNDARY CONDITIONS

At the inlet the number of particles, particle position, initial velocity, and the number
of releases is specified. An Inlet node can contribute with a Wall or Outlet node, so it is
possible to specify a behavior for particles that return to the inlet at a later time.

118 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
ABOUT THE OUTLET BOUNDARY CONDITIONS
At the outlets the particles either freeze or disappear.

Accumulator Theory: Domains

The Accumulator (Domain) node is used to transfer information from particles to the
domains they occupy or pass through. Each Accumulator defines a variable, called the
accumulated variable, in the selected domains. The accumulated variable is discretized
using constant shape functions, so its value is uniform over every mesh element and
may be discontinuous between adjacent mesh elements.

The name of the accumulated variable is specified in the Accumulated variable name edit
field in the Accumulator Settings section of the settings window. The default variable
name, rpd, will be used in the remainder of this section when referring to the
accumulated variable.

ACCUMULATOR TYPE
The options in the Accumulator type list are Density and Count. If Density is selected, the
source term is divided by the area or volume of the mesh element when calculating
each particle’s contribution to the accumulated variable. If Count is selected, no
division by the area or volume of the mesh element occurs.

The equations in the following section are valid for the Density type. The
corresponding value of the accumulated variable for the Count type is

rpd count = rpd density × V

where V is the mesh element volume (in 3D) or area (in 2D).

ACCUMULATION OVER ELEMENTS

When Elements is selected from the Accumulate over list, the value of the accumulated
variable in a mesh element is the sum of the source terms Ri evaluated for all particles
in that mesh element:

N
1
rpd = ----
V  Ri
i=1

where N is the total number of particles in the element and V is the area or volume of
the mesh element. In other words, the contribution of each particle to the accumulated

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 119
 variable is distributed uniformly over the mesh element the particle is in, regardless of
the particle’s exact position within the element.

If Elements and time is selected from the Accumulate over list, then the sum of the
source terms for particles in the mesh element is used to define the time derivative of
the accumulated variable, rather than its instantaneous value:

N
d ( rpd )- 1
------------------
dt
= ----
V  Ri
i=1

Thus the value of the accumulated variable depends on the time history of particles in
the mesh element, instead of the instantaneous positions of the particles. As each
particle propagates, it will leave behind a trail based on its contributions to the
accumulated variables in mesh elements it has traversed. The algorithm for
accumulating over time takes into account the fraction of a time step taken by the
solver that the particle spends in each mesh element, even if it crosses between
elements during the time step.

Accumulator Theory: Boundaries

The Accumulator (Boundary) feature transfers information from particles to the
boundaries they hit or pass through. Each Accumulator defines a variable, called the
accumulated variable, on the selected boundaries. The accumulated variable is
discretized using constant shape functions, so its value if uniform over every mesh
element and may be discontinuous between adjacent mesh elements.

The name of the accumulated variable is specified in the Accumulated variable name edit
field in the Accumulator Settings section of the settings window. The default variable
name, rpb, will be used in the remainder of this section when referring to the
accumulated variable.

The options in the Accumulator type list are Density and Count. If Density is selected, the
source term is divided by the surface area or length of the boundary mesh element
when calculating each particle’s contribution to the accumulated variable. If Count is
selected, no division by the surface area or length of the boundary element occurs.

The equations in the following section are valid for the Density type. The
corresponding value of the accumulated variable for the Count type is

rpb count = rpb density × V

120 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
where V is the boundary element surface area (in 3D) or length (in 2D).

When Particles in boundary elements is selected from the Accumulate over list, the
accumulated variable in a boundary element gets incremented by the source term R
whenever a particle freezes or sticks to the boundary:

R
rpb new = rpb + ----
V

where division by the mesh element area or length occurs because the accumulator is
assumed to be of type Density. Thus the source term evaluated for an incident particle
is uniformly distributed over the boundary element it freezes or sticks to.

If instead Particle-wall interactions is selected from the Accumulate over list, then the
accumulated variable gets incremented regardless of what type of particle-wall
interaction occurs. Thus, it is possible for the same particle to increment the
accumulated variable in many different boundary elements, or even in the same
element multiple times.

BUILT-IN GLOBAL VARIABLES

By default, the boundary Accumulator defines the following global variables:
TABLE 3-3: BUILT-IN GLOBAL VARIABLES FOR BOUNDARY ACCUMULATORS

NAME DESCRIPTION

<scope>.<name>_ave Average of accumulated variable

<scope>.<name>_int Integral of accumulated variable
<scope>.<name>_max Maximum of accumulated variable
<scope>.<name>_min Minimum of accumulated variable
<scope>.<name>_sum Sum of accumulated variable over elements

Here, <scope> includes the physics interface name and the names the Accumulator
and parent feature. For example, the average of the accumulated variable over a
boundary may be called pt.wall1.bacc1.rpb_ave, where pt is the name of the
Mathematical Particle Tracing interface, wall1 is the name of the parent Wall node,
bacc1 is the name of the Accumulator node, and rpb is the accumulated variable
name. These variables are all available in the Add/Replace Expression menus during
results evaluation.

T H E O R Y F O R T H E M A T H E M A T I C A L P A R T I C L E TR A C I N G I N T E R F A C E | 121
 These global variables are computed by defining a set of component couplings on the
selection of the parent physics feature, such as the Wall feature to which the
Accumulator is added. The following expressions for the global variables are used.

TABLE 3-4: BUILT-IN GLOBAL VARIABLE DEFINITIONS FOR BOUNDARY ACCUMULATORS

NAME EXPRESSION

<scope>.<name>_ave <wscope>.aveop(<scope>.<name>)
<scope>.<name>_int <wscope>.intop(<scope>.<name>)
<scope>.<name>_max <wscope>.maxop(<scope>.<name>)
<scope>.<name>_min <wscope>.minop(<scope>.<name>)
<scope>.<name>_sum <wscope>.intop(<scope>.<name>/<scope>.meshVol)

Here, <wscope> is the scope of the parent boundary feature, for example, pt.wall1.

References for the Mathematical Particle Tracing Interface

1. L.D. Landau and E.M. Lifshitz, Mechanics, 3rd ed., Elsevier, 1976.

122 | C H A P T E R 3 : M A T H E M A T I C A L P A R T I C L E TR A C I N G
 4

Charged Particle Tracing

This chapter describes the Charged Particle Tracing interface, which is found
under the AC/DC branch ( ) when adding a physics interface.

In this chapter:

• The Charged Particle Tracing Interface

• Theory for the Charged Particle Tracing Interface
See The Electromagnetics Interfaces in the COMSOL Multiphysics Reference
Manual for other AC/DC interface and feature node settings.

123
 The Charged Particle Tracing
Interface
The Charged Particle Tracing (cpt) interface ( ), found under the AC/DC>Particle
Tracing branch ( ) when adding a physics interface, is used to model charged particle
orbits under the influence of electromagnetic forces. In addition, it can also model
two-way coupling between the particles and fields. Some typical applications are
particle accelerators, vacuum tubes and ion implanters. The physics interface supports
time-domain modeling only in 2D and 3D.

The physics interface solves the equation of motion for charged particles. It is possible
to add forces to ions and electrons, including predefined electric, magnetic, and
collisional forces. Collisions can be modeled as a continuous force or using a Monte
Carlo model in which particles can undergo elastic or inelastic collisions with a rarefied
gas, including charge exchange and ionization reactions.

When this physics interface is added, these default nodes are also added to the Model
Builder: Wall and Particle Properties. Then, from the Physics toolbar, add other nodes
that implement, for example, boundary conditions and release features. You can also
right-click Charged Particle Tracing 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 cpt.

PARTICLE RELEASE AND PROPAGATION

The Formulation, Relativistic Correction, Store extra time steps for wall interactions, and
Maximum number of secondary particles settings are the same as for The Mathematical
Particle Tracing Interface, except that only the Newtonian, Newtonian, first order, and
Massless formulations are available.

124 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Particle Release Specification
Select a Particle release specification—Specify release times (the default) or Specify
current. If Specify release times is selected, then each model particle is treated as the
instantaneous position of one or more charged particles for the purpose of modeling
particle-field interactions. This means, for example, that if the Space Charge Density
Calculation node is used, the space charge density is only nonzero in mesh elements
that are currently occupied by particles.

If Specify current is selected, then for the purpose of modeling particle-field

interactions, each model particle traces a path that is followed by a number of charged
particles per unit time. This means that the charge density computed by the Space
Charge Density Calculation feature is nonzero in all mesh elements that the particle
trajectories pass through, not just at the instantaneous positions of the particles. In
other words, the particles leave behind a trail of space charge as they propagate.

The Specify current option is primarily used to model charged particle beams in which
the charge and current density do not change over time. Changing the Particle release
specification affects some inputs in the settings windows for release features such as the
Release and Inlet nodes. In addition, the Surface Charge Density node is only available
with the Specify release times option, while the Current Density, Heat Source, and
Etch nodes are only available with the Specify current option.

ADDITION AL VAR IA BL ES
Except for the option described below, these settings are the same as for The
Mathematical Particle Tracing Interface.

For 2D axisymmetric components an azimuthal particle velocity gives rise

to a force in the radial direction, the centrifugal force. The out-of-plane
velocity thus has an effect on the in-plane motion of the particles. In 2D
components it is still possible to compute the out-of-plane position and
velocity, but the effect on the in-plane trajectories is not as pronounced as
in the axisymmetric case.

Select the Include out-of-plane degrees of freedom check box to define

additional degrees of freedom for each particle for the out-of-plane
position and velocity.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 125
 The extra degrees of freedom that are created by selecting the Include
out-of-plane degrees of freedom check box are considered auxiliary
dependent variables and are solved for using coupled first-order
differential equations. If, in addition, Newtonian is selected from the
Formulation list, then the resulting system of equations includes both first-
and second-order equations.

Mixed first- and second-order equations systems are not supported for all
solver configurations. For example, it is not possible to use explicit time
stepping methods such as Dormand-Prince 5. To use explicit
time-stepping methods with out-of-plane degrees of freedom, consider
using the Newtonian, first order formulation instead.

ADVANCED SETTINGS
These settings are the same as for The Mathematical Particle Tracing Interface.

DEPENDENT VARIABLES
The dependent variables (field variables) are the Particle position, Particle position
components, Particle velocity, and Particle velocity components. The degrees of freedom
for particle velocity are only used if Newtonian, first order is selected from the
Formulation list. The name can be changed but the names of fields and dependent
variables must be unique within a model.

• Domain, Boundary, Pair, and Global Nodes for the Charged Particle
Tracing Interface
• Theory for the Mathematical Particle Tracing Interface
• Theory for the Charged Particle Tracing Interface

• With the AC/DC Module see Magnetic Lens: Application Library

path ACDC_Module/Particle_Tracing/magnetic_lens and Quadrupole Mass
Filter: Application Library path
ACDC_Module/Particle_Tracing/quadrupole_mass_filter.
• With the Plasma Module see Ion Energy Distribution Function:
Application Library path
Plasma_Module/Inductively_Coupled_Plasmas/ion_energy_distribution_fu
nction.

126 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Domain, Boundary, Pair, and Global Nodes for the Charged Particle
Tracing Interface
The Charged Particle Tracing Interface has these domain, boundary, pair, and global
nodes 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.

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

• Attachment • Nonresonant Charge Exchange

• Collisions • Nuclear Stopping
• Current Density • Particle Beam
• Elastic • Particle-Matter Interactions
• Electric Force • Particle Properties
• Etch • Resonant Charge Exchange
• Excitation • Space Charge Density Calculation
• Friction Force • Surface Charge Density
• Heat Source • Symmetry
• Ionization • Thermionic Emission
• Ionization Loss • User Defined
• Magnetic Force

In the COMSOL Multiphysics Reference Manual see Table 2-3 for links
to common sections and Table 2-4 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.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 127
 These nodes and subnodes are described for The Mathematical Particle Tracing
Interface (listed in alphabetical order):

• Accumulator (Boundary) • Particle-Particle Interaction

• Accumulator (Domain) • Periodic Condition
• Auxiliary Dependent Variable • Release
• Force • Release from Data File
• Inlet • Release from Edge
• Nonlocal Accumulator • Release from Grid
• Outlet • Release from Point
• Override Properties • Secondary Emission
• Particle Continuity • Velocity Reinitialization
• Particle Counter • Wall (the default boundary
condition)

Collisions
Use the Collisions node model the interaction between model particles and a rarefied
background gas using a Monte Carlo model. Particles have a random chance to be
deflected by gas molecules during each time step taken by the solver, based on the gas
density and collision cross-section or collision frequency. A variety of different built-in
collisions are available, each with predefined expressions for the post-collision velocity
of the model particles.

By itself, the Collisions node does not cause any collisions to occur. It is necessary to
add one or more subnodes, which correspond to different collision types. The
following subnodes are available from the context menu (right-click the parent node)
or from the Physics toolbar, Attributes menu: Elastic, Excitation, Attachment,
Ionization, Resonant Charge Exchange, Nonresonant Charge Exchange, and User
Defined.

When the Collisions node and subnodes (the reaction set) are used,
collisions with a background gas occur randomly and the particle velocity
can change discontinuously at discrete times. The friction model, a
deterministic force, can be accessed using a dedicated Friction Force
node.

128 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
FLUID PROPERTIES
Enter the Background number density Nd (SI unit: 1/m3). The default is
1 × 1020 1/m3.

Enter the Background gas molar mass Mg (SI unit: kg/mol). The default is
0.04 kg/mol.

Enter coordinates for the Velocity field u (SI unit: m/s) based on space dimension. If
another physics interface is present which computes the velocity field then this can be
selected from the list.

To model the case of a stagnant background gas, leave all the velocity
components as zero. If a nonzero velocity field is specified, the
background gas molecules will follow a drifting Maxwellian velocity
distribution.

Enter values or expressions for the Temperature T (SI unit: K) of the background gas.
The default is 293.15 K.

COLLISION STATISTICS
Select the Count all collisions check box to allocate an auxiliary dependent variable that
records the number of times each particle collides with the background gas. If the
Collisions node has multiple subnodes, the variable created by the Count all collisions
check box is updated for collisions caused by any of these subnodes. In addition, the
settings window for each collision type includes a Count collisions check box that can
be used to count the number of times a specific type of collision occurs.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
This section is available if an Auxiliary Dependent Variable has been added to the
model. The settings are the same as described for the Wall node. The auxiliary
dependent variables for the particle get updated every time a collision occurs.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 129
 AFFECTED PARTICLES
Select an option from the Particles to affect list to apply the force to specific particles:
All (the default), Single species, or Logical expression. The settings are described for the
Force node.

Collisional Force Theory

Elastic
The Elastic subnode is available from the context menu (right-click the Collisions
parent node) or from the Physics toolbar, Attributes menu. Use it to model
momentum-conserving collisions between model particles and background gas
molecules.

COLLISION FREQUENCY
Select an option from the Specify list: Cross section (the default) or Collision frequency.

• For Collision frequency, enter a value or expression for the Collision frequency ν
(SI unit: Hz). The default is 2 × 106 Hz.
• For Cross section enter a value or expression for the Collision cross section σ
(SI unit: m2). The default is 3 × 10-19 m.

For either choice:

• Select the Number density specification — From parent (the default) or User defined.
For User defined, enter a Background number density Nd (SI unit: 1/m3). The
default is1 × 1020 1/m3. If From parent is selected, the number density is inherited
from the Collisions node.
• Select the Molar mass specification — From parent (the default) or User defined. For
User defined, enter the Background gas molar mass Mg (SI unit: kg/mol). The default

130 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 is 0.04 kg/mol. If From parent is selected, the background gas molar mass is
inherited from the Collisions node.

If the molar mass of the background gas is similar to the molar mass of the
particles then the particles tend to form a Maxwellian velocity distribution
function very rapidly. If the mass of the particles is very different to that
of the background gas then the particles tend to have a non-Maxwellian
distribution function.

If the geometry is axisymmetric and Include out-of-plane degrees of freedom is selected

in the physics interface Advanced Settings section, enter a value or expression for the
Cutoff radial coordinate for out-of-plane momentum rc (SI unit: m). The default value
is 0. If a particle undergoes a collision while its radial coordinate is less than the
specified cutoff, the azimuthal component of its reinitialized velocity is set to zero. Use
this setting to prevent particles from being subjected to inordinately large centrifugal
forces if they undergo collisions very close to the axis of symmetry.

COLLISION STATISTICS
Select the Count collisions check box to count the number of times each particle collides
with the background gas. The collision counter is only incremented by this instance of
the Elastic node, not by any other collision type or by other Elastic nodes. To define a
variable that counts all types of collisions, select the Count all collisions check box in the
settings window for the parent Collisions node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
This section is available if an Auxiliary Dependent Variable has been added to the
model. For each auxiliary dependent variable, select an option from the New value of
auxiliary dependent variable list: From parent (the default) or User defined. If From
parent is selected, the value of the auxiliary dependent variable is reinitialized according
to the settings for the parent Collisions node.

If User defined is selected, select the Assign new value to auxiliary variable check box to
assign a new value to the auxiliary dependent variable when an elastic collision occurs.
Then enter the new value or expression in the field. For example, if there is an auxiliary
variable, psi, then enter a value for psinew in the field. So, to increment the value of
psi by 1 when a particle undergoes an elastic collision, enter psi+1 in the text field
for psinew.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 131
 Excitation
The Excitation subnode is available from the context menu (right-click the Collisions
parent node) or from the Physics toolbar, Attributes menu. The Excitation subnode
causes particles to undergo inelastic collisions with the background gas, in which some
amount of kinetic energy is expended to raise the background gas molecule or ion to
an excited state.

COLLISION FREQUENCY
These settings are the same as for the Elastic subnode with the addition of entering a
value for the Energy loss ΔE (SI unit: J). The default value is 0. Typical values for
electron impact collisions are between 6 eV and 10 eV.

COLLISION STATISTICS
These settings are described for the Elastic node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
These settings are described for the Elastic node.

Attachment
The Attachment subnode is available from the context menu (right-click the Collisions
parent node) or from the Physics toolbar, Attributes menu. The Attachment subnode
causes particles to become attached to background gas particles, removing them from
the simulation. Optionally, the attached particle can then be released. Typically, the
primary particles subjected to this collision type are electrons and the attached particle
is a neutralized or ionized species from the background gas.

COLLISION FREQUENCY
These settings are the same as for the Elastic subnode with an additional option to
select the Release attached particle check box. When this check box is cleared, collisions
only cause the primary particle velocity and auxiliary dependent variables to be
reinitialized. When this check box is selected, a secondary particle is released during
each collision.

ION PROPERTIES
This section is available when the Release attached particle check box is selected. Select
an option from the Ion properties list when either a Particle Properties or Override

132 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Properties node is added to the physics interface. This controls the type of species that
is introduced via secondary particle emission during the attachment reaction.

COLLISION STATISTICS
These settings are described for the Elastic node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
These settings are described for the Elastic node.

IN IT IA L VA LUE OF AUXILIAR Y DEPEN DENT VARIABLES

This section is available when the Release attached particle check box is selected. For
each auxiliary dependent variable that is defined in the model, enter the initial value of
the auxiliary dependent variable for the attached particle. The default value is 0.

Ionization
The Ionization subnode is available from the context menu (right-click the Collisions
parent node) or from the Physics toolbar, Attributes menu. The Ionization subnode
causes particles to ionize background gas molecules, releasing secondary electrons into
the simulation. Optionally, the ionized background gas molecule can also be released.
Typically, the primary particles subjected to this collision type are electrons and the
ionized particle is taken from the background gas.

COLLISION FREQUENCY
These settings are the same as for the Elastic subnode with additional options to enter
a value for the Energy loss ΔE (SI unit: J) and to select the Release ionized particle check
box. Typical energy loss values for electron impact collisions are between 12 eV and
24 eV.

ELECTRON PROPERTIES
Select an option from the Electron properties list, which may be any Particle Properties
or Override Properties node that has been added to the physics interface.

ION PROPERTIES
This section is available when the Release ionized particle check box is selected. Select
an option from the Ion properties list when either a Particle Properties or Override
Properties node is added to the physics interface.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 133
 COLLISION STATISTICS
These settings are described for the Elastic node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
These settings are described for the Elastic node.

I NI TI AL VAL UE O F AUX I L IA R Y D EP E ND EN T VAR IA BL E S

For each auxiliary dependent variable that is defined in the model, enter the initial
value of the auxiliary dependent variable for the electron and ionized particle that are
released as a result of the ionization reaction. The default value is 0.

Resonant Charge Exchange

The Resonant Charge Exchange subnode is available from the context menu (right-click
the Collisions parent node) or from the Physics toolbar, Attributes menu. The Resonant
Charge Exchange subnode causes particles to undergo charge exchange reactions with a
background gas. Resonant charge exchange occurs between atoms of the same element
or molecules of the same substance that differ in charge. The total internal energy of
the particles does not change as a result of the collision. The ionized particle,
neutralized particle, or both ionized and neutralized particles can be released.

COLLISION FREQUENCY
These settings are the same as for the Elastic subnode with additional options to enter
a value for the Scattering angle in the center of mass system χ (SI unit: rad, default 0)
and to select one of the following from the Species to release list — Ion (the default),
Neutral particle, or Ion and neutral particle.

NEUTRAL PROPERTIES
This section is available when Neutral particle or Ion and neutral particle is selected from
the Species to release list. Select an option from the Neutral properties list, which may
be any Particle Properties or Override Properties node that has been added to the
physics interface.

COLLISION STATISTICS
These settings are described for the Elastic node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
These settings are described for the Elastic node.

134 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
IN IT IA L VA LUE OF AUXILIAR Y DEPEN DENT VARIABLES
This section is available when Neutral particle or Ion and neutral particle is selected from
the Species to release list. For each auxiliary dependent variable that is defined in the
model, enter the initial value of the auxiliary dependent variable for the released neutral
particle. The default value is 0.

Nonresonant Charge Exchange

The Nonresonant Charge Exchange subnode is available from the context menu
(right-click the Collisions parent node) or from the Physics toolbar, Attributes menu.
The Nonresonant Charge Exchange subnode causes particles to undergo charge
exchange reactions with a background gas. Unlike the Resonant Charge Exchange
node, it is possible for the particle that undergoes the collision and the ionized and
neutralized particles that result from it to be three different species. The ionized
particle, neutralized particle, or both ionized and neutralized particles can be released.

COLLISION FREQUENCY
These settings are the same as for the Elastic subnode with additional options to enter
a value for the Energy loss ΔE (SI unit: J), enter a value for the Scattering angle in the
center of mass system χ (SI unit: rad), and to select one of the following from the
Species to release list: Ion (the default), Neutral particle, or Ion and neutral particle.

NEUTRAL PROPERTIES
This section is available when Neutral particle or Ion and neutral particle is selected from
the Species to release list. Select an option from the Neutral properties list, which may
be any Particle Properties or Override Properties node that has been added to the
physics interface.

ION PROPERTIES
This section is available when Ion or Ion and neutral particle is selected from the Species
to release list. Select an option from the Ion properties list, which may be any Particle
Properties or Override Properties node that has been added to the physics interface.

COLLISION STATISTICS
These settings are described for the Elastic node.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
These settings are described for the Elastic node.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 135
 I NI TI AL VAL UE O F AUX I L IA R Y D EP E ND EN T VAR IA BL E S
For each auxiliary dependent variable that is defined in the model, enter the initial
value of the auxiliary dependent variable for the released neutral particle. The default
value is 0.

User Defined
The User Defined subnode is available from the context menu (right-click the Collisions
parent node) or from the Physics toolbar, Attributes menu. The User Defined node
allows the post-collision particle velocity to be a user-defined expression.

COLLISION FREQUENCY
These settings are the same as for the Elastic subnode.

R E I N I T I A L I Z E D P A R T I C L E VE L O C I T Y
Enter values or expressions for the Reinitialized particle velocity vr (SI unit: m/s) based
on space dimension.

COLLISION STATISTICS
These settings are the same as for the Elastic subnode.

INCLUDE SECONDARY EMISSION

Select the Include secondary emission check box to release a secondary particle when the
collision occurs.

SECONDARY EMISSION CONDITION

This section is available when the Include secondary emission check box is selected.

Select a Secondary emission condition: None (the default), Probability, or Expression. For
Probability enter the Probability γ (dimensionless) for secondary emission to occur. For
Expression enter an Evaluation expression e (dimensionless) that must be nonzero for
secondary emission to occur.

SECONDARY PARTICLES
This section is available when the Include secondary emission check box is selected.

In the Number of secondary particles field Ns, specify the number of secondary particles
to release per primary particle. The default value is 1.

Select an Initial velocity — User defined (the default) or Constant speed, spherical.

136 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
• Constant speed, spherical releases the secondary particles with a constant speed and
spherical distribution of velocity directions. Enter a Speed v0 (SI unit: m/s) with
which all secondary particles are released.
• User Defined allows for an arbitrary velocity vector to be set for the secondary
particles. Enter values or expressions for the Initial particle velocity v0 (SI unit: m/s)
based on space dimension.

Select the Offset initial position check box to offset the positions of the secondary
particles before releasing them. Then select an Offset method — Using initial velocity
(the default) or Isotropic sphere.

• For Using initial velocity, enter a Time interval Δt (SI unit: s). The default is 0. Each
secondary particle is then displaced by the product of its initial velocity and this time
interval before being released.
• For Isotropic sphere, enter a Particle displacement magnitude Δr (SI unit: m). The
default is 0. All secondary particles are then moved a distance equal to the
displacement magnitude before they are released. If the offset would cause
secondary particles to be placed outside of the modeling domain, they are instead
released at the location of the primary particle.

INHERIT PROPERTIES
This section is available if the Include secondary emission check box is selected. Select
an option from the Inherit Properties list. The Particle Properties node (default) is
always available. If any Override Properties nodes have been added to the model, they
can also be selected. Use this setting to determine what set of properties is given to the
secondary particles that are released by this feature.

N EW VA L U E O F A U X I L I A R Y D E P E N D E N T VA R I A BL E S
These settings are described for the Elastic node.

IN IT IA L VA LUE OF AUXILIAR Y DEPEN DENT VARIABLES

For each auxiliary dependent variable that is defined in the model, enter the initial
value of the auxiliary dependent variable for the secondary particles that are released as
a result of the User Defined collision. The default value is 0.

Friction Force
Use the Friction Force node to model collisions between particles and a background gas
as a friction force that is proportional to the relative speed of the particles. Unlike the

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 137
 collision types that are available with the Collisions node, the Friction Force is
deterministic and does not cause the particle velocity to change discontinuously over
time.

COLLISION FREQUENCY
Choose an option from the Specify list: Cross section and number density or Collision
frequency.

If Cross section and number density is selected, enter the following:

• Collision cross section σ (SI unit: m2). The default is 3 × 10-19 m2.
• Background number density Nd (SI unit: 1/m3). The default is 1 × 1020 1/m3.

If Collision frequency is selected, enter the Collision frequency ν (SI unit: 1/s). The
default value is 2 ×106 Hz.

FLUID PROPERTIES
Enter coordinates for the Velocity field u (SI unit: m/s) based on space dimension.

To model the case of a stagnant background fluid, leave all the velocity
components as zero.

AFFECTED PARTICLES
Select an option from the Particles to affect list to apply the force to specific particles:
All (the default), Single species, or Logical expression. The settings are described for the
Force node.

Collisional Force Theory

Particle-Matter Interactions
Use the Particle-Matter Interactions node to model the propagation of energetic ions
through solid material. The Ionization Loss and Nuclear Stopping subnodes are
available from the context menu (right-click the parent node) or from the Physics
toolbar, Attributes menu. In order to model the interaction of particles with the target
material, at least one subnode must be added.

138 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
TA RG E T M A T E R I A L P RO P E R T I E S
By default, the Refractive index, real part n (dimensionless) is taken From material. For
User defined enter another value or expression.

By default, the Density ρ (SI unit: kg/m2) is taken From material. For User defined enter
another value or expression.

Enter a value or expression for the Atomic mass A (SI unit: kg). The default value is
28.0855 amu, the relative atomic mass of silicon. Then enter a value or expression for
the Atomic number Zm (dimensionless). The default value is 14.

PARTICLE PROPERTIES
Enter a value or expression for the Atomic number Zp (dimensionless). The default
value is 1. The particle atomic number is used to define screened Coulomb potentials
for the Nuclear Stopping node.

CUTOFF ENERGY
Select the Stop particles with energy below cutoff to cause particles to stop moving when
their energy is less than a specified value. This check box is selected by default. Then
enter a value or expression for the Cutoff energy Ec (SI unit: J). The default value is
8.6 eV.

In general, the Stop particles with energy below cutoff check box should
always be selected if it is possible for ions in the material to approach
extremely low energies. With the built-in ionization loss and nuclear
scattering models, accurate data is not available for energies significantly
less than the default value, 8.6 eV.

AFFECTED PARTICLES
Use this section to determine which particles are subjected to the particle-matter
interaction. By default, all particles in the model are affected. The settings are
described for the Force node.

Ionization Loss
The Ionization Loss subnode is available from the context menu (right-click a
Particle-Matter Interactions parent node) or from the Physics toolbar, Attributes menu.
Use this subnode to model the deceleration of ions by electrons in the target material.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 139
 The deceleration is treated as a continuous braking force that acts opposite the
direction of motion of the particles.

LOSS MECHANISM
Select an option from the Electronic Stopping Power list: Protons on silicon, Alpha
particles on silicon, or User defined. For Protons on silicon and Alpha particles on silicon
the force that opposes the particle motion is expressed as a function of the particle
kinetic energy based on empirical data tabulated in Ref. 1. For User defined enter a
value or expression for the stopping power Se (SI unit: m4/s2). The default value is
1 MeV-cm2/g.

Nuclear Stopping
The Nuclear Stopping subnode is available from the context menu (right-click a
Particle-Matter Interactions parent node) or from the Physics toolbar, Attributes menu.
Use this subnode to model the deflection of ions by nuclei in the target material. The
deflection is modeled using a Monte Carlo model in which the particle velocity can
change discontinuously in any time step.

LOSS MECHANISM
Select an option from the Screening function list: None (the default), Bohr, Moliere,
Lenz-Jensen, or Universal. The screening function is the ratio of the electric potential of
the target atom to the potential of an unscreened nucleus.

Enter a value or expression for the Cutoff scattering angle χc (SI unit: rad). The default
is 0.1 deg. Any interactions with target nuclei with a scattering angle less than the
specified cutoff will be neglected. By specifying a larger cutoff scattering angle, it is
possible to ignore interactions that have a small effect on particle momentum, allowing
a larger time step to be taken by the solver while still accounting for the more
important nuclear collisions (which usually have larger scattering angles).

COLLISION STATISTICS
Select the Count collisions check box to assign an auxiliary dependent variable for the
number of collisions with the target material that each particle undergoes. The name
of this collision counter is Nc preceded by the scope of the Nuclear Stopping node, for
example cpt.pmi1.ns1.Nc.

140 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 Particle-Matter Interaction Theory

Particle Beam
Use the Particle Beam node to release a nonlaminar beam of particles normal to a
boundary with a specified distribution in phase space.

The Nonlocal Accumulator subnode is available from the context menu (right-click
the parent node) or from the Physics toolbar, Attributes menu.

For 2D axisymmetric models, always select the Include out-of-plane

degrees of freedom check box in the physics interface Advanced Settings
section when modeling particle beams, since the out-of-plane position
and velocity are crucial in ensuring that the released beam has the
specified emittance and other beam parameters.

Go to Release for information about the following sections: Release Times, Release
Current Magnitude, Inherit Properties, Initial Value of Auxiliary Dependent Variables, and
Advanced Settings.

INITIAL POSITION
Enter the Number of particles per release N (dimensionless). The default is 1000.

Select a Beam position — From coordinates (the default), Centroid of selection, or

Selected point. This option determines how the position of the center of the beam is
computed. For 2D axisymmetric model components, an additional option is available
— Axis of symmetry (the default).

For From coordinates enter coordinates for the Beam center location rc (SI unit: m).
When Selected point is selected, the Reference Point Selection section is shown. Add at
least one point to the selection to specify the center of the beam.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 141
 I N I T I A L TR A N S VE R S E VE L O C I T Y
The Equation section of the Particle Beam feature contains images indicating how the
phase space ellipses are constructed for the different settings available. To enable this,
use Show>Equation Sections in the Model Builder toolbar.

• Select a Sampling from phase space ellipse: KV, Waterbag, Parabolic or Gaussian (the
default). The KV distribution places particles uniformly on the surface of a
hyperellipsoid in 4-dimensional phase space. In 3D, this results in a particle
distribution which is uniform in physical space, i.e. the distance between particles is
roughly the same throughout the beam cross-section. The Waterbag distribution
places particles uniformly inside the volume of a hyperellipsoid in 4-dimensional
phase space. This results in a distribution in physical space in which particles are
placed preferentially towards the center of the beam. The Parabolic distribution is
similar to the Waterbag distribution, but more particles are placed towards the
center of the beam in physical space. The Gaussian distribution fills the volume of a
hyperellipsoid with a normal distribution density profile in 4-dimensional phase
space. Similar to the Waterbag and Parabolic distributions, particles are preferentially
placed closer to the center of the beam, although in this case, some particles can be
placed a significant distance from the center
• In 3D, select a Beam symmetry: Symmetric (the default) or Asymmetric. If Asymmetric
is selected, two different sets of beam parameters, such as emittance and Twiss
parameters, can be specified for the two transverse directions. If Symmetric is
selected, the same beam parameters are used for both transverse directions.
• Select a Beam orientation: Upright (the default) or Not upright. Upright means that
the semimajor and semiminor axes of the phase space ellipses are parallel to the
phase space coordinate axis. For a beam that is not subjected to any forces, Upright
means that the beam is released at the location of the beam waist. Not upright allows
the phase space ellipse to be rotated about its centroid.
• Select a Transverse velocity distribution specification: Specify emittance and Twiss
parameters (the default) or Specify phase space ellipse dimensions. Twiss parameters
are generally favored by accelerator physicists, as they provide a standard method of
defining the geometric properties of the phase space ellipse. Specifying phase space
ellipse dimensions is often favored by spectrometer designers, as it allows the
dimensions of the phase space ellipse and rotation angle to be specified directly.
• In some cases it is possible to specify the Brightness B (SI unit: A/m2) instead of
emittance. The default value is 1 mA/m2. The beam emittance is then computed
from the brightness and the specified Release current magnitude. It is possible to

142 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 enter a value or expression for the brightness of the beam if the following conditions
are met:
- The model component is 3D or 2D axisymmetric.
- Specify current is selected from the Particle release specification list in the physics
interface Particle Release and Propagation section.
- Symmetric is selected from the Beam symmetry list.
- Specify emittance and Twiss parameters is selected from the Transverse velocity
distribution specification list.
- Specify brightness is selected from the Emittance specification list, which is only
shown if the preceding conditions are met.

• For 3D components, when the Beam symmetry is set to Asymmetric, it is possible to

specify different phase space distributions in two orthogonal transverse directions.
To decide how these transverse directions are oriented, select a Transverse direction
specification: User defined or Parallel to reference edge. When Parallel to reference
edge is selected, the Reference Edge Selection section is shown. Add an edge to this
selection to define the coordinate system with respect to which the beam emittance
and Twiss parameters are defined. Quantities with subscript 1 (εrms,1, β1, etc.) are
used to determine the distributions of the position and velocity components parallel
to the selected edge. Quantities with subscript 2 correspond to the position and
velocity components perpendicular to the selected edge and to the direction of
beam propagation. If User defined is selected, enter components of the Transverse
beam direction t1 (dimensionless) directly. The default is the positive y-direction.

Depending on the options selected above, it is possible to enter values for different
physical quantities which describe the phase space ellipse dimensions. These are
summarized below and in the following table:

• The Twiss parameter beta β (SI unit: m). The default value is 1 m. The product of
this and the emittance give an indication of the transverse size of the particle beam.
For a given value of the beam emittance, increasing β causes the particle distribution
in transverse position space to be broadened and the distribution in transverse
velocity space to be tightened.
• The 1-RMS beam emittance εrms (SI unit: m). The default is 1 mm. The emittance is
a measure of the average spread of the particles in phase space. Qualitatively, a larger
emittance means that the area occupied by the distribution of particles in phase
space is larger.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 143
 • The Brightness B (SI unit: A/m2). The default is 1 mA/m2. For symmetric beams
the brightness can used instead of specifying the emittance. This option is only
available when a certain combination of settings are applied, as explained in the
previous list.
• The Twiss parameter alpha α (dimensionless). The default is 0. When this value is
non-zero, the phase space ellipse becomes rotated about the position — velocity
axis. When Upright is selected from the Beam orientation list, α = 0. The remaining
Twiss parameter γ (SI unit: 1/m) is not specified because it is derived from α and β.
• The Maximum transverse displacement xm (SI unit: m). The default is 0.1 m.
Typically, this value should be the same as the initial beam radius.
• The Maximum relative transverse velocity xm’ (dimensionless). The default is 0.04.
Larger values indicate that the particles will have higher velocities in the transverse
direction with respect to the longitudinal direction.
• The Rotation angle θ (SI unit: rad). The default is 0. This rotates the phase space
ellipse about the position-velocity axis.

Two separate values for all of the above options can be entered in 3D when the Beam
symmetry is set to Asymmetric.

144 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
TABLE 4-1: SUMMARY OF INITIAL TRANSVERSE VELOCITY OPTIONS

SAMPLING ORIENTATION TRANSVERSE IMAGE AVAILABLE

VELOCITY INPUT
SPECIFICATION

KV Upright Twiss β, εrms

β, B

Waterbag Upright Twiss β, εrms

β, B

Parabolic Upright Twiss β, εrms

β, B

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 145
 TABLE 4-1: SUMMARY OF INITIAL TRANSVERSE VELOCITY OPTIONS

SAMPLING ORIENTATION TRANSVERSE IMAGE AVAILABLE

VELOCITY INPUT
SPECIFICATION

Gaussian Upright Twiss β, εrms

β, B

KV Not upright Twiss β, εrms, α

β, B, α

Waterbag Not upright Twiss β, εrms, α

β, B, α

146 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
TABLE 4-1: SUMMARY OF INITIAL TRANSVERSE VELOCITY OPTIONS

SAMPLING ORIENTATION TRANSVERSE IMAGE AVAILABLE

VELOCITY INPUT
SPECIFICATION

Parabolic Not upright Twiss β, εrms, α

β, B, α

Gaussian Not upright Twiss β, εrms, α

β, B, α

KV Upright Dimensions xm, xm’

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 147
 TABLE 4-1: SUMMARY OF INITIAL TRANSVERSE VELOCITY OPTIONS

SAMPLING ORIENTATION TRANSVERSE IMAGE AVAILABLE

VELOCITY INPUT
SPECIFICATION

Waterbag Upright Dimensions xm, xm’

Parabolic Upright Dimensions xm, xm’

Gaussian Upright Dimensions xm, xm’

148 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
TABLE 4-1: SUMMARY OF INITIAL TRANSVERSE VELOCITY OPTIONS

SAMPLING ORIENTATION TRANSVERSE IMAGE AVAILABLE

VELOCITY INPUT
SPECIFICATION

KV Not upright Dimensions xm, xm’, θ

Waterbag Not upright Dimensions xm, xm’, θ

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 149
 TABLE 4-1: SUMMARY OF INITIAL TRANSVERSE VELOCITY OPTIONS

SAMPLING ORIENTATION TRANSVERSE IMAGE AVAILABLE

VELOCITY INPUT
SPECIFICATION

Parabolic Not upright Dimensions xm, xm’, θ

Gaussian Not upright Dimensions xm, xm’, θ

I N I T I A L L O N G I T U D I N A L VE L O C I T Y
Select a Longitudinal velocity specification: Specify kinetic energy (the default) or Specify
velocity. This determines the physical quantity of the remaining inputs in this section,
which are used to initialize the longitudinal component of the particle velocity. Select
a Longitudinal velocity distribution: None (the default), Normal, Uniform or List of values.
The following options are available.

• For None, enter either a Kinetic energy E (SI unit: J), default 1 eV or a Velocity
magnitude V (SI unit: m/s), default 1 m/s. This option releases one particle for each
initial position.
• For Normal, enter either a Mean kinetic energy Em (SI unit: J), default 1 eV and a
Kinetic energy standard deviation σ (SI unit: J), default 0.1 eV, or a Mean velocity
magnitude Vm (SI unit: m/s), default 1 m/s and Velocity magnitude standard
deviation σ (SI unit: m/s), default 0.1 m/s. In both cases, also enter the Number of
velocity values. The default is 1, and this option determines the number of particles
released for each initial position, meaning this value multiplies the total number of

150 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 particles released. The longitudinal component of the velocity is sampled from a
normal distribution, meaning that roughly 68% of the particles will have an initial
velocity within one standard deviation, and 95% within two standard deviations.
• For Uniform, enter either a Minimum kinetic energy Emin (SI unit: J), default 1 eV and
a Maximum kinetic energy Emax (SI unit: J), default 2 eV, or a Minimum velocity
magnitude Vmin (SI unit: m/s), default 1 m/s and Maximum velocity magnitude Vmax
(SI unit: m/s), default 2 m/s. In both cases, also enter the Number of velocity values.
The default is 1, and this option determines the number of particles released for each
initial position, meaning this value multiplies the total number of particles released.
The uniform distribution will always include the minimum and maximum velocity
or energy values if the Number of velocity values is greater than 1. If the Number of
velocity values is 1, the velocity magnitude is simply the mean of the minimum and
maximum values.
• For List of values, enter either a list of Kinetic energy values (SI unit: J) or a list of
Velocity values (SI unit: m/s). The length of this list determines how many particles
are released per initial position.

Select the Reverse direction check box to reverse the longitudinal direction of the
released particles. This is useful when the feature selection is an interior boundary, and
the normal direction is ambiguous. The direction of beam propagation is indicated by
arrows on the selected boundaries in the Graphics window.

Particle Beam Theory

Thermionic Emission

The Thermionic Emission node is only available in 3D.

It is also only available when Newtonian or Newtonian, first order is selected

from the Formulation list and Specify current is selected from the Particle
release specification list in the physics interface Particle Release and
Propagation section.

Use the Thermionic Emission node to model the release of electrons from a hot cathode.
The properties specified in the Particle Properties node should be appropriate for
electrons, that is, a mass of me_const and a charge number of -1.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 151
 Go to Release for information about the following sections: Inherit Properties, Initial
Value of Auxiliary Dependent Variables, and Advanced Settings.

SURFACE PROPERTIES
Enter a value or expressions for the Temperature T (SI unit: K) of the cathode. If the
temperature is computed by another physics interface then it can be selected from the
list.

Enter a value or expression for the Metal work function Φ (SI unit: V). The default
expression is 4.5[V].

Enter a value or expression for the Effective Richardson constant A∗

(SI unit: A/(m2 K2)). The default expression is 110[A/(K*cm)^2].

Enter a value for the Number of particles per release N (dimensionless). The default is
100.

I N I T I A L VE L O C I T Y
Enter the Number of particles in velocity space Nvel (dimensionless). The default is 200.
This determines the number of particle velocity vectors to sample at each release point.
Therefore the total number of particles released by the feature is usually N × Nvel, but
may be greater if any auxiliary dependent variables are also sampled from distributions.

The thermionic emission of electrons from the surface follows an axially symmetric
probability distribution function of particle velocities that is centered about the surface
normal. Select the Reverse direction check box to reverse the direction of the surface
normal for the purpose of initializing the particle velocities. This is useful when
releasing particles from internal boundaries where the default surface normal might
not point into the desired adjacent domain. The surface normals on selected
boundaries are indicated by arrows in the Graphics window.

Select an option from the Weighting of macroparticles list: Uniform current (the default),
Uniform speed intervals, or Uniform energy intervals. Each model particle released by the
Thermionic emission feature is a macroparticle representing a certain number of emitted
electrons per unit time.

When Uniform current is selected, each model particle represents the same number of
electrons and the initial particle speeds are sampled from a probability distribution
function based on a population of electrons following a Maxwell-Boltzmann
distribution.

152 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
When Uniform speed intervals is selected, particles are sampled from uniform intervals
from 0 to a maximum speed, which is expressed in terms of the average thermal energy
of the electrons. Specify the Maximum multiple of thermal energy n (default 10).

When Uniform energy intervals is selected, particles are sampled from uniform intervals
from 0 to a maximum energy, which is a multiple of the average thermal energy of the
electrons. Specify the Maximum multiple of thermal energy n (default 10).

For Uniform speed intervals or Uniform energy intervals each model particle can be
weighted differently so that it represents a unique number of electrons per unit time.
Particles that are initialized with a more probable initial speed or energy are assigned a
greater weight for the purpose of computing charge density on domains and current
density at boundaries.

Thermionic Emission Theory

Planar Diode: Application Library path

Particle_Tracing_Module/Charged_Particle_Tracing/planar_diode

Particle Properties
Use the Particle Properties node to specify the particle mass and charge number when
using the Newtonian or Newtonian, first order formulation. Specify the particle velocity
when using the Massless formulation.

PARTICLE MASS
This section is shown when Newtonian or Newtonian, first order is selected as the
Formulation.

Enter a value or expression for the Particle mass mp (SI unit: kg). The default
expression is me_const, which is a predefined constant within COMSOL Multiphysics
for the electron mass, 9.10938188 × 10−31 kg.

If the Relativistic correction check box is selected in the physics interface Advanced
Settings section, instead specify the Particle rest mass mr (SI unit: kg). The default
expression is me_const.

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 153
 CHARGE NUMBER
This section is shown when Newtonian or Newtonian, first order is selected as the
Formulation.

Enter a value for the Charge number Z (dimensionless). The default is −1 corresponding
to the charge number of an electron.

P A R T I C L E VE L O C I T Y
This section is shown when Massless is selected as the Formulation.

Enter a vector for the Particle velocity v (SI unit: m/s) based on space dimension. The
Massless formulation means that the particles follow streamlines of the particle velocity
expression.

Symmetry
Use the Symmetry node to define a plane of symmetry at selected boundaries in the
geometry.

For particle tracing models, the physical interpretation of plane symmetry is as follows:
for every particle that exits the domain through a boundary assigned a Symmetry node,
a new particle enters the modeling domain at the same location at the same time, with
an incoming velocity that mirrors the outgoing velocity of the exiting particle. Thus
the model particle is specularly reflected as if it hit a Wall with the Bounce condition.

Electric Force

The Electric Force node is also only available when Newtonian or

Newtonian, first order is selected as the Formulation in the physics interface
Particle Release and Propagation section.

Use the Electric Force node to define the electric part of the Lorentz force. The
particles are accelerated in the parallel or antiparallel to the electric field depending on
their charge. The force is specified via an electric potential or the electric field. For cases
where the field was computed in the frequency domain, the force can be computed by
multiplying the field by the phase angle. Additionally, piecewise polynomial recovery
can be used, which can give a more accurate representation of the specified electric
field.

154 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
ELECTRIC FORCE
Select an option from the Specify force using list — Electric field (the default) or Electric
potential.

• For Electric field enter values or expressions in the table for the Electric field E (SI
unit: V/m) based on space dimension. If the electric field is computed by another
physics interface then it can be selected from the list.
• For Electric potential enter a value or expression for Electric potential V (SI unit: V).
If the electric potential is computed by another physics interface then it can be
selected from the list.

ADVANCED SETTINGS
Select the Multiply force by phase angle check box in cases where the field is computed
in the frequency domain. Then set the Initial phase angle φ0 (SI unit: rad). The default
value is 0. When the particle trajectories are computed, the value of the field used to
compute the force is computed from the complex field via:

˜
E = real ( E exp ( j ( ωt + φ 0 ) ) )
˜
where E is the electric field value used to compute the force, E (SI unit: V/m) is the
complex-valued electric field, and ω (SI unit: rad/s) is the angular frequency. If the
field is computed using a different study type, such as a Stationary or Time Dependent
study, this option has no effect.

Select the Use piecewise polynomial recovery on field check box to smooth the electric
field using piecewise polynomial recovery. This can give a much more accurate
representation of the electric field as it uses information on adjacent mesh elements to
reconstruct the field. If a coarse mesh is used to compute the field then this option can
be especially useful.

Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

Electric Force Theory

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 155
 Magnetic Force

The Magnetic Force node is only available when Newtonian or Newtonian,

first order is selected as the Formulation in the physics interface Particle
Release and Propagation section

Use the Magnetic Force node to define the magnetic component of the Lorentz force.
This causes the particle to curve perpendicularly to the particle velocity and magnetic
field. A magnetic force alone does no work on the particles, so in the absence of any
other external forces, the particle retains its original energy. Some small variation in the
particle kinetic energy can be expected, but it is just due to numerical error and can
usually be reduced by adjusting the solver settings. The force is specified via a magnetic
flux density.

MAGNETIC FORCE
Enter values or expressions in the table for the Magnetic flux density B (SI unit: T)
based on space dimension, or, if the magnetic flux density is computed by another
physics interface, select another option from the list.

Computing a Force due to Earth’s Magnetic Field

Select Earth’s magnetic field from the Magnetic flux density list to use measurements of
the Earth’s magnetic field to compute the magnetic force. The external data is from
the International Geomagnetic Reference Field (IGRF; Ref. 2). Select an option from
the Location specification list: From particle positions (default) or Geographic location.
Enter a value or expression for the Epoch E (dimensionless) to determine which
measurements are used to compute the Earth’s magnetic field. The default value is
2015. The IGRF data is tabulated in 5-year increments, and data from 1950 to 2015
is included for the Magnetic Force node.

If From particle positions is selected, the magnetic force is computed by treating the
particle coordinates as position vector components with respect to the center of the
Earth, such that the geographic North pole points in the positive z-direction and the
Prime Meridian intersects the positive x-axis.

If Geographic location is selected the magnetic field components are assigned constant
values based on the specified location. This is appropriate when the modeling domain
is very small relative to the length scales over which the Earth’s magnetic field changes
significantly. By default, the magnetic field is determined based on the assumption that
the positive x-axis points to the North, the positive y-axis points to the West, and the

156 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
positive z-axis points upward. Select an option from the Location defined by list: City
(the default) or Coordinates. If City is selected, choose one of several major cities
available in the City list. The default is Las Vegas, USA. If Coordinates is selected, enter
the Latitude Φ and Longitude Θ. Positive values indicate latitude North of the equator
and longitude East of the Prime Meridian, respectively. The default values are 36.1°N,
115.2°W.

For the Latitude and Longitude it is important to indicate that the entered
quantities are in degrees, usually by including [deg] after the numeric
value. Otherwise, numeric values with no specified unit will be interpreted
in the default plane angle unit, which is typically the radian.

For both options in the Location defined by list, enter the Altitude h (SI unit: m). The
default is 0.

ADVANCED SETTINGS
Select the Multiply force by phase angle check box in cases where the field is computed
in the frequency domain. Then set the Initial phase angle φ0 (SI unit: rad). The default
value is 0 radians. When the particle trajectories are computed, the value of the field
used to compute the force is computed from the complex field via:

˜
B = real ( B exp ( j ( ωt + φ 0 ) ) )
˜
where B is the magnetic flux density value used to compute the force, B is the
complex-valued magnetic flux density, and ω is the angular frequency. If the field is
computed using a different study type, such as a Stationary or Time Dependent study,
this option has no effect.

Select the Use piecewise polynomial recovery on field check box to smooth the magnetic
flux density using piecewise polynomial recovery. This can give a much more accurate
representation of the magnetic flux density as it uses information on adjacent mesh
elements to reconstruct the field. If a coarse mesh is used to compute the field then
this option can be especially useful.

Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

Magnetic Force Theory

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 157
 Space Charge Density Calculation
Use the Space Charge Density Calculation node if space charge effects are important.
This node defines a dependent variable for the space charge density in the selected
domains. Optionally, the current density can also be computed.

The Space Charge Density Calculation node computes space charge density
and current density in the same way as the Electric Particle Field
Interaction and Magnetic Particle Field Interaction Multiphysics nodes,
respectively, but does not automatically include the accumulated variables
as source terms when computing the fields. For modeling bidirectionally
coupled particle-field interactions, consider using The Particle Field
Interaction, Non-Relativistic Interface or The Particle Field Interaction,
Relativistic Interface instead of the Space Charge Density Calculation node.

For more information about the calculation of the space charge density,
see Space Charge Density Calculation in the section Theory for the
Particle Field Interaction, Non-Relativistic Interface. For more
information about the calculation of current density, see Current Density
Calculation in the section Theory for the Magnetic Particle Field
Interaction, Relativistic Interface.

SPACE CHARGE DENSITY CALCULATION

If Specify release times is selected from the Particle release specification list in the physics
interface Particle Release and Propagation section, enter a value or expression for the
Charge multiplication factor n (dimensionless). The default is 106. The Charge
multiplication factor can be used to represent each particle as a group of n particles that
follow the same trajectory. This means that the magnitude of the contribution to the
space charge density by each particle is multiplied by n. It is thus possible to model
space charge effects without allocating degrees of freedom for an overwhelming
number of particles.

If Specify current is selected from the Particle release specification list in the
physics interface Particle Release and Propagation section, the Charge
multiplication factor n cannot be specified because the number of charged
particles represented by each model particle is instead controlled by the
Release current magnitude, which is specified in the settings for release
features such as theRelease and Inlet nodes.

158 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Select the Compute current density check box to allocate additional degrees of freedom
for the components of the current density.

The following features are specific applications of the core functionality

used in the Accumulator (Boundary) subnode, described in The
Mathematical Particle Tracing Interface.

Surface Charge Density

When the Particle release specification is set to Specify release times, the Surface Charge
Density subnode is available from the context menu (right-click the Wall parent node)
or from the Physics toolbar, Attributes menu. The Surface Charge Density subnode
defines an accumulated variable on each boundary element in the selection of the
parent node that calculates the charge density of deposited particles.

Surface Charge Density Theory

Current Density
When the Particle release specification is set to Specify current, the Current Density
subnode is available from the context menu (right-click the Wall parent node) or from
the Physics toolbar, Attributes menu. The Current Density subnode uses accumulated
variables on each boundary element in the selection of the parent node to calculate the
current density due to the impact of charged particles.

CURRENT DENSITY
Select a Type: Current density or Normal current density. For Current density
accumulated variables are defined for the components of the current density vector.
For Normal current density one accumulated variable is defined for the normal current
density in each boundary element.

Current Density Theory

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 159
 Heat Source
When the Particle release specification is set to Specify current, the Heat Source subnode
is available from the context menu (right-click the Wall parent node) or from the
Physics toolbar, Attributes menu. The Heat Source subnode defines an accumulated
variable on the selection of the parent node that calculates a boundary heat source
using the kinetic energy of the incident particles.

Heat Source Theory

Etch
When the Particle release specification is Specify current, the Etch subnode is available
from the context menu (right-click the Wall parent node) or from the Physics toolbar,
Attributes menu, which computes the etch rate due to the impact of particles.

ETCH
Select an Angular dependence model: Argon on polysilicon (the default) or Expression.
Argon on polysilicon defines the dependence of the etch rate on the angle of incidence
by using an empirical model for the impact of argon ions on polysilicon. For Expression
enter an Angular dependence function Y (dimensionless). The default is 1.

Enter a Threshold energy Eth (SI unit: J). The default is 50[eV].

Enter a Slope of etch yield curve Y (SI unit: J). The default is 625[eV].

Enter a Plasma type: Collisionless (default) or Collisional. For Collisional enter a value or
expression for the Normal ion current density n ⋅ J i (SI unit: A/m2). The default is 0.
If the normal ion current density is computed by another physics interface then it can
be selected from the list.

SURFACE PROPERTIES
The following settings are required if the Angular dependence model is set to Expression.

Enter the Molecular weight of surface material Ms (SI unit: kg/mol). The default is
0.028[kg/mol].

Enter the Mass density of surface ρ (SI unit: kg/m3). The default is 2320[kg/m^3].

160 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Etch Theory

T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 161
 Theory for the Charged Particle
Tracing Interface
The Charged Particle Tracing Interface theory is described in this section:

• Introduction to the Charged Particle Tracing Interface Theory

• Electric Force Theory
• Magnetic Force Theory
• Collisional Force Theory

Introduction to the Charged Particle Tracing Interface Theory

Motion of charged particles in electromagnetic field is best understood by starting with
the Lagrangian for a charge in an electromagnetic field:

2 2
L = – mp c 1 – v ⋅ v ⁄ c + ZeA ⋅ v – ZeV

where

• mp (SI unit: kg) is the particle mass,

• c = 2.99792458 × 108 m/s is the speed of light in a vacuum,
• v (SI unit: m/s) is the particle velocity,
• Z (dimensionless) is the particle charge,
• e = 1.602176565 × 10-19 C is the elementary charge,
• A (SI unit: Wb/m) is the magnetic vector potential, and
• V (SI unit: V) is the electric scalar potential.

For low velocities, after subtracting the rest energy, the Lagrangian becomes:

2
mp v
L = -------------- + ZeA ⋅ v – ZeV
2

The equations of motion are given by the Lagrange equation:

d  ∂L = ∂L
d t  ∂ v ∂q

162 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
where q (SI unit: m) is the particle position vector. The right-hand side of the
Lagrange equation is:

∂L
= ∇L = Ze ( ∇( A ⋅ v ) – ∇V )
∂q

Using some elementary vector calculus, this can be rewritten as:

∂L
= Ze ( v ⋅ ∇ )A + Ze ( v × ∇×A ) – Ze ∇V
∂q

So, the Lagrange equation becomes:

d m
( v + ZeA ) = Ze ( v ⋅ ∇ )A + Ze ( v × ∇×A ) – Ze ∇V (4-1)
dt p

The second term on the left-hand side of Equation 4-1 represents the total differential
which can be expressed as:

d ( ZeA ) = Ze  ∂A + ( v ⋅ ∇ )A (4-2)

dt ∂t 

Inserting Equation 4-2 into Equation 4-1 results in:

d ( m v ) = – Ze ∂A – Ze ∇V + Ze ( v × ∇×A ) (4-3)
dt p ∂t

Defining the electric field as:

∂A
E = – – ∇V
∂t

and the magnetic flux density as:

B = ∇×A

the equation of motion for a charged particle in an electromagnetic field can be written
as:

d
( m v ) = – ZeE + Ze ( v × B ) (4-4)
dt p

The term on the right-hand side of Equation 4-4 is called the Lorentz force. So far,
only the Electric Force and Magnetic Force have been considered. Additional forces,
such those introduced by the Friction Force and Ionization Loss features, can be added

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 163
 to the right-hand side if necessary.

Electric Force Theory

The Electric Force feature adds the following contribution to the total force Ft:

F t = ZeE

where E (SI unit: V/m) is the electric field. When an electric potential V (SI unit: V)
is used to specify the electric field, the following is used:

E = – ∇V

When the electric field is computed in the frequency domain it is complex valued. The
field must be cast into a real value which depends on the angular frequency and the
simulation time:

˜
E ( t ) = real ( E exp ( j ( ωt + φ 0 ) ) )
˜
where E is the complex-valued electric field, ω is the angular frequency, φ0 is the initial
phase angle, and t is time.

Magnetic Force Theory

The Magnetic Force feature adds the following contribution to the total force Ft:

F t = Ze ( v × B )

where B (SI unit: T) is the magnetic flux density. When the magnetic flux density is
computed in the frequency domain it is complex valued. The field must be cast into a
real value which depends on the angular frequency and the simulation time:

˜
B ( t ) = real ( B exp ( j ( ωt + φ 0 ) ) )
˜
where B is the complex-valued electric field, ω is the angular frequency, φ0 is the initial
phase angle, and t is time.

Computing Earth’s Magnetic Field

It is possible to exert a magnetic force on particles based on the Earth’s magnetic field.
The data for Earth’s magnetic field is the International Geomagnetic Reference Field
(IGRF; Ref. 2) and is based on the Geomag 7.0 software. The IGRF data consists of

164 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
sets of Gauss coefficients that define a spherical harmonic expansion for the magnetic
scalar potential:

L l
l–1
R
-----e-
 
m m m
V ( r, Φ , Θ ) = R e ( g l cos ( mΦ ) + h l sin ( mΦ ) )P l cos ( Θ )
 r
l = 1 m = –l

where

• r (SI unit: m) is the radial distance from Earth’s center,

• L (dimensionless) is the maximum degree of the expansion,
• Φ (SI unit: rad) is the East longitude,
• Θ (SI unit: rad) is the co-latitude (polar angle),
• Re = 6371.2 km is Earth’s radius,
m m
• g l and h l (SI unit: T) are Gauss coefficients, and
m
• P l cos ( Θ ) are the Schmidt normalized associated Legendre functions of degree l
and order m.

Collisional Force Theory

It is possible to apply collisional forces that are deterministic or random. A
deterministic force based on a friction model can be applied using the Friction Force.
Random collisional forces can be applied using the Collisions node, which includes
several different forces for modeling the interaction of particles with a background gas.

FRICTION FORCE
The Friction Force feature adds the following contribution to Ft:

Ft = –mp ν ( v – u )

The limitations of this option are described in the Introduction to the

Charged Particle Tracing Interface Theory section.

where v (SI unit: m/s) is the particle velocity and u (SI unit: m/s) is the fluid velocity.
The collision frequency ν (SI unit: 1/s) is either specified directly, or via a collision
cross section σ (SI unit: m2) and background number density Nd (SI unit: 1/m3):

ν = Nd σ v – u

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 165
 RANDOM COLLISIONAL FORCES
The Collisions node may cause the particle velocity to change during a time step of size
Δt taken by the solver, with a collision probability P defined as:

P = 1 – exp ( – νΔt )

If P is greater than a uniformly distributed random number between 0 and 1 then the
particle velocity is reinitialized. The collision frequency ν is the sum of the individual
collision frequency values νj for each subnode of the Collisions node, which
corresponds to a specific type of collision:

ν =  νj
j=1

where N is the total number of subnodes. If two or more subnodes are present and a
collision occurs, one of the collision types is chosen at random. The probability pj of a
specific collision type occurring is

νj
p j = ----
ν

The collision probability for each collision type is

νj = Nd σ g

where g = v – vg is the relative velocity in the center of mass reference frame.

The Collision cross section σ (SI unit: m2) is defined for each subnode to the Collisions
node. The Background number density Nd (SI unit: 1/m3) can either be defined
individually for each subnode or inherited from the parent node.

For the purpose of computing the collision probability and post-collision particle
velocity, the background gas is assumed to follow a drifting Maxwellian velocity
distribution:

2
1 mg vg – u
f ( v g ) = -------------------------------------------
3⁄2
- exp – -----------------------------
-
( 2πk B T ⁄ m g ) 2k B T

where:

• vg (SI unit: m/s) is the velocity of a gas molecule,

• u (SI unit: m/s) is the average gas velocity,

166 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
• mg (SI unit: kg) is the mass of a gas molecule,
• T (SI unit: K) is the gas temperature, and
• kB = 1.3806488 × 10-23 J/K is the Boltzmann constant.

The mass mg can be expressed in terms of the molar mass Mg (SI unit: kg/mol):

Mg
m g = --------
NA

where NA = 6.02214129 × 1023 1/mol is the Avogadro constant. The molar mass can
either be defined individually for each subnode or inherited from the parent node.

Although all collision types use the same expression for the background gas velocity
distribution and the collision probability, they generally have different expressions for
the post-collision particle velocity. In addition, some collision types produce secondary
particles. The available collision types are described in the following sections.

ELASTIC
The Elastic collision force causes a particle to collide with a background gas molecule
in such a way that the total energy of the system is conserved. The post-collision
velocity of the particle is defined by the expression:

v' = v – m g ⁄ ( m p + m g ) ( g – g′ )

where g = v – vg is the relative velocity in the center of mass reference frame, mg is

the mass of the background gas atoms or molecules. The post-collision relative velocity
g′ is g′ = g R where R is a uniformly distributed random unit vector. No
secondary particles are produced.

EXCITATION
The Excitation collision force causes a particle to collide with a background gas
molecule in such a way that the total energy of the system is not conserved. The
post-collision velocity of the particle is defined by the expression:

v' = v – m g ⁄ ( m p + m g ) ( g – g′ )

where g′ = g′ R and where

2ΔE ( m p + m g )
g′ = g ⋅ g – ---------------------------------------
mp mg

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 167
 where R is a uniformly distributed random unit vector, and ΔE (SI unit: J) is the
kinetic energy lost as a result of the collision. No secondary particles are produced.

ATTACHMENT
The Attachment node causes the model particle to be annihilated, as in the Disappear
condition for the Wall node. Optionally, the attached species can be released as a
secondary particle that is initially at rest.

IONIZATION
The Ionization node causes an electron to collide with a background gas molecule in
such a way that the total energy of the system is not conserved. The post-collision
velocity of the particle is defined by the expression:

collision velocity of the particle is defined by the expression:

v' = v – m g ⁄ ( m p + m g ) ( g – g′ )

where g′ = g′ R and where

2ΔE ( m p + m g )
g′ = g ⋅ g – ---------------------------------------
mp mg

where R is a uniformly distributed random unit vector, and ΔE (SI unit: J) is the
kinetic energy lost as a result of the collision. No secondary particles are produced.

The background gas molecule is assumed to be ionized; that is, one electron collides
with the background gas molecule and causes a secondary electron to be released. For
the results to be physical, this particle should inherit its properties from a Particle
Properties or Override Properties node that is appropriate for electrons.

It is also possible to release the ionized particle as a secondary particle. Both secondary
particles, the ion and the electron, are initially assumed to be at rest.

CHARGE EXCHANGE COLLISIONS

The Resonant Charge Exchange and Nonresonant Charge Exchange nodes are used
to model collisions in which charge is exchanged between a model particle and a
background gas particle. It is possible to compute the post-collision trajectories of
either or both products of the charge exchange reaction. The post-collision velocity of
the ionized species v' and neutralized species v n' are defined by the expressions:

v' = v g + m g ⁄ ( m p + m g ) ( g – g′ )

168 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 v n' = v – m g ⁄ ( m p + m g ) ( g – g′ )

where

g' = g ⊥ cos ( χ ) + g || sin ( χ )

and χ (SI unit: rad) is the scattering angle in the center of mass coordinate system. The
vectors g || and g ⊥ are defined by the expressions:

g'
g || = ------- g
g

g ⊥ = g' ( e 1 cos ( φ ) + e 2 sin ( φ ) )

where φ (SI unit: rad) is a uniformly distributed angle between 0 and 2π. The unit
vectors e 1 and e 2 are defined so that they form an orthonormal basis with the
normalized pre-collision relative velocity g ⁄ g .The post-collision relative velocity
magnitude is defined by the expression:

2ΔE ( m p + m g )
g′ = g ⋅ g – ---------------------------------------
mp mg

where ΔE (SI unit: J) is the energy loss. During resonant charge exchange collisions
the energy loss is 0 and g' = g .

Particle-Matter Interaction Theory

The Particle-Matter Interactions node is used to model interaction of energetic ions
with solid material. The interaction of energetic ions with the target material is divided
into two main interaction types: ionization losses and nuclear stopping.

IONIZATION LOSS
The Ionization Loss node models the interaction of ions with the electrons in the
target material as a continuous braking force:

v
F = – S e ρ ------
v

where

• F (SI unit: N) is the force on the ion,

• Se (SI unit: m4/s2) is the electronic stopping power,

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 169
 • ρ (SI unit: kg/m3) is the mass density of the target material, and
• v (SI unit: m/s) is the ion velocity.

Thus the force always acts opposite the direction of particle motion. For built-in
ionization loss models empirical data from Ref. 1 is used to generate a 1D
interpolation function, from which the stopping power is expressed as a function of the
particle kinetic energy.

NUCLEAR STOPPING
The Nuclear Stopping node models the interaction of ions with the nuclei in the target
material. Unlike ionization losses, which are treated as a force that is continuous as a
function of time, interactions with target nuclei are treated as distinct events that occur
instantaneously with a given probability during each time step. In addition, nuclear
interactions may change the direction of the ion velocity as well as its magnitude.

During each time step taken by the solver, a value of the scattering angle χ
(SI unit: rad) is computed for each particle using the expression

∞
b
χ = π–2  --------------------------------------------- dξ
2
(4-5)
2 Φ(ξ) b
ξ min ξ 1 – ------------ – -----2-
ξε ξ

which is a dimensionless version of the expression for the scattering angle as given in
Ref. 3, in which b (dimensionless) is the reduced impact parameter and ξ is the
dimensionless energy defined as

r
ξ = -----
aI

where r (SI unit: m) is the radial distance from the particle trajectory to the target
nucleus and aI (SI unit: m) is the screening length. The definition of the screening
length changes depending on the option selected from the Screening function list; see
Table 4-2 below.

The reduced energy ε (dimensionless) is defined by the expression

4πε 0
ε = --------------------2- a I E cm
Zp Zm e

170 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Where

• Zp (dimensionless) is the atomic number of the propagating ions,

• Zm (dimensionless) is the atomic number of the material,
• ε0 = 8.854187817 × 10-12 F/m is the permittivity of vacuum,
• e = 1.602176565 × 10-19 C is the elementary charge, and
• Ecm (SI unit: J) is the kinetic energy in the center-of-mass coordinate system.

The lower limit of integration ξmin (dimensionless) is the largest positive root of the
equation

2 Φ ( ξ min ) 2
– ξ min + ---------------------- ξ min + b = 0
ε

The screening function Φ(χ) (dimensionless) changes depending on the option

selected from the Screening function list. The available options are tabulated below.
TABLE 4-2: SCREENING LENGTH AND SCREENING FUNCTION DEFINITIONS

SCREENING SCREENING SCREENING FUNCTION EXPRESSION Φ(ξ)

FUNCTION LENGTH
EXPRESSION aI

None 0.8853a 0 1
--------------------------------
-
0.23 0.23
Zp + Zm

Bohr a0 exp(−ξ)
--------------------------------
-
0.23 0.23
Zp + Zm

Moliere 0.8853a 0 0.35exp(−0.3ξ) + 5.5exp(−1.2ξ) + 0.1exp(−6ξ)

------------------------------
-
2⁄3 2⁄3
Zp + Zm

Lenz-Jensen 0.8853a 0 0.7466exp(−1.038ξ) + 0.2433exp(−0.3876ξ)

------------------------------
- + 0.01818exp(−0.206ξ)
2⁄3 2⁄3
Zp + Zm

Universal 0.8853a 0 0.1818exp(−3.2ξ) + 0.5099exp(−0.9423ξ)

--------------------------------
- +0.2802exp(−0.4028ξ) + 0.2817exp(−0.2016ξ)
0.23 0.23
Zp + Zm

Information on the various screening functions and supplemental references can be

found in Ref. 4.

The value of the reduced impact parameter b is sampled from a Rayleigh probability
distribution using the expression

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 171
 1 ln ( U )-
b = ----- –------------------
a I πLN

where U is a dimensionless random number sampled with uniform probability from

the interval (0,1), L is the distance traveled by the particle during the time step, and N
is the number density of particles in the target material. Thus it is clear that if very small
time steps are taken by the solver, the value of b is typically quite large. This in turn
means that the scattering angle χ is very close to 0, so most collisions tend not to have
a large effect on the particle trajectory. If the computed value of χ is less than the
specified Cutoff scattering angle χc, then the collision is deemed insignificant and the
particle velocity is not reinitialized during that time step.

If the collision is considered significant, that is if χ ≥ χ c , then in addition to being

deflected by the scattering angle, the reinitialized particle trajectory is also rotated by
an azimuthal angle φ that is sampled at random from the interval [0,2π]. A fraction of
the particle energy is also lost to the surroundings; the kinetic energy E of the particle
decreases by the recoil energy T, defined by the expression

4m 1 m 2 χ
T = -----------------------------2- E sin  ---
( m1 + m2 ) 2

where m1 and m2 (SI unit: kg) are the ion mass and the atomic mass of the target
material.

To avoid the computational cost of evaluating the integral in Equation 4-5 at every
time step for every particle, the value of this integral is tabulated for a range of values
of b and ε, then imported into models as a set of 2D interpolation functions. Thus,
there is a finite range of values in which the nuclear stopping data is computed
accurately, corresponding to the interval defined by the inequalities

– 5 ≤ log ( ε ) ≤ 15

– 25 ≤ log ( b ) ≤ 10

outside of this range, the value of the scattering angle is computed by extrapolation
and may be less accurate compared to values that are computed within this range.

Particle Beam Theory

The Particle Beam feature can be used to release nonlaminar particle beams with
specified emittance values. The classification of a particle beam as laminar or

172 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
nonlaminar depends on the distribution of transverse position and velocity within the
beam. Further discussion of laminar and nonlaminar beams can be found in Ref. 5.

A beam is laminar if there is a one-to-one relationship between transverse position and

velocity. The volume occupied by the beam particles in phase space is essentially zero.
In addition, the transverse velocity of particles in a laminar beam must be linearly
proportional to the distance from the axis of beam symmetry; otherwise it is possible
for particles with unequal transverse velocity components to cross at a later point.

If the beam is nonlaminar, it is possible for the trajectories of particles with different
transverse position and velocity components to intersect. Often, such beams occupy
nonzero areas in phase space. The area a beam occupies in phase space is related to a
quantity known as the emittance that can be defined in several different ways.

The particle beam is released normal to a surface and includes several options for
specifying the transverse velocity distribution of beam particles.

DEFINITIONS OF BEAM PROPERTIES

If at least one Particle Beam node is present, additional global variables are defined for
certain beam properties. Quantities that indicate the distribution of transverse beam
position and velocity are defined in a coordinate system that is centered at the average
position of beam particles, qav. In 2D, the transverse direction is orthogonal to the
average particle velocity vav; in 3D, the two transverse directions are orthogonal to vav
and to each other.

Let the transverse displacement from the beam center be denoted x, and the transverse
velocity be expressed using the dimensionless variable x' , which is the ratio of
transverse velocity to axial velocity. For 3D, the following discussion can be extended
to consider two distinct transverse displacement components in orthogonal directions,
called for example x1 and x2.

The area of the phase space ellipse is indicated by the beam emittance. In general,
beams may exhibit a gradual fall-off in particle number density, so that it is not always
possible to define a sharp boundary where the phase space distribution begins and
ends. Instead of being treated as the area of a clearly defined geometric entity, the
phase space ellipse can be defined in a statistical sense. The 1-rms emittance ε1,rms
(SI unit: m) is defined as:

ε 1, rms =  x 2  ( x' ) 2 –  xx' 2

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 173
 Where the brackets represent an arithmetic mean over all particles. In addition, the
4-rms emittance ε4,rms is frequently reported because it corresponds to the area of an
ellipse if the distribution of particles in phase space is uniform:

ε 4, rms = 4ε 1, rms

In addition to the size of the phase space ellipse, several characteristics of its shape can
be described using the Twiss parameters α, β, and γ, defined as:

 xx'
α = – -----------------
ε 1, rms

 x 2
β = -----------------
ε 1, rms

 ( x' ) 2
γ = -----------------
ε 1, rms

In 3D, it is also possible to define the hyperemittance:

εh = ε1 ε2

Depending on the way in which the hyperemittance is defined, either the 1-rms or
4-rms emittance may be used.

RELEASING PARTICLE BEAMS

The Particle Beam node includes built-in options for releasing distributions of particles
in velocity space by sampling from uniform or Gaussian distributions in each transverse
velocity direction. These distributions can be upright if the initial value of the Twiss
parameter α is set to 0. In this case, for the elliptical distributions of particles in phase
space, the semimajor and semiminor axes of the ellipse are initially parallel to the x- and
x' -axes. The initial values of the Twiss parameters must fulfill the Courant-Snyder
condition,

2
γβ – α = 1

Thus, out of the three Twiss parameters, it is only necessary to specify β and α.

Given the initial value of the Twiss parameter β, the 1-rms beam emittance ε and the
Twiss parameter α, the initial distribution of particles in phase space depends on the
option selected from the Sampling from phase space ellipse list.

174 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Sampling from Phase Space
The initial particle positions and velocities can be generated by sampling from the
following function, as described in Ref. 6:

x 2 ax' – a'x 2 y 2 by' – b'y 2

A =  --- +  --------------------- +  --- +  ---------------------
2
 a  εx   b  εy 

where a and b are the semi-major and semi-minor axis of the ellipse in physical space,
a' and b' are the envelope angles which are related to the Twiss parameters according
to:

a = 2 ε x β, b = 2 εy β

ε ε
a' = – 2α ----x-, b' = – 2α ----y- .
β β

The amplitude of A2 can be resolved into two components:

2 2 2
A = Ax + Ay

where:

2 ax' – a'x 2
A x =  --- +  ---------------------
2 x
a εx

y 2 by' – b'y 2
A y =  --- +  --------------------- .
2
 b  εy 

Algorithm for generating initial positions and velocities

As described in Ref. 6, start by defining two uniformly-distributed random numbers
uˆ A ∈ [ 0, 1 ] and uˆ φ ∈ [ 0, 1 ] , then define:

A x = A uˆ φ, A y = A 1 – uˆ φ

with the definition of A depends on the type of distribution and follows later. Now
define two additional uniformly-distributed random numbers β̂ x ∈ [ 0, 1 ] and
β̂ y ∈ [ 0, 1 ] then the relative initial positions are the particles are given by:

x = aA x cos ( 2πβˆ x ), y = bA y cos ( 2πβˆ y )

and the relative initial transverse velocities by:

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 175
 4ε x
x' = A x  a' cos ( 2πβˆ x ) – -------- sin ( 2πβˆ x )
 a 

4ε y
y' = A y  b' cos ( 2πβˆ y ) – -------- sin ( 2πβˆ y )
b

2 2
z' = 1 – x' – y' .

The definition of A changes depending on the requested distribution:

TABLE 4-3: DEFINITION OF VARIABLE A FOR DIFFERENT DISTRIBUTION TYPES

KV WATERBAG PARABOLIC

A = 1 3
A = --- uˆ A
2 η – 2π
A = 1 – 2 cos  ----------------
 3 

η = acos ( 1 – 2uˆ A )

The Gaussian distribution is generated in a more straightforward way. Let

gˆ x, gˆ y, gˆ x', gˆ y' be normally distributed random numbers with zero mean and standard
deviation one. The beam position and velocity can be defined as:

a b
x = --- gˆ x, y = --- gˆ y
2 2

and

a' 4ε x b' 4ε y
x' = ---- x + -------- gˆ x', y' = ---- y + -------- gˆ y',
2 2
z' = 1 – x' – y' .
a 2a b 2b

Once the relative initial positions and velocities are generated, they are converted to
global coordinates using the following:

r = r c + xt 1 + yt 2

v = x'Vt 1 + y'Vt 2 + z'Vn

where rc is the beam centroid, t1 and t2 are the two tangent vectors on the surface, n
is the surface normal and V is the velocity magnitude.

176 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
When the Beam symmetry is Symmetric, the same initial values of the Twiss parameter
β, Twiss parameter α and beam emittance ε are applied to each of the two transverse
directions. When the Beam symmetry is Asymmetric, distinct values of these parameters
can be assigned to each transverse direction.

When the Particle release specification property is Specify current, the Beam symmetry is
Symmetric, the Transverse velocity distribution specification is Specify emittance and Twiss
parameters and the Emittance specification is Specify brightness, the Brightness is entered
instead of the emittance. In this case, the 1-rms emittance is computed from the
Brightness B (SI unit: A/m2) and the Release current magnitude I (SI unit: A) using

I- .
ε rms = ---
B

Converting from Phase Space Ellipse Dimensions to Twiss Parameters

When the Transverse velocity distribution specification is set to Specify phase space ellipse
dimensions, the Twiss parameters are computed from the Maximum transverse
displacement, xm, the Maximum relative transverse velocity, x’m and the Rotation angle,
θ using the following

xm
β = --------
x' m

and

x m x' m
ε rms = ---------------- .
4

The Twiss parameter α is computed by first defining a rotation matrix

M = cos θ – sin θ = m 11 m 12
sin θ cos θ m 21 m 22

then defining the semimajor and semiminor axis using

xm
x m, 1 = -------------------------- ε rms
2 2
x m + x' m

x' m
x' m, 1 = -------------------------- ε rms
2 2
x m + x' m

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 177
 which results in

x' m, 1 x m, 1
α = – ------------- m 12 m 22 – ------------- m 11 m 21 .
x m, 1 x' m, 1

Thermionic Emission Theory

The Thermionic Emission feature is used to release electrons from a hot cathode. It
releases model particles from a boundary such that the emitted current density
magnitude Jth (SI unit: A/m2) is (Ref. 7)

J th = A T exp  – -----------
∗ 2 eΦ
 k B T

where

• T (SI unit: K) is the temperature of the cathode,

• Φ (SI unit: V) is the work function of the metal,
• A∗ (SI unit: A/(m2 K2)) is the effective Richardson constant,
• kB = 1.3806488 × 10-23 J/K is the Boltzmann constant, and
• e = 1.602176565 × 10-19 C is the elementary charge.

The total current Ith (SI unit: A) released from the surface is

I th =   Jth dΩ
Ω

where the surface integral is taken over the selected boundaries.

I N I T I A L VE L O C I T Y D I R E C T I O N S
The initial particle velocity components are

v t1 = V sin θ cos φ

v t2 = V sin θ sin φ

v n = V cos θ

where vn is the velocity component normal to the cathode surface and vt1 and vt2 are
the two orthogonal velocity components parallel to the surface. The azimuthal angle
φ is uniformly distributed in the interval [0, 2π]. The polar angle θ is

178 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 θ = asin U 1

where U1 is a uniformly distributed random number in the interval [0, 1]. The particle
speed V (SI unit: m/s) can be sampled from a probability distribution function in
several different ways, which are described in the following section.

INITIAL SPEED
The probability distribution function of the initial electron speed is more easily
expressed in terms of the normalized kinetic energy W (dimensionless), defined as

me
W = --------------- V 2 (4-6)
2k B T

where me = 9.10938291 × 10-31 kg is the electron mass. The thermal electrons are
assumed to be nonrelativistic. The probability distribution function of the normalized
kinetic energy is

f ( W ) = W exp ( – W ) (4-7)

The way in which the initial particle kinetic energy is sampled from this probability
distribution function is controlled by the Weighting of macroparticles list in the physics
feature Initial Velocity section.

Uniform Current
For Uniform Current the values of the normalized kinetic energy for the particles are
sampled from Equation 4-7 using the method of inverse transform sampling, in which
the probability distribution function (PDF) is used to compute a cumulative
distribution function (CDF) which is then normalized to unity and inverted (inverse
normal CDF). When uniformly distributed random numbers in the interval [0,1] are
used as input to the inverse normal CDF, the resulting set of values follows the PDF.

The CDF of Equation 4-7 is

c( W) =  W' exp ( –W' ) dW' = 1 – ( W + 1 ) exp ( – W )

0

which is already normalized, i.e.

lim W → ∞ ( 1 – ( W + 1 ) exp ( – W ) ) = 1

The inverse normal CDF is

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 179
 W = – LambertW ( – 1, ( U 3 – 1 ) exp ( – 1 ) ) – 1

where LambertW(−1, x) is the −1 branch of the Lambert W function or product log,

defined as the real root of the equation

LambertW ( – 1, x ) exp ( LambertW ( – 1, x ) ) = x

for which

LambertW ( – 1, x ) ≤ – 1

U3 is a uniform random number in the interval [0,1] that is uncorrelated with the
random numbers used to initialize the polar angle θ and azimuthal angle φ.

The PDF and CDF are plotted in Figure 4-1.

Figure 4-1: Normalized probability distribution function (PDF) and cumulative

distribution function (CDF) for the normalized particle kinetic energy.

Because the initial particle energy is sampled from the probability distribution function
given in Equation 4-7, each particle carries the same weight; that is, each model
particle is a macroparticle representing an equal number of electrons emitted per unit
time. This macroparticle weighting is represented by a static degree of freedom stored
for all model particles called the effective frequency of release frel,

f rel = const

To yield the specified total current Ith, the effective frequency of release of the ith
model particle must be

180 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 I th
f rel, i = -------
eN

where N (dimensionless) is the total number of model particles released by the feature.

For more information on the effective frequency of release and its effect
on space charge density calculation, see Space Charge Density Calculation
in Theory for the Particle Field Interaction, Non-Relativistic Interface.

Uniform Energy Intervals

When Uniform energy intervals is selected, the initial normalized kinetic energy of each
particle is uniformly sampled from the interval [0, n], where n is a user-defined
proportionality factor. Because the energy is uniformly sampled instead of following
the probability distribution given by Equation 4-7, the probability distribution must
instead be incorporated into the effective frequency of release:

f rel ∝ W exp ( – W ) (4-8)

Thus the effective frequency of release of the ith particle is

I th W i exp ( – W i )
f rel, i = ------- -----------------------------------------

e W exp ( – W )
j j

where the sum over j in the denominator is taken over all model particles released by
the feature.

Equation 4-7 has a local maximum at W = 1, after which it gradually decreases. Some
representative values of the CDF are given in Table 4-4.
TABLE 4-4: SELECTED VALUES OF THE CDF FOR NORMALIZED ELECTRON ENERGY

W CDF

3.89 0.9
6.64 0.99
9.23 0.999

Therefore a value of n = 10, for example, will encompass more than 99.9% of the
cumulative distribution function.

Uniform Speed Intervals

When Uniform speed intervals is selected, the initial speed of each particle is uniformly
sampled from the interval [0, Vmax], where

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 181
 2k B T
V max = n --------------
-
me

where n (dimensionless) is a user-defined proportionality factor.

Differentiating both sides of Equation 4-6 and rearranging yields

kB T
dV = - dW
----------------
2m e W

Substituting into Equation 4-7 yields

me 2m e W
f ( W )dW = f  --------------- V 2 ----------------
- dV
2k B T kB T

Therefore, compared to Equation 4-8 the proportionality factor is multiplied by an

additional term proportional to W :

f rel ∝ W 3 / 2 exp ( – W ) (4-9)

Thus the frequency of release of the ith particle is

I th W i3 / 2 exp ( – W i )
f rel, i = ------- -----------------------------------------------

e W 3 / 2 exp ( – W )
j j

where again the sum over j in the denominator is taken over all model particle emitted
by the feature.

When to Sample from Uniform Energy or Speed Intervals

The main motivation for selecting Uniform energy intervals or Uniform speed intervals is
that these options devote a disproportionately large number of degrees of freedom to
electrons at the extreme ends of the probability distribution function. For example,
when sampling from Uniform energy intervals with n > 8, equal representation would
be given to particles in the neighborhood of W = 1 and W = 7.64, despite W = 1 having
a probability density 100 times more greater than W = 7.64.

It can be useful to devote additional degrees of freedom to particles with extremely

high and low energy because these particles may have a greater influence on the
behavior of a system than electrons of the most probable energy, and therefore they
might require finer discretization in velocity space to yield a statistically converged
solution. For example, in plane parallel vacuum diodes in the space charge limited

182 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
regime, the presence of electrons between the electrodes creates a potential barrier
which only the most energetic electrons can bypass. Devoting more degrees of
freedom to the most energetic electrons in the distribution, despite the low percentage
of the overall number of electrons that they represent, can lead to a more robust and
reproducible solution because a larger sample of model particles is used to compute
the charge density in the region beyond the potential barrier.

Surface Charge Density Theory

The Surface Charge Density feature adds an accumulated variable to the boundary
elements in the selection of its parent node. The variable defined on one of these
boundary elements is incremented by the charge of each particle that hits it, divided
by the surface area of the element,

1
σ i = ------
Ai  eZ
where

• σi (SI unit: C/m2) is the surface charge density in boundary element i,

• Ai (SI unit: m2) is the area of the boundary element,
• e = 1.602176565 × 10-19 C is the elementary charge, and
• Z (dimensionless) is the charge number.

The sum is taken over all particles that hit the boundary element.

Current Density Theory

The Current Density feature adds one or more accumulated variables to each
boundary element in the selection of its parent node.

If the Type is set to Normal current density, the normal current density in each boundary
element is

1
J n, i = ------
Ai  frel eZ cos φ
where

• Jn,i (SI unit: A/m2) is the normal current density in boundary element i,
• Ai (SI unit: m2) is the area of the boundary element,

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 183
 • e = 1.602176565 × 10-19 C is the elementary charge,
• Z (dimensionless) is the charge number,
• φ (SI unit: rad) is the acute angle of incidence measured from the surface normal,
and
• frel (SI unit: 1/s) is the effective frequency of release.

The sum is taken over all particles that hit the boundary element. A particle’s effective
frequency of release is determined by the current of the particle’s release feature and
the total number of particles released by the feature,

I
( f rel ) = ----------------
i N tf eZ

If the Type is set to Current density, the components of the current density vector Js,i
are then calculated in each boundary element:

1 v
J s, i = ------
Ai  frel eZ ---V-
where v/V is a unit vector in the same direction as the velocity of the incident particle.

Heat Source Theory

The Heat Source feature adds an accumulated variable to the boundary elements in
the selection of its parent node. This accumulated variable represents a boundary heat
source equal to the kinetic energy lost by incident particles per unit time, divided by
the surface area of the boundary element:

1
Q i = ------
Ai  frel E
where

• Qi (SI unit: W/m2) is the heat source per unit area in boundary element i,
• Ai (SI unit: m2) is the surface area of the element,
• frel (SI unit: 1/s) is the effective frequency of release of the incident particle, and
• E (SI unit: J) is the particle kinetic energy.

The sum is taken over all particles that hit the boundary element.

184 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
Etch Theory
The Etch feature uses accumulated variables to calculate the etch rate due to physical
sputtering by energetic particles on a surface. The dependence of the sputtering yield
on the energy of particles and their angle of incidence has been described by Yin and
Sawin (Ref. 8) and by Guo and Sawin (Ref. 9).

If the Plasma type is Collisionless, a single accumulated variable is defined on each

boundary element in the selection of the parent node. The etch rate (SI unit: m/s) is
defined as

Ms  E E th
E E, i = ------------------
ρN A A i  frel Y ( φ )  -
------ – --------
Ys Ys 

where

• Ms (SI unit: kg/mol) is the molar mass of the surface material,

• ρ (SI unit: kg/m3) is the density of the surface material,
• NA = 6.02214129 × 1023 1/mol is Avogadro’s constant,
• Ai (SI unit: m2) is the surface area of the boundary element,
• frel (SI unit: 1/s) is the effective frequency of release,
• Y(φ) (dimensionless) is a function that indicates the dependence of the etch rate on
the angle of incidence φ,
• E (SI unit: J) is the particle kinetic energy,
• Eth (SI unit: J) is the threshold energy, and
• Ys (SI unit: J) is the slope of the etch yield curve.

The sum is taken over all particles that hit the boundary element.

If the Plasma type is Collisional, the effective frequency of release is no longer a reliable
indicator of the incident charged particle current. Instead, incident particles in each
boundary element are assigned a weighting factor Γi (SI unit: 1/(m2s)), defined as

n ⋅ Ji
Γ i = ---------------------------
 eZ sin φ

where the numerator is the normal current of charged particles at the surface. The etch
rate is then defined as

T H E O R Y F O R T H E C H A R G E D P A R T I C L E TR A C I N G I N T E R F A C E | 185
 Γi Ms E th
 Y ( φ ) 
E E, i = ------------- E
---- – --------
-
ρN A Y Y 

A typical angular dependence function is given by Guo and Sawin (Ref. 9) as

Y ( φ ) = – 141.3cos 6 φ + 641.1cos 5 φ – 1111.3cos 4 φ

+ 944.6cos 3 φ – 422.0cos 2 φ + 95.31cosφ – 5.460

where the angle of incidence φ is measured from the surface normal. This angular
dependence function is appropriate for physical sputtering of polysilicon by argon ions.

References for the Charged Particle Tracing Interface

1. NIST Stopping-Power and Range Tables for Electrons, Protons, and Helium Ions,
http://www.nist.gov/pml/data/star/index.cfm.

2. International Geomagnetic Reference Field (IGRF),

http://www.ngdc.noaa.gov/IAGA/vmod/igrf.html.

3. L.D. Landau and E.M. Lifshitz, Mechanics, 3rd ed., Elsevier, 1976.

4. J. F. Ziegler, J. P. Biersack, and M. D. Ziegler, The Stopping and Range of Ions and
Matter, SRIM, 2015.

5. S. Humphries, Charged Particle Beams, Dover, 2013.

6. S. M. Lund, T. Kikuchi, R. C. Davidson, “Generation of initial Vlasov distributions

for simulation of charged particle beams with high space-charge intensity,” Physical
Review Special Topics--Accelerators and Beams, 2007.

7. P. T. Kirstein, G. S. Kino, and W. E. Waters, Space charge flow, McGraw-Hill, 1967.

8. Y. Yin and H.H. Sawin, “Surface Roughening of Silicon, Thermal Silicon Dioxide,
and Low-k Dielectric Coral Films in Argon Plasma,” J. Vac. Sci. Technol. A, vol. 26,
no. 1, pp. 151–160, 2008.

9. W. Guo and H.H. Sawin, “Modeling of the Angular Dependence of Plasma

Etching,” J. Vac. Sci. Technol. A, vol. 27, no. 6, pp. 1326–1336, 2009.

186 | C H A P T E R 4 : C H A R G E D P A R T I C L E TR A C I N G
 5

Particle Tracing for Fluid Flow

This chapter describes the Particle Tracing for Fluid Flow interface found under
the Fluid Flow branch ( ) when adding a physics interface.

In this chapter:

• The Particle Tracing for Fluid Flow Interface

• Theory for the Particle Tracing for Fluid Flow Interface
See The Single-Phase Flow, Laminar Flow Interface in the COMSOL Multiphysics
Reference Manual for other Fluid Flow interface and feature node settings.

187
 T he P a r ti c le Trac i n g for Fl u i d Fl ow
Interface
The Particle Tracing for Fluid Flow (ftp) interface ( ), found under the Fluid
Flow>Particle Tracing branch ( ) when adding a physics interface, is used to compute
the motion of particles in a background fluid. Particle motion can be driven by drag,
gravity, electric, magnetic, and acoustophoretic forces. User-defined forces can be
added. It is also possible to compute the particle mass and temperature as well as
particle-fluid interactions.

When this physics interface is added, these default nodes are also added to the Model
Builder—Wall and Particle Properties. Then, from the Physics toolbar, add other nodes
that implement, for example, boundary conditions and volume forces. You can also
right-click Particle Tracing for Fluid Flow 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 fpt.

PARTICLE RELEASE AND PROPAGATION

The Formulation, Relativistic Correction, Store extra time steps for wall interactions, and
Maximum number of secondary particles settings are the same as for The Mathematical
Particle Tracing Interface, except that only the Newtonian, Newtonian, first order, and
Massless formulations are available.

Particle Release Specification

Select an option from the Particle release specification list—Specify release times (the
default) or Specify mass flow rate. If Specify release times is selected, then each model
particle is treated as the instantaneous position of one or more particles for the purpose
of modeling fluid-particle interactions. This means, for example, that if the Volume

188 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Force Calculation node is used, the volume force on the fluid is only nonzero in mesh
elements that are currently occupied by particles.

If Specify mass flow rate is selected, then for the purpose of modeling fluid-particle
interactions, each model particle traces a path that is followed by a number of particles
per unit time. This means that the volume force computed by the Volume Force
Calculation node is nonzero in all mesh elements that the particle trajectories pass
through, not just at the instantaneous positions of the particles. In other words, the
model particles leave behind a trail of nonzero force components in the mesh elements
they pass through.

The Specify mass flow rate option is primarily used to model streams of particles under
steady-state conditions. Changing the particle release specification affects some inputs
in the settings windows for release features such as the Release and Inlet nodes. In
addition, the Mass Deposition subnode to the Wall node is only available with the
Specify release times option, while the Boundary Load and Mass Flux subnodes are
only available when Specify mass flow rate is selected.

ADDITION AL VAR IA BL ES
The options Store particle status data and Store particle release statistics are the same
as for The Mathematical Particle Tracing Interface. The option Include out-of-plane
degrees of freedom is the same as for The Charged Particle Tracing Interface.

Compute Particle Temperature

Select the Compute particle temperature check box to compute particle temperatures
(the default is to not compute particle temperatures). When this option is activated the
temperature of the particle is computed by solving an additional ordinary differential
equation per particle. Thermal properties for the particles can be specified in the
Settings window for the Particle Properties node.

By selecting this check box it is possible to add Heat Source, Convective

Heat Losses and Radiative Heat Losses nodes, which are available from
the context menu (right-click the parent node, then view the Thermal list)
or from the Physics toolbar, Domains menu.

Computing Particle Temperature in the theory section.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 189
 Compute Particle Mass
Select the Compute particle mass check box to compute particle masses (the default is
to not compute particle masses). When this option is activated an additional equation
for the particle mass is computed by solving an additional ordinary differential
equation per particle. The accretion rate of each particle can be specified in the Settings
window for the Particle Properties node.

The Compute particle mass check box must be selected in order to use the
Droplet Breakup and Nozzle features, which are available from the
context menu (right-click the parent node). The Droplet Breakup feature
is also accessible from the Physics toolbar, Domains menu. The Nozzle
feature is accessible from the Physics toolbar, Points and Global menus.

Computing Particle Mass in the theory section.

Enable Macroparticles
Some fluid-particle systems contain such a large number of particles or droplets that
modeling each entity individually is not feasible. In such cases, it is often convenient to
introduce the concept of a macroparticle, or a single model particle that can represent
a larger number of real particles.

Select the Enable macroparticles check box to allocate an auxiliary dependent variable
for a dimensionless multiplication factor, allowing the number of real particles
represented by each model particle to be stored.

The Enable macroparticles check box must be selected in order to use the
Nozzle feature, which is available from the context menu (right-click the
parent node) or from the Physics toolbar, Points and Global menus.

The Enable macroparticles check box is only available if Specify release

times is selected from the Particle release specification list. Otherwise, the
weighting of each macroparticle is determined by the Mass flow rate,
which is specified in the release features.

Include Rarefaction Effects

Select the Include rarefaction effects check box to apply correction factors to the forces
defined by the Drag Force and Thermophoretic Force nodes. These correction factors

190 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
improve the accuracy of the drag and thermophoretic forces when the particle
Knudsen number is significantly large. This can be used to model the motion of
particles in a rarefied gas flow.

Several options, such as Include out-of-plane degrees of freedom, Compute

particle mass, and Compute particle temperature, cause auxiliary
dependent variables to be allocated for the particles. These auxiliary
dependent variables are solved for by defining first-order differential
equations. If, in addition, Newtonian is selected from the Formulation list,
then the resulting system of equations includes both first- and
second-order equations.

Mixed first- and second-order equations systems are not supported for all
solver configurations. For example, it is not possible to use explicit time
stepping methods such as Dormand-Prince 5. To use explicit
time-stepping methods with auxiliary dependent variables, consider using
the Newtonian, first order formulation instead.

ADVANCED SETTINGS
These settings are the same as for The Mathematical Particle Tracing Interface.

DEPENDENT VA RIA BLES

The dependent variables (field variables) are the Particle position, Particle position
components, Particle velocity, Particle velocity components, Particle temperature, Particle
mass, and Number multiplication factor. Note that not every dependent variable is
needed for every combination of physics interface settings; for example, the Particle
velocity and Particle velocity components are only used if Newtonian, first order is
selected from the Formulation list. The names can be changed but the names of fields
and dependent variables must be unique within a model.

• Domain, Boundary, Pair, and Global Nodes for the Particle Tracing for
Fluid Flow Interface
• Theory for the Mathematical Particle Tracing Interface
• Newtonian Formulation and Massless Formulation
• Theory for the Particle Tracing for Fluid Flow Interface

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 191
 • Particle Trajectories in a Laminar Static Mixer: Application Library
path Particle_Tracing_Module/Fluid_Flow/laminar_mixer_particle
• With the CFD Module see Particle Tracing in a Micromixer:
Application Library path
CFD_Module/Particle_Tracing/micromixer_particle_tracing.
• With the Acoustics Module see Acoustic Levitator: Application
Library path Acoustics_Module/Nonlinear_Acoustics/acoustic_levitator.

Domain, Boundary, Pair, and Global Nodes for the Particle Tracing
for Fluid Flow Interface
The Particle Tracing for Fluid Flow Interface has these domain, boundary, pair, and
global nodes 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). The context menu includes dedicated menus for Forces and Thermal features.

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.

In the COMSOL Multiphysics Reference Manual see Table 2-3 for links
to common sections and Table 2-4 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.

For a theoretical background to the forces, see Theory for the Particle
Tracing for Fluid Flow Interface.

192 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
These nodes and subnodes are described in this section (listed in alphabetical order):

• Acoustophoretic Force • Magnetic Force

• Boundary Load • Magnetophoretic Force
• Brownian Force • Mass Deposition
• Convective Heat Losses • Mass Flux
• Dielectrophoretic Force • Nozzle
• Drag Force • Particle Properties
• Droplet Breakup • Radiative Heat Losses
• Electric Force • Shell
• Erosion • Symmetry
• Gravity Force • Thermophoretic Force
• Heat Source • Volume Force Calculation
• Lift Force

These nodes and subnodes are described for The Mathematical Particle Tracing
Interface (listed in alphabetical order):

• Accumulator (Boundary) • Particle-Particle Interaction

• Accumulator (Domain) • Periodic Condition
• Auxiliary Dependent Variable • Release
• Axial Symmetry • Release from Data File
• Force • Release from Edge
• Inlet • Release from Grid
• Nonlocal Accumulator • Release from Point
• Outlet • Secondary Emission
• Override Properties • Velocity Reinitialization
• Particle Continuity • Wall (the default boundary
• Particle Counter condition)

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 193
 Particle Properties
Use the Particle Properties node to specify the particle properties, charge number, and
particle velocity based on whether a Newtonian formulation or massless formulation is
selected.

PARTICLE PROPERTIES
This section is shown when Newtonian or Newtonian, first order is selected as the
Formulation for the physics interface. If the Compute particle mass check box is cleared
in the physics interface Additional Variables section, select the Particle property
specification: Specify particle density and diameter (the default), Specify particle mass and
density, or Specify particle mass and diameter. Then enter the applicable values or
expressions for the following:

• Particle density ρp (SI unit: kg/m3). The default is 2200 kg/m3.

• Particle diameter dp (SI unit: m). The default is 1E-6 m.
• Particle mass mp (SI unit kg). The default is 1E-6 kg.

If the Compute particle mass check box is selected, enter a value or expression for the
Particle density.

Select an option from the Particle type list—Solid particles (the default) or Liquid
droplets/bubbles. If Liquid droplets/bubbles is selected, enter values or expressions for
the following:

• Particle dynamic viscosity μp (SI unit: Pa·s) the default is 1[mPa*s].

• Particle surface tension σp (SI unit: N/m). The default is 0.0729[N/m].

Liquid droplets/bubbles must be selected for at least one species in the

model in order to use the Droplet Breakup and Nozzle features.

CHARGE NUMBER
Enter a value or expression for the Charge number Z (dimensionless). The default is 0.

P A R T I C L E VE L O C I T Y
This section is shown when Massless is selected as the Formulation for the physics
interface. Enter a vector for the Particle velocity v (SI unit: m/s) based on space
dimension. The Massless formulation means that the particles follow streamlines of the
particle velocity expression.

194 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
THERMAL PROPERTIES
This section is shown when the Compute particle temperature check box is selected in
the physics interface Additional Variables section. Enter a value or expression for the
Particle specific heat capacity Cp (SI unit: J/(kg K)). The default value is
2200[J/(kg*K)]. Then enter a value or expression for the Particle latent heat hp
(SI unit: J/kg). The default value is 42000[J/kg].

ACCRETION RATE
This section is shown when the Compute particle mass check box is selected in the
physics interface Additional Variables section. Enter a value for the Accretion rate R
(SI unit: kg/s). The default is 0. The Accretion rate is the time derivative of the particle
mass.

Symmetry
Use the Symmetry node to define a plane of symmetry at selected boundaries in the
geometry.

For particle tracing models, the physical interpretation of plane symmetry is as follows:
for every particle that exits the domain through a boundary assigned a Symmetry node,
a new particle enters the modeling domain at the same location at the same time, with
an incoming velocity that mirrors the outgoing velocity of the exiting particle. Thus
the model particle is specularly reflected as if it hit a Wall with the Bounce condition.

The main advantage of using the Symmetry condition instead of a Wall

feature with the Bounce condition is that the Symmetry boundaries are
ignored for the purposes of computing the distance and direction of each
particle from the nearest wall, which is used for wall induced lift, drag,
and turbulent dispersion by the Drag Force and Lift Force features.

Drag Force

The Drag Force node is only available when Newtonian or Newtonian, first
order is selected as the Formulation on the Settings window for The
Particle Tracing for Fluid Flow Interface.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 195
 Use the Drag Force node to select a drag law, specify the velocity field, enter material
properties, and optionally apply random perturbations to account for particle motion
in a turbulent flow.

DRAG FORCE
Select a Drag law: Stokes (the default), Schiller-Naumann, Haider-Levenspiel, Oseen
correction, Hadamard-Rybczinski, or Standard drag correlations.

Stokes drag and Oseen correction are applicable for particles that have a
relative Reynolds number much less than one. Either should be a good
enough approximation in most cases. If the particle Reynolds number is
greater than one, then select Schiller-Naumann. If, in addition, the particles
are non-spherical, select Haider-Levenspiel. If the particles are very pure gas
bubbles or liquid droplets, select Hadamard-Rybczinski. The Standard drag
correlations are a set of piecewise-continuous functions that are applicable
over a wide range of relative Reynolds numbers.

• For all choices, enter coordinates for the Velocity field u (SI unit: m/s) based on
space dimension. If another physics interface is present which computes the velocity
field then this can be selected from the list.
• For all choices, the Dynamic viscosity μ (SI unit: Pa·s) is taken From material. For User
defined enter another value or expression.
• If Stokes is selected as the Drag law, select a Wall Correction.to include a correction
factor to the drag force for particles near walls.
• If Schiller-Naumann, Haider-Levenspiel, Oseen correction, Hadamard-Rybczinski, or
Standard drag correlations is selected as the Drag law, or if the Include rarefaction
effects check box is selected in the physics interface Particle Release and Propagation
section, the fluid Density ρ (SI unit: kg/m3) is taken From material. For User defined
enter another value or expression.
• For Haider-Levenspiel enter a value or expression for the particle Sphericity Sp
(dimensionless). The default is 1.
• For Hadamard-Rybczinski enter a value or expression for the Particle dynamic viscosity
μp (SI unit: Pa·s). The default value is 1.8e-5[Pa*s].

Particle Motion in a Fluid in Theory for the Particle Tracing for Fluid
Flow Interface.

196 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
RAREFACTION EFFECTS
This section is only available if the Include rarefaction effects check box is selected in the
physics interface Particle Release and Propagation section. It defines a correction factor
that is multiplied by the drag force to account for a large particle Knudsen numbers.

Select an option from the Rarefaction effects list: Basset, Epstein, Phillips,
Cunningham-Millikan-Davies (the default), or User defined.

The Basset model is appropriate for relative Knudsen numbers much less
than one, while the Epstein model is appropriate for free molecular flows.
The Phillips model is usable at intermediate values of the relative Knudsen
number and shares the same asymptotic behavior as the Epstein and Basset
numbers at very large and small Knudsen numbers, respectively. The
Cunningham-Millikan-Davies model includes three tunable parameters that
can be used to fit the drag force correction factor to empirical data.

Select an option from the Mean free path calculation list—Ideal gas, hard sphere collisions
(the default) or User defined correction. For User defined correction enter a Mean free
path correction λ’/λ (dimensionless). The default value is 1.

• If Basset, Epstein, or Phillips is selected from the Rarefaction effects list, enter an
Accommodation coefficient σR (dimensionless). The default value is 1. The
accommodation coefficient may be interpreted as the fraction of gas molecules that
undergo diffuse reflection at the particle surface.
• If Cunningham-Millikan-Davies is selected from the Rarefaction effects list, enter the
three dimensionless coefficients C1, C2, and C3. The default values are 2.514, 0.8,
and 0.55, respectively.

Enter a value or expression for the Pressure p (SI unit: Pa). If a physics interface is
present that computes the pressure, it can be selected directly from the list.

Drag Force Correction Factors in Theory for the Particle Tracing for
Fluid Flow Interface.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 197
 INCLUDE WALL CORRECT IONS
This section is only available if Stokes is selected as the Drag law in the Drag Force
section. Select the Include wall corrections box to define a correction factor that is
multiplied by the drag force as particles approach a wall.

Wall Corrections in Theory for the Particle Tracing for Fluid Flow
Interface.

TU R B U L E N T D I S P E R S I O N
If particles are moving in a turbulent flow, this section can be used to apply random
perturbations to the drag force to account for the turbulence.

Select an option from the Turbulent dispersion model list—None (the default), Discrete
random walk, Discrete random walk, variable time step, or Continuous random walk.

• If None is selected, no turbulent dispersion term is applied.

• If Discrete random walk is selected, a random term is added to the background fluid
velocity when computing the drag force at every time step taken by the solver. The
random perturbation is regenerated at every time step.
• If Discrete random walk, variable time step is selected, a random term is added to the
background fluid velocity when computing the drag force at every time step taken
by the solver. Unlike Discrete random walk, the random perturbation term is only
regenerated at time intervals greater than or equal to the eddy lifetime of the flow.
Usually, this improves the statistical convergence of the particle distribution as the
time step size is made arbitrarily small.
• If Continuous random walk is selected, a random perturbation is applied to each
particle by integrating a Langevin equation. Unlike Discrete random walk, the
perturbation of the background velocity that is applied to each particle depends on
the time history of all previously applied perturbations.

If Discrete random walk, Discrete random walk, variable time step, or Continuous random
walk is selected, enter a value or expression for the Turbulent kinetic energy k (SI unit:
m2/s2), which determines the magnitude of the turbulent dispersion term. If a physics
interface is present that computes the turbulent kinetic energy, it can be selected
directly from the list.

If Discrete random walk, variable time step or Continuous random walk is selected, enter
a value or expression for the Turbulent dissipation rate ε (SI unit: m2/s3), which is
related to the lifetime of eddy currents in the flow and thus determines how frequently

198 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
the random perturbation term is sampled. If a physics interface is present that
computes the turbulent dissipation rate, it can be selected directly from the list. Then
enter a value or expression for the Lagrangian time scale coefficient CL (dimensionless).
The default value is 0.2.

If Continuous random walk is selected, it is possible to adjust the random contributions

to the fluid velocity to account for anisotropic turbulence, which is often significant
near boundaries. Select the Include anisotropic turbulence in boundary layers check box
to compute different random contributions to the fluid velocity field based on the
streamwise, spanwise, and wall normal directions in the flow, which are computed
using the fluid velocity direction and the direction to the nearest point on an adjacent
wall. When searching for the nearest wall, boundaries using the Symmetry, Inlet, and
Outlet features are ignored.

Particle Motion in a Turbulent Flow in Theory for the Particle Tracing

for Fluid Flow Interface.

ADVANCED SETTINGS
Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

If any option except None is selected from the Turbulent dispersion model list, the Drag
Force feature creates random numbers. If, in addition, the Arguments for random
number generation setting is User defined in the physics interface Advanced Settings
section, enter the Additional input argument to random number generator. The default
value is 1.

Lift Force

The Lift Force node is only available when Newtonian or Newtonian, first
order is selected as the Formulation on the Settings window for The
Particle Tracing for Fluid Flow Interface.

Use the Lift Force node to exert a lift force on particles in a fluid. Lift forces are
applicable when particles move through a fluid in which the velocity field is
nonuniform.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 199
 LIFT FORCE
Select a Lift law: Saffman (the default) or Wall induced.

Saffman and Wall induced lift forces are applicable for particles traveling
through a creeping flow. The Saffman lift force is applicable for particles
far from walls. The Wall induced lift force is a specialized formulation to
account for the effects of nearby walls as particles move through pipes or
channels.

Enter coordinates for the Velocity field u (SI unit: m/s) based on space dimension. If
another physics interface is present which computes the velocity field then this can be
selected from the list.

The Dynamic viscosity μ (SI unit: Pa·s) is taken From material. For User defined enter
another value or expression.

If Saffman is selected from the Lift law list it is only necessary to specify the set of
domains in which the lift force is exerted. If Wall induced is selected, the Parallel
Boundary 1 and Parallel Boundary 2 selections are also shown. These two boundary
selections are typically two parallel surfaces on either side of a parabolic flow profile.

Particle Motion in a Shear Flow in Theory for the Particle Tracing for
Fluid Flow Interface.

Brownian Force

The Brownian Force node is only available when Newtonian or Newtonian,

first order is selected as the Formulation on the Settings window for The
Particle Tracing for Fluid Flow Interface.

Use the Brownian Force node to account for diffusion of suspended particles in a fluid
according to Einstein’s theory. When using this feature, a random force is applied on
each particle at each time step. This often forces the solver to take very small time steps

200 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
because the default relative and absolute tolerances solver settings are quite strict (10−5
and 10−6, respectively).

In order to make the model solve efficiently, it might be necessary to relax

these tolerances. The solver needs to take enough time steps so that the
random force is applied enough times to be meaningful, but not so many
time steps that the model takes a very long time to solve. As a general rule
of thumb, it is usually best to relax both the tolerances to 10−3 and then
perform a convergence study on the tolerances.

BROWNIAN FORCE
Enter a value or expression for the fluid Temperature T (SI unit: K). The default is
293.15 K. If another physics interface that computes the fluid temperature is present
in the model, then this can be selected from the list as a model input.

Enter a value or expression for the Dynamic viscosity μ (SI unit: Pa·s). If a physics
interface is present that computes the dynamic viscosity it can be selected directly from
the list, or the value or expression can come from the selected material.

ADVANCED SETTINGS
Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

The random number generator used in COMSOL Multiphysics uses seeding so that
the same results are obtained every time the model is solved. When adding forces that
are random in nature, this is not always desirable. When the Arguments for random
number generation setting is User defined, you can use the Additional input argument to
random number generator field in conjunction with a Parametric Sweep to perform a
Monte Carlo simulation. For example, use the following steps:

1 From the Home toolbar click Parameters ( ). Or in the Model Builder, right-click
Global Definitions ( ) and select Parameters.
2 Add a parameter with Name, for example, ds and set the Expression to 1.
3 In the Brownian Force feature, set the Additional input argument to random number
generator to ds.
4 Add a Parametric Sweep to the current Study and set the Parameter name to ds.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 201
 5 Set the Parameter values to be an array whose length is equal to the number of times
the model should be solved. For example, to solve the model 5 times, set the
Parameter values to 1 2 3 4 5.

When 2 (the default) is selected from the Wall accuracy order list in the
physics interface Advanced Settings section, the Brownian force may cause
particles to undergo extremely large accelerations after they are released
or when they interact with walls. This is due to the second-order
Runge-Kutta extrapolation using the random force to predict the particle
position at the end of the current time step. To prevent this non-physical
behavior, consider selecting 1 from the Wall accuracy order list when
modeling Brownian motion.

• About Parameters, Variables, and Expressions and Parametric Sweep in

the COMSOL Multiphysics Reference Manual
• Additional Forces in the theory section.

Brownian Motion: Application Library path

Particle_Tracing_Module/Tutorials/brownian_motion

Gravity Force

The Gravity Force node is only available when Newtonian or Newtonian,

first order is selected as the Formulation in the physics interface Particle
Release and Propagation section.

Use the Gravity Force node to define the gravity vector and density. The gravity force
causes particles with a suitably high mass and density to move in the same direction as
the orientation of the gravity vector. For submicron particles, the drag and other
external forces can dominate and gravity can have little effect on the particle
trajectories.

GRAVITY FORCE
Enter coordinates based on space dimension for the Gravity vector g
(SI unit: m/s2).The default magnitude for the gravity vector is g_const, which is a

202 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
built-in physical constant equal to g = 9.80665 m/s2 corresponding to the standard
acceleration due to gravity on Earth.

The default direction for the gravity vector is in the y-direction for 2D
components.

The default direction for the gravity vector is in the z-direction for 3D and
2D axisymmetric components.

The default Density ρ (SI unit: kg/m3) is taken From material, from a physics interface
that defines or computes the density, or select User defined to enter another value or
expression.

AFFECTED PARTICLES
Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

Particle Motion in a Fluid in the theory section.

Acoustophoretic Force

The Acoustophoretic Force node is only available when Newtonian or

Newtonian, first order is selected as the Formulation on the Settings window
for The Particle Tracing for Fluid Flow Interface.

Use the Acoustophoretic Force node to add forces on small particles due to acoustic
radiation. This force only adds equation contributions if the acoustic pressure was
computed from a Frequency Domain study type. This is because the magnitude of the
acoustophoretic force depends on the angular frequency.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 203
 ACOUSTIC PROPERTIES
Enter a value or expression for the following:

• Pressure p (SI unit: Pa). This should be computed from another physics interface
using a Frequency Domain study type.
• Speed of sound c (SI unit: m/s) of the surrounding fluid. This can either be taken
From material or User defined, in which case the default is 343 m/s.
• Density ρ (SI unit: kg/m3) of the surrounding fluid. This can either be taken From
material or User defined, in which case the default is 1 kg/m3.

ACOUSTOPHORETIC FORCE
Select an option from the Specify list: Particle bulk modulus (the default), Particle
compressibility, or Compressional speed of sound. Then, if required change the defaults
and enter a value or expression the available fields as follows:

• Particle bulk modulus Kp (SI unit: Pa). The default is 2.2 GPa (2.2 × 109).
• Particle compressibility βp (SI unit: 1/Pa). The default is 1/(2.2 × 109) 1/Pa.
• Compressional speed of sound cp (SI unit: m/s). The default is 3000 m/s.

ADVANCED SETTINGS
Select the Use piecewise polynomial recovery on field check box to smooth the radiation
pressure using piecewise polynomial recovery. This can give a much more accurate
representation of the radiation pressure because it uses information on adjacent mesh
elements to reconstruct the field. If a coarse mesh is used to compute the field then
this option can be especially useful.

• Studies and Solvers and Frequency Domain in the COMSOL

Multiphysics Reference Manual
• Additional Forces in the theory section.

Electric Force

The Electric Force node is only available when Newtonian or Newtonian,

first order is selected as the Formulation on the Settings window for The
Particle Tracing for Fluid Flow Interface.

204 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Use the Electric Force node to exert an electric force to the particles. The force is
specified via an electric potential or the electric field. For cases where the field was
computed in the frequency domain, the force can be computed by multiplying the field
by the phase angle. Additionally, piecewise polynomial recovery can be used which can
give a more accurate representation of the specified electric field.

The settings for this feature node are the same as for the Electric Force node described
for The Charged Particle Tracing Interface.

Additional Forces in the theory section.

Magnetic Force

The Magnetic Force node is only available when Newtonian or Newtonian,

first order is selected as the Formulation on the Settings window for The
Particle Tracing for Fluid Flow Interface.

Use the Magnetic Force node to exert a magnetic force to the particles. A magnetic force
alone does no work on the particles, so in the absence of any other external forces, the
particle retains its original energy. The force is specified via a magnetic flux density.

The settings for this feature node are the same as for the Magnetic Force node
described for The Charged Particle Tracing Interface.

Additional Forces in the theory section.

Dielectrophoretic Force

The Dielectrophoretic Force node is only available when Newtonian or

Newtonian, first order is selected as the Formulation on the Settings window
for The Particle Tracing for Fluid Flow Interface.

Use the Dielectrophoretic Force node to exert a dielectrophoretic force to the particles.
The force is specified via an electric potential or the electric field. The force has a

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 205
 different meaning depending on whether the source electric field is computed in a
Stationary or Frequency Domain study.

Additionally, piecewise polynomial recovery can be used which can give a more
accurate representation of the specified electric field. The influence of the
dielectrophoretic force on the particles depends on the difference in permittivity
between the particles and the fluid. When the particle relative permittivity is greater
than the relative permittivity of the fluid, the particles are attracted to regions where
the absolute electric field is strong. When the particle relative permittivity is less than
the relative permittivity of the fluid, the particles are attracted to regions where the
absolute electric field is weak.

The Shell subnode is available from the context menu (right-click the parent node) or
from the Physics toolbar, Attributes menu. Use it to model dielecrophoresis of particles
with thin dielectric shells. If no Shell subnodes are added, the particle is treated as a
homogeneous sphere of uniform permittivity and conductivity for the purpose of
computing the dielectrophoretic force.

DIELECTROPHORETIC FORCE
Select an option from the Specify force using list: Electric field (the default) or Electric
potential.

• For Electric field enter values or expressions in the table for the Electric field E (SI
unit: V/m) based on space dimension. If the electric field is computed by another
physics interface then it can be selected from the list.
• For Electric potential enter a value or expression for Electric potential V (SI unit: V).
If the electric potential is computed by another physics interface then it can be
selected from the list.

PARTICLE PROPERTIES
Enter a value or expression for the Particle relative permittivity εr,p (dimensionless).
The default value is 1. The value must be greater than or equal to one. Then enter a
value or expression for the Particle electrical conductivity σp (SI unit: S/m). The
default value is 0.

FLUID PROPERTIES
By default the Relative permittivity εr (dimensionless) is taken From material. Select
User defined from the list to enter a value or expression.

By default the Electrical conductivity σ (SI unit: S/m) is taken From material. Select
User defined from the list to enter a value or expression.

206 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
ADVANCED SETTINGS
Select the Use piecewise polynomial recovery on field check box to smooth the electric
field using piecewise polynomial recovery. This can give a much more accurate
representation of the electric field as it uses information on adjacent mesh elements to
reconstruct the field. If a coarse mesh is used to compute the field then this option can
be especially useful.

Select an option from the Angular frequency list—From solution or User defined. For
From solution the dielectrophoretic force is computed using the angular frequency of
the electric field computed in a previous study step. For User defined enter the Angular
frequency ω (SI unit: rad/s). The default value is 0.

Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

• Studies and Solvers in the COMSOL Multiphysics Reference Manual

• Additional Forces in the theory section.

Shell
The Shell subnode is available from the context menu (right-click the
Dielectrophoretic Force parent node) or from the Physics toolbar, Attributes menu.
Add Shell subnodes to a Dielectrophoretic Force node to compute an equivalent
Clausius-Mossotti factor using the electrical properties of the interior of the particle
and one or more thin layers on its surface.

If multiple Shell subnodes are present, their order in the Model Builder determines
how the nested shells are modeled, with the first subnode constituting the innermost
layer.

SHELL PROPERTIES
Enter the Shell thickness ts (SI unit: m). The default is 1 × 10-7 m. Enter a value for
the Shell relative permittivity εr,s (dimensionless). The default is 1. Enter a Shell
electrical conductivity σs (SI unit: S/m). The default is 1 S/m.

Dielectrophoretic Force in the theory section.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 207
 Magnetophoretic Force

The Magnetophoretic Force node is only available when Newtonian or

Newtonian, first order is selected as the Formulation on the Settings window
for The Particle Tracing for Fluid Flow Interface.

Use the Magnetophoretic Force node to include forces on particles due to a difference
in permeability between the particles and background fluid. The force is specified via
a magnetic field and the permeability of the particle and the fluid.

Additionally, piecewise polynomial recovery can be used, which can give a more
accurate representation of the specified magnetic field.

MAGNETOPHORETIC FORCE
Enter a value or expression for the Magnetic field H (SI unit: A/m) based on space
dimension, or if the field is computed by another physics interface then it can be
selected from the list. Unlike the Dielectrophoretic Force the Magnetophoretic Force
always treats the model particles as homogeneous spheres.

The magnetophoretic force is a function of the spatial derivatives of the

magnetic field components. For formulations based on the magnetic
scalar potential A, these derivatives of H may not be available. See
Additional Forces in the theory section for more details.

PARTICLE PROPERTIES
Enter a value or expression for the Particle relative permeability μr,p (dimensionless).
The value must be greater than or equal to one.

FLUID PROPERTIES
Enter a value or expression for the Fluid relative permeability μr,f (dimensionless). The
value must be greater than or equal to one.

ADVANCED SETTINGS
Select the Use piecewise polynomial recovery on field check box to smooth the magnetic
field using piecewise polynomial recovery. This can give a much more accurate
representation of the magnetic field as it uses information on adjacent mesh elements
to reconstruct the field. If a coarse mesh is used to compute the field then this option
can be especially useful.

208 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

Additional Forces in the theory section.

Thermophoretic Force

The Thermophoretic Force node is only available when Newtonian or

Newtonian, first order is selected as the Formulation on the Settings window
for The Particle Tracing for Fluid Flow Interface.

Use the Thermophoretic Force node to include forces on particles due to a temperature
gradient in the background fluid. The force is specified via the background
temperature and other properties of the fluid.

Additionally, piecewise polynomial recovery can be used which can give a more
accurate representation of the specified temperature.

THERMOPHORETIC FORCE
If the Include rarefaction effects check box is selected in the physics interface Particle
Release and Propagation section, select a Thermophoretic force model—Epstein,
Waldmann, Talbot (the default), or Linearized BGK. Use these settings to improve the
accuracy of the thermophoretic force calculation in fluids with high Knudsen numbers.
When the Include rarefaction effects check box is cleared, the fluid is treated as a
continuum and only the Epstein model is used.

Enter a value or expression for the Temperature T (SI unit: K), or if the temperature is
computed by another physics it can be selected from the list. The default is 293.15 K.

The following fluid parameters all take default values From material or can be
computed from another physics interface. Or select User defined to enter a value or
expression for the:

• Dynamic viscosity μ (SI unit: Pa·s). The default is 2E-5 Pa·s.

• Density ρ (SI unit: kg/m3). The default is 1 kg/m3.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 209
 • Thermal Conductivity k (SI unit: W/(m·K)). For User defined select Isotropic,
Diagonal, Symmetric, or Anisotropic and then enter values in the field or matrix.
• Heat capacity at constant pressure Cp (SI unit: J/(kg·K)). The default is 1000.

If the Waldmann, Talbot, or Linearized BGK model is used, enter the Background gas
molar mass Mg (SI unit: kg/mol). The default is 0.04[kg/mol].

If the Waldmann, Talbot, or Linearized BGK model is used, select an option from the
Mean free path calculation list: Ideal gas, hard sphere collisions or User-defined correction.
For User-defined correction enter a Mean free path correction λ’/λ by which the mean
free path differs from the value for hard spheres in an ideal gas.

If the Linearized BGK model is used, enter an Energy accommodation coefficient αE

(dimensionless) and a Momentum accommodation coefficient αM (dimensionless). The
default values are 1.

PARTICLE PROPERTIES
Enter a value or expression for the:

• Particle thermal conductivity kp (SI unit: W/(m·K)). The default is 1 W/(m·K).

• Thermophoretic correction factor Cs (dimensionless). The default is 1.17.
• If the Talbot model is used, enter two additional dimensionless thermophoretic
correction factors Cm and Ct. The default values are 1.146 and 2.2, respectively.

ADVANCED SETTINGS
Select the Use piecewise polynomial recovery on field check box to smooth the
temperature using piecewise polynomial recovery. This can give a much more accurate
representation of the temperature as it uses information on adjacent mesh elements to
reconstruct the field. If a coarse mesh is used to compute the field then this option can
be especially useful.

Use the Particles to affect list to apply the force to specific particles. The available
settings are the same as for the Force node.

Additional Forces in the theory section.

210 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Erosion
The Erosion subnode is available from the context menu (right-click the Wall parent
node) or from the Physics toolbar, Attributes menu. Use it to calculate the rate of
erosive wear or the total mass removed due to the impact of particles on a boundary.

• When Specify release times is selected from the Particle release specification list in the
physics interface Particle Release and Propagation section, the Erosion node calculates
the total mass lost per unit area (SI unit: kg/m2).
• If Specify mass flow rate is selected from the Particle release specification list, the
Erosion node calculates the rate of erosive wear (SI unit: kg/(m2s)).

EROSION MODEL
Select an Erosion model: Expression, Finnie (default), E/CRC, Oka, or DNV. Then enter the
following settings:

Expression
• If Specify release times is selected from the Particle release specification list in the
physics interface Particle Release and Propagation section, enter the Mass removed by
particle Δmi (SI unit: kg). The default is 1e-15 kg.
• If Specify release times is selected from the Particle release specification list, enter the
Erosive wear of particle EM,i (SI unit: kg/s). The default is 1e-15 kg/s.

Finnie
• Fraction of particles cutting in an idealized manner ci (dimensionless). The default is
0.1.

• Ratio of normal and tangential forces K (dimensionless). The default is 2.

• Surface hardness HV (SI unit: N/m). The default is 2e9 Pa.
• Mass density of surface ρ (SI unit: kg/m3). The default is 7.5e3[kg/m^3].

E/CRC
• E/CRC model coefficient C (dimensionless).
• Brinell hardness of surface material BH (dimensionless). The default is 200.
• Particle shape coefficient Fs (dimensionless). The default is 0.2.
• E/CRC model exponent n (dimensionless). The default is 2.41.

Oka
• Oka model coefficient K (SI unit: m3/kg). The default is 65[mm^3/kg].
• Reference diameter dref (SI unit: m). The default is 326[um].

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 211
 • Reference velocity vref (SI unit: m/s). The default is 104[m/s].
• Each of the following dimensionless Oka model coefficient:
- k1 (default is -0.12) and k2 (default is 0.19).
- q1 default is 0.14) and q2 (default is -0.94).
- s1 (default is 0.71) and s2 (default is 2.4).
• Surface hardness HV (SI unit: N/m2). The default is 2e9[N/m^2].
• Mass density of surface ρ (SI unit: kg/m3). The default is 7.5e3[kg/m^3].

DNV
• DNV model coefficient K (dimensionless). The default is 2e-9.
• DNV mode exponent n (dimensionless). The default is 2.6.

If Specify release times is selected from the Particle release specification list in the physics
interface Particle Release and Propagation section, it is possible to multiply the eroded
mass by another scale factor. Select an option from the Number multiplication factor
specification list—From physics (the default) or User defined.

For From physics the extra scale factor is usually equal to 1. However, if the Enable
macroparticles check box is selected in the physics interface Additional Variables section,
the scale factor is instead taken from the built-in auxiliary dependent variable for the
particle number multiplication factor, which can be initialized in the Settings windows
for most particle release features.

For User defined, specify the extra scale factor directly. Enter a value or expression for
the Number multiplication factor nn (dimensionless). The default is 1.

Pipe Erosion due to Contaminant Particles: Application Library path

Particle_Tracing_Module/Fluid_Flow/pipe_elbow_erosion

Mass Deposition
If Specify release times is selected from the Particle release specification list in the physics
interface Particle Release and Propagation section, the Mass Deposition subnode is
available from the context menu (right-click the Wall parent node) or from the Physics
toolbar, Attributes menu. The Mass Deposition subnode uses an accumulated variable
on each boundary element in the selection of the parent node to calculate the mass of
particles deposited on the surface, expressed in mass per unit area.

212 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Boundary Load
If Specify mass flow rate is selected from the Particle release specification list in the
physics interface Particle Release and Propagation section, the Boundary Load subnode
is available from the context menu (right-click the Wall parent node) or from the
Physics toolbar, Attributes menu. The Boundary Load subnode uses accumulated
variables on each boundary element in the selection of the parent node to calculate the
force due to the impact of particles.

BOUNDARY LOAD
Select a Type—Pressure or Force per unit area. For Force per unit area accumulated
variables are defined for components of the boundary load vector. For Pressure one
accumulated variable is defined for the normal force per unit area in each boundary
element.

Mass Flux
If Specify mass flow rate is selected from the Particle release specification list in the
physics interface Particle Release and Propagation section, the Mass Flux subnode is
available from the context menu (right-click the Wall parent node) or from the Physics
toolbar, Attributes menu. The Mass Flux subnode uses accumulated variables on each
boundary element in the selection of the parent node to calculate the mass flux of
particles at a boundary.

MASS FLUX
Select a Type—Mass flux or Normal mass flux. For Mass flux accumulated variables are
defined for components of the mass flux vector. For Normal mass flux one accumulated
variable is defined for the normal mass flux in each boundary element.

Volume Force Calculation

Use the Volume Force Calculation node if the particles exert a significant force on the
surrounding fluid. This node defines dependent variables for the components of the
volume force; the net volume force is equal in magnitude and opposite in direction to
the total

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 213
 on the particles.

To compute the volume force exerted by the particles on the fluid, at least
one Drag Force node must be present.

The Volume Force Calculation node computes the volume force in the same
way as the Fluid-Particle Interaction Multiphysics node but does not
automatically include the accumulated variables as source terms when
computing the fluid velocity and pressure. For modeling bidirectionally
coupled fluid-particle interactions, consider using The Fluid-Particle
Interaction Interface instead of the Volume Force Calculation node.

For more information about the calculation of the volume force, see
Volume Force Calculation in the section Theory for the Fluid-Particle
Interaction Interface.

VO L U M E F O R C E C A L C U L A T I O N
This section is available if Specify release times is selected from the Particle release
specification list in the physics interface Particle Release and Propagation section. Select
an option from the Force multiplication factor specification list—From physics (the
default) or User defined.

• If From physics is selected and the Enable macroparticles check box is cleared in the
physics interface Additional Variables section, the multiplication factor is 1.
• If From physics is selected and the Enable macroparticles check box is selected in the
physics interface Additional Variables section, the multiplication factor is based on the
auxiliary dependent variable for the multiplication factor, which is typically specified
in release feature settings. If the physics interface has name fpt, this variable has
name fpt.nn.
• If User defined is selected, enter a value or expression for the Force multiplication
factor n (dimensionless). The default is 1.

214 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
The Force multiplication factor can be used to represent each particle as a group of n
particles that follow the same trajectory. This means that the magnitude of the
contribution to the volume force by each particle is multiplied by n.

If Specify mass flow rate is selected from the Particle release specification
list in the physics interface Particle Release and Propagation section, the
Force multiplication factor n cannot be specified because the number of
particles represented by each model particle is instead controlled by the
Mass flow rate, which is specified in the settings for release features such
as theRelease and Inlet nodes.

Convective Heat Losses

The Convective Heat Losses node is available when the Compute particle temperature
check box is selected in the physics interface Additional Variables section. Use the
Convective Heat Losses node to model the heating or cooling of a particle due to
convective heat exchange with the surrounding fluid. The temperature within the
particle is assumed to be uniform; that is, heat transfer by conduction within the
particle takes place on a much shorter time scale than heat transfer by convection at
the surface.

CONVECTIVE HEAT LOSSES

Enter a Heat transfer coefficient h (SI unit: W/(m2·K)). The default is 0. Enter a value
or expression for the Temperature, T (SI unit: K), or if the temperature is computed by
another physics it can be selected from the list. The default is 293.15 K.

Computing Particle Temperature in the theory section.

Radiative Heat Losses

The Radiative Heat Losses node is available when the Compute particle temperature
check box is selected in the physics interface Additional Variables section. Use the
Radiative Heat Losses node to model the heating or cooling of a particle due to radiative
heat exchange with its surroundings according to the Stefan-Boltzmann law for diffuse
gray surfaces.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 215
 RADIATIVE HEAT LOSSES
Enter a Particle emissivity εp (dimensionless). The default is 1. Enter a value or
expression for the Temperature, T (SI unit: K), or if the temperature is computed by
another physics it can be selected from the list. The default is 293.15 K.

Computing Particle Temperature in the theory section.

Heat Source
The Heat Source node is available when the Compute particle temperature check box is
selected in the physics interface Additional Variables section. Use the Heat Source node
to apply a source term that affects the particle temperature.

HEAT SOURCE
Enter a Heat source, Q (SI unit: W). The default is 0.

Computing Particle Temperature in the theory section.

Droplet Breakup
Use the Droplet Breakup node to model the breakup of liquid droplets or parcels of
droplets.

To use the Droplet Breakup feature, the following prerequisites must be met:

• Newtonian or Newtonian, first order must be selected from the Formulation list in the
physics interface Particle Release and Propagation section.
• The Compute particle mass check box must be selected in the physics interface
Additional Variables section.
• Liquid droplets/bubbles must be selected from the Particle type list in the settings
window for either the Particle Properties node or at least one instance of the
Override Properties feature.
• At least one instance of the Drag Force feature is needed. The valid selection for the
Droplet Breakup feature comprises all domains where drag forces are defined.

216 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
The Kelvin-Helmholtz Breakup Model and Rayleigh-Taylor Breakup Model subnodes
are available from the context menu (right-click the Droplet Breakup parent node) or
from the Physics toolbar, Attributes menu. Only a single instance of each feature can be
created. The Kelvin-Helmholtz Breakup Model is created by default.

To model droplet breakup, unreleased secondary particles must be

available. To ensure that all child droplets are released properly, a
sufficiently large value must be specified for the Maximum number of
secondary particles in the physics interface Particle Release and Propagation
section.

Kelvin-Helmholtz Breakup Model

Use the Kelvin-Helmholtz breakup model to account for the effect of Kelvin-Helmholtz
instability on the breakup of liquid droplets. This causes small child droplets to be
stripped away from larger droplets over time.

BREAKUP MODEL
Enter a value or expression for the Kelvin-Helmholtz model coefficient BKH
(dimensionless). The default value is 10.

If the Enable macroparticles check box is selected in the physics interface Additional
Variables section, enter a value or expression for the Maximum number of particles
released per breakup Nmax (dimensionless). The default is 10. This prevents an
inordinate amount of child droplets from being released. If more child droplets than
the maximum specified number would be released, then the multiplication factor of
each particle is adjusted to compensate for the limited number of model particles.

GAS PROPERTIES
The Density ρ (SI unit: kg/m3) is taken From material. For User defined enter another
value or expression. The default is 1[kg/m^3].

Enter the components of the Velocity field u (SI unit: m/s) based on space dimension.
If another physics interface is present which computes the velocity field then this can
be selected from the list.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 217
 Rayleigh-Taylor Breakup Model
Use the Rayleigh-Taylor breakup model to account for the effect of Rayleigh-Taylor
instability on the breakup of liquid droplets. Rayleigh-Taylor breakup causes the parent
droplet to be completely broken up into smaller child droplets, and is usually
associated with droplets undergoing large accelerations.

BREAKUP MODEL
Specify the Rayleigh-Taylor coefficient CRT (dimensionless). The default value is 5.

If the Enable macroparticles check box is selected in the physics interface Additional
Variables section, enter a value or expression for the Maximum number of particles
released per breakup Nmax (dimensionless). The default is 10. This prevents an
inordinate amount of child droplets from being released. If more child droplets than
the maximum specified number would be released, then the multiplication factor of
each particle is adjusted to compensate for the limited number of model particles.

Droplet Breakup Theory

Nozzle
Use the Nozzle feature to release a spray of liquid droplets.

To use the Nozzle feature, the following prerequisites must be met:

• Newtonian or Newtonian, first order must be selected from the Formulation list in the
physics interface Particle Release and Propagation section.
• Specify release times must be selected from the Particle release specification list in the
physics interface Particle Release and Propagation section.
• The Compute particle mass and Enable macroparticles check boxes must be selected
in the physics interface Additional Variables section.
• Liquid droplets/bubbles must be selected from the Particle type list in the settings
window for either the Particle Properties node or at least one instance of the
Override Properties feature.

The Nozzle feature is supported on two geometric entity levels. If it is added as a global
feature (similar to the Release from Grid feature), the initial position of the injected
droplets can be specified by entering a list of coordinates. If it is added as a 0D feature

218 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
(similar to the Release from Point feature), the initial positions of the injected droplets
correspond to the selected points in the geometry.

Go to Release for information about the following sections: Inherit Properties, Initial
Value of Auxiliary Dependent Variables, and Advanced Settings.

RELEASE TIMES
Enter values for the Release times (SI unit: s). The default expression is 0 1[ms]. At
least two release times must be specified, and the release times must increase
monotonically. This is because the time intervals between successive release times are
used to assign appropriate number multiplication factors to the released particles so
that the mass flow rate is constant in a finite time interval.

I N I T I A L VE L O C I T Y
Select an option from the Velocity specification list—From mass flow rate (the default),
From change in pressure, From injection pressure, or User defined.
·
For From mass flow rate enter a value or expression for the Mass flow rate m
(SI unit: kg/s). The default is 1e-3[kg/s].

For From change in pressure enter a value or expression for the Change in pressure Δp
(SI unit: Pa). The default value is 1[MPa]. Then enter a value or expression for the
Discharge coefficient cd (dimensionless). The default value is 0.7.

For From injection pressure enter a value or expression for the Injection pressure pinj
(SI unit: Pa). The default value is 1[MPa]. Then enter a value or expression for the
Discharge coefficient cd (dimensionless). The default value is 0.7.

For User defined enter a value or expression for the Initial droplet velocity magnitude U0
(SI unit: m/s). The default value is 100 m/s.

INJECTION PARAMETERS
Specify the initial droplet positions based on space dimension (qx,0, qy,0, and qz,0 for
3D components) or click the Range button ( ) to select and define a range of specific
coordinates.

Select an option from the Injector Location list: All combinations (the default) or
Specified combinations. If Specified combinations is selected, the number of initial
coordinates entered for each space dimension must be equal, and the total number of
release positions is equal to the length of one of the lists of initial coordinates. If All
combinations is selected, the total number of release positions is equal to the product
of the lengths of each list of initial coordinates.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 219
 For example, suppose a 2D component includes a Nozzle node with the following
initial coordinates:

• qx,0 = range(0,1,3)
• qy,0 = range(2,2,8)

If All combinations is selected, a total of 16 initial positions will be released, including

every possible combination of the initial x- and y-coordinates. If Specified combinations
is selected, particles will only be released at 4 initial positions (0,2), (1,4), (2,6), and
(3,8).

The position for any particles with initial coordinates outside the geometry are set to
NaN, so the particles do not appear when plotted during postprocessing.

If the Nozzle feature is added as a 0D feature (similar to the Release from Point feature),
the Injector location list and the edit fields for the initial coordinates are not shown, and
the initial coordinates are instead specified by selecting points in the geometry.

Enter coordinates for the Injector axis r (dimensionless) based on space dimension. By
default the injector axis points in the positive x-direction.

Enter a value or expression for the Nozzle exit radius r0 (SI unit: m), which determines
the initial size of released droplets.

Enter a value or expression for the Number of parcels per release N0 (dimensionless).

The Number of parcels per release does not actually restrict the number of
droplets or the total mass of fluid that can be released by the Nozzle
feature; it only restricts the number of model particles that can be
released. Each model particle represents a parcel representing an arbitrary
number of droplets. The number of droplets per model particle is always
available as an auxiliary dependent variable because the Enable
macroparticles check box must be selected in the physics interface
Additional Variables section in order to use this feature.

MAXIMUM CONE ANGLE

Select an option from the Maximum cone angle specification list—User defined (the
default) or From Kelvin-Helmholtz instability model.

If User defined is selected, enter a value or expression for the Maximum cone angle Θ
(SI unit: rad). The default value is 10[deg].

220 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
If From Kelvin-Helmholtz instability model is selected, specify the Spray angle constant A
(dimensionless). The default value is 0.188.

Nozzle Theory

Nozzle Domain
The Nozzle Domain feature is automatically added as a subnode to the Nozzle feature. It
can never be added or removed. Use the Nozzle Domain feature to specify which fluid
domain the droplets are being injected into, and to specify material properties in this
domain.

The Density ρ (SI unit: kg/m3) is taken From material. For User defined enter another
value or expression.

Enter a value or expression for the Pressure p (SI unit: Pa). If another physics interface
is present which computes the pressure then this can be selected from the list.

To use the Nozzle release feature, at least one domain must be in the
selection of the Nozzle Domain subnode.

T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 221
 T he Dr o ple t S p rays i n Fl u i d Fl ow
Interface
The Droplet Sprays in Fluid Flow interface ( ), found under the Fluid Flow>Particle
Tracing branch ( ) when adding a physics interface, creates an instance of the Particle
Tracing for Fluid Flow interface with specialized default settings and features that are
often used to model sprays of liquid droplets.

All available physics features and settings are the same as for The Particle Tracing for
Fluid Flow Interface, with the exception that the following defaults are used:

• The Compute particle mass check box is selected in the physics interface Additional
Variables section,
• Liquid droplets/bubbles is selected from the Particle type list in the settings window
for the default Particle Properties node,
• The Drag Force node is added to all domains by default, and
• The Droplet Breakup node is added to all domains by default. The
Kelvin-Helmholtz Breakup Model node is added as a default subnode to the Droplet
Breakup node.

222 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Theory for the Particle Tracing for
Fluid Flow Interface
The Particle Tracing for Fluid Flow Interface theory is described in this section:

• Particle Motion in a Fluid

• Particle Motion in a Turbulent Flow
• Particle Motion in a Shear Flow
• Additional Forces
• References for the Particle Tracing for Fluid Flow Interface

Particle Motion in a Fluid

The particle momentum comes from Newton’s second law, which states that the net
force on a particle is equal to its time rate of change of its linear momentum in an
inertial reference frame:

d m
( v ) = F D + F g + F ext
dt p

When the particle mass is being computed, such that accretion or evaporation can take
place, the mass is moved outside the time derivative to prevent non-physical
acceleration of the particles:

dv
mp = F D + F g + F ext
dt

The assumption is that any mass lost by the particles continues to move with the
particle velocity and doesn’t cause the particle to decelerate.

DRAG FORCE
In the above equation, FD is the Drag Force is defined as:

F D =  ----- m p ( u – v )
1
τp

where

• mp is the particle mass (SI unit: kg)

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 223
 • τp is the particle velocity response time (SI unit: s)
• v is the velocity of the particle (SI unit: m/s)
• u is the fluid velocity (SI unit: m/s)
• Fg is the gravitational force vector (SI unit: N), and
• Fext is any other external force (SI unit: N).

A large number of expressions for the particle response time are available; selecting an
appropriate drag law requires knowledge of the relative Reynolds number Rer
(dimensionless) of particles in the flow. The relative Reynolds number is given by the
expression

ρ u – v dp
Re r = -----------------------------
μ

Many drag laws and the relative Reynolds numbers at which they are applicable are
given in Ref. 4.

Stokes
The particle velocity response time for spherical particles in a laminar flow is defined as:

2
ρp dp
τ p = -------------
18μ

where μ is the fluid viscosity (SI unit: Pa s), ρp is the particle density (SI unit: kg/m3),
and dp is the particle diameter (SI unit: m). This is frequently known as Stokes drag
law. The Stokes drag law is applicable for creeping flows, Re r « 1 .

Schiller-Naumann
When the drag law is set to Schiller-Naumann, the particle velocity response time is
redefined as:

2
4ρ p d p
τ p = -------------------------
3μC D Re r

where:

24 0.687
C D = --------- ( 1 + 0.15Re r )
Re r

The Schiller-Naumann drag law is applicable at moderate relative Reynolds numbers,

Re r < 800 .

224 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Haider-Levenspiel
When the drag law is set to Haider-Levenspiel, the drag coefficient is given by

B ( Sp ) C ( Sp )
C D = ---------  1 + A ( S p )Re r  + ---------------------------------------
24 -
Re r   1 + D ( S p ) ⁄ Re r

where A, B, C and D are empirical correlations of the particle sphericity. The sphericity
is defined as the ratio of the surface area of a volume equivalent sphere to the surface
area of the considered non-spherical particle

A sphere
0 < S p = ------------------ ≤ 1
A particle

The correlation coefficients are given by

A ( S p ) = exp ( 2.3288-6.4581S p + 2.4486S p2 )

B ( S p ) = 0.0964+0.5565S p
C ( S p ) = exp ( 4.905-13.8944S p + 18.4222S p2 – 10.2599S p3 )
D ( S p ) = exp ( 1.4681 + 12.2584S p – 20.7322S p2 + 15.8855S p3 )

The diameter used in the Reynolds number is that of the volume equivalent sphere.

Oseen Correction
When the drag law is set to Oseen correction, the drag coefficient is given by

C D = ---------  1 + ------ Re r
24 3
Re r  16 

Like Stokes drag law, the Oseen correction is applicable at low relative Reynolds
numbers, typically Rer < 0.1. The Oseen correction is determined by simplifying, rather
than neglecting, the inertia term in the Navier-Stokes equation, resulting in drag
coefficients that are slightly greater than those based on the Stokes drag law.

Hadamard-Rybczynski
The Hadamard-Rybczynski drag law is based on an analytical solution for creeping flow
past a spherical liquid drop or gas bubble. The interface between the two fluids is
assumed to be completely free of surface-active contaminants. The drag coefficient is
given by the expression

C D = ---------  ----------------
8 2 + 3κ
Re r  1 + κ 

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 225
 where κ = μp/μ is the ratio of the particle viscosity and fluid viscosity. For arbitrarily
large values of μp this expression asymptotically approaches the Stokes law.

The Hadamard-Rybczynski drag law is only applicable if the droplets and the
surrounding fluid are extremely pure and free from contaminants. If the fluids are not
sufficiently pure, the drag force on the bubble or droplet is more likely to be accurately
determined by the Stokes drag law.

When the drag law is set to Standard drag correlations, the drag coefficient is defined
as a piecewise function of the relative Reynolds number:

RANGE CORRELATION

Re r ≤ 0.01
C D = ---------  1 + ------ Re r
24 3
Re r  16 

0.01 < Re r ≤ 20 24 0.82 – 0.05w

C D = --------- ( 1 + 0.1315Re r )
Re r

20 < Re r ≤ 260 24 0.6305

C D = --------- ( 1 + 0.1935Re r )
Re r

260 < Re r ≤ 1500 log CD = 1.6435 − 1.1242w + 0.1558w2

1500 < Re r ≤ 1.2 × 10

4 log CD=−2.4571+2.5558w−0.9295w2+0.1049w3

4
1.2 × 10 < Re r ≤ 4.4 × 10
4 log CD = −1.9181 + 0.6370w − 0.0636w2

4
4.4 × 10 < Re r ≤ 3.38 × 10
5 log CD = −4.3390 + 1.5809w − 0.1546w2

5 5 CD = 29.78 − 5.3w
3.38 × 10 < Re r ≤ 4 × 10

5 6 CD = 0.1w − 0.49
4 × 10 < Re r ≤ 1 × 10

6
1 × 10 < Re r 8 × 10 4
C D = 0,19 – -------------------
Re r

where w = log Rer and the base 10 logarithm has been used.

The Standard drag correlations can be used when the relative Reynolds number is
expected to change by several orders of magnitude during a simulation. It is also
applicable at significantly higher relative Reynolds numbers than the Schiller-Naumann
drag law.

226 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
The drag coefficient as a function of the relative Reynolds number is shown in
Figure 5-1.

Figure 5-1: The piecewise function used to compute the drag coefficient when Standard
Drag Correlations are used.

DRAG FORCE CORRECTION FACTORS

When the Include rarefaction effects check box is selected in the physics interface
Particle Release and Propagation section, a correction factor is applied to the drag force
to account for the effects of a nonzero particle Knudsen number:

C D, C
C D = -------------
S

Where CD (dimensionless) is the drag coefficient, CD,C (dimensionless) is the drag

coefficient under conditions of continuum flow (the limit as the particle Knudsen
number Kn approaches zero), and S (dimensionless) is the correction factor that
accounts for non-continuum flow.

For the following definitions of the drag force correction factor, the particle Knudsen
number is defined as:

λ
Kn = ------
dp

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 227
 where λ is the mean free path of gas molecules:

π
λ = ----------μ
2pρ

where

• p is the gas pressure (SI unit: Pa),

• ρ is the gas density (SI unit: kg/m3),
• dp is the particle diameter (SI unit: m), and
• μ is the gas viscosity (SI unit: Pa s).

It is also possible to multiply this definition of the mean free path by a user-defined
coefficient.

Use extra caution when interpreting expressions in terms of the Knudsen

number from the literature, since other characteristic lengths may be used
instead of the particle diameter; for example, the particle radius is
sometimes used.

If the Basset model is used, the drag force correction factor is defined as:

1
S = ----------------------------------------
Kn
1 – ( 2 – σ R ) --------
σR

where σR (dimensionless) is the accommodation coefficient, the fraction of gas

molecules that are assumed to reflect diffusely from the particle surface. The Basset
model is appropriate for near-continuum flows.

If the Epstein model is used, the correction factor is defined as:

18Kn
S = --------------------
8 + πσ R

The Epstein model is appropriate for free molecular flows.

If the Phillips model is used, the correction factor is defined as:

15 + 12c 1 Kn + 9 ( c 12 + 1 )Kn 2 + 18c 2 ( c 12 + 2 )Kn 3

S = -------------------------------------------------------------------------------------------------------------------------------
-
15 – 3c 1 Kn + c 2 ( 8 + πσ R ) ( c 12 + 2 )Kn 2

228 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
where c1 and c2 (dimensionless) are functions of the accommodation coefficient:

2 – σR
c 1 = ----------------
σR

1
c 2 = ----------------
2 – σR

The Phillips model has the same asymptotic behavior as the Basset and Epstein models,
and is also usable at intermediate values of the particle Knudsen number. This
asymptotic behavior is shown in Figure 5-2.

Figure 5-2: Comparison of the Basset, Epstein, and Phillips models for drag coefficient
corrections in rarefied gases.

The Cunningham-Millikan-Davies model uses three dimensionless, user-defined

coefficients to define the correction factor:

C3
S = 1 + Kn  C 1 + C 2 exp  – -------- 
  Kn 

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 229
 WA LL COR RECTIONS
When the Include wall corrections check box is selected, the drag force is given by Ref. 5
as

F WC = MF D

 
M =  1 + -------------------------------------------------------------------------------- ( I – P ( n ) ) + ---------------------------------- P ( n ) (5-1)
1 1
 9 1 45 1  9 1
 1 – ------ α + --- α – ---------- α – ------ α 
3 4 5 1 – --- α + --- α 3
16 8 256 16 8 2

rp
α = ------
lW

where

• I is the identity matrix

• P ( n ) is the projection operator onto the wall normal, n
• rp is the radius of the particle (SI unit: m)
• lW is the distance from the center of the particle to the nearest wall (SI unit: m)

The first term in Equation 5-1 applies to the component of the relative particle velocity
parallel to the nearest wall; the second term applies to the component of the relative
velocity normal to this wall. The correction factors become more prominent as the wall
distance becomes comparable to the particle radius.

GRAVITY FORCE
The Gravity Force is given by:

( ρp – ρ )
F g = m p g --------------------
ρp

where ρ is the density of the surrounding fluid (SI unit: kg/m3) and g (SI unit: m/s2)
is the gravity vector.

Particle Motion in a Turbulent Flow

When particles move in a high Reynolds number flow, they may interact with a large
number of eddies that perturb their trajectories. The fluid velocity field used to perturb
the particle trajectories is generally computed using one of the following numerical
methods (Ref. 6): Reynolds-averaged Navier Stokes (RANS), direct numerical
simulation (DNS), or large eddy simulation (LES).

230 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Although DNS (and, to some degree, LES) are capable of resolving the fluid velocity
field on sufficiently small length scales so that the particle trajectories can be computed
deterministically, the relatively high computational cost of these numerical methods
often makes them impractical for simulation of high Reynolds number flows,
particularly in complicated geometries. Instead, in the following analysis, it will be
assumed that the fluid velocity field has been computed using the RANS model, in
which the mean flow is simulated and variations in fluid velocity are described via the
turbulent kinetic energy k (SI unit: m2/s2). When solving for the particle trajectories,
the drag force is computed by first adding a term to the mean flow field based on the
local value of the turbulent kinetic energy.

The built-in turbulent dispersion models are dependent on the turbulent kinetic
energy and on the turbulent dissipation rate ε (SI unit: m2/s3). Therefore, the
accompanying fluid flow interface should use one of the following turbulence models,
for which these variables are defined:

• k-ε
• k-ω
• SST
• Low Reynolds number k-ε

For more information on the turbulence models see Theory for the
Turbulent Flow Interfaces in the CFD Module User’s Guide.

The following sections describe the available turbulent dispersion models. Each of
these models can be used by selecting the appropriate option from the Turbulent
Dispersion section of the settings window for the Drag Force node.

CONTINUOUS RAND OM WA LK MODEL

In the Continuous random walk model (Ref. 6,9,10), the evaluation of the position and
velocity of a fluid element chosen at random is a Markov process. For a time step dt
(SI unit: s) taken by the Time-Dependent Solver, the general form for the change in
the ith fluid velocity component dui is

du i = a i ( x, u, t )dt + b ij ( x, u, t )dξ j

where ai and bij are coefficients yet to be specified, x and u are the position and
velocity of the fluid element, and dξj are the increments of a vector-valued Weiner
process with independent components, which are uncorrelated Gaussian random

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 231
 numbers with zero mean and variance dt (Ref. 9). A more in-depth discussion of the
relevance of Weiner processes in turbulent dispersion is given in Ref. 11.

In homogeneous turbulence, the velocity perturbations dui are computed by solving

a classical Langevin equation (Ref. 9,10):

uf 2
du f = – ----- dt + σ -----dξ
τL τL

where s (SI unit: m/s) is the fluctuating rms of the velocity perturbation in any
direction. For isotropic turbulence computed using the k-ε model,

2k-
σ = u f2, 1 = u f2, 2 = u f2, 3 = ------
3

In general, the Lagrangian integral time scale τL (SI unit: s) is given by (Ref. 6)

f, i ( t )u f, i ( t + s)
∞u
τ L, i = 0 -----------------------------------------
u (t)
2
- ds
f, i

Essentially, the Lagrangian time scale characterizes the time interval size over which
successively sampled random numbers should become uncorrelated. Typically τL is
estimated using the functional form

k
τ L = C L ---
ε

where the dimensionless coefficient CL is on the order of unity. Typical values of CL

have been reported in the interval [0.2,0.96] (Ref. 6) and as low as 1/7 (Ref. 10).

It is convenient to express the Langevin equation in terms of a dimensionless velocity

perturbation of the form

uf
u' f = -----
σ

The components of u' f are stored for each model particle as auxiliary dependent
variables, requiring one additional degree of freedom per particle per space dimension
(including out-of-plane DOFs). The normalized Langevin equation then takes the
form

u' f 2
du' f = – ------ dt + -----dξ
τL τL

232 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
A drift correction term is added to the normalized Langevin equation to reduce
non-physical diffusion of particles:

u' f 2
du' f = – ------ dt + -----dξ + δu' f
τL τL

For isotropic turbulence, the following expression for the velocity perturbation,
including the drift term, is given (Ref. 10):

u' f 2 1 ∂k dt
du' f = – ------ dt + -----dξ + ------- ------ ----------------
τL τL 3σ ∂x 1 + St

The particle Stokes number St (dimensionless) is defined as

τp
St = -----
τL

where the particle relaxation time scale τp (SI unit: s) or particle velocity response time
depends on the drag law being used (see the Drag Force section) and on rarefaction
effects applied to the drag force, if any (see the Drag Force Correction Factors section).

Anisotropic Turbulence
If the Continuous random walk model is used and the Include anisotropic turbulence in
boundary layers check box is selected, the velocity perturbations are computed using
different expressions in the streamwise, spanwise, and wall normal directions at the
position of each particle.

In the following, the subscripts 1, 2, and 3 refer to the streamwise, spanwise, and wall
normal directions, respectively. The streamwise direction is parallel to the fluid velocity.
The wall normal direction points away from the nearest surface with a Wall boundary
condition. Boundaries with the Inlet, Outlet, and Symmetry boundary conditions are
not considered walls for the purpose of computing the wall normal direction. The
spanwise direction is orthogonal to the wall normal and streamwise directions. Thus
the streamwise, spanwise, and wall normal directions form an orthonormal basis,

n3 = n1 × n2
n2 = n3 × n1
n1 = n2 × n3

The anisotropic turbulence model differs from the isotropic Continuous random walk
model when the particles are sufficiently close to walls such that y+ < 100, where y+ is
the wall distance in dimensionless units,

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 233
 u*
y + = x 3 -------
ρμ

x3 (SI unit: m) is the distance to the nearest wall, ρ (SI unit: kg/m3) is the fluid
density, μ (SI unit: Pa s) is the fluid dynamic viscosity, and u∗ (SI unit: m/s) is the
friction velocity at the nearest wall. Usually u∗ is computed by one of the turbulent
flow physics interfaces.

The normalized Langevin equations in the streamwise, spanwise, and normal

directions, respectively, are

u f1 u f1 1
d  ------------------- = – ------------------- ----- dt + -----dξ 1
2
u f1, rms u f1, rms τ L τL

u f2 u f2 1
d  ------------------- = – ------------------- ----- dt + -----dξ 2
2
u f2, rms u f2, rms τ L τL

u f3 u f3 1 du f3, rms 1
d  ------------------- = – ------------------- ----- dt + -----dξ
2 + ----------------------- ---------------- dt
 u f3, rms u f3, rms τ L τL 3 dx 3 1 + St

where the rms velocity perturbations (Ref. 10) are

u f1, rms 0.40y +

------------------- = -------------------------------------------------------
-
u
* 1.0 + 0.0239 ( y + ) 1.496

u f2, rms 0.19y +

------------------- = -------------------------------------------------------
-
u
* 1.0 + 0.0361 ( y + ) 1.496

u f3, rms 0.0116 ( y + ) 2

------------------- = -----------------------------------------------------------------------------------
-
u
* 1.0 + 0.23y + + 0.00140 ( y + ) 2.421

The Lagrangian time scale τL (SI unit: s) is a piecewise function of the wall distance,


10ρμ
-------------- y+ ≤ 5
 u*
τL = 
 -------------
10ρμ-
( 7.122 + 0.5731y + – 0.00129 ( y + ) 2 ) y+ > 5
 ( u* )2

The dimensionless velocity perturbations are plotted in the anisotropic region in

Figure 5-3. The Lagrangian time scale is plotted in Figure 5-4.

234 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
In all other respects, the effect of these velocity perturbation terms on the particle
trajectories is the same as for the isotropic Continuous random walk model.

Figure 5-3: Dimensionless rms velocity perturbations in the streamwise, spanwise, and
wall normal directions for modeling anisotropic turbulence.

Figure 5-4: Lagrangian time scale as a function of wall distance for modeling anisotropic
turbulence.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 235
 DISCRETE RANDOM WALK MODEL
In the Discrete random walk model (Ref. 6,12,13), the components of the velocity
perturbation are not computed by solving a Langevin equation, but are instead given
instantaneous values

2k-
u f, i = ζ i ------
3

where ζi (dimensionless) are uncorrelated Gaussian random numbers with unit

variance (Ref. 13).

If Discrete random walk is selected from the Turbulent dispersion model list, then the ζi
are recomputed at every time step taken by the solver, whenever particles are released,
and whenever particles interact with boundaries. If the time step taken by the solver is
made arbitrarily small, the amount of turbulent dispersion in the model will decrease
to non-physically small values.

To avoid such non-physical behavior, when using the Discrete random walk option it is
recommended to specify a time step size approximately equal to the average interaction
time between a particle and an eddy in the flow. One estimate of this interaction time
is the eddy lifetime τe (SI unit: s), given by (Ref. 6)

τ e = 2τ L

where τL is the Lagrangian response time. If the option Discrete random walk, variable
time step is selected, then the ζi are only recomputed whenever a particle has existed
for an integer number of eddy lifetimes. This is accomplished by allocating a
dimensionless auxiliary dependent variable N with initial value 0. The value of N is
integrated over time using the equation

dN 1
-------- = ----
dt τe

Then floor(N) is used instead of the instantaneous time t as an argument in the

random number seed used to compute the Gaussian random numbers ζi. The resulting
distribution of particle velocity values will then statistically converge and have nonzero
variance, even if the time step taken by the solver is arbitrarily small.

Particle Motion in a Shear Flow

Particles in a non-uniform velocity field are subjected to lift and drag forces. The drag
force acts in the direction opposite the relative velocity of the particle with respect to

236 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
the fluid. The lift force acts along the direction of the gradient of the fluid velocity.
These forces change significantly as the particle approaches a wall.

Use the Lift Force feature to exert a lift force on the particles. Two formulations of the
lift force are available. Both expressions for the lift force are valid when the angular
velocity of the particle can be neglected and a no-slip boundary condition applies at
the particle surface.

The following sections describe the available lift force models. Each of these models
can be used by selecting the appropriate option from the Lift law list in the settings
window for the Lift Force node. Use the Saffman lift force when particles are in a
non-uniform velocity field far from walls. The Wall induced lift force applies to particles
close to the walls of a flow channel.

SAFFMAN LIFT FORCE

The Saffman lift force FL (SI unit: N), described in Ref. 7, is given by

u–v
F L = – 81,2r p2 L v μρ ----------------
Lv

Lv = ( u – v ) × [ ∇ × ( u – v ) ]

where

• rp is the particle radius (SI unit: m)

• μ is fluid viscosity (SI unit: Pa s)
• v is the velocity of the particle (SI unit: m/s)
• u is the fluid velocity (SI unit: m/s)

The Saffman lift force is only nonzero for particles that have not yet been entrained in
the surrounding fluid. Once a particle has reached equilibrium with the surrounding
fluid velocity, the Saffman lift force is zero. To model the inertial lift force on particles
in a flow channel, which is typically nonzero even after the particles become entrained
in the flow, consider using the Wall induced lift force described in the following section.

WAL L I ND UC ED L IF T FO RCE

The Wall induced lift force model described in Ref. 8 uses a first order correction to the
velocity profile at the channel walls and a second order correction at the surface of the
particle in order to account for higher-order derivatives of the fluid velocity, compared
to the Saffman model. It is given by

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 237
 r p4
F L = ρ ------2- β ( βG 1 ( s ) + αG 1 ( s ) )n
D

β = D ( n ⋅ ∇ )u p

D2
γ = ------- ( n ⋅ ∇ ) 2 u p
2

u p = ( I – ( n ⊗ n ) )u

where

• rp is the particle radius (SI unit: m)

• μ is fluid viscosity (SI unit: Pa s)
• u is the fluid velocity (SI unit: m/s)
• I is the identity matrix (dimensionless),
• n is the wall normal at the nearest point on the reference wall (dimensionless)

Here the term “reference wall” refers to the set of boundaries selected in
the Parallel Boundary 1 section in the settings window for the Lift Force
node.

• D is the distance between the channel walls (SI unit: m)

• s is the nondimensionalized distance from the particle to the reference wall, divided
by D so that 0 < s < 1 for particles in the channel
• G1 and G2 are functions of nondimensionalized wall distance s as given in Ref. 8
(dimensionless). They are plotted in Figure 5-5 and Figure 5-6 below.

238 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
Figure 5-5: G1 coefficient for the lift force.

Figure 5-6: G2 coefficient for the lift force.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 239
 Figure 5-7: Dimensions used to define the wall induced lift force. Here the bottom wall has
been selected in the Parallel Boundary 1 section and the top wall has been selected in the
Parallel Boundary 2 section.

Additional Forces
In addition to the Drag Force and Gravity Force described in Particle Motion in a
Fluid, the following predefined forces can be added to influence particle motion.

• Brownian Force • Acoustophoretic Force

• Electric Force • Thermophoretic Force
• Magnetic Force • Magnetophoretic Force
• Dielectrophoretic Force

BROWNIAN FORCE
The Brownian Force feature adds the following contribution to Fext:

12πk B μTr p
F ext = ζ ------------------------------- (5-2)
Δt

240 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
where

• Δt (SI unit: s) is the time step taken by the solver

• rp (SI unit: m) is the particle radius
• T (SI unit: K) is the absolute fluid temperature
• μ (SI unit: Pa s) is the fluid dynamic viscosity,
• kB = 1.3806488 × 10-23 J/K is the Boltzmann constant, and
• ζ is a normally distributed random number with a mean of zero and unit standard
deviation.
As explained in Ref. 16, independent values for ζ are chosen in all directions. A
different value of ζ is created for each particle, at each time step for each component
of the Brownian force. The Brownian force leads to spreading of particles from
regions of high particle density to low density.

The fluid viscosity required by the Brownian force must be consistent

with the one supplied in the Drag force. Furthermore, the specific form
of the Brownian force given above is only consistent if the Drag law is set
to Stokes.

ELECTRIC FORCE
The Electric Force feature adds the following contribution to Fext:

F ext = eZE

where e = 1.602176565 × 10-19 C is the elementary charge, Z is the charge number

(dimensionless), and E is the electric field (SI unit: V/m). The electric field can be
given directly or computed from an electric potential:

E = – ∇V

When the electric field is computed in the frequency domain it is complex valued. The
field must be cast into a real value that depends on the angular frequency and the
simulation time:

˜
E ( t ) = real ( E exp ( j ( ωt + φ 0 ) ) )
˜
where E is the complex-valued electric field, ω is the angular frequency, φ0 is the initial
phase angle, and t is time.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 241
 MAGNETIC FORCE
The Magnetic Force feature adds the following contribution to Fext:

F ext = eZ ( v × B )

When the magnetic flux density is computed in the frequency domain it is complex
valued. The field must be cast into a real value that depends on the angular frequency
and the simulation time:

˜
B ( t ) = real ( B exp ( j ( ωt + φ 0 ) ) )
˜
where B is the complex-valued electric field, ω is the angular frequency, φ0 is the initial
phase angle, and t is time.

DIELECTROPHORETIC FORCE
The Dielectrophoretic Force feature adds the following contribution to Fext for the
case of a static electric field:

 ε p* – ε * 
- ∇ E 2
F ext = 2πr p3 ε 0 real ( ε * ) real  --------------------
 ε p* + 2ε *

where ε * (dimensionless) is the complex relative permittivity of the fluid, ε p*

(dimensionless) is the complex relative permittivity of the particle, and E (SI unit:
V/m) is the electric field. When computing the dielectrophoretic force exerted by a
stationary electric field, the complex relative permittivity becomes the real-valued
relative permittivity. In the case that the electric field is computed in the frequency
domain, the following is added:

 ε p* – ε * 
- ∇ E rms 2
F ext = 2πr p3 ε 0 real ( ε * ) real  --------------------
 ε p* + 2ε *

where Erms denotes the root mean square electric field. For fields that are computed
in the frequency domain, the complex permittivity can be expressed as

iσ
ε * = ε – -----
ω

where ε (SI unit: F/m) is the permittivity, σ (SI unit: S/m) is the electrical
conductivity, and ω (SI unit: rad/s) is the angular frequency of the electric field.

242 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
The Shell feature can be added to the Dielectrophoretic Force node to model the
dielectrophoretic force on particles with thin shells. The complex permittivity of the
shell can differ from the permittivity of the rest of the particle. When computing the
dielectrophoretic force, the complex permittivity of the particle is replaced by the
equivalent complex relative permittivity ε eq
* of a homogeneous particle comprising

both the shell and the interior of the particle:

 r----0- + 2  --------------------
εp – εs 
3 * *
-
 ri   ε p* + 2ε s*
ε eq
* = ε s ---------------------------------------------------
*

 r----0- –  --------------------
ε p* – ε s* 
3
-
 ri   ε p* + 2ε s*

where r0 and ri (SI unit: m) are the outer and inner radii of the shell, respectively; ε p*
(dimensionless) is the complex relative permittivity of the particle, and ε s*
(dimensionless) is the complex relative permittivity of the outer shell. If multiple shells
are added, the shells are applied in the order in which they appear in the Model Builder,
with the first shell being the innermost. The present treatment of spherical particles
with thin shells is described in Ref. 3.

ACOUSTOPHORETIC FORCE
The Acoustophoretic Force is a special subclass of the CFD forces, which are the
acoustic radiation forces on small particles. The present theory is derived for inviscid
fluids and small particles. This implies that the expressions are valid for particles of
diameter dp = 2rp much smaller than the acoustic wavelength λ and of size larger than
the acoustic boundary layer thickness λbl. The last condition is not necessary for
buoyant particles (of same density ρp as the fluid density ρ0). The present expressions
are based on the work of Gorkov (Ref. 14) which is reviewed by Bruus (see Ref. 15).

Following the references the radiation force on a particle is given as

rad
F aco = – ∇U

where Urad is the radiation potential given by

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 243
 rad 1 2 3 2
U = V p f 1 --------------2-  p in – f 2 --- ρ 0  v in
2ρ 0 c 0 4

2
ρ0 c0 βp K0
f 1 = 1 – -----------2- = 1 – ------ = 1 – -------
ρp cp β 0 K p

2 ( ρp – ρ0 )
f 2 = --------------------------
2ρ p + ρ 0

where

• pin (SI unit: Pa) is the incident (external) pressure field,

• vin (SI unit: m/s) is the incident particle velocity amplitude,
• c0 (SI unit: m/s) is the speed of sound in the fluid,
• cp (SI unit: m/s) is the compressional speed of sound in the particle, and
• βp is the compressibility of the particle (reciprocal of bulk modulus = (Kp)−1).

The   terms are time averaged over one period.

In the case of an acoustic field derived from a time-harmonic analysis

(frequency-domain analysis) where the dependent variables are complex valued, the
time-averaged expressions simplify as

2 1
 p in = --- ( p in p in
* )
2
2 1 * ) = 1
 v in = --- ( v in v in --- ( v in ⋅ v in
* )
2 2

where * is the complex conjugate. For the case of pressure acoustics the velocity is
simply given as

1
v in = – ------------ ∇p
iωρ 0

Thus the present acoustic radiation forces are only space dependent and not
time-dependent as they are time averaged.

THERMOPHORETIC FORCE
The Thermophoretic Force accounts for a force on a particle due to gradients in the
temperature of the background fluid. The thermophoretic force is defined as:

244 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
 2
6πd p μ C s Λ ∇T
F tp = – ----------------------------------------
ρ ( 2Λ + 1 )T

k
Λ = ------
kp

where

• k (SI unit: W/(m K)) is the thermal conductivity of the fluid,

• kp (SI unit: W/(m K)) is the particle thermal conductivity,
• T (SI unit: K) is the fluid temperature,
• dp (SI unit: m) is the particle diameter,
• ρ (SI unit: kg/m3) is the fluid density, and
• Cs is a dimensionless constant equal to 1.17.

This expression for the thermophoretic force is valid for continuum flows, in which the
particle Knudsen number is very small. The thermophoretic force makes particles
move from hotter to cooler regions. This is why dust tends to settle in a corner of the
kitchen, furthest away from the oven.

RAREFACTION EFFECTS IN THERMOPHORESIS

If the Include rarefaction effects check box is selected in the physics interface Particle
Release and Propagation section, other expressions for the thermophoretic force are
available. The Epstein model is equivalent to the expression for continuum flow.

If the Waldmann model is selected, the thermophoretic force is defined as:

8 2πM
F tp = – ------ --------------g-kr p2 ∇T
15 k B T

where

• Mg (SI unit: kg/mol) is the gas molar mass,

• kB = 1.3806488 × 10-23 J/K is the Boltzmann constant, and
• rp (SI unit: m) is the particle radius.

The Waldmann model is appropriate for free molecular flows.

If the Talbot model is selected, the thermophoretic force is defined as:

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 245
 2
6πd p μ C s ( Λ + C t Kn ) ∇T
F tp = – -------------------- ----------------------------------------------------------------------------------- --------
ρ ( 1 + 3C m Kn ) ( 1 – 2Λ + 2C t Kn ) T

If the Linearized BGK model is selected, the thermophoretic force is expressed using
the analysis of the linearized Bhatnagar-Gross-Krook (BGK) and S model kinetic
equations by Beresnev and Chernyak (Ref. 17). The thermophoretic force is defined
in terms of the free-molecular limit,

15 Λ
α E + ------ π ( 1 – α E ) ---- φ 1 + a E Λφ 2
8 R
F tp = ----------------------------------------------------------------------------------------------------------------------------------------- F W
15 Λ 5 Λ
α E + ------ π ( 1 – α E ) ---- φ 3 + α E + --- π ( 3 – α E ) ---- φ 4
8 R 8 R

where FW (SI unit: N) is the thermophoretic force as calculated in the Waldmann

model, R is a dimensionless number defined as

π
R = -------------------
3PrKn

where Pr (dimensionless) is the fluid Prandtl number,

μC p
Pr = -----------
k

and αE (dimensionless) is the energy accommodation coefficient, with αE = 1 and

αE = 0 corresponding to diffuse and specular reflection of all gas molecules at the
surface, respectively. The dimensionless coefficients φi are defined as

φ i = f i1 + ( 1 – α T )f i2

where the dimensionless terms fij are defined as interpolation functions of R (Ref. 17).

R f11 f12 f21 f22

-5 -5 -9
1000 -1.324 × 10 1.317 × 10 1.366 × 10 4.924 × 10-8
800 -2.067 × 10-5 2.055 × 10-5 2.677 × 10-9 9.601 × 10-8
-5 -5 -9
600 -3.673 × 10 3.645 × 10 6.379 × 10 2.270 × 10-7
400 -8.256 × 10-5 8.159 × 10-5 2.175 × 10-8 7.619 × 10-7
200 -3.291 × 10-4 3.215 × 10-4 1.791 × 10-7 5.997 × 10-6
100 -1.307 × 10-3 1.248 × 10-3 1.502 × 10-6 4.643 × 10-5
-3 -3 -6
80 -2.036 × 10 1.921 × 10 2.987 × 10 8.917 × 10-5

246 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
R f11 f12 f21 f22
60 -3.598 × 10-3 3.331 × 10-3 7.225 × 10-6 2.055 × 10-4
40 -8.004 × 10-3 7.139 × 10-3 2.439 × 10-5 6.551 × 10-4
20 -3.115 × 10-2 2.505 × 10-2 1.066 × 10-4 4.417 × 10-3
10 –0.1109 7.525 × 10-2 -1.852 × 10-3 2.639 × 10-2
8 –0.1636 0.1039 -6.174 × 10-3 4.513 × 10-2
6 –0.2626 0.1524 -2.244 × 10-2 8.749 × 10-2
4 –0.4780 0.2443 –0.1013 0.2089
2 –1.058 0.4507 –0.7320 0.7674
1 –1.749 0.6623 –3.077 2.247
0.8 –1.952 0.7205 –4.501 3.051
0.6 –2.181 0.7851 –7.034 4.424
0.4 –2.439 0.8559 -1.237 × 101 7.214
0.2 –2.717 0.9310 -2.889 × 101 1.562 × 101
0.1 –2.857 0.9678 -6.220 × 101 3.238 × 101
0.08 –2.884 0.9748 -7.885 × 101 4.073 × 101
0.06 –2.910 0.9816 -1.066 × 102 5.463 × 101
0.04 –2.835 0.9881 -1.620 × 102 8.239 × 101

R f31 f32 f41 f42

1000 1 –0.9947 –1.004 0.9991
800 1 –0.9934 –1.006 0.9989
600 1 –0.9912 –1.007 0.9985
400 1 –0.9868 –1.011 0.9978
200 1 –0.9739 –1.022 0.9955
100 1 –0.9489 –1.044 0.9909
80 1 –0.9367 –1.055 0.9885
60 1 –0.9169 –1.074 0.9846
40 1 –0.8794 –1.110 0.9765
20 1 –0.7812 –1.218 0.9512
10 1 –0.6360 –1.415 0.8994
8 1 –0.5819 –1.504 0.8737
6 1 –0.5101 –1.635 0.8319
4 1 –0.4098 –1.847 0.7535

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 247
 R f31 f32 f41 f42
2 1 –0.2570 –2.226 0.5676
1 1 –0.1421 –2.533 0.357
0.8 1 –0.1142 –2.609 0.2962
0.6 1 -8.453 × 10-2 –2.691 0.2265
0.4 1 -5.344 × 10-2 –2.781 0.1482
0.2 1 -2.237 × 10-2 –2.881 6.437 × 10-2
0.1 1 -8.413 × 10-3 –2.937 2.470 × 10-2
0.08 1 -5.993 × 10-3 –2.945 1.767 × 10-2
0.06 1 -3.793 × 10-3 –2.961 1.123 × 10-2
0.04 1 -1.900 × 10-3 –2.973 5.648 × 10-3
Note that this data is tabulated assuming that the fluid Prandtl number is 2/3.

MAGNETOPHORETIC FORCE
The Magnetophoretic Force causes motion of permeable particles toward regions
where the magnetic field is stronger. The magnetophoretic force is applicable for
particles which are charge neutral and have a different relative permeability than the
background fluid. The magnetophoretic force is defined as:

3 2
F m = 2πr p μ 0 μ r K ∇ H (5-3)

where H is the magnetic field, μr is the fluid relative permeability, μr,p is the particle
relative permeability, and K is defined as:

μ r, p – μ r
K = --------------------------
μ r, p + 2μ r

The force is only valid for nonconducting particles, so additional force contributions
due to eddy currents induced in particles are not taken into account.

Numerical Differentiation of Magnetic Field Components

As shown in Equation 5-3, the magnetophoretic force is a function of derivatives of
the components of the magnetic field H. Depending on the physics interface used to
compute the magnetic field, its derivatives may not be available.

For example, in instances of the Magnetic Fields interface, the degrees of freedom
correspond to components of the magnetic vector potential A (SI unit: Wb/m). The
magnetic field can be derived from the magnetic vector potential using the relations

248 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
 B = ∇×A

B = μ0 ( H + M )

where B (SI unit: T) is the magnetic flux density and M (SI unit: A/m) is the
magnetization vector. Thus, derivatives of the magnetic field components correspond
to second derivatives of the degrees of freedom, the components of A. However, the
second derivative is not defined for the vector (curl) elements that are used to
discretize the components of A in 3D or the in-plane components of A in 2D. In these
situations, the derivatives of the magnetic field components will be evaluated as zero
everywhere, resulting in zero magnetophoretic force.

Therefore, depending on the formulation used to compute the magnetic potential, it

may be necessary to take additional steps to ensure that the derivatives of the magnetic
field components are evaluated correctly. A typical approach involves the definition of
an intermediate PDE interface to store the values of the magnetic field components
using scalar basis functions.

Erosion Theory
The Erosion feature calculates the rate of erosive wear or the total mass removed per
unit area due to the impact of particles on a surface. It includes the following models:

FINNIE
Finnie (Ref. 19) defined the volume removed from a surface as:

cMU 2 P
V = ---------------------------------- [ ( cos α ) 2 ] tan α > ----
mr 2 2
4p  1 + -----------
 I 

cMU 2 2 ( sin α ) 2 P
V = ---------------------------------- ---- sin ( 2α ) – 2 -------------------- tan α ≤ ----
mr P 2 P 2
4p  1 + -----------
 I 

The parameters are defined as follows:

• c (dimensionless) is the fraction of particles cutting in an idealized manner.

• M (SI unit: kg) is the total mass of eroding particles.
• U (m/s) is the magnitude of the incident particle velocity.
• p (Pa) is the Vickers hardness of the material.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 249
 • m (SI unit: kg) is the mass of an individual particle hitting the surface.
• r (SI unit: m) is the average particle radius.
• I (SI unit: kg-m2) is the moment of inertia of an individual particle about its center
of mass. For an isotropic sphere, I = 2mr2/5.
• α (rad) is the angle of incidence, with α = 0 tangent to the surface and α = π/2
normal to the surface.
• P is a dimensionless parameter, defined as P = K/(1+mr2/I), where
K (dimensionless) is the ratio of vertical and horizontal forces acting on the particle.

In the Finnie model, particles are assumed to remove mass from the surface via an
idealized cutting mechanism. It does not predict any erosive wear by particles at
normal incidence to a surface, and is recommended for modeling erosion of ductile
materials by particles at small angles of incidence.

E/CRC
The E/CRC model defines the erosion rate in terms of the ratio of mass lost by the
surface to mass of incident particles:

-0.59  v - n
E = CF s ( BH ) ------------------- F( α)
 1 [ m ⁄ s ]

2 3 4 5
F ( a ) = 5.40a – 10.11a + 10.93a – 6.33a + 1.42a

where C is an erosion model coefficient, Fs is the particle shape coefficient, and BH is

the Brinell hardness of the wall material (all dimensionless). The angle of incidence α
is measured in radians.

OKA
The Oka model defines the volume of surface material removed per unit mass of
incident particles (in units of mm3/kg) as:

E ( α ) = g ( α )E 90

where E(α) is the etch rate, E90 is the etch rate at normal incidence, and g(α) is an
angular dependence function:

Hv n2
g ( a ) = ( sin α ) n 1 1 + --------------------- ( 1 – sin α )
1 [ GPa ]

250 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
 k1 v k2 D k3
E 90 =  ---------------------  ----  -----
Hv
 1 [ GPa ]  v'  D'

where v’ is the reference velocity; D’ is the reference diameter; the dimensionless

parameters n1, n2, and k2 are defined as

q1
n 1 = s 1  ---------------------
Hv
 1 [ GPa ]

q2
n 2 = s 2  ---------------------
Hv
1 [ GPa ]

0.038
k 2 = 2.3  ---------------------
Hv
 1 [ GPa ]

and the remaining dimensionless parameters s1, s2, q1, q2, K, k1, and k3 are constants.

DNV
The DNV model defines the erosion rate in terms of the ratio of mass lost by the
surface to mass of incident particles:

–n
E = K  --------------------
v
F(α)
 1 [ m ⁄ s ]

2 3 4
F ( a ) = 9.370α – 42.295α + 110.864α – 175.804α
+ 170.137α 5 – 98.398α 6 + 31.211α 7 – 4.170α 8

where K and n are dimensionless constants that depend on the surface material.

Droplet Breakup Theory

It is possible to model the breakup of liquid droplets using the Droplet Breakup
feature. This feature supports two subfeatures, Kelvin-Helmholtz Breakup Model
(added by default) and Rayleigh-Taylor Breakup Model, which correspond to different
mechanisms for breakup. It is possible to include either breakup model separately or
to use both models at once.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 251
 DIMENSIONLESS NUMBERS
When analyzing the mechanisms of droplet breakup, several dimensionless parameters
are useful. These parameters are summarized in Table 5-1.
TABLE 5-1: DIMENSIONLESS QUANTITIES USED IN DROPLET BREAKUP MODELS

SYMBOL NAME EXPRESSION

Weg Gas Weber number ρU rel 2 r

p
We g = ---------------------
σp

Wel Liquid Weber number ρ p U rel 2 r

p
We l = ------------------------
σp

Rel Droplet Reynolds number ρ p r p U rel

Re l = ------------------------
μp

Z Ohnesorge number We l
Z = ---------------
Re l

T Taylor number T = Z We g

where

• ρ (SI unit: kg/m3) is the gas density,

• ρp (SI unit: kg/m3) is the droplet density,
• Urel (SI unit: m/s) is the relative speed between the droplet and gas,
• rp (SI unit: m) is the droplet radius,
• σp (SI unit: N/m) is the surface tension, and
• μp (SI unit: Pa s)is the droplet dynamic viscosity.

KELVIN-HELMHOLTZ BREAKUP
The Kelvin-Helmholtz Breakup Model is based on a first-order linear analysis of
Kelvin-Helmholtz instability in a cylindrical jet of fluid. For a parent droplet of radius
r0 (SI unit: m), the wavelength ΛKH (SI unit: m) of the fastest-growing disturbance
on the surface of the particle is

9.02r 0 ( 1 + 0.45 Z l ) ( 1 + 0.4T 0.7 )

Λ KH = ----------------------------------------------------------------------------------------
-
( 1 + 0.865We g1.67 ) 0.6

The corresponding growth rate ΩKH (SI unit: 1/s) of the fastest-growing disturbance
is

252 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
 0.34 + 0.385We g 1.5
σ
Ω KH = ------------------------------------------------------
- -----------
-
( 1 + Z l ) ( 1 + 1.4T 0.6 ) ρ p r 03

The characteristic breakup time τKH (SI unit: s) can be derived from the wavelength
and frequency of the fastest-growing disturbance,

3.788B KH r 0
τ KH = ---------------------------------
Λ KH Ω KH

where BKH is a dimensionless constant. The value of BKH can vary considerably,
depending on the mechanism for spray formation. Typical values are between 3
and 60.

Kelvin-Helmholtz breakup is likely to occur when the parent droplet is sufficiently

large and the time since previous breakup is comparable to or greater than τKH. The
approximate size of the parent droplet over time is computed by allocating an effective
Kelvin-Helmholtz radius rKH (SI unit: m) as an auxiliary dependent variable for all
particles. The effective radius is initially equal to the parent droplet radius r0 and
decreases over time according to the expression

dr KH r ch – r KH
-------------- = ------------------------
dt τ KH

Where rch (SI unit: m) is the child droplet radius,

B 0 Λ KH B 0 Λ KH ≤ r KH

 2 U  1 / 3 3πr 2 U
r ch =   3πr KH 1/3
 min  ------------------------------ ,  ------------------------------
rel KH rel
B 0 Λ KH > r KH
  2Ω KH  4

where B0 = 0.61 is a dimensionless constant.

The criterion for breakup is that, at a time step taken by the solver, there exists a
positive integer nch such that

m 0 ≥ m + n ch r ch

When this criterion is true, a number of child droplets equal to nch is released at the
parent droplet’s position. The new mass of the parent droplet mnew (SI unit: kg) is

m new = m 0 – n ch r ch

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 253
 R AY L E I G H - TAY L O R B RE A K UP
The Rayleigh-Taylor Breakup Model is a significant breakup mechanism for droplets
that undergo rapid accelerations. Compared to the Kelvin-Helmholtz model, in which
child droplets are continually stripped away from a liquid jet, the Rayleigh-Taylor
model describes the complete annihilation of the parent droplet as it breaks into a large
number of child droplets.

According to the Rayleigh-Taylor model, a droplet breaks up when the wavelength of

a disturbance on the droplet’s surface exceeds a critical value, which depends on the
droplet’s radius.

Following Ref. 21, the wavelength of the fastest-growing disturbance ΛRT

(SI unit: m) is

3σ-
Λ RT = 2π ---------
aρ p

The frequency of the fastest-growing wave ΩRT (SI unit: 1/s) is

2a aρ
Ω RT = ------- ---------p-
3 3σ

where a (SI unit: m/s2) is the magnitude of the droplet acceleration. The time
constant for breakup τRT (SI unit: s) is

3- 3σ
τ RT = ------ ----------
2a ρ p a

The radius of the child droplets rch (SI unit: m) is

C RT Λ RT
r ch = ----------------------
2

where CRT is a dimensionless constant, with typical values ranging from 2.5 to 5.

When the Rayleigh-Taylor model is used, the time since previous Rayleigh-Taylor
breakup tRT (SI unit: s) is defined as an auxiliary dependent variable for all particles.
Whenever t RT > τ RT at the beginning of a time step taken by the solver, the parent
droplet is completely broken up into child droplets. The number of child droplets nch
is

254 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
  r 03 
n ch = floor  -------
3
-
 r ch 

The floor operator is used to ensure that the number of child droplets is an integer,
even though the resulting droplets may have radii different from rch.

USING MULTIPLE DROPLET BREAKUP MODELS

It is possible to include the Kelvin-Helmholtz and Rayleigh-Taylor breakup models in
the same simulation. In this case, separate auxiliary dependent variables are allocated
for the Kelvin-Helmholtz effective radius rKH and the time since previous
Rayleigh-Taylor breakup tRT.

If the condition for Kelvin-Helmholtz breakup is met, then for the parent droplet and
all child droplets tRT is set to zero.

If the condition for Rayleigh-Taylor breakup is met, then for all child droplets the
Kelvin-Helmholtz radius rKH is set equal to the child droplet radius.

Nozzle Theory
The Nozzle feature is used to release a spray of liquid droplets, as if they were released
from a nozzle. It computes the initial size and velocity of the released droplets using
the so-called Blob model (Ref. 22). This model requires the following assumptions:

• The inlet is circular and produces a round liquid jet of density ρp (SI unit: kg/m3).
• The gas in the adjacent domain is incompressible and initially quiescent, vg,0 = 0.
• The liquid velocity vl (SI unit: m/s) is constant inside the jet.
• Both the injection pressure pinj (SI unit: Pa) and the ambient pressure pamb
(SI unit: Pa) are constant.
• The initial parent drop radius is constant and equal to the injector radius r0
(SI unit: m).
• The maximum cone angle Θ (SI unit: rad) is constant and stays constant as the
injection progresses.

The following discussion uses several dimensionless variables. Their

definitions are listed in Table 5-1 in the Droplet Breakup Theory section.

The initial speed of the released droplets is

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 255
 2Δp 1 ·
U0 = ----------- c d = ---------------2- m
ρp ρ p πr 0
·
where cd (dimensionless) is the discharge coefficient and m (SI unit: kg/s) is the mass
flow rate.

The droplets are released in a cone with maximum spray angle Θ. The spray angle can
be specified directly, or it can be estimated from Kelvin-Helmholtz instability theory:

Θ ΛΩ
tan  ---- = A ---------
2 U0

where Λ (SI unit: m) and Ω (SI unit: 1/s) are, respectively, the wavelength and growth
rate of the fastest-growing disturbance on the surface of a liquid jet:

9.02r 0 ( 1 + 0.45 Z l ) ( 1 + 0.4T 0.7 )

Λ KH = ----------------------------------------------------------------------------------------
-
( 1 + 0.865We g1.67 ) 0.6

0.34 + 0.385We g1.5 σ

Ω KH = ------------------------------------------------------
- -----------
-
( 1 + Z l ) ( 1 + 1.4T 0.6 ) ρ p r 03

The number multiplication factor n, activated by the Enable macroparticles check box
in the physics interface Additional Variables section, is a dimensionless quantity that
indicates the number of droplets that are represented by each model particle.

Suppose the range of specified release times is t 1, t 2, t 3, …, t Nt , where Nt is the

number of release times.

The droplets that are released at time ti are assigned initial multiplication factor ni:

·
3m Δt i
n i = --------------------------3-
4πN 0 ρ p r 0

where N0 is the number of model particles per release and the time intervals Δti are
defined as

256 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
 t2 – t1
 --------------- i = 1
 2

 ti + 1 – ti – 1
Δt i =  ---------------------------- 1<i<N
 2
 t –t
 ------------------------
N N–1
- i = N
 2

where N is the number of release times. Note that, based on the previous expressions,

Np
4 ·
 --3- πρp nr03 = m ( tN – t1 )
i=1

where N p = N × N 0 is the total number of released model particles. That is, the sum
of the masses of all released droplets, scaled by their multiplication factors, equals the
total mass released by the nozzle, assuming that the mass flow rate is constant between
the initial and final release times and zero otherwise.

Computing Particle Temperature

Select the Compute Particle Temperature check box in the physics interface Advanced
Settings section to define an auxiliary dependent variable for particle temperature. The
particle temperature Tp (SI unit: K) is then computed along each particle trajectory by
integrating the first-order equation

dT p
m p C p ----------- = Q (5-4)
dt

where

• mp (SI unit: kg) is the particle mass,

• Cp (SI unit: J/(kg K)) is the particle specific heat capacity,
• t (SI unit: s) is the time, and
• Q (SI unit: W) is the total heat source at the particle surface.

The particle temperature is treated as a single value for each particle, not as a
temperature distribution throughout the particle’s volume. Therefore, the
temperature computation is only valid when the temperature throughout the particle
can be considered uniform; that is, the heat transfer resistance within the particle is

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 257
 negligibly small compared to the heat transfer resistance at the surface of the particle.
This is typically true for small particles with high thermal conductivity.

CONVECTIVE HEAT LOSSES

Use the Convective Heat Losses feature to apply convective heat transfer at the surface
of the particles. This feature creates the following contribution to the total heat source
in Equation 5-4:

Q = hA p ( T – T p )

where

• h (SI unit: W/(m2 K)) is the heat transfer coefficient,

• Ap (SI unit: m2) is the particle surface area, and
• T (SI unit: K) is the temperature of the surrounding fluid at the particle’s position.

The Biot number Bi (dimensionless) can be used to determine whether the particle
temperature can be treated as a uniform value. The Biot number is defined as

hL C
Bi = -----------
kp

where LC (SI unit: m) is a characteristic length, typically the ratio of particle volume
to particle surface area, and kp (SI unit: W/(m K)) is the particle thermal conductivity.
If the Biot number is very small, much less than unity, then the conductive heat transfer
within the particle takes place on a much shorter time scale than convective heat
transfer at the surface of the particle, so the particle temperature can be treated as a
uniform value.

RADIATIVE HEAT LOSSES

Use the Radiative Heat Losses feature to make the particles undergo radiative heat
exchange with their surroundings. This feature creates the following contribution to
the total heat source in Equation 5-4:

Q = ε p σA p ( T 4 – T p4 )

where

• εp (dimensionless) is the particle emissivity,

• σ = 5.670373 × 10 – 8 W/(m2 K4)] is the Stefan-Boltzmann constant,

258 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
• Ap (SI unit: m2) is the particle surface area, and
• T (SI unit: K) is the temperature of the enclosure or ambient surroundings.

USER-DEFINED HEAT SOURCE

Use the Heat Source feature to create a user-defined heat source for the particles. This
feature creates a user-defined contribution to the total heat source Q in Equation 5-4.

Computing Particle Mass

Select the Compute Particle Mass check box in the physics interface Advanced Settings
section to allocate an auxiliary dependent variable for the mass of each particle. You
can specify the Accretion rate R (SI unit: kg/s), the time derivative of the particle mass,
in the settings window for the Particle Properties feature.

Velocity of Mass Gained or Lost

When the Compute particle mass check box is selected, the interpretation of Newton’s
second law of motion for each particle is

m p ------  ------- = F t
d dq
(5-5)
dt  dt 

instead of the more customary, momentum-conserving expression

d- 
m ------- = F t
dq
----- (5-6)
dt  p dt 

The physical interpretation of Equation 5-5 is that the mass gained or lost by the
model particle is moving with the particle’s velocity, whereas the physical
interpretation of Equation 5-6 is that the mass gained or lost by the model particle is
stationary with respect to the coordinate system in which the geometry is defined.

When Compute particle mass is selected, Equation 5-5 is solved. To instead solve
Equation 5-6, so that mass gained or lost by the particle affects its velocity such that
the momentum of the model particle is conserved, clear the Compute particle mass
check box and instead either enter a time-dependent expression for the particle mass
directly or express the particle mass in terms of a user-defined auxiliary dependent
variable.

References for the Particle Tracing for Fluid Flow Interface

1. E.M. Lifshitz and L.P. Pitaevskii, Physical Kinetics, Butterworth, 1981.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 259
 2. E.M. Lifshitz and L.P. Pitaevskii, Fluid Mechanics, Butterworth, 1987.

3. T. B. Jones, Electromechanics of Particles, Cambridge, 1995.

4. R. Clift, J.R. Grace, and M.E. Weber, Bubbles, Drops, and Particles, Dover, 1978.

5. Happel, John, and Howard Brenner. Low Reynolds number hydrodynamics: with
special applications to particulate media. Vol. 1. Springer Science & Business Media,
2012.

6. L. Tian, G. Ahmadi, “Particle deposition in turbulent duct flows–comparisons of

different model predictions”, Aerosol Science, vol. 38, 2007, pp. 377-397.

7. Saffman, P. G. T, “The lift on a small sphere in a slow shear flow”, J. Fluid Mech.
22, no. 02 (1965), pp. 385-400.

8. Ho, B. P., and L. G. Leal, “Inertial migration of rigid spheres in two-dimensional

unidirectional flows”, J. Fluid Mech. 65, no. 02 (1974), pp. 365-400.

9. D.J. Thomson, “Criteria for the selection of stochastic models of particle trajectories
in turbulent flows”, J. Fluid Mech., vol. 180, 1987, pp. 529-556.

10. A. Dehbi, “Turbulent particle dispersion in arbitrary wall-bounded geometries: A

coupled CFD-Langevin-equation based approach”, International Journal of
Multiphase Flow, vol. 34, 2008, pp. 819-828.

11. P. A. Durbin, “Stochastic Differential Equations and Turbulent Dispersion”,

NASA Reference Publication 1103, 1983.

12. P. Hutchinson, G.F. Hewitt, and A.E. Dukler, “Deposition of liquid or solid
dispersions from turbulent gas streams: a stochastic model”, Chemical Engineering
Science, vol. 26, 1971, pp.419-439.

13. B. Zhao, C. Yang, X. Yang, and S. Liu, “Particle dispersion and deposition in
ventilated rooms: Testing and evaluation of different Eulerian and Lagrangian
models”, Building and Environment, vol. 43, 2008, pp. 388-397.

14. L.P. Gorkov, Sov. Phys. Doklady, vol. 6, p. 773, 1962.

15. H. Bruus, Lecture Notes: Microfluidics and ultrasound acoustophoreses.

http://www.nanotech.dtu.dk/Research/Research%20groups/TMF.aspx

16. M. Kim and A.L. Zydney, “Effect of Electrostatic, Hydrodynamic, and Brownian
Forces on Particle Trajectories and Sieving in Normal Flow Filtration,” J. Colloid and
Interface Science, vol. 269, pp. 425–431, 2004.

260 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
17. S. Beresnev and V. Chernyak, “Thermophoresis of a Spherical Particle in a Rarefied
Gas: Numerical Analysis Based on the Model Kinetic Equations,” Phys. Fluids, vol. 7,
pp. 1743-1756, 1995.

18. F. Zheng, “Thermophoresis of Spherical and Non-Spherical Particles: a Review of

Theories and Experiments,” Advances in Colloid and Interface Science, vol. 97,
pp. 255-278, 2002.

19. I. Finnie, “Some Observations on the Erosion of Ductile Metals,” Wear, vol. 19,
pp. 81–90, 1972.

20. Y. Zhang, E.P. Reuterfors, B.S. McLaury, S.A. Shirazi, and E.F. Rybicki,
“Comparison of computed and measured particle velocities and erosion in water and
air flows,” Wear, vol. 263, pp. 330–338, 2007.

21. M. A. Patterson and R. D. Reitz, Modeling the Effects of Fuel Spray

Characteristics on Diesel Engine Combustion and Emission, SAE Paper 980131,
1998.

22. R. D. Reitz, Modeling Atomization Processes in High-Pressure Vaporizing

Sprays, Atomization and Spray Technology, vol. 3, pp. 309-337, 1987.

T H E O R Y F O R T H E P A R T I C L E TR A C I N G F O R F L U I D F L O W I N T E R F A C E | 261
262 | C H A P T E R 5 : P A R T I C L E TR A C I N G F O R F L U I D F L O W
 6

Multiphysics Interfaces

The Particle Tracing Module contains predefined multiphysics interfaces to

facilitate easy setup of models with the most commonly occurring settings. This
chapter describes the multiphysics interfaces that are available with the Particle
Tracing Module. Two of these multiphysics interfaces can be found under the AC/DC
branch ( ) when adding a physics interface — Particle Field Interaction,
Non-Relativistic and Particle Field Interaction, Relativistic. The Fluid Particle
Interaction interface is found under the Fluid Flow branch ( ).

In this chapter:

• The Particle Field Interaction, Non-Relativistic Interface

• Theory for the Particle Field Interaction, Non-Relativistic Interface
• The Particle Field Interaction, Relativistic Interface
• Theory for the Magnetic Particle Field Interaction, Relativistic Interface
• The Fluid-Particle Interaction Interface
• Theory for the Fluid-Particle Interaction Interface

See The Electromagnetics Interfaces in the COMSOL Multiphysics Reference

Manual for other AC/DC interface and feature node settings.

263
 T he P a r ti c le F i el d In t erac t i on ,
Non-Relativistic Interface
The Particle Field Interaction, Non-Relativistic ( ) multiphysics interface combines the
Charged Particle Tracing interface with the Electrostatics interface. The Electric
Particle Field Interaction multiphysics coupling feature is added automatically. The
Particle Field Interaction, Non-Relativistic interface is used to model beams of charged
particles at non-relativistic speeds. The particles generate a space charge density term
as they propagate through domains. The space charge density is then used as a source
term in the Electrostatics interface, and the resulting electric force on the particles is
computed.

When a predefined Particle Field Interaction, Non-Relativistic interface is added from the
AC/DC>Particle Tracing branch ( ) of the Model Wizard or Add Physics windows,
Electrostatics and Charged Particle Tracing interfaces are added to the Model Builder. A
Multiphysics node is also added, which automatically includes the Electric Particle Field
Interaction multiphysics coupling.

On the Constituent Physics Interfaces

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.

The Charged Particle Tracing interface is used to model charged particle orbits under
the influence of electromagnetic forces. In addition, it can also model two-way
coupling between the particles and fields. Some typical applications are particle
accelerators, vacuum tubes, and ion implanters. The physics interface supports
time-domain modeling only in 2D and 3D. The physics interface solves the equation
of motion for charged particles subjected to electromagnetic forces.

SETTINGS FOR PHYSICS INTERFACES AND COUPLING FEATURES

When physics interfaces are added using the predefined couplings — for example,
Particle Field Interaction, Non-Relativistic — specific settings are included with the

264 | CHAPTER 6: MULTIPHYSICS INTERFACES

physics interfaces and the coupling features. However, if physics interfaces are added
one at a time, followed by the coupling features, these modified settings are not
automatically included.

For example, if single Electrostatics and Charged Particle Tracing interfaces are added,
COMSOL Multiphysics adds an empty Multiphysics node. You can choose from the
available coupling features but the modified settings are not included.

Coupling features are available from the context menu (right-click the
Multiphysics node) or from the Physics toolbar, Multiphysics menu.

TABLE 6-1: MODIFIED SETTINGS FOR A PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE

PHYSICS INTERFACE OR MODIFIED SETTINGS (IF ANY)

COUPLING FEATURE

Electrostatics No changes.
Charged Particle Tracing For the Charged Particle Tracing interface, under
Particle Release and Propagation, the Particle release
specification is set to Specify current.
The Electric Force node is added to the Charged
Particle Tracing interface. The Domain Selection is the
same as that of the Charged Particle Tracing interface.
Electric Particle Field The Domain Selection is the same as that of the
Interaction participating physics interfaces.
The corresponding Electrostatics and Charged Particle
Tracing interfaces are preselected in the Electric
Particle Field Interaction section.

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 Electric Particle Field Interaction and Space Charge Limited Emission coupling
feature nodes are described in this section.

THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 265

 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 Charged Particle Tracing Interface are listed
in the section Domain, Boundary, Pair, and Global Nodes for the Particle Tracing
for Fluid Flow Interface.
• The available physics features for The Electromagnetics Interfaces are listed in the
section Domain, Boundary, Edge, Point, and Pair Nodes for the Electrostatics
Interface in the COMSOL Multiphysics Reference Manual.

Electric Particle Field Interaction

The Electric Particle Field Interaction multiphysics coupling ( ) computes the space
charge density due to particles and assigns it to a dependent variable that is defined on
the domain mesh elements that contain the particles. It also applies the accumulated
space charge density as a source when computing the electric potential.

The computation of the space charge density is controlled by the Particle release
specification list in the settings window for the Charged Particle Tracing interface. If
Specify release times is selected, each particle contributes to the space charge density
based on its instantaneous location. If Specify current is selected, each model particle is
treated as representing a number of charged particles per unit time, leaving behind a
contribution to the space charge density in mesh elements it has previously passed
through.

SETTINGS
The Label is the default multiphysics coupling name.

The Name is used primarily as a scope prefix for variables defined by the coupling node.
Refer to such 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.

266 | CHAPTER 6: MULTIPHYSICS INTERFACES

The default Name (for the first multiphysics coupling in the model) is epfi1.

CHARGE MULTIPLICATION FACTOR

Enter a Charge multiplication factor n (dimensionless). The default value is 1 × 106. If
the Particle release specification in the settings window for the Charged Particle Tracing
interface is set to Specify release times, the Charge multiplication factor indicates the
number of real particles represented by every model particle for the purpose of
computing the space charge density.

CONTINUATION SETTINGS
Select the Use cumulative space charge density check box to enable the following inputs,
which are used to accelerate the convergence of models with bidirectionally coupled
particle-field interactions.

The options in the Continuation Settings section only have an effect on the
solution if the model uses the Bidirectionally Coupled Particle Tracing
study step. This is a specialized study step that should be used to model
bidirectionally coupled particle-field interactions, in which the
contribution of the charged particles to the space charge density in the
surrounding domain is sufficiently large to significantly perturb the
electric field, which in turn modifies the particle trajectories.

Enter the Number of iterations β (dimensionless). The default value is 1. During the
iterative solver loop set up by the Bidirectionally Coupled Particle Tracing study step,
the contribution of the particles to the space charge density will be linearly ramped up
for iteration numbers less than β. This improves the probability of stable convergence
for models in which overestimation of the space charge density may lead to
non-physical particle motion. This is often true when using the Space Charge Limited
Emission feature, since overestimation of the space charge density in the vicinity of the
cathode may cause a non-physical reversal in the direction of the initial particle velocity.

Select an option from the Weights for subsequent iterations list: Uniform (the default),
Arithmetic sequence, or Geometric sequence. For Geometric sequence enter the Common
ratio r (dimensionless). The default value is 1.5. The contribution of charged particles
to the space charge density is treated as a cumulative average of the solutions for all
iteration numbers greater than β, and these settings control how the cumulative
average is computed.

THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 267

 For more information about the options in the Continuation Settings
section, see Stabilization of the Space Charge Density Calculation in
Theory for the Particle Field Interaction, Non-Relativistic Interface.

COUPLED INTERFACES
This section defines the physics involved in the multiphysics coupling. By default, the
applicable physics interface is selected in the Source and Destination lists.

You can also select None from either list to uncouple the node from a physics interface.
If the physics interface is removed from the Model Builder then the applicable list
defaults to None as there is nothing to couple to.

If a physics interface is deleted and then added to the model again, and in
order to re-establish the coupling, you need to choose the physics
interface again from the Source or Destination lists. This is applicable to all
multiphysics coupling nodes that would normally default to the once
present physics interface. See Multiphysics Modeling Approaches in the
COMSOL Multiphysics Reference Manual.

Space Charge Limited Emission

Use the Space Charge Limited Emission multiphysics coupling ( ) to model the space
charge limited emission of electrons from a surface.

SETTINGS
The Label is the default multiphysics coupling name.

The Name is used primarily as a scope prefix for variables defined by the coupling node.
Refer to such 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 in the model) is scle1.

ELECTRIC POTENTIAL
Enter an Electric potential at emission surface V0 (SI unit: V). The default value is 0.

268 | CHAPTER 6: MULTIPHYSICS INTERFACES

NUMBER OF PARTICLES
Enter the Number of particles per release N (dimensionless). The default value
is 1 × 103.

EXTRA DIMENSION SETTINGS

Enter the Position offset os (SI unit: m). The default value is 1 mm.

Select the Initial value of electric potential in extra dimension—Automatic (the default)
or User defined. For User defined enter the value for Vxd,0 (SI unit: V). The default value
is 1 V.

Select the Mesh settings: Automatic (the default) or User defined. For User defined enter
values for the following settings:

• Number of elements (dimensionless). The default value is 50.

• Element ratio (dimensionless). The default value is 10.

SPACE CHARGE LIMITED EMISSION

This section defines the physics involved in the multiphysics coupling. By default, the
applicable physics interface is selected in the Source and Destination lists.

You can also select None from either list to uncouple the node from a physics interface.
If the physics interface is removed from the Model Builder then the applicable list
defaults to None as there is nothing to couple to.

If a physics interface is deleted and then added to the model again, and in
order to re-establish the coupling, you need to choose the physics
interface again from the Source or Destination lists. This is applicable to all
multiphysics coupling nodes that would normally default to the once
present physics interface. See Multiphysics Modeling Approaches in the
COMSOL Multiphysics Reference Manual.

THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 269

 Theory for the Particle Field
Interaction, Non-Relativistic
Interface
The Particle Field Interaction, Non-Relativistic Interface combines charged particle
tracing with electrostatics to model non-relativistic beams of charged particles that can
create significant space charge density distributions in the domains that contain the
particles. The space charge density may, in turn, exert a significant electric force on the
particles, resulting in a bidirectional coupling between the particle trajectories and the
electric potential.

If all charged particle beams are released at constant current, it is possible to

significantly reduce the simulation time and computational cost by combining a
time-domain calculation of the particle trajectories with a Stationary solver for the
calculation of the electric potential. The two calculations can then be performed using
an iterative procedure that alternates between them until a self-consistent solution is
attained.

An iterative solver loop that consists of a time-dependent solver for computing particle
trajectories and a stationary solver for computing all other dependent variables can be
set up automatically using the Bidirectionally Coupled Particle Tracing study step,
described in the COMSOL Multiphysics Reference Manual.

Electrostatics and Charged Particle Tracing Equations

Under static conditions, Gauss’ law can be written as a variant of Poisson’s equation

– ∇ ⋅ ( ε 0 ∇V – P ) = ρ

where

• ε0 = 8.854187817 × 10-12 F/m is the permittivity of vacuum,

• V (SI unit: V) is the electric potential,
• P (SI unit: C/m2) is the polarization, and
• ρ (SI unit: C/m3) is the charge density.

270 | CHAPTER 6: MULTIPHYSICS INTERFACES

The electric field E (SI unit: V/m) can be expressed in terms of the electric potential
V (SI unit: V) using the relationship

E = – ∇V

The equation of motion for a charged particle in an electromagnetic field can be

written as

d-
----- ( m v ) = – ZeE + Ze ( v × B )
dt p

where

• mp (SI unit: kg) is the particle mass,

• v (SI unit: m/s) is the particle velocity,
• e = 1.602176565 × 10-19 C is the elementary charge,
• Z (dimensionless) is the particle charge number, and
• B (SI unit: T) is the magnetic flux density.

For non-relativistic particles, the contribution of the particles to the magnetic force is
assumed to be negligibly small compared to the contribution of particles to the electric
force; thus, the self-magnetic field is neglected. In the next section the contribution to
the electric force based on particle position is considered.

Space Charge Density Calculation

Given an array of idealized point sources such that the position vector of the ith source
is denoted qi (SI unit: m), the space charge density at position r is

ρ(r) =  qi δ ( r – qi ) (6-1)
i=1

THEORY FOR THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 271

 where δ is the Dirac delta function, qi is the charge of the ith particle, and N is the total
number of particles. The equation for the charge density is unusable in this form,
however, for the following reasons:

• The number of particles may be extraordinarily large, making it impractical to model

all of them.
• The charge density becomes infinite at the location of each idealized point source,
making the calculation of the electric potential infeasible with most numerical
methods.

In the following sections we discuss solutions to each of these problems.

MODELING A REPRESENTATIVE SAMPLE OF PARTICLES

Because the number of real particles, such as ions or electrons, may be too large for
every particle to be modeled individually, a practical numerical approach is to release a
representative sample of model particles, allowing each model particle to make the
same contribution to the space charge density as an equivalent number of real particles.

For example, instead of allocating degrees of freedom for 1012 electrons, it will often
suffice to model 104 particles, each of which has a Charge multiplication factor of 108,
meaning that it represents 108 electrons.

Simplification for Constant-Current Beams

If a beam of particles is released at constant current, then a full time-domain calculation
of the coupled particle trajectories and electric fields may require particles to be
released at a large number of time steps until a stationary solution for the electric
potential is reached. This can be needlessly memory-intensive and time-consuming. An
alternate approach is to release particles at time t = 0 and to allow each model particle
to represent a continuous stream of real particles per unit time. The number of real
particles per unit time represented by each model particle is denoted the effective
frequency of release, frel.

The charged particles will tend to contribute to a greater space charge contribution in
regions in which they are moving slowly, causing particles that are released at successive
times to be closer together. This behavior can be conveniently reproduced by defining
an expression for the time derivative of the charge density, rather than the charge
density itself:

N
dρ ( r )-
--------------
dt
=  frel, i qi δ ( r – qi ) (6-2)
i=1

272 | CHAPTER 6: MULTIPHYSICS INTERFACES

The charge density can then be computed by integrating over time, as long as sufficient
time is given so that the particle trajectories can be traced completely through the
modeling domain.

The frequency of release can be computed using the current and number of model
particles that are specified in release feature settings. For example, for an Inlet node
with release current magnitude I (SI unit: A) and number of particles per release N
(dimensionless), the effective frequency of release is

I
f rel = -----------
qN

When particle beams are assumed to have constant current, then the space charge
density at the last time step includes contributions from particles at every point along
their trajectories in the modeling domain. Thus, it can be applied as the space charge
density term when computing the electric potential.

The treatment of particle trajectories as paths in a constant-current beam is determined

by the Particle release specification setting in the settings window for the Charged
Particle Tracing physics interface. If Specify release times is selected, the charge density
is computed using Equation 6-1 and is determined by the instantaneous positions of
all model particles. Thus, it is necessary to solve for the particle trajectories and electric
potential in the time domain. If Specify current is selected, the charge density is
computed using Equation 6-2 and is determined by the time history of the model
particle positions.

The difference between the Specify current and Specify release times particle release
specification is thus analogous to the difference between integration over Elements and
time and integration over Elements as described for the Accumulator (Domain) node.

At this point, the effect of a bidirectional coupling between the particle trajectories and
fields has not been considered. If Specify release times is selected from the Particle
release specification list, this does not require special consideration because the
trajectories and fields are computed simultaneously. If Specify current is selected,
however, the trajectories and fields are computed using different study types, and an
additional feedback mechanism is needed. The Bidirectionally Coupled Particle
Tracing study step can be used to generate a solver sequence that does the following:

1 Set the space charge density contribution due to the particles to zero.
2 Compute the electric potential and other field variables using a Stationary solver,
using the value of the space charge density computed in the previous step.

THEORY FOR THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 273

 3 Compute the particle trajectories and the resulting space charge density in the time
domain, using the field variables computed in the previous step.
4 Repeat steps 2 and 3 until a specified number of iterations has been reached.

If the number of iterations taken by the solver sequence is sufficiently large, the
resulting solution will fully account for the bidirectional coupling between the particle
trajectories and stationary fields.

AVOIDING INFINITELY LA RGE VALUES OF TH E SPACE CHARGE DENS IT Y

The Electric Particle Field Interaction node defines a variable for the contribution to
the space charge density by particles in each mesh element. This variable is discretized
using constant shape functions that are, in general, discontinuous across boundaries
between elements. For a mesh element j with volume Vj, and with the Particle release
specification set to Specify release times, the average space charge density ρj is

N
1
ρ j = -----
Vj  n i q i  δ ( r – q i ) dV
i=1

where ni is the charge multiplication factor of the ith model particle. The integral on
the right-hand side is a volume integral over element j. The resulting charge density is
the average charge density over the mesh element, which may be written more
concisely as

Nj
1
ρ j = -----
Vj  ni qi
i=1

where the sum is taken over all particles that are within mesh element j.

If instead the Particle release specification is Specify current, each model particle
represents a number of particles per unit time which follow along the same path,
determined by the effective frequency of release frel. The space charge within the mesh
element can then be expressed as the solution to the first-order equation

Nj
dρ 1
---------j = -----
dt Vj  frel, i qi
i=1

274 | CHAPTER 6: MULTIPHYSICS INTERFACES

Stabilization of the Space Charge Density Calculation
When the Use cumulative space charge density check box is cleared, the contribution of
the charged particles to the space charge density in the surrounding domain is
overwritten at each iteration of the Bidirectionally Coupled Particle Tracing study step.
If this check box is selected, then instead the space charge density contribution is
computed as a cumulative average over successive iterations of the solver sequence.
This often leads to more robust and consistent statistical convergence of the
bidirectionally coupled model.

The process of attaining statistical convergence of a bidirectionally coupled space

charge model can be separated into two steps: the ramping-up step and the cumulative
averaging step.

RAMPING-UP STEP
While the number of iterations taken by the solver sequence is less than or equal to the
specified Number of iterations β, then the contribution of the particles to the space
charge density is updated at each iteration by scaling the newly computed charge
density contribution by a factor less then 1. This reduces the probability that the
charge density will be overestimated. During the ramping-up iterations, the new value
of the cumulative space charge density ρ s is

iter
ρ s, prev + ---------- ρ s
β
ρ s = ------------------------------------------
2

where

• ρ s, prev (SI unit: C/m3) is the stored value of the cumulative space charge density
from the previous iteration,
• ρ s (SI unit: C/m3) is the contribution of the particles to the space charge density
in the current iteration,
• iter (dimensionless) is the iteration number, and
• β (dimensionless) is the maximum number of ramping-up iterations.

CUMULATIVE AVERAGING STEP

After the ramping-up iterations are complete (iter > β), the cumulative space charge
density at each subsequent iteration is

THEORY FOR THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 275

 –1
 i   i – 1  
ρ s =   w j 
 w ρ
, + w ρ 
 j s prev i s
j = 1  j = 1  

i = iter – β + 1

where wj is the weight of each iteration. The weights are determined by the option
selected from the Weights for subsequent iterations list:

• For Uniform wj = 1.
• For Arithmetic sequence wj = j.
• For Geometric sequence wj = rj for a user-defined Common ratio r.

The benefit of using an Arithmetic sequence or Geometric sequence is that the early
iterations have a reduced impact on the solution. Therefore, if the first few iterations
have very large relative error, this early error will attenuate more quickly than when
using Uniform weighting.

Theory for the Space Charge Limited Emission Node

The space charge density of an electron beam can generate significant electric forces
that slow or deflect the beam electrons. The self-repulsion of electrons in the beam
results in a maximum allowed value for the current that is emitted from the surface.
For example, the space charge limited current J in a plane-parallel vacuum diode is
given by Child’s Law,

4ε 0 2e V 3 / 2
J = --------- ------
- ------------
9 me d2

where

• ε0 = 8.854187817 × 10-12 F/m is the permittivity of vacuum,

• e = 1.602176565 × 10-19 C is the elementary charge,
• me = 9.10938291 × 10-31 kg is the permittivity of vacuum,
• V (SI unit: V) is the potential difference across the diode, and
• d (SI unit: m) is the gap thickness.

The space charge limited current may vary depending on the shape of the geometry,
but the underlying principle remains the same: excessive charge accumulation in the

276 | CHAPTER 6: MULTIPHYSICS INTERFACES

domain surrounding the emitting surface will repel the released electrons and prevent
them from propagating.

EXTRA DIMENSION DEFINITION

The assumption that electron current is space charge limited at a cathode surface
imposes the condition

n⋅D = 0

at the boundary. The space charge limited current that satisfies this condition is often
not known in advance but can be conveniently computed by creating a fictitious
surface, called the emission surface, to release particles a short distance away from the
cathode. The particle velocity at the emission surface is nonzero and can be computed
by solving Poisson’s Equation in the thin region between the emission surface and the
cathode. To avoid the explicit creation of geometric entities to represent this narrow
buffer region, the gap between the emission surface and the cathode is treated as a 1D
extra dimension of length σs (SI unit: m). The electric potential Vxd in the product
space of this 1D extra dimension and the selection of the Space Charge Limited
Emission node is then computed using Poisson’s Equation:

∂ 2 V xd ρ
---------------- = – -----
∂x
2 ε 0

where x is the position in the extra dimension, that is, the distance from the cathode.
If the current density in the gap is J0, the charge density can be denoted

J0
ρ = ------
v

where v is the particle speed,

2eV xd
v = ----------------
me

Thus the electric potential can be obtained by solving the equation

∂ 2 V xd J0
---------------- = – -------------------------------------
-
2
∂x ε 0 2eV xd ⁄ m e

for which the following boundary conditions are enforced. At the cathode (x=0),

THEORY FOR THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 277

 V xd = 0

– n ⋅ ε 0 ∇V xd = 0

At the emission surface (x=σs),

V xd = V

Where V is the electric potential that is computed in the base geometry. Although J0
is currently unknown, it can act as a Lagrange multiplier during the computation of
Vxd, which would otherwise be overconstrained.

STABILIZATION OF THE SPACE CHARGE LIMITED CURRENT

CALCULATION
If the magnitude of the gradient of the electric potential V at the emission surface
becomes too large, it is possible for the electric potential in the extra dimension Vxd to
become negative at some point. This causes the expression for the particle velocity in
the extra dimension to become complex. To prevent this occurrence, the effect of the
space charge density on the electric potential V in the base geometry should not all be
applied at once, but should instead be increased gradually, to allow the space charge
limited current to increase in a stable manner.

The options to gradually increase the contribution to the space charge density are
shown in the Continuation Settings section of the settings window for the Electric
Particle Field Interaction node. The corresponding theory is outlined in the
Stabilization of the Space Charge Density Calculation section.

ADDITIONAL SETTINGS FOR SPACE CHARGE LIMITED EMISSION

The Extra Dimension Settings section in the settings window for the Space Charge
Limited Emission node can be used to improve the accuracy and stability of the space
charge limited current calculation.

The Position offset σs is the width of the extra dimension in which Vxd is computed.
As σs becomes smaller relative to the size of the rest of the geometry, the aspect ratio
of the extra dimension increases and the accuracy of the current calculation increases
while the stability of the calculation is reduced. In general, σs should be small relative
to the geometry size and to the radius of curvature of any curved boundaries in the
selection of the Space Charge Limited Emission node.

278 | CHAPTER 6: MULTIPHYSICS INTERFACES

Use the Initial value of electric potential in extra dimension setting to define the initial
value of Vxd. If Automatic is selected, the initial value increases linearly from 0 at the
cathode to 1 V at the emission surface.

Use the Mesh settings to control the mesh in the extra dimension. The Automatic
settings result in a mesh that consists of 50 elements which increase according to an
arithmetic sequence with an element ratio of 10. As a result, the default mesh is very
fine close to the cathode and coarser at the emission surface.

To maximize the stability of the space charge limited current calculation, the following
should typically be used:

• The Number of iterations β must be less than the Number of iterations specified in the
Bidirectionally Coupled Particle Tracing study step; otherwise, the effect of the
space charge density will not be fully taken into account.
• As the thickness of the extra dimension σs is reduced, it is often necessary to increase
the Number of iterations β.
• The domain adjacent to the selection for the Space Charge Limited Emission node
should be meshed using a boundary layer mesh.

THEORY FOR THE PARTICLE FIELD INTERACTION, NON-RELATIVISTIC INTERFACE | 279

 T he P a r ti c le F i el d In t erac t i on ,
Relativistic Interface
The Particle Field Interaction, Relativistic ( ) multiphysics interface combines the
Charged Particle Tracing, Electrostatics, and Magnetic Fields physics interfaces. The
Electric Particle Field Interaction and Magnetic Particle Field Interaction multiphysics
coupling features are added automatically. The Particle Field Interaction, Relativistic
interface is used to model beams of relativistic charged particles. The particles generate
space charge density and current density terms as they propagate through domains.
The space charge density and current density are then used to compute electric and
magnetic forces, respectively, which are exerted on the particles. This physics interface
requires both the AC/DC Module and the Particle Tracing Module.

When a predefined Particle Field Interaction, Relativistic interface is added from the
AC/DC ( )>Particle Tracing branch ( ) of the Model Wizard or Add Physics windows,
Electrostatics, Magnetic Fields, and Charged Particle Tracing interfaces are added to the
Model Builder. A Multiphysics node is also added, which automatically includes the
multiphysics coupling features Electric Particle Field Interaction and Magnetic Particle
Field Interaction.

On the Constituent Physics Interfaces

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.

The Magnetic Fields interface is used to compute magnetic field and induced current
distributions in and around coils, conductors, and magnets. Depending on the licensed
products, stationary, frequency-domain, small-signal analysis and time-domain
modeling are supported in 2D and 3D. Note that the frequency- and time-domain
formulations become ill-posed when approaching the static limit. You may extend the
useful frequency range downward by adding a low conductivity. The physics interface
solves Maxwell’s equations formulated using the magnetic vector potential and,
optionally for coils, the scalar electric potential as the dependent variables.

280 | CHAPTER 6: MULTIPHYSICS INTERFACES

The Charged Particle Tracing interface is used to model charged particle orbits under
the influence of electromagnetic forces. In addition, it can also model two-way
coupling between the particles and fields. Some typical applications are particle
accelerators, vacuum tubes, and ion implanters. The physics interface supports
time-domain modeling only in 2D and 3D. The physics interface solves the equation
of motion for charged particles subjected to electromagnetic forces.

SETTINGS FOR PHYSICS INTERFACES AND COUPLING FEATURES

When physics interfaces are added using the predefined couplings, for example Particle
Field Interaction, Relativistic, specific settings are included with the physics interfaces
and the coupling features. However, if physics interfaces are added one at a time,
followed by the coupling features, these modified settings are not automatically
included.

For example, if single Electrostatics, Magnetic Field, and Charged Particle Tracing
interfaces are added, COMSOL Multiphysics adds an empty Multiphysics node. You
can choose from the available coupling features but the modified settings are not
included.

Coupling features are available from the context menu (right-click the
Multiphysics node) or from the Physics toolbar, Multiphysics menu.

TABLE 6-2: MODIFIED SETTINGS FOR A PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE

PHYSICS INTERFACE OR MODIFIED SETTINGS (IF ANY)

COUPLING FEATURE

Electrostatics No changes.
Magnetic Fields No changes.
Charged Particle Tracing For the Charged Particle Tracing interface, under
Particle Release and Propagation, the Particle release
specification is set to Specify current, and the
Relativistic correction check box is selected.
The Electric Force and Magnetic Force nodes are added
to the Charged Particle Tracing interface. The Domain
Selection for each node is the same as that of the
Charged Particle Tracing interface.

THE PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE | 281

 TABLE 6-2: MODIFIED SETTINGS FOR A PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE

PHYSICS INTERFACE OR MODIFIED SETTINGS (IF ANY)

COUPLING FEATURE

Electric Particle Field The Domain Selection is the same as that of the
Interaction participating physics interfaces.
The corresponding Electrostatics and Charged Particle
Tracing interfaces are preselected in the Electric
Particle Field Interaction section.
Magnetic Particle Field The Domain Selection is the same as that of the
Interaction participating physics interfaces.
The corresponding Magnetic Fields and Charged
Particle Tracing interfaces are preselected in the
Electric Particle Field Interaction section.

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 Magnetic Particle Field Interaction coupling is described in this section. The
Electric Particle Field Interaction and Space Charge Limited Emission coupling
feature nodes are described for The Particle Field Interaction, Non-Relativistic
Interface.

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 Charged Particle Tracing Interface are listed
in the section Domain, Boundary, Pair, and Global Nodes for the Particle Tracing
for Fluid Flow Interface.

282 | CHAPTER 6: MULTIPHYSICS INTERFACES

• The available physics features for The Electromagnetics Interfaces are listed in the
section Domain, Boundary, Edge, Point, and Pair Nodes for the Electrostatics
Interface in the COMSOL Multiphysics Reference Manual.
• The available physics features for The Magnetic Fields Interface are listed in the
section Domain, Boundary, Point, and Pair Nodes for the Magnetic Fields Interface
in the COMSOL Multiphysics Reference Manual.

This physics interface requires the addition of the AC/DC Module. This
means that additional feature nodes are available for the Electrostatics and
Magnetic Fields interfaces. These are described in the AC/DC Module
User’s Guide and the information is most easily available from the online
Help found when working in COMSOL Multiphysics.

Magnetic Particle Field Interaction

The Magnetic Particle Field Interaction multiphysics coupling ( ) computes the
current density due to the motion of charged particles and assigns the current density
components to a set of dependent variables that are defined on the domain mesh
elements that contain the particles. It also applies the accumulated current density as a
source when computing the magnetic vector potential.

The computation of the current density is controlled by the Particle release specification
list in the settings window for the Charged Particle Tracing interface. If Specify release
times is selected, each particle contributes to the current density based on its
instantaneous location. If Specify current is selected, each model particle is treated as
representing a number of charged particles per unit time, leaving behind a
contribution to the current density in mesh elements it has previously passed through.

SETTINGS
The Label is the default multiphysics coupling name.

The Name is used primarily as a scope prefix for variables defined by the coupling node.
Refer to such 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 in the model) is mpfi1.

THE PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE | 283

 MAGNETIC PARTICLE FIELD INTERACTION
This section defines the physics involved in the multiphysics coupling. By default, the
applicable physics interface is selected in the Source and Destination lists.

You can also select None from either list to uncouple the node from a physics interface.
If the physics interface is removed from the Model Builder then the applicable list
defaults to None as there is nothing to couple to.

If a physics interface is deleted and then added to the model again, and in
order to re-establish the coupling, you need to choose the physics
interface again from the Source or Destination lists. This is applicable to all
multiphysics coupling nodes that would normally default to the once
present physics interface. See Multiphysics Modeling Approaches in the
COMSOL Multiphysics Reference Manual.

284 | CHAPTER 6: MULTIPHYSICS INTERFACES

Theory for the Magnetic Particle
Field Interaction, Relativistic
Interface
The Particle Field Interaction, Relativistic Interface combines the Charged Particle
Tracing, Electrostatics, and Magnetic Fields interfaces to model relativistic beams of
charged particles that can create significant space charge density and current density
distributions in the domains that contain the particles. The space charge density may,
in turn, exert a significant electric force on the particles, whereas the current density
contributes to the magnetic field in the surrounding domains, exerting a magnetic
force on the particles.

If all charged particle beams are released at constant current, it is possible to

significantly reduce the simulation time and computational cost by combining a
time-domain calculation of the particle trajectories with a Stationary solver for the
calculation of the electric potential and magnetic vector potential. The two calculations
can then be performed using an iterative procedure that alternates between them until
a self-consistent solution is attained.

An iterative solver loop that consists of a time-dependent solver for computing particle
trajectories and a stationary solver for computing all other dependent variables can be
set up automatically using the Bidirectionally Coupled Particle Tracing study step,
described in the COMSOL Multiphysics Reference Manual.

Electrostatics and Charged Particle Tracing Equations

Under static conditions, Gauss’ law can be written as a variant of Poisson’s equation

– ∇ ⋅ ( ε 0 ∇V – P ) = ρ

where

• ε0 = 8.854187817 × 10-12 F/m is the permittivity of vacuum,

• V (SI unit: V) is the electric potential,
• P (SI unit: C/m2) is the polarization, and
• ρ (SI unit: C/m3) is the charge density.

THEORY FOR THE MAGNETIC PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE | 285
 The electric field E (SI unit: m/s) can be expressed in terms of the electric potential V
(SI unit: V) using the relationship

E = – ∇V

The equation for the magnetic vector potential A (SI unit: Wb/m) may be written as

∂A
σ ------- + ∇ × H = J e
∂t

B = ∇×A

where σ (SI unit: S/m) is the electrical conductivity, H (SI unit: A/m) is the magnetic
field, B (SI unit: T) is the magnetic flux density, and Je (SI unit: A/m2) is the external
current density. The constitutive relationship between B and H can be written

B = μ0 ( H + M )

where μ0 (SI unit: H/m) is the permeability of free space and M (SI unit: A/m) is the
magnetization.

The equation of motion for a charged particle in an electromagnetic field can be

written as

d-
----- ( m v ) = – ZeE + Ze ( v × B )
dt p

where

• mp (SI unit: kg) is the particle mass,

• v (SI unit: m/s) is the particle velocity,
• e = 1.602176565 × 10-19 C is the elementary charge,
• Z (dimensionless) is the particle charge number, and
• B (SI unit: T) is the magnetic flux density.

For relativistic particles, the particle mass is expressed in terms of the rest mass mr
(SI unit: kg):

mr
m p = ----------------------------------
1 – v ⋅ v ⁄ c2

where c = 2.99792458 × 108 m/s is the speed of light in a vacuum. For non-relativistic
particles, the particle mass can be replaced by the rest mass and the self-imposed

286 | CHAPTER 6: MULTIPHYSICS INTERFACES

magnetic force of the beam is often negligibly small compared to the electric force,
allowing The Particle Field Interaction, Non-Relativistic Interface to be used
effectively.

The calculation of the space charge density term and the resulting effect on the electric
force is explained in Space Charge Density Calculation in Theory for the Particle Field
Interaction, Non-Relativistic Interface. The contribution of particle motion to the
magnetic force is considered in the following section.

Current Density Calculation

Given an array of idealized point sources such that the position vector of the ith source
is denoted qi (SI unit: m), the contribution to the current density by particles js at
position r is

js ( r ) =  Zi evi δ ( r – qi ) (6-3)
i=1

where

• δ is the Dirac delta function,

• Zi (dimensionless) is the charge number of the ith particle,
• e = 1.602176565 × 10-19 C is the elementary charge,
• vi (SI unit: m/s) is the velocity of the ith particle, and
• N (dimensionless) is the total number of particles.

The equation for the current density is unusable in this form, however, because the
number of particles involved may be extraordinarily large, and because the transfer of
information from the point particles to degrees of freedom defined on a finite element
mesh introduces some discretization error. In the following sections we discuss
solutions to these problems.

MODELING A REPRESENTATIVE SAMPLE OF PAR TICLES

Because the number of real particles, such as ions or electrons, may be too large for
every particle to be modeled individually, a practical numerical approach is to release a
representative sample of model particles, allowing each model particle to make the
same contribution to the current density as an equivalent number of real particles.

THEORY FOR THE MAGNETIC PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE | 287
 For example, instead of allocating degrees of freedom for 1012 electrons, it will often
suffice to model 104 particles, each of which has a Charge multiplication factor of 108,
meaning that it represents 108 electrons.

Simplification for Constant-Current Beams

If a beam of particles is released at constant current, then a full time-domain calculation
of the coupled particle trajectories and electric and magnetic fields may require
particles to be released at a large number of time steps until a stationary solution for
the electric potential and magnetic vector potential is reached. This can be needlessly
memory-intensive and time-consuming. An alternate approach is to release particles at
time t = 0 and to allow each model particle to represent a continuous stream of real
particles per unit time. The number of real particles per unit time represented by each
model particle is denoted the effective frequency of release, frel.

The charged particles will contribute to the current density along their entire
trajectories, not just at their instantaneous positions. This behavior can be conveniently
reproduced by defining an expression for the time derivative of the current density,
rather than the current density itself:

N
dj s ( r )
---------------
dt
- =
 frel, i qi vi δ ( r – qi ) (6-4)
i=1

The current density can then be computed by integrating over time, as long as
sufficient time is given so that the particle trajectories can be traced completely
through the modeling domain.

The frequency of release can be computed using the current and number of model
particles that are specified in release feature settings. For example, for an Inlet node
with release current magnitude I (SI unit: A) and number of particles per release N
(dimensionless), the effective frequency of release is

I
f rel = -----------
qN

When particle beams are assumed to have constant current, then the current density at
the last time step includes contributions from particles at every point along their
trajectories in the modeling domain. Thus, it can be applied as the current density term
when computing the magnetic vector potential.

The treatment of particle beams as constant-current beams is determined by the

Particle release specification list in the settings window for the Charged Particle Tracing

288 | CHAPTER 6: MULTIPHYSICS INTERFACES

physics interface. If Specify release times is selected, the charge density is computed
using Equation 6-3 and is determined by the instantaneous positions of all model
particles. Thus, it is necessary to solve for the particle trajectories, electric potential,
and magnetic vector potential in the time domain. If Specify current is selected, the
current density is computed using Equation 6-4 and is determined by the time history
of the model particle positions.

The difference between the Specify current and Specify release times particle release
specifications is thus analogous to the difference between integration over Elements
and time and integration over Elements as described for the Accumulator (Domain)
node.

At this point, the effect of a bidirectional coupling between the particle trajectories and
fields has not been considered. If Specify release times is selected from the Particle
release specification list, this does not require special consideration because the
trajectories and fields are computed simultaneously. If Specify current is selected,
however, the trajectories and fields are computed using different study types, and an
additional feedback mechanism is needed. The Bidirectionally Coupled Particle
Tracing study step generates a solver sequence that does the following:

1 Set the current density contribution due to the particles to zero.

2 Compute the magnetic vector potential and other field variables using a Stationary
solver, using the value of the current density computed in the previous step.
3 Compute the particle trajectories and the resulting current density in the time
domain, using the field variables computed in the previous step.
4 Repeat steps 2 and 3 until a specified number of iterations has been reached.

Given a sufficient number of iterations, the resulting solution will fully account for the
bidirectional coupling between the particle trajectories and stationary fields.

AVOIDING INF IN IT EL Y LA RGE VALUES OF TH E CURREN T DENSIT Y

The Magnetic Particle Field Interaction node defines a variable for each component of
the contribution to the current density by particles in each mesh element. This variable
is discretized using constant shape functions. For a mesh element j with volume Vj,
and with the Particle release specification set to Specify release times, the average current
density ρj is

N
1
j s, j = -----
Vj  n i q i v i  δ ( r – q i ) dV
i=1

THEORY FOR THE MAGNETIC PARTICLE FIELD INTERACTION, RELATIVISTIC INTERFACE | 289
 where ni (dimensionless) is the charge multiplication factor of the ith model particle.
The integral on the right-hand side is a volume integral over element j. The resulting
current density is the average current density over the mesh element, which may be
written as

Nj
1
j s, j = -----
Vj  ni qi vi
i=1

where the sum is taken over all particles that are within mesh element j.

If instead the Particle release specification is Specify current, each model particle
represents a number of particles per unit time which follow along the same path,
determined by the effective frequency of release frel. Then the time derivative of the
current density can be expressed as

Nj
dj s, j 1
-----------
dt
- = -----
Vj  frel, i qi vi
i=1

290 | CHAPTER 6: MULTIPHYSICS INTERFACES

T he Flui d- P a r ti cl e In t eract i on
Interface
The Fluid-Particle Interaction ( ) multiphysics interface combines the Particle
Tracing for Fluid Flow interface coupled with the Laminar Flow interface. The Fluid
Particle Interaction multiphysics coupling feature is added automatically. The Fluid
Particle Interaction interface is used to model the motion of particles in a fluid. As the
particles are accelerated or decelerated by the drag force exerted by the fluid, the
corresponding reaction force is applied to the fluid.

When a predefined Fluid-Particle Interaction interface is added from the Fluid

Flow ( )>Particle Tracing branch ( ) of the Model Wizard or Add Physics windows,
Laminar Flow and Particle Tracing for Fluid Flow interfaces are added to the Model
Builder. A Multiphysics node is also added, which automatically includes the
multiphysics coupling feature Fluid-Particle Interaction.

On the Constituent Physics Interfaces

The equations solved by the Laminar Flow interface are the Navier-Stokes equations
for conservation of momentum and the continuity equation for conservation of mass.
A Fluid Model is active by default on all the interface selection. The flow interface
domain selection may be edited if the model contains solid domains.

The Particle Tracing for Fluid Flow interface is used to compute the motion of
particles in a background fluid. Particle motion can be driven by drag, gravity, and
electric, magnetic, and acoustophoretic forces. User-defined forces can be added. It is
also possible to compute the particle mass and temperature.

SETTINGS FOR PHYSICS INTERFACES AND COUPLING FEATURES

When physics interfaces are added using the predefined couplings specific settings are
included with the physics interfaces and the coupling features. However, if physics
interfaces are added one at a time, followed by the coupling features, these modified
settings are not automatically included.

For example, if single Laminar Flow and Particle Tracing for Fluid Flow interfaces are
added, COMSOL adds an empty Multiphysics node. You can choose from the available
coupling features but the modified settings are not included.

THE FLUID-PARTICLE INTERACTION INTERFACE | 291

 Coupling features are available from the context menu (right-click the
Multiphysics node) or from the Physics toolbar, Multiphysics menu.

TABLE 6-3: MODIFIED SETTINGS FOR A FLUID-PARTICLE INTERACTION INTERFACE

PHYSICS INTERFACE OR MODIFIED SETTINGS (IF ANY)

COUPLING FEATURE

Laminar Flow No changes.

Particle Tracing for Fluid Flow For the Particle Tracing for Fluid Flow interface, under
Particle Release and Propagation, the Particle release
specification is set to Specify mass flow rate.
The Drag Force node is added to the Particle Tracing
for Fluid Flow interface. The Domain Selection is the
same as that of the Particle Tracing for Fluid Flow
interface.
Fluid-Particle Interaction The Domain Selection is the same as that of the
participating physics interfaces.
The corresponding Laminar Flow and Particle Tracing
for Fluid Flow interfaces are preselected in the
Fluid-Particle Interaction section.

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 Fluid-Particle Interaction coupling feature node is described in this 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.

292 | CHAPTER 6: MULTIPHYSICS INTERFACES

• The available physics features for The Particle Tracing for Fluid Flow Interface are
listed in the section Domain, Boundary, Pair, and Global Nodes for the Particle
Tracing for Fluid Flow Interface.
• The available physics features for The Single-Phase Flow, Laminar Flow Interface
are listed in the section Domain, Boundary, Pair, and Point Nodes for Single-Phase
Flow in the COMSOL Multiphysics Reference Manual.

Fluid-Particle Interaction
The Fluid-Particle Interaction multiphysics coupling ( ) computes a volume force
that is equal in magnitude and opposite in direction to the total drag force exerted on
particles in each mesh element in the selected domains. This volume force contributes
to the total force acting on the fluid in the Laminar Flow interface.

SETTINGS
The Label is the default multiphysics coupling name.

The Name is used primarily as a scope prefix for variables defined by the coupling node.
Refer to such 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 in the model) is fpi1.

FORCE MULTIPLICATION FACTOR

Use this section to specify a proportionality factor to multiply by the change in particle
momentum when computing the volume force.

Select an option from the Force multiplication factor specification list — From physics
(the default) or User defined.

Is Specify mass flow rate is selected from the Particle Properties section in the Settings
window for The Particle Tracing for Fluid Flow Interface, this section has no effect
because the proportionality factor is based on the mass flow rate of particles, which is
specified in the particle release features. Otherwise, the following conditions apply:

• If From physics is selected and the Enable macroparticles check box is cleared in the
physics interface Additional Variables section, the multiplication factor is 1.
• If From physics is selected and the Enable macroparticles check box is selected in the
physics interface Additional Variables section, the multiplication factor is based on the

THE FLUID-PARTICLE INTERACTION INTERFACE | 293

 auxiliary dependent variable for the multiplication factor, which is typically specified
in release feature settings. If the physics interface has name fpt, this variable has
name fpt.nn.
• If User defined is selected, enter a value or expression for the Force multiplication
factor n (dimensionless). The default is 1.

COUPLED INTERFACES
This section defines the physics involved in the multiphysics coupling. By default, the
applicable physics interface is selected in the Source and Destination lists.

You can also select None from either list to uncouple the node from a physics interface.
If the physics interface is removed from the Model Builder then the applicable list
defaults to None as there is nothing to couple to.

If a physics interface is deleted and then added to the model again, and in
order to re-establish the coupling, you need to choose the physics
interface again from the Source or Destination lists. This is applicable to all
multiphysics coupling nodes that would normally default to the once
present physics interface. See Multiphysics Modeling Approaches in the
COMSOL Multiphysics Reference Manual.

294 | CHAPTER 6: MULTIPHYSICS INTERFACES

Theory for the Fluid-Particle
I nte r a c t i o n Interfac e
The Fluid-Particle Interaction Interface combines the Particle Tracing for Fluid Flow
and Laminar Flow interfaces to model the motion of particles in a fluid in which the
acceleration or deceleration of particles creates a significant volume force that affects
the motion of the fluid. The Fluid-Particle Interaction node computes a volume force
that is equal in magnitude and opposite in direction to the total drag force that the
fluid exerts on particles.

If particles are released into the fluid at a constant mass flow rate, it is possible to
significantly reduce the simulation time and computational cost by combining a
time-domain calculation of the particle trajectories with a Stationary solver for the
calculation of the pressure and fluid velocity. The two calculations can then be
performed using an iterative procedure that alternates between them until a
self-consistent solution is attained.

An iterative solver loop that consists of a time-dependent solver for computing particle
trajectories and a stationary solver for computing all other dependent variables can be
set up automatically using the Bidirectionally Coupled Particle Tracing study step,
described in the COMSOL Multiphysics Reference Manual.

Laminar Flow and Particle Tracing Equations

For an incompressible single-phase fluid in the laminar flow regime, the Navier-Stokes
equations can be reduced to the following:

∂u T
ρ ------- + ρ ( u ⋅ ∇ )u = ∇ ⋅ [ – pI + μ ( ∇u + ( ∇u ) ) ] + F
∂t

ρ∇ ⋅ u = 0

where

• u (SI unit: m/s) is the fluid velocity,

• p (SI unit: Pa) is the pressure,
• ρ (SI unit: kg/m3) is the density,

THEORY FOR THE FLUID-PARTICLE INTERACTION INTERFACE | 295

 • μ (SI unit: Pa·s) is the dynamic viscosity, and
• F (SI unit: N) is the total volume force.

In the following, let FV denote the contribution of particle motion to the total volume
force acting on the fluid.

The equation of motion of particles in the fluid can be written as

d-
----- ( m v ) = F D + F g + F ext
dt p

where

• mp is the particle mass (SI unit: kg)

• v is the velocity of the particle (SI unit: m/s)
• FD is the drag force (SI unit: N),
• Fg is the gravitational force (SI unit: N), and
• Fext is any other external force (SI unit: N).

Several different expressions for the drag force are available; these are described in the
section Particle Motion in a Fluid in Theory for the Particle Tracing for Fluid Flow
Interface.

Volume Force Calculation

Given an array of idealized point masses such that the position vector of the ith particle
is denoted qi (SI unit: m), the volume force at position r is

FV ( r ) = –  F D, i δ ( r – q i ) (6-5)
i=1

where δ is the Dirac delta function, FD,i is the drag force exerted on the ith particle,
and N is the total number of particles. The equation for the volume force is unusable
in this form, however, for the following reasons:

• The number of particles may be very large, making it impractical to model all of
them.
• The magnitude of the volume force becomes infinite at the location of each
idealized point mass, making the calculation of the volume force infeasible with
most numerical methods.

296 | CHAPTER 6: MULTIPHYSICS INTERFACES

In the following sections we discuss solutions to each of these problems.

MODELING A REPRESENTATIVE SAMPLE OF PAR TICLES

Because the number of real particles may be too large for every particle to be modeled
individually, a practical numerical approach is to release a representative sample of
model particles, allowing each model particle to make the same contribution to the
volume force as an equivalent number of real particles.

For example, instead of allocating degrees of freedom for 107 small particles, it will
often suffice to model 104 particles, each of which has a Force multiplication factor of
103, meaning that it exerts a volume force that is 103 times greater in magnitude than
the drag force that acts on it.

Simplification for Constant Mass Flow Rate

If particles are released into the fluid at a constant mass flow rate, then a full
time-domain calculation of the coupled particle trajectories and field variables may
require particles to be released at a large number of time steps until a stationary
solution is reached. The calculation of the fluid velocity and pressure at each time step
can be needlessly memory-intensive and time-consuming. An alternate approach is to
release particles at time t = 0 and to allow each model particle to represent a
continuous stream of real particles per unit time. The number of real particles per unit
time represented by each model particle is denoted the effective frequency of release,
frel.

The behavior of a continuous stream of particles can be conveniently modeled by

defining an expression for the time derivative of the volume force, rather than the
volume force itself:

N
dF V ( r )
-------------------- = –
dt  frel, i FD, i δ ( r – qi ) (6-6)
i=1

The charge density can then be computed by integrating over time, as long as sufficient
time is given so that the particle trajectories can be traced completely through the
modeling domain.

The frequency of release can be computed using the current and number of model
particles that are specified in release feature settings. For example, for an Inlet node
·
with release current magnitude m (SI unit: kg/s) and number of particles per release
N (dimensionless), the effective frequency of release is

THEORY FOR THE FLUID-PARTICLE INTERACTION INTERFACE | 297

 ·
m
f rel = -------------
mp N

When the mass flow rate can be assumed to be constant, then the volume force at the
last time step includes contributions from particles at every point along their
trajectories in the modeling domain. Thus, it can be applied as the volume force term
when computing the fluid pressure and velocity.

The treatment of the constant mass flow rate is determined by the Particle release
specification list in the settings window for the Particle Tracing for Fluid Flow interface.
If Specify release times is selected, the volume force is computed using Equation 6-5
and is determined by the instantaneous positions of all model particles. Thus, it is
necessary to solve for the particle trajectories, fluid velocity, and pressure in the time
domain. If Specify mass flow rate is selected, the volume force is computed using
Equation 6-6 and is determined by the time history of the model particle positions.

The difference between the Specify mass flow rate and Specify release times option in
the Particle release specification list is thus analogous to the difference between
integration over Elements and time and integration over Elements as described for the
Accumulator (Domain) node.

At this point, the effect of a bidirectional coupling between the particle trajectories and
fields has not been considered. For the Specify release times option, this does not
require special consideration because the trajectories and fields are computed
simultaneously. For the Specify mass flow rate option, however, the trajectories and
fields are computed using different study types, and an additional feedback mechanism
is needed. The Bidirectionally Coupled Particle Tracing study step can be used to
generate a solver sequence that does the following:

1 Set the volume force exerted by the particles on the fluid to zero.
2 Compute the fluid velocity and pressure using a Stationary solver, using the value of
the volume force computed in the previous step.
3 Compute the particle trajectories and the resulting volume force in the time
domain, using the field variables computed in the previous step.
4 Repeat steps 2 and 3 until a specified number of iterations has been reached.

If the number of iterations taken by the solver sequence is sufficiently large, the
resulting solution will fully account for the bidirectional coupling between the particle
trajectories and stationary fields.

298 | CHAPTER 6: MULTIPHYSICS INTERFACES

A VO I D I N G I N F I N I T E L Y L A R G E VA L U E S O F T H E VO L U M E F O R C E
The Fluid-Particle Interaction node defines variables for each component of the
volume force exerted by particles on the surrounding fluid. These variables are
discretized using constant shape functions that are, in general, discontinuous across
boundaries between elements. For a mesh element j with volume Vj, and with the
Particle release specification set to Specify release times, the average volume force FV,j is

N
1
F V, j = – -----
Vj  ni FD, i  δ ( r – qi ) dV
i=1

where ni is the force multiplication factor of the ith model particle. The integral on the
right-hand side is a volume integral over element j. The resulting volume force is the
average volume force over the mesh element, which may be written more concisely as

Nj
1
F V, j = – -----
Vj  ni FD, i
i=1

where the sum is taken over all particles that are within mesh element j.

If instead the Particle release specification is Specify mass flow rate, each model particle
represents a number of particles per unit time which follow along the same path,
determined by the effective frequency of release frel. The volume force within the mesh
element can then be expressed as the solution to the first-order equation

Nj
dF V, j 1
--------------- = – -----
dt Vj  frel, i FD, i
i=1

THEORY FOR THE FLUID-PARTICLE INTERACTION INTERFACE | 299

300 | CHAPTER 6: MULTIPHYSICS INTERFACES
 7

Glossary

This Glossary of Terms contains modeling terms in a particle tracing context. 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.

301
 Glossary of Terms
Brownian motion The random drifting of particles suspended in a fluid.

Coulomb force The force between charged particles which is inversely proportional
to the square of the distance between the particles.

dispersed flow A fluid-particle system where particle-fluid and particle-particle

interactions need to be accounted for in the model.

Hamiltonian A convenient way of describing how a system of particles interact with

surrounding fields. The Hamiltonian is usually defined as the sum of the kinetic and
potential energy.

Lagrangian A convenient way of describing how a system of particles interact with

surrounding fields. The Lagrangian is usually defined as the kinetic energy minus the
potential energy.

Maxwellian velocity distribution Describes the probability that a velocity is near a given
value as a function of the temperature of the system.

residence time The average amount of time that a particle spends in a particular
system. The residence time can be computed by adding Auxiliary dependent variables.

space charge effects When the number density of charged particles is sufficiently high,
they can affect the field in which they are placed. This is often referred to as a “space
charge effect.”

sparse flow A particle laden flow is described as a sparse flow when the particles have
no appreciable effect on the motion of the fluid.

transmission probability The probability that a particle transmits from a given

selected boundary or domain to another.

302 | CHAPTER 7: GLOSSARY

I n d e x
A accumulator (node), boundaries 71 C centrifugal force 80, 110
accumulator (node), domains 82 charged particle tracing 21
accumulator theory, boundaries 120 charged particle tracing interface 124
accumulator theory, domains 119 theory 162
accuracy order, of time stepping 64 collision diameter 85
acoustophoretic force (node) 203 collisions (node) 128
acoustophoretic force, theory 243 common settings 14
Application Libraries window 16 continuous random walk 231
application library examples convective heat losses (node) 215
Brownian force 202 Coriolis force 80, 110
charged particle tracing 126 Coulomb force 22, 113
for-end for loop 48 Coupling Operators 42
mathematical particle tracing 65 current density (node) 159
Monte Carlo modeling 45
D degrees of freedom, out-of-plane 125
newtonian, first order 33
dense flow 23
particle field interaction 22
dielectrophoretic force (node) 205
particle release feature 39
dielectrophoretic force, theory 242
particle tracing 12
diffuse scattering, wall condition 68, 117
particle tracing for fluid flow 192
dilute flow 24
particle-particle interactions 86
disappear, wall condition 68, 116
sparse flow 24, 212
discrete random walk 236
thermionic emission 153
dispersed flow 25
transmission probability 40
documentation 15
attachment (node) 132
drag force (node) 196
auxiliary dependent variable 70, 116
drag force, theory 223
auxiliary dependent variable (node) 100
droplet breakup 216, 251
auxiliary dependent variables, initializing
droplet sprays in fluid flow interface 222
50
E elastic (node) 130
azimuthal particle velocity 125
electric force (node) 154, 205
B bndenv operator 12
electric force, theory 241
bounce, wall condition 67, 117
electric particle field interaction (node),
boundary conditions
multiphysics 266
theory 116
emailing COMSOL 17
boundary load (node) 213
env operator 12
Brownian force (node) 200
erosion (node) 211
Brownian force, theory 240

INDEX| 303
 etch (node) 160 lift force, theory 237
Euler force 80, 110 London dispersion force 115
excitation (node) 132 Lorentz force 163

F filtering particles to view 53 M magnetic force (node) 156, 205

fluid flow, particle tracing for 23 magnetic force, theory 242
fluid-particle interaction (node), multi- magnetic particle field interaction
physics 293 (node), multiphysics 283
fluid-particle interaction interface 291 magnetophoretic force, theory 248
force (node) 79 mass deposition (node) 212
forces mass flux (node) 213
on particles in fluids 223 massless formulation 78
freeze, wall condition 67, 117 massless particle 109
friction force (node) 137 mathematical particle tracing interface 60
theory 106
G general reflection, wall condition 68, 118
Maxwellian initial velocity distribution
gravity force (node) 202
111
gravity force, theory 230
mixed diffuse and specular reflection,
H Hamiltonian 78
wall condition 68, 118
Hamiltonian formulation 108
Monte Carlo modeling 44
heat source (node) 160, 216
Monte Carlo simulation 201
I initial conditions, particle position 110 MPH-files 16
initial velocity of particles 111 multiphysics
initializing 50 electric particle field interaction 266
inlet (node) magnetic particle field interaction 283
particle tracing 94 space charge limited emission 268
internet resources 15 multiphysics coupling 293
ionization (node) 133
N Newton’s second law 106, 223
ionization loss (node), particle-matter in-
nodes, common settings 14
teractions 139
nonlocal accumulator (node) 98
K Kelvin-Helmholtz breakup model 217, nonresonant charge exchange (node)
252 135
knowledge base, COMSOL 17 Nozzle 221
nozzle 218, 255
L Lagrangian 78
nozzle domain 221
Lagrangian formulation 108
nuclear stopping (node), particle-matter
Lennard-Jones force 85
interactions 140
Lennard-Jones potential 114
lift force (node) 199 O operators

304 | I N D E X
 env and bndenv 12 S Saffman lift force 237
outlet (node) secondary emission (node) 73
particle tracing 99 shell (node) 207
out-of-plane degrees of freedom 125 space charge density calculation (node)
override properties (node) 103 158
space charge effects 21
P particle beam (node) 141
space charge limited emission (node),
particle continuity (node) 100
multiphysics 268
particle field interaction, non-relativistic
sparse flow 23
interface 264
standard settings 14
particle field interaction, relativistic in-
stick, wall condition 68, 117
terface 280
surface charge density (node) 159
particle index 39
symmetry (node)
particle mass 77, 153
particle tracing 154, 195
particle motion in fluids 223
particle properties (node) 77, 153, 194 T technical support, COMSOL 17
particle release time 39 theory
particle tracing for fluid flow interface charged particle tracing 162
188, 222 mathematical particle tracing 106
theory 223 particle tracing for fluid flow interface
particle velocity 78 223
particle-matter interactions (node) 138 thermionic emission (node) 151
particle-particle interaction (node) 84 thermionic emission, theory 178
particle-particle interactions 113 thermophoretic force (node) 209
pass through, wall condition 68 thermophoretic force, theory 244
Péclet number 27 total number of particles 40
periodic condition (node) 70 transmission probability 40
physics interfaces, common settings 14 U user defined (node), collisions 136
R radiative heat losses (node) 215 V velocity reinitialization (node) 81
Rayleigh-Taylor breakup model 218, 254 volume force calculation (node) 213
release (node) 86
W wall (node)
release from data file (node) 103
particle tracing 67
release from edge (node) 101
wall correction (drag force), theory 230
release from grid (node) 102
websites, COMSOL 17
release from point (node) 101
residence time 48
resonant charge exchange (node) 134
rotating frame (node) 80
rotating frame, theory 109

INDEX| 305
306 | I N D E X