Discover millions of ebooks, audiobooks, and so much more with a free trial

Only $11.99/month after trial. Cancel anytime.

Numerical Methods in Electromagnetism
Numerical Methods in Electromagnetism
Numerical Methods in Electromagnetism
Ebook999 pages5 hours

Numerical Methods in Electromagnetism

Rating: 0 out of 5 stars

()

Read preview

About this ebook

Electromagnetics is the foundation of our electric technology. It describes the fundamental principles upon which electricity is generated and used. This includes electric machines, high voltage transmission, telecommunication, radar, and recording and digital computing. Numerical Methods in Electromagnetism will serve both as an introductory text for graduate students and as a reference book for professional engineers and researchers. This book leads the uninitiated into the realm of numerical methods for solving electromagnetic field problems by examples and illustrations. Detailed descriptions of advanced techniques are also included for the benefit of working engineers and research students.
  • Comprehensive descriptions of numerical methods
  • In-depth introduction to finite differences, finite elements, and integral equations
  • Illustrations and applications of linear and nonlinear solutions for multi-dimensional analysis
  • Numerical examples to facilitate understanding of the methods
  • Appendices for quick reference of mathematical and numerical methods employed
LanguageEnglish
Release dateNov 10, 1999
ISBN9780080512891
Numerical Methods in Electromagnetism

Related to Numerical Methods in Electromagnetism

Related ebooks

Mathematics For You

View More

Related articles

Reviews for Numerical Methods in Electromagnetism

Rating: 0 out of 5 stars
0 ratings

0 ratings0 reviews

What did you think?

Tap to rate

Review must be at least 10 words

    Book preview

    Numerical Methods in Electromagnetism - Sheppard Salon

    MAYERGOYZ

    PREFACE

    Since the formulation of electromagnetic theory, researchers and engineers have sought accurate solutions of the resulting boundary and initial value problems. Initially the focus of these efforts was to find closed form and analytical solutions. These were followed by analogue methods. Progress in digital computer hardware and software have now made numerical solutions, such as finite difference, finite element and integral equation formulations popular.

    Great advances have been made in field computation for static, steady state sinusoidal, and transient problems, first in two dimensions and more recently in three dimensions. Both linear and nonlinear cases are now treated routinely. Recent work has resulted in new formulations involving mixed scalar and mixed vector variables, especially for three dimensional problems. There has also been considerable progress in the area of numerical computation resulting in fast equation solvers.

    For finite difference, finite element and integral equations, the reader is presented here with a broad picture of the development and research in the area of numerical analysis applied to problems in electromagnetics. The book is suitable for a graduate course in the subject and this material has formed the basis of a graduate course presented at Rensselaer Polytechnic Institute. A valuable feature of this book is the large number of tutorial examples that are presented. These are small problems (a few finite elements or boundary elements) that are worked out in detail. With this feature the student can follow all of the steps involved in the different methods. This also allows the book to be used as a self-study guide. This work will also be of use to researchers and code developers in this area as it includes a great deal of reference material. Engineers and designers using these methods will also benefit as we present many examples and discuss the accuracy of the methods as well as their strong and weak points.

    The chapters contain principle categories—namely, introductory, historical, and concept development; advanced computational techniques; and specific and illustrative examples. Chapter 1 reviews the basic equations of electromagnetic theory, the different equations and formulations used in the remaining chapters, and contains discussions of materials, boundary conditions, and constraints. Chapter 2 presents an historical perspective of the different approaches that have been applied to the solution of electromagnetic problems. These range from classical closed form methods to analogue methods to modern digital methods. A brief discussion of the relative merits of these methods is presented. In Chapter 3 we introduce the method of finite differences. General methods of approximating differential equations by finite differences and the accuracy of these approximations is presented. Boundary conditions and materials are discussed. In examples, the method is applied to Poisson’s equation, the diffusion equation and the wave equation. In Chapter 4 we present two mathematical techniques that are important in the understanding of the finite element method and the integral equation method. These are the variational method and the Galerkin method. A general approach is taken and several Illustrative examples are used. In Chapter 5 the theory of shape functions is presented. This is a very complete treatment and is used in the development of the finite element and boundary element methods. In Chapter 6 the finite element method is presented. Examples given to illustrate the procedure include a magnetostatic problem, a problem with nonlinear materials, a problem with permanent magnets, and a problem using the isoparametric approach. A comparison of the system of equations obtained by finite elements and finite differences is also included. In Chapter 7 we present integral equation methods. These include the method of moments, the charge simulation method and the boundary element method. Examples illustrate the application of these methods. In Chapter 8 we discuss the use of finite elements in the solution of open boundary problems. The techniques discussed include analytic coupling, ballooning, infinitesimal scaling, infinite elements and the hybrid finite element boundary element method. In Chapters 9 and 10 we present applications of the finite element method in high frequency and low frequency problems. These include examples of waveguides and scattering and the coupling of low frequency fields to mechanical, circuit, and fluid problems. In chapter 11 we discuss methods of solution for large linear systems and the different formulations of nonlinear problems that are frequently used in numerical analysis. A number of appendices include reference material that elaborates some of the material presented in the chapters. For example, the reader will find shape functions and integration formulae for many finite elements.

    The authors are indebted to many people; far too many to acknowledge here. These include our colleagues at General Electric, Westinghouse, RPI, and Magsoft Corporation. We are also indebted to many talented graduate students whose work is represented here. We would like to particularly acknowledge Mark DeBortoli, who worked out several of the examples presented, Laurent Nicolas, who provided some of the waveguide and scattering figures, and Kiruba Sivasubramanaim who contributed many of the figures and helped in many ways. We also acknowledge the late Peter Silvester and the many discussions we had during his visits to RPI. We thank our friend and colleague Isaak Mayergoyz for his support, advice, and encouragement during the writing of this book and appreciate the work of the helpful staff at Academic Press. Finally, we could not have written this book without the support and encouragement of our wives, Padma and Corine, to whom we dedicate this work.

    1

    BASIC PRINCIPLES OF ELECTROMAGNETIC FIELDS

    1.1 INTRODUCTION

    This chapter is intended to introduce and to discuss the basic principles and concepts of electromagnetic fields, which will be dealt with in detail in succeeding chapters. The concept of flux and potential is discussed and Maxwell’s set of equations is presented. The description of electric and magnetic materials then follows. We then discuss the set of differential equations that are commonly used as a starting point for the numerical methods we will study. An explanation of some of the more popular choices of state variables (scalar and vector potential or direct field variables) is then presented.

    1.2 STATIC ELECTRIC FIELDS

    The electric charge is the source of all electromagnetic phenomena. Electrostatics is the study of electric fields due to charges at rest. In SI units the electric charge has the units of coulombs after Charles Augustin de Coulomb (1736–1806). The charge of one electron is 1.602 × 10−19 coulombs. In rationalized MKS units an electric charge of Q coulombs produces an electric flux of Q Coulombs. In Figure 1.1 we see the electric charge emitting flux uniformly in the radial direction in a spherical coordinate system with origin at the charge. If we consider a closed surface around the charge (for example, the sphere shown in Figure 1.1), then all of the flux must pass through the surface. This would be true for any number of charges in the volume enclosed by the surface. Using the principle of superposition, we conclude that the total flux passing through any closed surface is equal to the total charge enclosed by that surfaceas the local value of the flux per unit area then we may write

    Figure 1.1 Point Charge and Flux

    (1.1)

    where ρ is the charge density. This is the integral form of Gauss’ law. From the symmetry of Figure 1.1 we find immediately that for a point charge

    (1.2)

    where r is the distance from the charge (source point) to the location where we compute the field quantity (field point).

    The inverse square law of equation (1.2) is one of the most important results of electromagnetism. There are two consequences of the inverse square law that we shall encounter in subsequent chapters. One is that there is a second-order singularity which occurs at r = 0. The second is that the electric flux density goes to zero as r → ∞.

    . This force is proportional to the magnitude of the charges and depends on the medium. Coulomb’s law states that for two charges Q1 and Q2 in vacuum,

    (1.3)

    . The units of electric field are volts/meter. From (1.2) and (1.3) we see that

    (1.4)

    1.3 THE ELECTRIC POTENTIAL

    directly. One of the reasons is that we need to find three components at each point and all of the algebra is vectorial. For these reasons we find it simpler to compute scalar variables and then compute the field from these values. In the case of the electric field we use the electric potential, V. Recalling that the electric field is the force on a unit charge, we define the electric potential at a point in space as the work expended in moving a unit charge from infinity to that point. Consider is E sin θ perpendicular to the field.

    Figure 1.2 Work and Electric Field

    We write

    (1.5)

    in the presence of an electric field is ΔW so that

    (1.6)

    Therefore, if we can find V, a scalar, we can compute the electric field by taking the gradient. An important property of the scalar potential is that the work done to move a charge between points P1 and P2 is independent of the path taken by the charge. Fields having this property are called conservative. This is typical of lossless systems in which we have thermodynamically reversible processes. Systems with losses do not have this property, and this will have implications later in our discussion of variational principles. A corollary of this principle is that it takes no work to move a charge around a closed path in a static electric field. This is a very important property of the potential. Without this, the potential would not be uniquely defined at a point and the path would have to be specified. We will see in the next section that the magnetostatic field with current sources does not have this property and the scalar potential is of more limited use. We have therefore

    (1.7)

    another of Maxwell’s equations.

    A few examples of the electric field will be of interest to us.

    Potential Due to a Point Charge

    The electric field due to a point charge of magnitude Q is, from (1.3) and (1.4),

    (1.8)

    We find the potential by integrating equation (1.8) to obtain

    (1.9)

    We make note of some properties that will be of interest in later chapters. First, there is a singularity at r = 0, but it is of first order and not of second order as in the case of the field. Second, the potential goes to zero as r approaches infinity.

    The Logarithmic Potential

    coulombs/meter, then the application of Gauss’ law in Figure 1.3 gives a flux density of

    Figure 1.3 Logarithmic Potential

    (1.10)

    and an electric field of

    (1.11)

    Integrating from infinity to a point P gives the potential

    (1.12)

    This logarithmic potential is frequently found in two-dimensional problems and problems in cylindrical coordinates. The potential has a singularity at r = 0 but unfortunately does not go to zero as r goes to infinity. In fact, the potential slowly approaches infinity as r increases. The potential vanishes at r = 1.

    Potential of a Dipole

    The field and potential due to an electric dipole will be important in our discussion of integral equations. Consider two charges +Q and –Q separated by a distance d. We can find the potential by superposition. We will be interested in the potential at a distance from the charges much greater than their separation. The potential due to a single point charge is

    (1.13)

    For both charges (see Figure 1.4) we have

    Figure 1.4 Dipole Field and Potential

    (1.14)

    We see that

    (1.15)

    We make the approximation that for r > > d

    (1.16)

    so

    (1.17)

    This potential drops off faster than the potential of a point charge and the singularity at r = 0 is of second order.

    Potential Due to a Ring of Charge

    [1]. Consider the ring with center at the origin of a cylindrical coordinate system illustrated in Figure 1.5.

    Figure 1.5 Potential of Ring of Charge

    The potential is

    (1.18)

    where

    (1.19)

    and

    (1.20)

    This is equivalent to

    (1.21)

    where

    (1.22)

    and

    (1.23)

    equation (1.21) is an elliptic integral of the first kind. Results for this are tabulated in many standard references [2].

    1.3.1 Potential Energy and Energy Density

    Using our definition of electric potential, the work required to move a charge of magnitude qi from infinity to a point (x, y, z) is

    (1.24)

    For an ensemble of N − 1 charges producing the potential, we have [3]

    (1.25)

    where

    . The potential energy of the charge is then

    (1.26)

    The energy in all of the charges is

    (1.27)

    The j < i is required so that we do not count the contribution of each charge twice. We can also write

    (1.28)

    It is understood that i j in the summations. We can always express the charges as a distribution¹ to obtain

    (1.29)

    We now substitute the scalar potential into equation (1.29).

    (1.30)

    From Poisson’s equation (see Section 1.6), we express the charge density in terms of the Laplacian of the potential

    (1.31)

    We now integrate by parts over all space. The surface integral at infinity vanishes because both E and V go to zero, giving

    (1.32)

    1.4 ELECTRIC FIELDS AND MATERIALS

    If we dealt only with charge distributions in free space, we would have no need for the numerical methods presented in this book. Given the charge distribution, we could use Coulomb’s law and in principle find the electric field at any point in space by superposition. In reality, we must consider materials that affect and are affected by the electric field. There are two types of materials to consider, both idealized here, which we must deal with to solve realistic problems in electrostatics. They are conductors and insulators or dielectrics.

    1.4.1 Conductors

    An ideal conductor is a material with an unlimited number of free charges. When an electric field is applied to a conductor, the free charges move under the influence of the field. An equilibrium state occurs when the free charges adjust themselves in a distribution such that the net or total electric field is identically zero in the medium. If the field is not zero, the charges will continue to experience a force and to move. In materials that we consider good conductors (e.g., copper, aluminum, iron) this redistribution takes place very rapidly. In room temperature copper it is on the order of 10−18 seconds. Unless we are dealing with very high frequency effects, we can consider this redistribution of the charge to be instantaneous. In this case the net electric field in the conductor is zero, which means that at each point in the volume of the conductor the induced electric field due to the internal charges is exactly equal and opposite to the applied electric field. (See Figure 1.6.)

    Figure 1.6 Electric Field Cancellation in a Conductor

    1.4.2 Dielectrics

    Real materials contain both free and bound charges. We define an ideal dielectric as a material with no free charges. Under the influence of an electric field the bound charges, which are originally electrically neutral, will distort into dipoles as shown in Figure 1.7.

    Figure 1.7 Polarization by an Electric Field

    These internal dipoles produce a polarization, which for linear materials is proportional to the electric field. The proportionality constant depends on the material. We can write

    (1.33)

    where χ is the susceptibility of the material.

    The net flux density is the sum of the applied flux density and the polarization flux density

    (1.34)

    The product ∈0(1 + χ) is called the permittivity of the material. The complex interactions related to the bound charges occurring in the dielectric are thus all included in this macroscopic view of the permittivity. Note that a material with a high permittivity begins to behave like a conductor. The greater the polarization, the smaller the internal electric field. In modeling applications it is therefore possible to represent a conductor as a dielectric with a very high value of ∈.

    1.5 INTERFACE CONDITIONS ON THE ELECTRIC FIELD

    The electric fields and flux densities must satisfy certain interface conditions. A valid solution to the field equations will automatically result in these conditions being satisfied. We can find these conditions by considering the interface between two materials in Figure 1.8. Let the lower material have electric property ∈1 and the upper material have ∈2.

    Figure 1.8 The Electric Field at a Boundary

    We first consider the incremental volume represented by the disk in Figure 1.8. We can let the height of the disk approach zero so that no flux leaves the disk through the cylindrical side. Gauss’s law tells us that the total flux leaving the volume through the top and bottom caps on the cylinder is equal to the charge enclosed in the cylinder. The flux leaving the top surface is ϕt 2 = Dn2S, where Dn2 is the normal component of the flux density and S is the surface area. The flux leaving the bottom is similarly ϕb 1 = −Dn1S. The total flux leaving the volume is ϕ = (Dn2 – Dn1)S and this equals the charge enclosed by the volume, which is Q = ρsS where ρs is the surface charge density (C/m²). So we have

    (1.35)

    If there is no surface charge density, then the normal component of the flux density is continuous. If there is a surface charge density, the normal components are discontinuous by the magnitude of the surface charge density.

    Our second interface condition comes from considering the electric field around the closed path abcda in . As the height of the rectangular path goes to zero, there will be no contribution from the vertical sides. The contribution from a to b , where Et1 is the tangential component of the electric field. Similarly, the contribution along cd ,

    (1.36)

    So the tangential component of the electric field is continuous. This is true even if there is surface charge. These are our two interface conditions for electric fields. We have made no assumptions concerning the nature of the materials. They can be conductors or dielectrics and can have nonlinear, anisotropic properties. Note that these conditions are implied in Maxwell’s equations and therefore usually need not be specifically enforced. We will see that in numerical solutions one of these is satisfied exactly (in a strong sense) and the other is only approximately satisfied (in a weak sense).

    1.6 LAPLACE’S AND POISSON’S EQUATIONS

    Two of the most important equations in physics are Laplace’s and Poisson’s equations. These equations describe the continuity of the field and apply not only to electric fields but also to magnetic fields, heat flow, gravitation, pressure, seepage, and so on. It is instructive to derive Laplace’s equation from a differential point of view because we will be dealing with finite difference and finite element equations later on.

    Consider the differential volume element in Figure 1.9, with sides Δx, Δy, and Δz. We will assume that the electric field in the center of the volume is E (x, y, z). We first find the flux leaving the elemental volume. The flux leaving the volume on the right in the x direction, to a first-order approximation, is

    Figure 1.9 Differential Element for Laplace’s Equation

    (1.37)

    The flux entering on the left in the x direction is

    (1.38)

    The total flux leaving in the x direction is the difference of (1.37) and (1.38) or

    (1.39)

    Similar formulae in the y and z directions give the total flux leaving the volume as

    (1.40)

    By Gauss’ law this must be equal to the total charge in the volume or ρΔxΔyΔz. equation (1.40) becomes

    (1.41)

    This is the differential form of Gauss’ law, expressed as ∇ · D etc. we get

    (1.42)

    This is written in operator form as

    (1.43)

    or for homogeneous space,

    (1.44)

    Note that Poisson’s equation is a second-order equation and therefore requires two boundary conditions. Generally speaking we must specify the potential and its normal derivative on the boundary. If we let ρ = 0 in equation (1.44), we have Laplace’s equation.

    The Uniqueness Theorem

    If we have a set of conductors and specify either the potential or the total charge on each of these, we have completely specified the problem. Any solution to Laplace’s equation that satisfies the boundary conditions is the unique solution. Consider the set of N conductors in Figure 1.10. Let φ1 be a solution to Laplace’s equation. We now

    Enjoying the preview?
    Page 1 of 1