Numerical _ Techniques in Electromagnetics Second EditionNumerical — Techniques in Electromagnetics Second Edition Matthew N. O. Sadiku, Ph.D. cRcPrLibrary of Congress Cataloging Sadiku, Mathew N.O. ‘Numerical techniques in elecuomagnetes / Mathew NO, Sadiku—[2nd el. em Inclnds bibliographical references and inde. ISBN 0-8193.1595-3 (alk. paper) 1 Electromagnetism. 2 Numerical anlsis. 1. Tle creo 524-2000 537°01'S15—de21 026823 ar ‘Tis hook contains information obianed from auhenc and highly regarded sources. Reprinted material ‘is quoted with permission, and sources ar indieed. A wide variety of references are std, Reasonable effort have been made (a publish reliale das and information, bt the author andthe publisher canna sssume responsibilty Fr the validity ofall materials o forthe consequences of thei se. [Neier tis book nor any part may be reproduced or ansmited in any form or by any means, elcome for mechanical. including nhotoconvina. miroimng. and receadina. or by an infomation sage of ‘eames system, without price permission i wri from the publisher, ‘The consent of CRC Press LLC does no extend to copying for general distbuton, for promo, for creating new works, ofr rele, Specie ermision mast be absied in wring from CRC Press LLC for auch copying Direct all inauiies to CRC Press LLC. 2000 N.W, Corporate Bld, Boca Raton, Florida S383 ‘Trademark Notice: Product or corporate names may be uademarks of registered uadematks, and are used nly for dertincation and explanauon, witout tent to ntnnge. 7001 hy CRE Pre No claim to eiginal US. Government works Ixenatioal Standard Book Nomber 08493-15953 Library of Congress Card Number 00-026823 Priaud in he Usted Suu of America 123-4567 89.0 Primed on ache poperPreface ‘The art of computation of electromagnetic (EM) problems has grown exponentially for three decades due to the availability ot powertul computer resources. In spite of this, the EM community has suffered without a suitable text on computational techniques commonly used in solving EM-telated problems. Although there have been monographs on one particular technique or the other, the monographs are written for the experts rather than students. Only a few texts cover the major techniques and do that in a manner suitable for classroom use. It seems experts in this area are familiar with one or few techniques and not many experts seem to be familiar with all the common techniques. This text attempts to fil the gap, ‘The text is intended for seniors or graduate students and may be used for a one- semester or two-semester course. The main requirements for students taking a course based on this text are introductory EM courses and a knowledge of a high-level computor languaus, proferably FORTRAN ar C. Software packages cush as Matlab and Mathcad may be helpful tools, Although familiarity with linear algebra and numerical analysis is wsefal, it is not required. In writing this book, three major objectives were borne in mind. First, the book is intended to teach students how to pose, numerically analyze, and solve EM problems. ‘Second, itis designed to give them the ability to expand their problem solving skills using a variety of available numerical methods. Third, it is meant to prepare graduate students for research in EM. The aim throughout has been simplicity of presentation so that the text can be useful for both teaching and self-study. In striving after simplicity, however, the reader is referred to the references for mote information. ‘Toward the end of each chapter, the techniques covered in the chapter ate applied to real life problems. Since the application of the technique is as vast as EM and author's experience is limited, the choice of application is selective, Chapter 1 covers some fundamental concepts in EM. Chapter 2 is intended to put ‘numerical methods in a proper perspective. Analytical methods such as separation Lf variables aud seties expansion ave covered. Chiplet 3 discusses the Gite differ- fence methods and begins with the derivation of difference equation from a partial differential equation (PDE) using forward, backward, and cental differences, The finite-difference time-domain (FDTD) technique involving, Yee’s algorithm is pre-sented and applied to scattering problems. Numerical integration is covered using trapezoidal, Simpson’s, Newton-Cotes rules, and Gaussian quadratures. Chapter 4 on variational methods serves as a preparatory ground for the next two major topics: moment methods and finite element methods. Basic concepts such as inner product, self-adjoint operator, functionals, and Euler equation are covered. ‘Chapter 5 on moment methods focuses on the solution of integral equations. Chap- ter 6 on finite element method covers the basic steps involved in using the finite clement method. Solutions of Laplace’s, Poisson's, and wave equations using the finite element method are covered. Chapter 7 is devoted to transmission-line matrix or modeling (TLM). The method is applied to diffusion and scattering problems. Chapter 8 is on Monte Carlo methods, while Chapter @ ic an the method af fines ‘Since the publication of the first edition, there has been an increased awareness and tilization of numerical techniques. Many graduate curricula now include courses in numerical analysis of EM problems. However, not much has changed in compu: tational electromagnetics. A major noticeable change ie in the FDTD method. The ‘method seems to have attracted much attention and many improvements are being ‘made to the standard algorithm, This edition adds the noticeable change in incorpo- rating absorbing boundary conditions in FDTD, FEM, and TLM. Chapter 9 is a new cchaptcr'on the method of lines Acknowledgements Tam greatly indebted to Temple University for granting me a sabbatical in Fall 1998 during Which T was able to do most ofthe revision, T specifically would like to thank ‘my dean, Dr. Keya Sadeghipour, and my chairman, Dr.John Helferty, for their support. Special thanks are due to Raymond Garcia of Georgia Tech for writing Appendices C and D in C++. Tam deeply grateful to Dr. Arthur D. Snider of the University of South Florida and Mohammad R. Zunoubi of Mississippi State University for taking the time to send me the list of errors in the first edition. T thank Dr. Reinhold Pregla for helping in clarifying concepts in Chapter 9 on the method of lines, T express my deepest gratitude to my wife, Chris, and our daughters. Ann and Joyce. for their Patience, sacrifices, and prayers. A Note to Students Before you embark on writing your own computer programor using the ones in this text, you should try to understand all relevant theoretical backgrounds. A computeris no more than a tool used in the analysis of a program. For this reason, you should. bbe as clear as possible what the machine is really being asked to do before setting it off on several hours of expensive computations. It has been well said by A.C. Doyle that “It isa capital mistake to theorize before you have all the evidence. It biases the judgment.” Therefore, you should never trust the results of a numerical computation unless they are validated, atleast in part. You validate the results by comparing them with those obtained by previous investigators or with similar results obtained using a different approach which may be analytical ‘or numerical. For this reason, itis advisable that you become familiar with as many ‘numerical techniques as possible. ‘The references provided atthe end of each chapter are by no means exhaustive but are meant to serve as the starting point for further reading.Contents 1 Fundamental Concepts Al 12 14 Introduction Review of Blectromaynetic Theory 1.2.1 Electrostatic Fis 122 Magnetostatic Fields 1.23. Time-varying Fields 1.24 Boundary Conditions 125 Wave Equations 1.26 Time-varying Potentials 1.27 Time-harmonic Fields Classification of EM Problems 1.3.1 Classification of Solution Regions 132 Classification of Differential Equations Some Important Theorems 1.4.1 Superposition Principle 1.4.2 Uniqueness Theorem References. Problems ‘Analytical Methods Introduction Separation of Variables ‘Separation of Variables in Rectangular Coordinates 2.3.1 Laplace's Equations 2.3.2 Wave Equation2.4. Separation of Variables in Cylindrical Coordinates 24.1 Laplace's Equation 24.2 Wave Equation 2.5 Separation of Variables in Spherical Coordinates, 23.1 Laplace's Equation ‘Wave Equation 2.6 Some Useful Orthogonal Functions 2.7. Series Expansion 27.1 Poisson’s Equation in a Cube Poisson's Equation in a Cylinder Strip Transmission Line 28 Practical Applications 28.1 Scatteting by Dielecttie Sphere 282 Scattering Cross Sections 2.9 Attenuation Due to Raindrops 2.10 Concluding Remarks References Problems Finite Difference Methods 3.1 Introduction 3.2. Finite Difference Schemes 3 Finite Differencing of Parabolic PDEs 3.4 Finite Diffetencing of Hyperbolic PDEs 5 Finite Differencing of Elliptic PDEs 35.1 Band Matrix Method 35.2 Iterative Methods 3.6 Accuracy and Stahility of FD Solutions 3.7 Practical Applications I — Guided Structures 47.1 Transmission Lines 3.7.2 Waveguides 2.8 Practical Applications II — Wave Seattoring (FDTD) 38.1 Yee’s Finite Difference Algorithm 382 Accuracy and Stability 383 Lattice Truncation Conditions 384 Initial Fiolds 385 Programming Aspects 3.9 Absorbing Boundary Conditions for FDTD 3.10 Finite Differencing for Nonrectangular Systems 3.10.1, Cylindsical Coordinates 3.10.2 Spherical Coordinates 3.11 Numerical Integration 3.11.1 Euler’s Rule 3.11.2. Thapezoidal Rule 3.11.3 Simpson’s Rule3.11.4 Newton-Cotes Rules 3.11.5 Gaussian Rules 3.11.6 Multiple Integration 3.12. Concluding Remarks References. Problems ‘Variational Methods 4.1 Introduction 4.2. Operators in Linear Spaces 4.3 Calculus of Variations 44 Construction of Functionals from PDEs 45 Rayleigh-Ritz Method 4.6 Weighted Residual Method 46.1 Collocation Method 4.6.2 Subdomain Method 4.6.3 Galerkin Method 4.64 Least Squares Method 4.7 Bigenvalue Problems 48. Practical Applications 49 Concluding Remarks References. Problems ‘Moment Methods 3.1 Introduction 5.2. Integral Equations 5.2.1 Classification of Tntegral Equations 5.2.2 Connection Between Differential and Integral Equations 5.3. Green’s Functions 53.1 For Free Space 5.32. Far Domain with Conducting Boundaries SA Applications I— Quasi-Static Problems 5.5 Applications I — Seattering Problems. 55.1 Scattering by Conducting Cylinder 55.2. Seattering by an Arbitrary Array of Parallel Wires 5.6 Applications IIT — Radiation Problems 5.6.1 Hallen’s Integral Equation 5.6.2. Pocklington’s Integral Equation 5.6.3. Expansion and Weighting Functions 5.7 Applications TV — EM Absorption in the Human Body 5.7.1 Derivation of Integral Equations ‘Transformation to Matrix Equation (Diseretization) Evaluation of Matrix Elements Solution of the Matrix Equation5.8 Concluding Remarks References Problems Finite Klement Method 6.1 Introduction 6.2 Solution of Laplace’s Equation 62.1 Finite Element Diseretization 62.2 Element Governing Equations 62.3 Assembling of All Elements 624 Solving the Resulting Equations 6.3 Solution of Poisson’s Equation 63.1 Deriving Element-goveming Equations 63.2 Solving the Resulting Equations 64 Solution of the Wave Equation 65 Automatic Mesh Generation I — Rectangular Domains 66 Automatic Mesh Generation II — Arbitrary Domains 64.1 Definition of Blocks Subdivision of Bach Block Connection of Individual Blocks 6.7 Bandwidth Reduction 68 Higher Order Elements 638.1 Pascal Triangle 682 Local Coordinates 68.3 Shape Functions 684 Fundamental Matrinas 69. Three-Dimensional Elements 6.10 Finite Blement Methods for Exterior Problems 6.10.1. Tnfinite Element Method 6.10.2. Boundary Blement Method 6.103 Absorbing Boundary Conditions 6.11 Concluding Remarks References Problems ‘Transmission-line-matrix Method 7.1 Introduction 7.2. Transmission Hine Equations 73. Solution of Diffusion Equation 7A. Solution of Wave Equations 7A.1 Equivalence Between Network and Field Parameters 742. Dispersion Relation of Propagation Velocity 7A3- Scattering Matrix 74 Boundary Representation 74.5 Computation of Fields and Frequency Response7.4.6 Output Response and Accuracy of Results 7.5 Inhomogeneous and Lossy Media in TLM 75.1 General Two-Dimensional Shunt Node 75.2. Seattering Matrix 75.3 Representation of Lossy Boundaries 7.6 Thrce-Dimensional TLM Mesh 7.6.1 Series Nodes 7.6.2. Three-Dimensional Node 7.6.3 Boundary Conditions 7.7 Brror Sources and Correction 7.2.1 Truncation Error 7.2.2 Coarseness Error 7.7.3 Velocity Error 7.7.4. Misalignment Error 7.8 Absorbing Boundary Conditions 7.9 Concluding Remarks References. Problems ‘Monte Carlo Methods 8.1 Introduction 8.2. Generation of Random Numbers and Variables 8.3 Evaluation of Error 84 Numerical Integration 84.1 Crude Monte Carlo Integration 847 Monte Carlo Entagration with Antithotie Variates 84.3 Improper Integrals 85. Solution of Potential Problems 85.1 Fixed Random Walk 8.5.2 Floating Random Walk 85.3 Exodus Method 8.6 Regional Monte Carlo Methods 8.7 Concluding Remarks References Problems, ‘Method of Lines 9.1 Introdvetion 9.2 Solution of Laplace's Equation 9.2.1 Rectangular Coordinates 9.2.2 Cylindrical Coordinates 9.3 Solution of Wave Equation 9.3.1 Planar Microstrip Structures 9.3.2. Cylindsical Microstrip Suuctures 94° Time-Domain Solution9.5 Concluding Remarks References Problems A. Vector Relations ALL Vector Kdentities A2 Vector Theorems A3. Orthogonal Coordinates B_ Solving Electromagnetic Problems Using C++ B.L_ Introduction B.2 A Brief Description of C++ 1.3. Object-Orientation BA CH Object-Oriented Language Features 15 A Final Note References. Numerical Techniques in C++ D_ Solution of Simultaneous Equations D.1_ Elimination Methods D.1.1 Gauss’s Method 1.2 Cholesky's Method D.2 erative Methods 2.1 Tacobi’s Method 2.2 Gauss-Seidel Method D2.9 Retanation Methox! Gradient Methods . D.3. Matrix Inversion D4 Eigenvalue Problems Da.1 Iteration (or Power) Method DA2_Tacobi’s Method Answers to Odd-Numbered ProblemsTo my teacher Carl A. Ventrice and my parents Ayisat and Solomon SadikuChapter 1 Fundamental Concepts “Science knows no country because knowledge belongs to humanity and is the Lorch which illuminates the world. Science i the highest personification of the nation because that nation will remain the first which carries the furthest the works of thoughts and intelligence” is Paster 1.1. Introduction Scientists and engineers use several techniques in solving continuum or field prob- Jems. Loosely speaking, these techniques can be classified as experimental, analyti- cal, ornumerical. Experiments are expensive, time consuming, sometimes hazardous, id usually dy aot allow aiucle Meaibiity is parameter vatiativn. Toweves, every ar ‘merical method, as we shal see, involves an analytic simplification tothe point where itis easy co upply the numerical method, Nowithstanding this fact, dhe Following ‘methods are among the mast commonly used in electromagnetics (EM), A. Analytical methods (exact solutions) (1) separation of variables (2) series expansion, (3) conformal mapping (4) integral solutions, e-g., Laplace and Fourier transforms (3) perturbation methods B, Numerical methods (approximate solutions) (1) finite difference method 2) method of weighted residuals 3) moment method (4) finite element method mor cRcrRESSLLC(5) transmission-line modeling (6) Monte Carlo method (1) method of lines Application of these methods is not limited to EM-related problems: they find applications in other continuum problems such as in fluid, heat transfer, and acous- ties (1] As we shall see, some of the numerical methods are related and they all generally sive approximate solutions of sufficient accuracy for engineering purposes. Since ‘our objective isto study these methods in detail in the subsequent chapters, it may be ‘premature to say more than this at this point. ‘The need for numerical solution of electromagnetic problems is best expressed in the words of Paris and Hurd: “Most problems that can be solved formally (analyti- cally) have been solved." Until the 1940s, most EM problems were solved using the classical methods of separation of variables and integral equation solutions. Besides ‘the fact that a high degree of ingenuity, experience, and effort were required to apply tose methods, oly & at iow sa to the complex geometries defining the problems, "Numerical solution of EM problems started in the mid-1960s with the availability of modern high-speed digital computers. Since then, considerable effort has been ‘expended on solving practical, complex ’M-related problems for which closed form. analytical solutions are either intractable or do not exist. The numerical approach has. the advantage of allowing the actual work fo be carried out by operators without @ knowledge of higher mathematics or physics, with a resulting economy of labor on ‘the part of the highly trained personnel. Before we set out to study the various techniques used in analyzing EM problems, it is expedient to remind ourselves of the physical laws governing EM phenomena in general. This will be done in Section 1.2. In Section 1.3, we shall be acquainted with different ways EM problems are categorized. The principle of superposition and uniqueness theorem will be covered in Section 1.4. of practival problems could be investigated due 1.2 Review of Electromagnetic Theory ‘The whole subject of EM unfolds as a logical deduction from eight postulated ‘equations, namely, Maxwvell’s four field equations and four medium dependent equa tions [2]-4). Before we briefly review these equations, it may be helpful to state two important theorems commonly used in EM. These are the divergence (or Gauss's} Tac Floctromagnetc Theory; DT, Pais snd PK. Hurd, MeCraw-Hil, Now York, 196, p16. mor cRcrRESSLLCtheorem, an gras= fv fFed= [vxP-as a2 h s Perhaps the best way to review EM theory is by using the fundamental concept of electric charge. FM thenry can he regarded as the sty af fields prrdiced hy electric charges at rest and in motion, Blectrostatic fields are usually produced by static electric charges, whereas magnetostatic fields are die t0 motion of electric charges ‘with uniform velocity (direct current). Dynamic or time-varying fields are usually due to accelerated charges or time-varying currents and Stokes's theorem, 1.2.1 Electrostatic Fields clectrostatie fields are Gauze": law, fr-us— f noav as) In Bqs. (1.3) and (1.4), Dis the electric flux density (in coulombs/meter?), oy is the volume charge density (in coulombs/meter’), and E is the electric field intensity (in volts/meter). The integral form of the laws in Eqs. (1.3) and (1.4) can be expressed in the differential form by applying Eq. (1.1) to Bg. (1.3) and Eq, (1.2) to Eg. (1.4), We obtain v-D=p, as) and VxE=0 ‘The vector fields D and E are related as pack an white ¢ is the dickevtie pet electric potential V (in volts), ity Cin favadofineter) of dhe miediuun. Th tems of the (1s) mor cRcrRESSLLCas ‘Combining Eqs. (1.5), (1.7), and (1.8) gives Poisson’s equation: view (1.10) Fe is constant, 2.106) When p, = 0, Eq, (1.10) becomes Laplace’s equation: v-evv =0 (tay or for constant € 1b) 1.2.2. Magnetostatie Fields ‘The basic laws of magnetostatic fields are Ampere’s law pHa [sas (1.12) Which is related to Biot-Savart law, and the law of conservation of magnetic fux (also called Gauss"s law for magntostatics) fr where His dhe magnetic fed intensity (in amperes/meter), Je isthe rent density (in amperes/meter), and B is the magnetic flux density (in tesla or wwebersimeter?). Applying Eq. (1.2) 10 Bq, (1.12) and Eg. (1.1) 0 Bq, (1.13) yields ther differential form as =o aay VxH=J (aay and v-B 0 (as) ‘The vector fk Baul of the medium as ave telated dugugh the petmtcability 4 (in bemies/ancter) wh 1.16) mor cRcrRESSLLCAlso, related to E through the conductivity o (in mhos/meter) of the medium as J=ok ay ‘This is usually referred to as point form of Ohm's law. In terms of the magnetic vector potential A (in Wh/meter) R-vvA aig) Applying the vector identity YAW AR =V0-P)-F 19) 10 Bas. (1-14) and (1.18) and assuming Coulomb gauge condition (¥ - A = 0} leads to Poisson’s equation for magnetostatic fields VAs =nd 4.20 When J = 0, Eq. (1.20) becomes Laplace's equation WA =0) a2n 1.23 e-var ying Fields In this case, electric and magnetic fields exist simultaneously. Equations (1.5) and (1.15) remain the same whereas Fqs. (1 6) and (1.14) requite some modification for dynamic fields. Modification of Eq, (1.6) is necessary to incorporate Faraday’s law of indnetion, and that of Fa. (1.14) is warranted tallow for displacement current ‘The time-varying EM fields are governed by physical laws expressed mathematically (1.228) (1.22%) (1.20) (1.228) the magnetic conductive current density (in volts/aquare metc#) and ois the magnetic resistivity (in ohms/meter). ‘These equations are referred to as Maxwell's equations in the generalized form, ‘They are first-order linear coupled differential equations relating the vector field quan- mor cRcrRESSLLCtities to each other. The equivalent integral form of Eq. (1.22) is [roto (1.234) ‘ = [ (Bos) a (as wae [ (122) as iL Ss 7 oa fee ‘These are called constitutive relations for the medium in which the fields exist. Equa- ) au (2.24) for the eiylat postulated eyuations ou whic EM dreory ut folds itself. We must note that in the region where Maxwellian fields exist, the fields sare assumed 10 be: tious C1 1) ingle valved, (2) bounded, and (3) continuous functions of space andl time with continuous derivatives. tis worthwhile to mention two other fundamental equations that go hand-in-hand \With Maxwell's equations. One is the Lorentz force equation = QB +uxB) 125) where F is the force experienced by a particle with charge @ moving at velocity win ‘an EM hela; the Lorentz force equation constitutes a link between KM and mechanics, ‘The other isthe continuity equation x if at 4.26) Which expresses the conservation (or indestuiuctibilty) oPelectie charge. The conti- nuity equation is implicit in Maxwell's equations (see Example 1.2). Equation (1.26) js not peculiar to EM. In fluid mechanics, where J coresponds with velocity and py with mass, Bg, (1.26) expresses the law of conservation of mass. mor cRcrRESSLLC1.2.4 Boundary Conditions ‘The material medium in which an EM field exists is woually characterized by its constitutive parameters 0, ¢, and j1. The medium is said to be linear if 0, ¢, and are independent of E. and H of nonlinear otherwise. It is homogeneous if @, ¢, and are not functions of space variables or inhomogeneous otherwise. Itis isotropic if 6, €, and 2 are independent of direction (scalars) or anisotropic otherwi a WA Figure 1 Interface between two media. ‘The boundary conditions at the interface separating two different media 1 and 2, ‘with parameters (61, €1, jt) and (09, €2, 2) as shown in Fig. 1.1, are easily derived from the integral form of Maxwell’s equations. They are Ey or (Ey ~ Eg) X apr 27a) Hy, ~ Hoy = K ot (Hy — Hb) x ay = K (1.276) Diy ~ Diy = ps of (Dy = Dy) anna = ps (a2te) By — By By) -ania (278) whorw a2 ¢ a nit normal vector dieneted Fram movin Tt mortinm 9 eheorinte 1 and 2 denote fields in regions 1 and 2, and subscripts t and n, respectively, denote tangent and normal components of the fields, Equations (1.27a) and (1.274) state thatthe tangential components of E and the normal components of B are continuous across the boundary. Equation (1.27b) states that the tangential component of His discontinuous by the surface current density K on the boundary. Equation (1.27c) states that the discontinuity in the normal component of D isthe came asthe surface charge density p, on the boundary Tn practice, only two of Maxwell's equations are used (Egs. (1.222) and (1.22d)) ‘when a medium is souree-free (J = 0, py = 0), since the other two are implied (see Problem 1.3). Also, in practice, itis sufficient to make the tangential components ofthe fields satisfy the necessary boundary conditions since the normal components ‘implicitly satisfy their corresponding boundary conditions 1.25 Wave Equations Asmenlioncd cats, Malls equations ate coupled fionde differential use tions which ae dfficlt 1 apply when solving boundary-value problems. The diffi culty is overcome by decoupling the fist-order equations thereby obtaining the wave equation, a second-order differential equation whichis useful fr solving problems. mor cRcrRESSLLC‘To obtain the wave equation fora linear, isotropic, homogeneous, source-free medium (be = 0,1 = 0) from Eq, (1.22), we take the curl of both sides of Eq. (1.22c). This ives. Vx VxB=—p2(v xe) (1.28) x Vx B= 42 (Vx) 2 From (1.220), i ee since J = 0, so that Ba, (1.28) becomes aE i vxe pea (129) Applying the vector identity Vx Vx FaV0-F- VF (1.30) in Eq. (1.29, a Vv B) = PE = ~pe Since p, = 0, V -E = 0 from Eq, (1.22a),and hence we obtain, ve asp which isthe time-dependent vector Helmholtz equation or simply wave equation. If vwe had started the derivation with Bq. (1.22d), we would obtain the wave equation for Has, AL an WH pe 32) Equations (1.51) and (1.52) are the equations of motion of IM waves in the medium under consideration. The velocity (in mis) of wave propagation is (1.33) where y= 0 53x 108 mls in free space. Tt show he nated that each of the vector ‘equations in (1.31) and (1.32) has three scalar components, so that altogether we have six scalar equations for Fy, Fy, Fe, Hy, Hy, and H,, Thus each component of the ‘wave equations has the form (134) which isthe scalar wave equation. mor cRcrRESSLLC1.2.6 ‘Time-varying Potentials Although we are often interested in electric and magnetic eld intensities (E and H), which are physically measurable quantities, itis often convenient to use auxiliary functions in analyzing an EM field. These auxiliary functions are the scalar electric potential V and vector magnetic potential A. Although these potential functions are arbitrary, they are required to satisfy Maxwell's equations, Their derivation is based. ‘on two fundamental vector identities (see Prob. 1.1), VxVo=0 (1.35) and vevx ° 436) which an arbitrary sealar field @ and vector field F must satisfy. Maxwell’s equa- tion (1.22b) along with Eq, (1.36) is satisfied if we define A such that B=VxA 37) Substituting this into Eq, (1. ) gives ° Since this equation has to be compatible with Eq, (1.35), we can choose the scalar field V such that or 1.38) Thus, if we knew the potential functions V and A, the fields K and B could be vbuained fivint Bys. (1.37) and (1.38). However, we sill need (9 fi the solution for the potential functions. Substituting Eqs, (1.37) and (1.38) into Eq. (1.22d) and assuming linear, homogeneous medium, : Applying the vector identity in Bi. (1.30) leads to aa av nev 2.39) VIA-W(V-A) = nd +e. ae Substituting Eq, (1.38) into Eq, (1.228) gives 2 gay VA) é a VE mor cRcrRESSLLCa +aV-A 40) i € “According to the Helmholtz theorem of vector analysis, a vector is uniquely defined if and only if both its curl and diversence are specified. We have only specified the ccurl of A in Eg. (1.37); we may choose the divergence of A so that the differential ‘equations (1.39) and (1.40) have the simplest forms possible. We achieve this in the so-called Lorentz condition: aan ‘Incorporating this condition into Eqs. (1.39) and (1.40) results in PA : VA- ner (1.42) and (1.43) which are inhomogeneous wave equations. Thus Maxwell's equations in terms of the potentials V and A reduce tothe three equations (1.41) t0 (1.43). In other words, the three equations are equivalent to the ordinary form of Maxwell's equations in that putculialssatist ying thoze eatin always load ty a avlutivns uf Mlanwell’s ctativina for E and B when used with Eqs. (1.37) and (1.38). Integral solutions to Eqs, (1.42) aud (143) ate the so-called rewarded! potentials ude looldv and (4s) 4xeR \here Ris the distance from the source point tothe field point, and the square brackets denote py and J are specified at a time (ce)! carlce than for which A or ¥ is being determined. 1.27 Upto this point, we have considered the general case of arbitrary time variation of EM fields. In many practical situations, especially at low frequencies, itis sufficient to 0 (1.03) parabolic if? — 4ac = 0 ‘The terms hyperbolic, parabolic, and elliptic are derived from the fact that the {quadratic equation wet | day toy ide bey i f—0 represent a hyperbola, para, or ellipse if #2 — dnc is pacitive, ero, or negative, respectively. In each ofthese categories there are PDE that model certain physical ‘Phenomena, Such phenomena are not limited to EM tat extend to almost all areas of science and engineering. Thus the mathematical model specified in Eg. (1.61) aces in probleme involving hea transfer, boundary-layer flow, vibrations, laste, eleetrostati, wave propagation, and s0 on Elliptic PDEs are associated with steady-state phenomena, ie, boundary-value problems. Typical examples of this type of PDE include Laplace's equation Fo md (1.64) and Poisson’s equation #o | &o aa tar 8) (1.68) where in both cases a = b = 0, Anclliptic PDE usually models an interior problem, and hence the solution region is usually closed or bounded as in Fig. 1.3 (a). mor cRcrRESSLLCBoundary conditions sreseribeu here R ose region « setion a dha ee Pees ton x Inia conditions prescribed here Figure 1 (a) Elliptic, (b) parabolic, or hyperbolic problem. Hyperbolic PDEs arise in propagation problems. ‘The solution region 1s usually ‘open so that a solution advances outward indefinitely from initial conditions while always satistying specified boundary conditions, A typical example of hyperbolic PDE is the wave equation in one dimension ao _ 180 eee ea 1.669 where a = 02, b= 0, ¢ = —1. Notice that the wave equation in (1 $0) is not hyper bolic but elliptic, since the time-dependence has been suppressed and the equation is, merely the steady-state solution of Eq. (1.24), Parabolic PDEs are generally associated with problems in which the quantity of interest varies slowly in comparison with the random motions which produce the variations. The most common parabolic PDE is the diffusion (or heat) equation in cone dimension * ab ; : i «67 where a = 1,b = 0 = e. Like hyperbolic PDE, the solution region for parabolic PDE is usually open, asin Fig. 1.3 (b). The initial and boundary conditions typically associated with patabulic eyuations resemble dhuse fur hyperbolic problems except that only one initial condition at ¢ = 0 is necessary since Eq, (1.67) is only first onder intime. Also, parabolic and hyperbolic equations ate solved using similar techniques, ‘whereas elliptic equations are usually more difficult and require different techniques. mor cRcrRESSLLC[Note that: (1) since the coefficients a,b, and c are in general functions of x and y, the classification of Eg. (1.61) may change from point to point in the solution region, and (2) PDEs with more than two independent variables (x, v, z, ...) may not fit ‘as neatly into the classification above. A summary of our discussion so Farin this, section is shown in Table 1.1 ‘Table 1.1 Classification of Partial Differential Equations Te ‘Sen of Example ‘Sotudon region ? = ac Elliptic = Laplace's equation: Ciosed Oy FO) =0 Hyperbolic + Wisve equation: Open WOx¢ = Oe Parabolic 0 Diffusion equation; Open Ore =k; ‘The type of problem represented by Eq. (1.58) is said to be deterministic. since the ‘quantity of interest can be determined directly. Another type of problem where the ‘quantity is found inditectly is called nondeterministic ar eigenvalue. "The standard ceigenproblem is of the form. Lo =o (1.68) where the soures term in Eq, (1.58) has been replaced by A®, A more general version is the generalized eigenproblem having the form LO =iMo (1.69) where MY, like L, is a linear operator for EM problems. In Eqs, (1.68) and (1.69), ‘only some particular values of 2, called eigemalues are permissible; associated with these values are the corresponding solutions ¢ called eigenftarctions. Eigenproblems ‘are usually encountered in vibration and waveguide problems where the eigenval- vues correspond to physical quantities such as resonance and cutof frequencies, respectively. 3. Classification of Boundary Conditions ‘Our problem consists of finding the unknown function & of a partial differential ‘equation. In addition tothe fact that © satisfies Bq. (1.58) within a prescribed solution region R, & must satisfy certain conditions on S, the boundary of R. Usually these boundary conditions are of the Dirichlet and Neumann types. Where a boundary has both, a mixed boundary condition is said to exist (1) Dirichlet boundary condition: OP) =0, Fons (1.70)(2) Neumann boundary condition: Amery ea cemaon Si a7 on Le,, the normal derivative of vanishes on S. (3) Mixed boundary condition ab(r) BO Linaey=0 + ons an svhere he) isa known function and AB isthe directional derivative of & along the outward normal fo the boundary 8, ie, ao Vo-a, 73) mm where ay is a unit normal directed out of R, as shown in Fig, 1.2. Note that the Neumann boundary condition isa special case of the mixed condition with hr) — 0. ‘The conditions in Bg. (1.70) t0 (1.72) recalled homogeneous boundary conditions. ‘The more general ones are the inhomogenc Dirichlet: Or) = ple), Fon (74) 280) _ ae), von 8 75) Mixed 90) n(eyoie) = wie), on $ (1.76) mn where p(t), g(P and w(t) ate explicitly known funtion onthe boundary 8. For example, (0) — 1 is an inhomogeneous Dirichlet boundary condition, and the associated homogeneous counterpat is (0) = 0, Also O'(1) = 2 and &'() are, respectively, inhomogeneous and homogeneous Neumann boundary conditions In lectostatis, for example, if the value of electric potential is specified on S, we have Dirichlet boundary conition, whereas ithe surface charge (px = Dy is specified, the boundary condition is Neumann. The problem of finding & function 6 that i harmonic in a region i called Driller problem (or Newnann problem) 4 (or 3 is presribed on the boundary ofthe region Tris Worth observing that he term “homogeneous” has been used to mean diferent things. The solution region could be homogeneous meaning that ¢,e, and ju are constant within R; the PDE could be homogeneous if g so that L® = 0; and the ‘boundary conditions are homogeneous when pr) Example 14 ‘Classify these equations as elliptic, hyperbolic, or parabolic: mor cRcrRESSLLC(@) 404, +20, + bytx+y av atv axa State whether the equations are homogeneous or inhomogeneous. [] Solution (a) Inthis PDE, a = 4, {.e,, the PDE Is parabolic. Since g Is inhomogeneous. (b) For this PDE, a= e*, b ? = dac = cos? y + de" > 0 an the PDE hyperbole. Since ¢ =, the PDE is homogeneous, 1.4 Some Important Theorems ‘Two theorems are of fundamental importance in solving EM problems. These are the principle of superposition and the uniqueness theorem. 1.4.1 Superposition Principle ‘The principle of superposition is applied in several ways. We shall consider two Hi TFeach member oa setof faction Oy, = 12... Nisa soliton tthe PDE ei een oy = 00+ Dane am ny al Lo 78) mor cRcrRESSLLCsubject to the boundary conditions Ms) Ma My(s) (1.79) as long as L is linear, we may divide the problem into a series of problems as follows: Lo =8 Lo; =0 Loy =0 Ms) =0 Ms My(s) =0 Mas) My(s)=0 — My(s) =0 i My(s)=hy (1.80) where dp, 1,.... , dy are the solutions to the reduced problems, which are easier to solve than the original problem. The solution tothe original problem is given by o-9, ash 1.4.2 Uniqueness Iheorem. ‘This dheunem guarantees that dhe solution ublained for aPDE with seine prescsibed boundary conditions is the only one possible. For EM problems, the theorem may be stated as follows: If in any way a set of fields (E, ID) is found which satisfies simultaneously Maxwell's equations and the prescribed boundary conditions, this set is unique, Theiefure, « field is uniquely specified by the souves (py, J) ‘medium plus the tangential components of E or H over the boundary. ‘To prove the uniqueness theorem, suppose there existtwo solutions (with subscripts 1 and 2) that satisfy Maxwell’s equations in the Vicki? (1.82a) Vea = (1.82) Vx By = —n he (1820) a Vx Hig =J+oEia +6 (182a If we denote the difference of the two fields as AI and AH = Hp —Hy, mor cRcrRESSLLCAE and AH must satisfy the source-free Maxwell’s equations, ie., alt= 0 (133) ox «1x ox alaoans <28E asx Doin hho Ba, LR wh A es Using the vector identity A-(V x B)=B-(V XA)—V-(AX BD and Eq. (1.83¢), Eq. (1.84) becomes 2 ee ee ¥-cax aH) =—14 (aH? +elape) —o1a8 Integrating over volume v bounded by surface $ and applying divergence theorem to ea -4 {seer + duam’] a ~ [ oiaeiae (185) fon x a as 5 showing that AE. and AH satisfy the Poynting theorem just as Ei,» and Hi,2. Only the tangential components of AE and AH contribute tothe surface integral onthe left side of Eq (1.85) Therefore ifthe tangential components of Ey and Ea or Hy and Ha are equal over § (thereby satisfying Eq, (1.27), the tangential components of AK and ‘AH vanish on $. Consequently, the surface integral in Eq, (1.85) is identically zero, ‘and hence the right side of the equation must vanish also. Tt follows that AE = 0 > we) equation VE, = jonoks mor cRcrRESSLLC1.13 Show that in an inhomogeneous medium, the wave equations become Vx (GV x Es) + joeks vx (Vx H,) + jog, =0 1.14. Show thatthe time-harmonie potential function Vand A, satisfy the following inhomageneonis wave equation viv, 42y, = Pe VA + RAL = 1d, where k? = oye. 1.15. Classify the following PDEs as elliptic, parabolic, or hyperbolic. (a) Oy, +2bry + 5dyy ) UF FH FOP + Hey (©), = 2cosxd,, — 3 +sin?xIb,y — voy (70x, — 2xyry + Py +1, + yoy 1.16 Repeat Prob. 1.15 for the following PDEs: > 0 0d @ asap ltt which is called convective heat equation, ) Wp+io=0 which is the Helmholtz equation. © VO+R=pale which is the time-independent Schrodinger equation. (« B = constant) mor cRcrRESSLLCmor cRcrRESSLLCChapter 2 Analytical Methods “I've learned that about 90 percent of the things shat happen to me are good ‘and onty about 10 percent are bad. To be happy, just have to focus on the 90 percent Anonymous 2.1 Introduction ‘The most satisfactory solution of a field problem is an exact mathematical one. Although in many practical cases such an analytical solution cannot be obtained and we must resort to numerical approximate solution. analvtical solution is useful in checking solutions obtained from numerical methods. Also, one would hardly appreciate the need for numerical methods without first seeing the limitations of the classical analytical methods. Hence our objective in this chapter is to briefly ‘examine the common analytical methods and thereby put numerical methods in proper perspective. ‘The most commonly used analytical methods in solving EM-related problems in- clude: (2) separation of variables (2) series expansion 3) conformal mapping (4) integral methods Perhaps the most powerful analytical method is the separation of variables; it is the tell that will be emphasized in this chaplet, Siuve the application of wonfortal ‘mapping is restricted to certain EM problems, it will not be discussed here. ‘The interested reader is referred to Gibbs [1]. The integral methods will be covered in Chapter 5, and fully discussed in (2 mor cRcrRESSLLC2.2. Separation of Variables ‘The method of separation of variables (sometimes called the method of Fourier) is a convenient method for solving a partial differential equation (PDE). Basically, it entails seeking a solution which breaks up into a product of funetio f which involves only one of the variables. For example, if we are seeking a solution GC, ¥. 2.1) 1 some PDE, we require that it has dhe produet form each, er y.2.0) COYOZEITIO en A solution ofthe form in Eg (2.1) ssa to be separable in, y, ‘consider the runetions (1) xyz sin 108, Qxyt+ G3) x+y) (1) is completely separable, (2) p To determine whether the method of independent separation of variables can be ap- plied toa given physical problem, we must consider the PDE deseribing the problem, the shape of the solution region, and the boundary conditions — the three elements that uniquely define a problem, For example, to apply the method to a problem involving to variables x and y (or and g, etc), three things must be considered [3} ‘and. For example, is not separable, while (3) is separable only in z and () The differential operator L must be separable, ic., it must be a function of (x,y) such that LUXCO¥E) 0, DXOO) 18a sum of a function of x only and a tunction ot y only. iy an jal and boundary conditions must be an eonstant-conrdinate surfaces, (iii) The linear operators defining the boundary conditions at x = constant (or y constant) must involve no partial derivatives of with respect to y (or x), and their coefficient must be independent of y (or x), For example, the operator equation aay TE ioltes (Ite soltion region Bi nota eetangle with sid paral tthe x and ¥ axes, (ii violated, With a boundary condition @ = O ona pat of x =O and 34/00 = On anaes pat is violated With this preliminary discussion, we will now apply the method of separation of vaiables fo PDEs in rectangular, ctculareylindical, and spherical condinate systems, Tn each ofthese applications, we shall alvays take these thee major steps mor cRcrRESSLLC(1) separate the (independent) variables (2) find particular solutions of the separated cquations, which satisfy some of the Doundary conditions. (3) combine these solutions to satisfy the remaining boundary conditions We begin the application of separation of variables by finding the product colution of the homogeneous scalar wave equation ano vig- 42% 2.2) ee oS Solution to Laplace's equation can be derived asa special case of the wave equation. Difficion and heat equation can he hanaled in the same manner ac we will reat wave equation. To solve Eg, (2.2), it is expedient that we first separate the time dependence. We let Onn Substituting this in Eq. (2.2), oro 23) TVU Dividing by UT gives 4) oT ‘The left side is independent ofr, while the right side is independent of F; the equality ccan be true only if each side is independent of both variables. If we let an arbitrary 2 be te common value OF the Wo Sites, BY (24) TEUUCES 10 Tse 0 (25a) vu+u=0 2.50) “Thus we have been able to separate the space variable from the time variable ¢. The arbitrary constant —K? introduced in the course of the separation of variables is called. the separation constant, We shall see that in general the total number of independent separation constants in a given problem is one less than the number of independent variables involved, Equation (2.5a) is an ordinary differential equation with solution Tit) = aye! + aye I (26a) T(0) = by costekt) + by sin(cke) 2.60) Since the Ue depeilence docs uot change with a cuuilinate system, the time de- pendence expressed in Eq. (2.6) is the same for all coordinate systems. Therefore, ‘we shall henceforth 1estrict our effort to seeking solution to Eq. (2.50). Notice that it k = 0, the time dependence disappears and Eg, (2.5b) becomes Laplace's equation mor cRcrRESSLLC2.3. Separation of Variables in Rectangular Coordinates In order not to complicate things, we shal first consider Laplace's equation in two ‘dimensions and later extend the idea to wave equations in three dimensions. 2.3.1 Laplace's Equations ‘Consider the Dirichlet problem of an infinitely long rectangular conducting trough ‘whose cross section Is shown in Fig. 2.1. For simplicity, let three of its sides be insulting wap 0 bo Figure 2.1 Cross section of the rectangular conducting trough. ‘maintained at zero potential while the fourth side is at a fixed potential V,. This is a boundary value problem. The PDE to be solved is. ev en subject to Diriehiley boundary conditions vo,»=0 28a Via.y=0 (2.80) VG,0)=0 280) Ve.b)= Vo (28a We let Ver, 9) = XOYO) (29) into Eq, (2.7) and divide by XY. This leads to xe yt tye2.10) ‘where 4 is the separation constant. Thus the separated equations are X" aX =0 ey vr iav-o eng, ‘To solve the ordinary differential equations (2.11) and (2.12), we must impose the boundary conditions in Eq. (2.8). However, these boundary conditions must be trans- arated equations. Since V = XY, > x@=0 2.138) = xw=o 2.130) eG) 2.130) > xXwrH=% 2.134) Notice that only the homogeneous conditions are separable, To solve Eq. (2.11), we distinguish the three possible cases: 4 = 0. > O,and i <0, Case 1: If. = 0, Bq, (2.11) reduces to x” or eu) ‘hich has the solution X@)saxtar 15) ‘where a} and az are constants. Imposing the conditionsin Eq, (2.13a) and Eq. (2.136), (Ohms Oe hea X(a) 0 Cate Hence X x) = 0, trivial solution. This renders case A = 0 as unacceptable. Case 2, Bq, 2.1) bocomes ex=0 2.16) ‘withthe corresponding auxiliary equations m? — a? — 0 or m — a. Hence the general solution is Xa be + hye" ey) mor cRcrRESSLLCx sinh ex + bs coshax (218) ‘The boundary conditions ave applied to determine by and 24. OO) Orit ba 0 X@=0 + b=0 since sinhcex is never zero for a > 0, Hence X(x) = 0, a trivial solution, and we ‘conclude tat case > 0 iy ut vali. Case 3: IFA < 0, say 4 = — 6, X"+ px =0 2.19) with the auxiliary equation m? + £7 = 0 or m = jf. The solution to Hq, (2.19) 18 X= Aye + Apel «2.204 or X = Bi sin Bx + Byeos px 2.206) Again, > fh + sinfa =0=sinnx neh 23 22n ‘since By cannot vanish for nontrivial solutions, whereas sin fa can vanish without its, argument being zero, Thus we have found an infinite set of discrete values of 4 for which Eg. 2.11) has nontrivial solutions, | eee ‘Those are the eigenvalues of the problem and the corresponding eigenfunctions are int ea Xq{x) — sin px — From Bg, (2.22) note that itis not necessary to include neyative values of since they lead to the same set of eigenvalues, Also we exclude n = 0 since it yields the trivial mor cRcrRESSLLCsolution X = 0 as shown under Case 1 when i = 0. Having determined 2, we can solve Eq, (2.12) to find ¥,(y) corresponding to X,(x). That is, we solve 0, em ‘which is similar to Eg. 2.16), whose solution is in Eg. (2.18). Hence the solution to Bq. (2.24) has the form nay nay ¥,(09 och 2.25) 29 sinh Imposing the boundary condition in Bq. (2-13c), YO=0 + by=0 so that in sin = 2.26) 0) Substituting Egs. (2.23) and (2.26) into Eg. (2.9), we obtain Vola, 9) = Xn D¥n (9) = ay sin = 7 2 which satisfies Eg. (2.7) and the three homogeneous boundary conditions in Eqs. (2.8a), (2.8b), and (2.8c). By the superposition principle, a hnear combination of the solutions V,, each with different values of r and arbitrary coefficient dy, is also a solution of Kg. (2.7). ‘Thus we may represent the solution V of Ki. (2.7) as an mntinite series in the function Vp, ie. vony Soo sin 2 inn ME 0%) ‘We now determine the coetticient a,, by imposing the inhomogeneous boundary condition in Eq. (2.84) on Eg, (2.28). We get Sen sin ME cin BP 0%) Vorb = VM, ‘which is Fourier sine expansion of V,. Hence, aysih22 22 fYysin 2 ax = 22 cosn Tih = ) oe Mt oad lee ae 20 0 ca mor cRcrRESSLLCSuis tion of Eq. (2.30) into Eq, (2.28) gives the complete solution as, 4Vp sin sinh = Vanya yee 31a By replacing n by 2k — 1, (2.31a) may be written as ara Very = Se EE ned 2a1b) 2a sinh 2.3.2 Wave Equation ‘The time dependence has been taken care afin Section ?.2, We are left with solving the Helmholtz equation vu seu =0 2.50) In rectangular coordinates, Hg. (2.50) becomes uu Uy prt arigrthus0 We let Vex, 2) = XOOYOIZE 233) ing, By. (2.33) int Bi (2.32) and dividing by XYZ, we obra vo ora — oyK ztytst Each term must be equal to a constant since each term depends only on the corre: sponding variable; X on x, ete. We conclude that 234), i y . Sak, Taw 235) so that Ea, (2.34) reduces to 0236) Notice that there are four separation constants f, kr, ky, and kz since we have four variables ¢,.x, y, and z. But from Eq, (2.36), one is related to the other thrce so that only three separation constants are independent, As mentioned earlier, the number of Jnndepeudent sepatalion vonstanis is generally oe less than Khe nuiubes of iepeandenL variables involved, The ordinary differential equations in Eq, (2.35) have solutions X= Ayelet + Age the (2.37a) mor cRcrRESSLLC2.37) 4 Agel @3te ¥ = Bysinkyy + Bycoshyy, (2.378) Za Aselle + Age i (2.37) Z = Bssink.r + Bo cosk.z 23TH (2.38) then Ula, yz) = del 2.39) vir) = Ae 240) Introducing the time dependence of Fa. (2.68) gives Ke, ya = Alon) 241 where 4 — ke is the angular frequency of the wave and k is given by Bi, (2.36). ‘The solution in Eq, (2.41) represents a plane wave of amplitude A propagating in the direction of the wave vector k = bea, + yay + ka with velocity Example 2.1 In this example, we would like to show that the method of separation of variables is not limited to a problem with only one inhomogeneous boundary condition as presented in Section 2.3.1. We reconsider the problem of Fig. 2.1, but with four inhomogeneous boundary conditions as in Fig, 2.2(a. Solution ‘The problem can be stated as solving Laplace's equation (2.42) aeFigure 2.2 Applying the principle of superposition reduces the problem in (a) to those in (b). subject to the following inhomogencous Dirichlet conditions: Ve.O=% Verb = 5 VO») = Ve vay =" 2.43) Since Laplace's equation is linear homogeneous equation, the problem can be sim plified by applying the superposition principle. If we let VisVe+ Vir + Yur Viv, 24a) ‘ve may reduce the problem four simpler problems, each uf which iy associated with ‘one of the inhomogeneous conditions. The reduced, simpler problems are illustrated in Fig, 2.2 (b) aml stated as follows: (24s)subject 10 subject to Viitx.0) =0 Vri0,b) =0 ¥1110,9) =0 Vila, y) = Vas evi, PV aye 7° subject to Vu1le.0)=0 Vira, 6) = V5 Vi y= 0 Vinita,» =05 and iw 0 ae subject to Viv(e,0) =0 Viv(x,6) =0 Viv@.9) = Va Vivia, 9) =0 (2.46) ean (2.48) 2.49) est) 2.52) Itis obvious that the reduced problem in Eqs. (249) and (2.50) with solution Vj) is the same as that in Fig. 21. The other thee reduced problems are quite simi Hence the solutions Vs, V1, and Viy’ can be obtained by taking the same steps as inSection 2.3.1 or by a proper exchange of variables in Eq, (2.31). Thus yet yee @s3) ee es x smh sod ie realy inh Vg sin 252 sin west yt eee 2.56) = ou, nsinh =p We obiain the complete solution by substituting Eys. (2.53) w (2.56) in Eq, (2.44) Feample 2.2 Find the product solution of the diffusion equation a Dy, Oecd, 120 257) subject to the boundary conditions e000 OUD, 10 (2.58) and inital condition %(x,0)=Ssin2xx, O a, outside the cylinder, V must be finite asp > co so that Ay = 0 1m Eq, (2.70) for this ease. By imposing the boundary condition in q. (2.08) and following the same steps as for ease p 2B) OE pt a. op) trae tee where n = 0, 1,2,... and? is the separation constant, Thus Fler Fr so and pa gale Dividing both sides of Eq. 2.84) by p? leads to A (48). (pe 2 pRdp i dp ion ' where 1? is another separation constant. Hence and i rwetet Paw, te duce separated equations (2.83), (2.85), and (2.86) become F"srF=0, Bh RoR (i ‘The solution to Bq. (2.88) is given by F@) = cel" + eye F(G) = c3sinnd + cycosnd mor cRcrRESSLLC 2.83) 2.84) 2.85) 086) (2.87) 2.88) 2.89) 2.90) 2.914) 2.91)Similarly, Eg, (2.89) has the solution 22) = ese! 4 egeI™ (2.924) Z(2) = er Sinnyt + cg cosnge (2.92) To solve Ea, (2,90), we let x = hp and replace R by ys the equation becomes Py" tay +6? —Wy=0 (2.93) ‘This is called Bessel’s equation, Tt has a general solution of the form YO) = BLT) + EF VO) (2.94) where Jx(x) and Yo(x) are, respectively. Bessel functions of the first and second kinds of order and real argument x. Yj is also called the Newman function, IF in Eq, (2.93) is imaginary so that we may replace x by jx. the equation becomes Py" aay 0? +My =0 (2.95) which is called modified Bessel's equation. This equation has a solution of the form G0) = Daly (x) + baKy (2) (2.96) where Ja(8) an Ky (3) ate respectively modified Besset functions of the frst anxt second kind of order n, For small values of x, Fiz. 2.5 shows the sketch of some typical Bessel functions (or cylindrical functions) Jp(x), Yn(X). In (X), aNd Kp). ‘To obtain the Bessel functions from Eqs (2.93) and (2.95), the method of Frobenius 4s applied, detailed iscussion Is found It Rerstew (3) abl Mylni-U [3]. FOr tke Bessel function of the first kind, yes = rant eon arated ‘where (+ 1) = 1 is the Gamma function. This is the most useful of all Bessel functions. Some ofits important properties and identities are listed in Table 2.1. For the modified Bessel function of the second kind ee Ggyjayrtn i" nix) = J (2.98) I) mile tm + 1 For the Neumann funetion, when » > 0 ymcuyayntom mila m+ 1 [pm + pint] (2.99) m=(oh o.n=0,12 a0 corona Figure 2.5 Bessel functions, where y = 1.781 is Euler’s constant and to 0100) In=0, 2 ye 2 DMC" Yo) =A poyn 2 +2 EP vom 2.10 7 ao the modified Rescel fnction ofthe second kind SP Und + MoO] 2.102) mor cRcrRESSLLC‘Table 2.1 Properties and Identities of Bessel Functions! Jy (x) Ion) = (=D In 0) Jp(=2) = (=D Jnl) (©) Ina) = By) = Jy1) (recurrence formula) @) edu) = Hn 1) = Jnr) © LL" Jn] = 2"Jyae) © ALON = ny) @ Jabs) = 3 fF cosin® —xsin0)d0, 20 (h) Fourier Bessel expansion of f(x): f= uae ex), 220 Ak — AfO)nOw dx, O0, poet x Cote=m= Der phy 2 Lys pen em) 292] 0. tS 2 pou sm sm24] en00 anditn = fate in a yw cant 2.104) a Other functions closely related to Bessel functions are Hankel functions of the first and second kinds, defined respectively by HP (x) = Jnl) + Ya) (2.1058) Ha) = Ine) = Yn) 2.1050) ‘Hankel functions are analogous to functions exp(-4 jx) just as J, and ¥, are analogous mor cRcrRESSLLCto cosine and sine functions. This is evident from asymptotic expressions In) E00, \Zeoner-nny2— ai), (2.106a) HMG) x > os, (Zoic nx/2 2/4}, (2.106¢) HO) p00, (Zest-ite—m a1, (2.106d) I) 00, Raa : (2.106e) Ky) nee, = (2.106) With the time factore’®. HS? (x)and H'(x) represent inward and outward traveline ‘waves, respectively, while Jy(x) or ¥y(x) represents a standing wave, With the time factor 22%, the roles of Hi) (x) and H,°) (x) are reversed. For further treatment of Bessel and related functions, refer to the works of Watson [6] and Bell 7] ‘Any of the Bezzel functions or related funetions ean be a solution to Bq, (2.90) depending on the problem. If we choose R(p) = Jy(x) = Jp(p) with Eqs. (2.91) and (2.92) and apply the superposition theorem, the solution to Bg, (2.66) is VOW = LT AmaOvienLing L jue — 2.107) Introducing the time dependence of Eq, (2.68), we finally get (0.8.2.0 = OY Ami dnp) expe ind + juzteot), — (2.108) where o = ke Example 2.3 Consider the skin effect on a solid cylindrical conductor. The current density distri bution within a good conducting wire (@ /«ve >> 1) obeys the diffusion equation ‘We want to solve tis equ wire ofragioso, mor cRcrRESSLLCSolution ‘We may derive the diffusion equation directly from Maxwell’s equation, We recall, that VxHad+ds where J =o isthe conduction current density and J = 42 isthe displacement ‘current density. For oo >> 1, Ji is negligibly small compared with J. Hence vvHAS © 109) Also, ee Hex Since ¥ 2.110) Replacing E with J/o, Ba. (2.110) becomes VY = po Quy Janos which isthe diffusion equation, AAconming harmonic fold with time Factor 2/0, Joust enn For infinitely long wire, drical coordinates: (2.112 reduces toa one-dimensional problem in cylin- 1a (an (68) = jones pap Pap Oi +p, or I! — jouop?l, =0 2.113) ‘Comparing this with Eq. (2.95) shows that Eq, (2.113) is the modified Bessel equation ‘of zero order. Hence the solution to Eq, (2.113) is Jel) = clap) + erKo(AP) ay where ey ad e2 ale constants and d= Jono = j'” e115) mor cRcrRESSLLCand 5 s the skin depth, Constante must vanishif J: is tobe fnite at p =O. Ato Je(a) = ¢4lo(2a) > 1 = J-(a)/lo(2a) Thus 1000) Jp) = bee 2.116 0) = eye 2.1106) Tfwe let 2p 1/2, it is convenient to replace To(Ap) = Inj") = Jotxe!*/4) = heros) + fbeios) ean) ‘where bern and beig are ber and bei functions of zero order. Ber and ber functions are also known as Kelvin functions. For zero order, they are given by SS cost /2)(4/2)" berg(xy = Yo SGP” 2.118) 2.19 Using ber and bei functions, Eq (2.116) may be written as berate) + jeio®) Jeo) = Lea) OF Heeiol, 2.120) = HOF erainy+ ibeio(W) ani where x = V2p/8, y= 24/8. Example 24 [A semi-infntely long cylinder (z > 0) of radius @ as its end at z= 0 maintained at Vo(a? — p2), 0 © p Sa. Tind the potential distribution within the cylinder Solution ‘The problem is that of finding a function V(p,z) satisfving the PDE #vy lav ey ae * pap | at ¥ subject 1 the boundary conditions @ VAMC? =p), 2-0, Nepean, Gi) V + 0asz > 00, ie., V is bounded, (ii) V =00np mor cRcrRESSLLCGv) V is finite on p = 0. Let V = R(p)2(@) and obtain the separated equations Zo -3z=0 (2.1220) and ORT + eR +e OR (2.1220) ‘whore 4 isthe separated constant. The solution to Bq. (2.1224) is n= tee 123) ‘Comparing Eq. (2.12%) with Bg. (2.93) shows that n = 0 so that Eq, (2.122b) Bessel’s equation with solution R= cyalip) + cs¥o(A0) (2.124) ‘Condition (i) forces ey = 0, while condition (iv) implies cy = U, since Yo(2,p) blows up when p = 0. Hence the solution to Ey. (2.121) is Vip sd = Sone doe) en28) = where A, and 2» are constants to be determined using conditions (i) and (ii). Im- posing condition (ii) on Eq, (2.15) yields the transcendent equation Jo Gna) =0 (2.126) Thus dy ate the positive roots of Jona). If we take Ay a5 the first root, A2 as the second root, etc, must start from I in Eq, (2.125). Imposing condition (i) on Eq. (2.125), we obtain =A Gen nt ¥en. 0) = % (8 = which is simply the Fourier-Bessel expansion of V,(a? — p%), From Table 2.1, propest th, “ Vy (a? — p?) Jone) dy (2.127) aerial #1 0-6") 000 ao , “To evaluate se integral, we utlize property (ein Table 2.1 [eiercrasBy changing variables, x = 2, f Pn 1p )dp == Ina) 2.128) In f phoOp)dp = £1100) 2.129) lo x ‘Similarly, using property (e) in Table 2.1, we may write “y 2a pP'oorde = | “——IpnGprlde 5 bb x» Integrating the right-hand side by parts and applying Eq. (2.128), a PdeOnrd = —AOa)~> | o*N@rrde = [sia — Ja (2) can be expressed in terms of Jo(x) and Jy(x) using the recurrence relations, i.e. property (¢) in Table 2.1 2 Ja) = 0) = foo) Hence [Aernrion% (2-2) r005] en 2 2 dhe ‘Substitution of Eqs. (2.129) and (2.130) into Eq, (2.127) gives 2¥o [M5 (,,0) = Jo(ana) PU Onae LAY GPO 8¥5 aah Anat) since Joya) from Bag, (2.126). Thus the potential distribution is yiven by es dole) 8M Vin. = raGinay Example 25 ‘Aplane wave E = Fee! is incident onan infinitely long vonducting cylinder of radius a, Determine the scattered fieldIncident wave conductor Since the cylinder is infinitely long, the problem is two-dimensional as shown in iy. 2.6, We shall suppress the time factor e/ throughout the analysis, For the sake ‘of convenience, we need to express the plane wave in terms of cylindrical waves, We let ae 16% = Sap dipyer"® ey where dy are expansion coefficients to be determined. Since e!"® are orthogo- al Fanetions, snalipying tah sides of By. 2.191) by oP al negating wer 05 52m gives [f chrsetet = 2not | ‘Taking the mth derivative of both sides with respect to p and evaluating at p = 0 leads to Substituting this into Eq, (2.131), we obtain on (An altemative, easier way of obtaining this is using the generating function for Jy(x) in Table 2.7.) Thus the incident wave may be written as 2 Daltile eam)Since the scattered field £$ must consist of outgoing waves that vanish at infinity it contains Ink) — HYa (ko) = HE'D) Hence BE DD Aah ape!® 2.133) ‘The total field in medium tyne t es while the total field in medium 1 is £, — O since medium 1 is conducting. At p the boundary condition requires that the tangential components of By and 2 be equal Hence + EMD ‘Substituting Eqs. (2.132) and (2.133) into Eq, (2.134), Y [Ecorse + A HPce)] ei <0 From this, we obtain Fol=i)" Inka) Hy (ka) Finally, substituting A, into Eq, (2.133) and introducing the time factor leads to the scattered wave as Ay = Ingkay Hy (koje? i, Da ear wd Pan 2.5 Separation of Variables in Spherical Coordinates Spherical coordinates (r,0,q) may be defined as in Fig. 2.7, where 0 = r= 00, 0 <6 00, V = Ber 0086, it follows that vee) ‘Also since V = 0 when 3. Hence the complete solution is Verb) =P. (r= 5) cose asi) ‘The electric field intensity is given by 192) 2 Jones mor cRcrRESSLLC2.5.2 Wave Equation To solve the wave equation (2.135), we substitute U0r,6.6) = RDHOFEO) 2.153) into the equation. Multiplying the result by 7? s sin?@ 2 a pat we ae ("ar) | ao (ae) + RP sin? @ O/RHE gives 2.154) m=0,1,2,... ‘where m, the first separation constant, is chosen to be nonnegative integer such that U is petit in J, This reyuinement is necessity for physical reasons Ghat ill be evident later. Thus Eg. (2.154) reduces 10 1d (2dR) , La Ra a) TE = “amoae ‘where isthe second separation constant. Asin Eqs. (2.141)t0(2.144), 2 = n(n-+1) so that the separated equations are now on F@) = cre +207 ™® (2.1584) P(p) = c3sinmd + cycosmp (2.158b) Tee let R() = r-'/4¢7), Bg, (2.156) becomes n+ 1/292) we on" leeo Rries[e mor cRcrRESSLLC‘which has the solution RB Ae!Peq (kr) — DE npayakkr 2.159) Functions zp(x) are spherical Bessel functions and are related to ordinary Bessel functions Z,+1/2 according 10 ene Fezet2) 2.160) In Eg. (2.160), Z.41/2(2) may be any ofthe ordinary Bessel functions of half-integer orders Jrpnj200 Yoet/2000sInet/20)s Kne200, Ha). and Hn, ‘while 29 (1) may be any of the corresponding spherical Bessel functions j(¥).n(¥), in Gono (a3, and HP (a), Bessel fantons of factional vider ave, in gene, ssiven by Skike 40) = Va 2.161) 2 PRerod esp Jolx) e0s(ur) ~ J ¥,@) = Seder te 2.102) a sin(vm) LO) = DGD 016 2.164) =| ae where Jy ain Ly ae, respectively, obtained from Bys. (2.161) aid (2.163) by replacing v with -v. Although v in Eqs. (2.161) to (2.164) can assume any fractional value, in out spevific problem, v = A+ 1/2. Sinve Cama Function UF Hal ‘order is needed in Eq, (2.161), it is necessary to add that Qn! Sa nz0 rotate, 2.165) irzeaii emi en ‘Thus the lower order spherical Bessel functions are as follows: ioe = ee gy eae acy eee a ae hd iii im ion EE, race joe yey umes, 4 EA Don, AQ) = — mor cRcrRESSLLCOther z»(x) can be obtained from the series expansion in Eqs. (2.161) and (2.162) or the recurrence relations and properties of z,(x) presented in Table 2.3. ‘Table 2.3 Properties and Identities of Spherical Bessel Functions 0) feenle) = peglneeaa — + Dees] (©) Abs] = —nz9) +.xep-100) (DL 29] = X29 1) (0) ele Men 2] = eng fx Pagoda =x" engi) (@ fag = —x"-2,-10) oH) fe COP de = FA bene (e420) [By replacing H in Eq, (2.157) with y.cos@ by x, and making other substitutions as we did for Eq. (2.143), we obtain “which is Legendre’s associated differential equation, Its general solution is of the form 0.166) YO) = dye PEC +4 w QCD e167 ‘where Pit(x) and Q'"(x) are called associated Legendre functions of the first and second kind, respectively. Equation (2.146) is a special ease of Eq, (2.167) when m= 0, P(x) and Q!(x) can be obtained from ordinary Legendre functions P(x) and Q, (x) using 2.168) and 2.168) where =I << 1, Wennote that Phe) = Po) QB) = Onto) PE) =0 for m>n 2.170)‘Typical associated Legendre functions are: Ploy = 0-2)" asin, Phx) = 3x(1 = 2°)'? = 3c0s0 sind , Pky) = 30-22) = 3sin?@, (=P Sx =) PR) ol@=a-2)" [ gh=a-2)? [Fm n[3, lex , 2-30? 2 x32 [3 OF= 0-23) [3 rt | Higher-order associated Legendre functions can be obtained using Eqs. (2.168) and (2.169) along with the properties in Table 2.4. As mentioned earlier, Q2"(x) is un- bounded at x = +H, and hence itis only used when ¥ = +1 isexcluded., Substituting Eqs. (2.158), 2.159), and (2.167) into Eq. 2.153) and applying superposition theo- rem, we obtain U0.8,4.1)= LOL dnnccllner P2606) expsimd + jen) — AT sant os Nore thatthe produts (0) #(@) are known as spherical harmonies Example 2.6 A thin ring of radius a eatries charge of density . Find the potential at: (a) point P(O,0, 2) on the axis of the ring, (b) point P(r. 6, #) in space. Solution ‘Consider the thin ring as in Fig. 2.10. (a) From elementary electrostaties, at P(0, 0, 2) pdt V= | izer where dl = adg, R= VaT 7 eT. Hence pads ap. Fede + SNF Fea SVE erry‘Table 2.4 Properties and Identities of Associated Legendre Functions! @ PQ =O, mam (0) Pir) = EArt DD (reeurrne: eatin fo fixed mm) PIN) = BESS RNa) — nm + 2) m= DP? (recurrence relations for fixed n) i aim pay = Hatt?) (hn = 200K" @ Pra) = Ey Hie = Dn = 2k =m! ‘where [] isthe bracket or greatest integer function, e.g. [3.54] = 3, to ger oy = Sennen 0 PPO) = Herm + Din em PEHe) = Pe) © Lem) = MEO (] yxy (22m = Bho ea ee a S Kon = Bn = 2k = my! ty Pm) = — 0-2 Pm) (Wy the series expansion or (xy FEY — Do Ante. = 1 whee Ay = Sa f feeey de nim fe Pae FePats Sao ee eee q Tao PIX) P(x) dx Co eae of OER OE = ET re mm ves ehmeerihaeey= [° AE 1, Properties (b) and (c) are also valid for QP (x). mor cRcrRESSLLC(b) To find the potential at P(r, 4, #), we may evaluate the integral forthe potential as wwe did in part (a). However, it tums out that the boundary-value solution is simpler. So we solve Laplace’s equation V2V = 0 where V(0,0, 2) must conform with the result in part (a). From Fig. 2.10, itis evident that V is invariant with @. Hence the solution to Laplace’s equation is E[ Me 25] [A, Puta) + 8 2nt)] where 1 — C059, Since Qy is singular at 8 — 0, 7, Bj ~ 0. Thus D v= Seu + 25 ] Pw 173) Figure 2.10 Charged 1g of Example 2.6. For0 <1 Sa, Df =O since V must be finite at r = VEL or rc (174) ‘To determine the coeficients C,, we set @ = 0 and equate V to the result in part (a, But when @=0, w= 1, Py(l)'= land r =z. Hence 2 ap OS Cn (“ ¥0.0.2) = aa 3 or (2.175) Using the binomial expansion, the term [a? + <2]! can be written as‘Comparing this with the last term in Eq, (2.175), we obtain oe cash 0, bat bast Ga sGw G=% G=-74GR° or in general, Om Co = Rope Sbstiuug hese nto Bg, (2.178 gives ap (= Rn)n va BP POM Gap, (0s0), Osr a pap Sars aa tndlia adel 2.176) Porr > a, C), = 0 since V must be finite as r + oc, and veyp Few ear =" Again, when @ = 0,u= 1, Py(1) = V@.0,2) swig 2D oe ery Seay = Using the binomial expansion, the mudale term [a* + <*]7*"* can be wntten as fos] "th Comparing this with the last rerm in Eq, (2.178), we obtain Baie? + Rayos 3 Fair+| Do cor in general, pial 1 a err ‘Substituting these into Bg. (2.177) gives a (1M ve ep Der 2 [nin (a/r)" Pay(c0s), ra 2.179) mor cRcrRESSLLCWe may combine Eqs. (2.176) and (2.179) to get a enlr/ay" P(cose), Osrsa vel Fresa/nP™Pye0s0), r>0 im where aye oN map © Example 27 Accondlucting spherical shell of radius a is maintained at potential V,.cos 24; deter- imine the poteutial at any point inside dhe sphere. Solution Thesolationoths problems somewhat smilartothat ofthe previous problem except that V isa function of. Hence te solution to Laplace's equation for <7 as ofthe form V = SLY Coan cosmd + bn sind) (r/a)" P6082) Foto} Since cos and sinmp are orthogonal fonction, mn dag, #0. Hence atr nn except that Vo.c0s 2p = cos 2 Yara P2cos@) = Van P2), ox = Vo cos which is the Legendre expansion of V,.. Multiplying both sides by PZ (x) ives 2 n+2! f. a mest = Yo fred +1 0-2)! 1 Integrating by parts twiee yields In+1—2! f. an = Vo Pyode 2 2 FD! Ly Using the generating functions for P(x) (ee Table 27 and Example 2.10)it show that (enc = 2P(-1)— RSI (=Iaie ; [none fl mono ao by the orthogonality property of Py(x). Hence on + 2H Fy ay ana = YoOn + VEL [0] and . V = Vecos26 en + ef — IF] ray" PR(cose) Example 2.8 Express: (a) the plane wave /* and (b) the cylindrical wave Jo(p) in terms of spherical wave functions. Solution (a) Since e/* = e!” °° is independent of ¢ and finite at the origin, we let = Doawintr)Paleose) e180) s Where ap are the expansion coefficients. To determine dg, we multiply both sides of Eq, (2.180) by P,(c0s@) sin? and integrate over <6 7: irs in a0 = Sag te) [Pele Patadde [fete eters) sinodd = So acitr) [ PoerPaterd 0, nam srdnh(r), nem where the osthogonality property (i) of Table 2.2 has been utilized. Taking the nth « [co on, (cos sinod0 = = an Fin lo sh “he leichand side of Ba, (2.181) yields wf ge aren? 2" Py (x) dc = S— (2.182) # [noe Bem ore To evaluate the right-hand side of Fg. (2.181), we recall that V2e yet i= re V3 2 wait en = 3p EmEHence ” at no eee ers a dey ah) AV 2 Tas Dee Ont Hi C18) Substituting Eq. (2.182) and (2.183) into Eg 2.181) gives 4_ = j"Qn +1) Thus Jr08 = 5 an + 1 jg) ele088) css) = (b) Since Jp(p) = Jo(r sin8) is even, independent of g, and finite at the origin, Jo(p) = Str sin6) = Sb jt?) Phn(c0s0) ©. To determine the coefficients of expansion bp, we multiply both sides by Hy (cos #) sind and integrate over 0 <@ <2. We obtain « 0, m #20 Jo(r sin@) Py (cos®) sin@ 4 = } by ine ien(), m= 2n Differentiating boil siden 2n tines widh reaper ior aint setting + = O aves CD"tn 4 Dn — DF » Pain = Dt Hence S (=1"(4n + On = Dt FNM eyratcosey 3 2.6 Some Useful Orthogonal Functions Orthogonal functions are of great importance in mathematical physics and engi- neering. A system of real functions y(n = 0,1, 2,...) is said to be orthogonal ight w(x) on the interval (a,b) if » freon eae 86) mor cRcrRESSLLCfor every m Aan. For example, the system of functions cos(n.x) is orthogonal with ‘weight | on the interval (0, 1) since [comeconeds=0, mb Orthogonal functions usually arise in the solution of partial differential equations ‘governing the behavior of certain physical phenomena. These include Bessel, Leg- endre, Hermite, Laguerre, and Chebyshev functions. In addition to the orthogonality properties in Eg, (2.186), these functions have many other general properties, which ‘will be discussed briefly inthis section. ‘They are very useful in series expansion ‘of functions belonging to very general classes, e.g, Fourier-Bessel series, Legendre series. etc. Although Hermite, Laguerte, and Chebyshev functions are of less impor- tance in EM problems than Bessel and Legendre functions, they are sometimes useful and therefore deserve some attention, An arbitrary function / (2), defined over interval (a, b), can be expressed in terms of any complete, orthogonal set of functions: Fa) = Abo) e187) i= where the expansion coefficients are given by A= & [wo feoon()de 2.188) and the (weighted) norm N,, is defined as Nee f weyeandr 2.189) ‘Simple orthogonality results when w(x) = 1 in Eqs. (2.186) 10 (2.189). Perhaps the best way to briefly deseribe the orthogonal functions isin table form, This is done in Tables 2.5 to 2.7. ‘The differential equations giving rise to each function are provided in Table 2.5. The orthogonality relations in Table 2.6 are necessary for expanding a given arbitrary function f(x) in terms of the orthogonal functions as in Eqs. (2.187) to (2.189). Most of the properties of the orthogonal functions can be proved using the generating Functions of Table 2.7. To the properties An Tables 2.5 to 2.7 we may add the recurrence relations and series expansion formulas for calculating the funetions for specific argument x and order n, These have been provided for Jy(x) and Yp (x) in Table 2.1 and Egs. (2.97) and (2.99), for Pye) and (Qy(x) in Table 2.2 and Bgs. (2.147) and (2.148), for jn(x) and yp(x) in Table 2.3 and Eq, (2.160), and for Py" (x) and Q(x) in Table 2.4 and Bqs. (2.168) and (2.169). For Hermite polynomials, the series expansion formula is 2 ayknteny2e cnineny Hats) = Me 2.190) fm Rn 2k)‘Table 2.5 Differential Equations with Solutions ‘Solutions Jno) Bessel functions ofthe frstkind Yy(X) Bessel functions ofthe second kind HA (0) Hankel functions ofthe rst kind HE\(x) Hankel functions ofthe second Kind In(2) Modified Besse! functions ofthe fist kind Ky(x) Modified Bessel fanetions of he second kind ints) Spherical Bessel functions ofthe nest kind n(2) Spherical Bessel functions fie sewn ind (= 22)y"=2xy-t n(n Dy =O Pyle) Legendre polynomials Quox) Legendre functions of the second kind (= y = any Pipex) Associated Legendre + [nom +D- #5)» =0 polynomials OMCs) Associated Legendre functions of the second Kind Hy(x) Hermite polynomials ay” $y tay T(x) Laguere polynomials ay" $m 1a)! tn =0 LIN) Associated Laguere polynomials a T,X) Chebyshev polynomial ofthe first kind U2) Chebyshev polynomiaks ofthe second kind where [0/2] = N is the largest even integer <= 1 oF simply the greatest integer function. Thus, Ho) = 1, Hy@) = 2x, Hale) = 4x? ete. ‘The recurrence relations are Hos) 2a Hy (x) = 2m Hy -1(0) (2.191a) mor cRcrRESSLLC‘Table 2.6 _Orthogonality Relations Functions Relations esse fanetions fp xno de = Fl Iovasa Pj where and 2 a the rots of Jy (ha) Spherical Bessel functions [°5. jn(X) mx) dx = FTI mn Legene polynonals PAE nde = Pm ant Ph, PROP) de = EO in polynomials [i EOS ax = Le be Assoctated Legendre Hermite polynomials f°S.e78 Hy (x) Hm (x) abe = 2m VDInn Laguerre polynomials {5° e¥ L(x) Lag) = dan Associated Laguere f° e*NELA ADEA, yd = iy polynomials 0 msn Chebyshev polynomiats —f!, HOA de = 1/2, mon HO m, m=n=0 0, m pn UaleWa Leela dy = | /2, man AO 0 and. Hykx) = 2n Hy) 2.1916) For Laguerre polynomials, La) = 19) = so that Lo =1 Lie ate ‘The recurrence relations are Lan() =n +1 = 2)Ly (x) — Ly (2) (2.1938) mor cRcrRESSLLC‘Table 2.7 Generating Functions Functions Bessel function Legendre polynomial Associated Legendre polynomist ‘Hermite polynomial Laguerre polynomial Associated Laguerre polynomial ‘Chebyshev polynomial ‘Generating function 12x eet? e[5 (t= 1 ia ula eee omni -an2 emg ne serena) expetx 2) = Dow ex Hl-O1 2 FAL (2) = sx /-91 - 9 ota) LE =e) +2 PTO) and a 1 2 ' Bray =F [ota = e200] 2.103) For the associated Laguerre polynomials, an tmnt 1) = 0 Epes) =O te e194) ae Akt = bin +b so that 2 2 L@)=-x+2, GOd= > -3r43, Ye = 5-41 46,et0 [Note that Ly-(x) = 0, m > n. The recurrence relations are 1 Lae) = ay lOn tm + LUN) — fe +MLE eo] C.195) mor cRcrRESSLLCFor Chebyshev polynomials ofthe first kind, wa RR nee rece rE: a > 10 ee HESS (2.196) sorta To(x Ti) T(x) ete. ‘The ecuuente relation in Tuasho) = 2070) = Te 100) e107 For Chebyshev polynomials ofthe second kind, Nn $Date = -lsxrsi 198) where N= [24+] so that Vole) 1 Ue) ‘The recurrence relation is the same as that in Eq. (2.197). Tor example, ia function (4) isto be expanded on the interval (0, 2), Laguente functions can be used as the orthogonal functions with an exponential weighting function, ie., w(x) = ew". IF f(a) is to be expanded on the interval (=vv, 20), ‘we may use Hermite functions with w(x) As we have noticed earlier, if J) 1S denned on the interval (~1, 1), We May choose Legenare tuncuons with w(x) = 1. For more detailed treatment of these functions, see Bell [7] or Johnson and Johnson (8) Example 2.9 ‘Expand the function feb -1sxs1 ina series of Chebyshev polynomials. Solution ‘The given function can be written as -lsre0 fo= Oexrsl f0)= Yo ArT) = mor cRcrRESSLLC‘where A,, are expansion coefficients to be determined, Since f(x) isaneven function, the odd terms in the expansion vanish, Hence F (0) = Ao SY Aze Fan(s) If we multiply both sides by w(x) T terms in the summation vanish except when m =n, That is, ftom Table 2.6, the ‘orthogonality property of 7,(x) requires that and integrate over -1 < x < I, all 1 0 [Be aaa 10-2) , Hence 77 sont [LOB =? 7h 2) rynd fa aS f 7h a2) rh (a2 Since T(x) = cos(n cos” x), itis convenient to let x = cos @ so that 0 cos @ cos 2nt af? l a Srj2 ‘sin® a ££ Noon we een - yma = 22 zh 3 ras ves f@) ! ange 2.10 Phx) Evaluate E> at = Land x 0 sotmon This example serves to illustrate how the generating functions are useful in deriving, some properties of the corresponding orthogonal functions, Since Pi) PA) sind i=”direct substitution of x = 1 or x = —1 gives 0/0, which is indeterminate. But P(x) = (1 —x2)!2.4 P,, by definition. Hence Play _ od sine ~ dx ive, te problem iy reduced to evaluating oP ele at 4 = $1. We uve the generat function for P,, namely, (1-20 +e) Differentiating both sides with respect to x, 2.199) When (2.200) But (98 = 1436467 +108 +154 + Yeotpe! 20 Comparing this with By, (2.200) clearly shows that a > ln] snort D2 ‘Similarly, when x = —1, Eq, (2.199) becomes 2.202) But (es 1-3-4607 = 1008 + 1574 = DeGias dang mi Hence =CD nas p20 mor cRcrRESSLLCExample 2.11 ‘Write a pmgram to uenerate Hermite fmetions Hy (x) for any argument x and order ‘n, Use the series expansion and recurrence formulas and compare your results, Take 05020215 Solution ‘The program is shown in Fig, 2.11, Equation 2.190) is use forthe series expan ‘method, while Bq, (2.191a) with Hg() = Land H) (x) = 2v isused forthe recurrence formul, Note that inthe program, we have repkiced m by n — 1 in Eq. (2.191) so that Hy) De Hy) — 2 — 1) Hy-209) ‘The result of the computation is in Table 2.8. Tn this case, the two methods give results. In general, the series expansion method gives results of greater accuracy since error in one computation is not propagated to the next as is the case when using recurrence relations. Table 2.8 Results of the Program in Fig. 2.11 ‘Values of HC) fr 3, O, 2 rr, IPFERENCE /) waurre(e) 10 60 H20 AMAL sun = 0.0 SRE GheaTESt(n2) Cain ricronraLon, em) CALL FACTORIALCR 0) Fe CEE eater e CG anew 9/C HE FRGFM RAY BE 700 LARGE TF 9 1S LARGE sca) = sum mca) = 1.0 Tra) 40,40,20, Bn. 30 Tea ERG) = 2'OexemR(r-1) ~ 2 +rLoarcr-a)eanca-a) CoNTENUE WRITE(6,S0) Hy ESCH) ACB), DIFFERENCE PRINT #7185 (0D, ARGH) DIFFERENCE Ponway(dx 22,9x,F12-2/98,752-2,98.719.2./) Eorrtnue arret6,0) Program for Hermite function H, (x) (Continued). mor cRcrRESSLLCoo 003 oot SuMMUUtIaE PACAUNIALGR ER) 006 me 10 coor TFG_29.0) @o 10 20 08 Dodo Toa a 008 Peo reerioarcD) cot 20 (RETURN ool =D 002 ¢. SunnGUTINg FoR cALOWLATING THE GREATEST 1 coos © © M= (A) WERE T= a/2 IN THIS PARTIGULAR CASE 2008 ome mae S00 as maxye 008 TF) 10,20,20 0009 100 en out > Figure 2.11 (Cont.) Program for Hermite function Hy(x)- 2.7. Series Expansion As we have noticed in earlier sections, partial differential equations can be solved with the aid of infinite series and, more generally, with the aid of series of orthogonal functions. In this section we apply the idea of infinite seties expansion to those PDEs in which the independent variables are not separable or, if they are separable, the boundary conditions are not satisfied by the particular solutions. We will illustrate the technique in the following three examples. 2.7.1 Poisson’s Equation in a Cube Consider the problem av av ey et ans 2.203) Vv = Set art Gra iene 2.203) subject to the boundary conditions VO, y,2) = Via, y,2) = V(x, 0,2) =0 V(x, 6,2) = VOx,¥,0) = Vr, ¥,6) =O (2.208) where fC, 9,2), the source ferm, is given. We should note that the indepen- dent variables in Eq. (2.208) are not separable. However, in Laplace's equation, mor cRcrRESSLLC(2, y52) = 0, and the variables are separable. Although the problem defined by Eqs. (2.203) and (2.204) can be solved in several ways, we stress the use of series ‘expansion in this section, Lt the solution be of the form mx ney pee Vony an sin sin ED in PRE 2.205) ‘where the triple sine series is chosen so thatthe individual terms and the entire series ‘would satis the boundary conditions of Eq, (2.208). However, the individual terms > b, region 1 is similar to parallel-plate problem. Thus, we have a one~ «dimensional problem similar to Eq. (2.14) with solution Vy =ary +a Since Vi(y =0) =O and V2(y = —b/2) = Vo, a2 = 0, a) = -2Vo/b. Hence Vice, ») ay (2.215) the series expansion solution is ofthe form Vals, y= Yo An sin 2 sin (a2 — 29, (2.216) As ‘which satisfies Laplace’s equation and the boundary condition along the box, No- tice that the even-numbered terms could not be included because they do not sat~ isfy the boundary condition requirements about line y = 0, ic, Ey(s = 0) =AV2/4Y|yag 0. To determine the expansion coefficients Ay in Bq. (2.216), We utilize the fact that V must be continuous at the interface x = w/2 between regions 1 and 2,40. Vile = w/2, y) = Vale = w/2, 99 which is Fourier series. Thus, BV, sin = An sinh ta — Henee 8V, sin 2 a e2rm sinh Fa = w) is instructive to find the capacitance per unit length C of the trip line using the fact, that the energy stoned per length is related to © according 40 w—tev3 2.218) where wed foowan= Sef wean ea mor cRcrRESSLLCFor region 1, bation Ont na Ve a eave 7 Tae tid Hence Leff? Me ayuen ie ayax =e 2.220) 28s haan OPM Forregion 2 6 TF Ansinh (a/2— xh 605 7* id ne nay FF Ancosh 7en =x) sin sin =e ee [smn a2 — ain? a2 — pcos ca Te in EP) ardy tent? 7 (a/9— x) cosh? SE (a /2.— x) sin ‘whore the doula summation ie wed to shaw that w= ara multiplying twa serias which ‘may have different indices mand n, Due to the orthogonality properties of sine and cosine functions, all terms vanish except when m =n. Thus Fi Bea we pe cme athe i be + cost? (a/2 sae Sh nta2a nx in Te cosy (aw) sinh aw) be Fup» roid Substituting for Ay gives Ds ae coth (a — w) 221) ner W, ‘The total energy in the four quadrants is W = 40) +) mor cRcrRESSLLC‘Thus (2.222) wpedunce of Une losses line an Clip oe oe CC € Cle given by where ¢ = > 108 ms, the speed of light in vacuum, and jy = is assumed. Frample 2.12 Solve the two-dimensional problem vv where Ps —x(y 1) nC fm? subject to Vo.0)=0, VOrb)=Ve, — VO.y) 0 Solution Trwe let 2.2048 subject to Vix.0=0, VAle)= Vo, V0.9) = Via.y) (2.224) and (2.225a) mor cRcrRESSLLCsubject 10 Vax. 0) 0 vate) 0 vay Via,y) (2.2250) By the superposition principle. the solution to the given problem is, =VtNn (2.226) 24) is already found in Section 2.3.1, Wo sit Mens ‘The solution to Ba. sinh 2 ae (2227) ‘The solution to Eq, (2.225) isa special case ofthat of Eq. (2.205). The only difference between this problem and that of Eqs. (2.203) and (2.204) is that this problem is two- «dimensional while that of Eqs. (2.203) and (2.204) is three-dimensional, Hence Sh in I jn MEY a DOD doe sin sin 2.228 VaGe 9 os whe mig (2200 ay fa cant Ane = [inna + oni mith nvr nmy of ff sys an day Lh tenes (y= Deo nCim?, But f(x,y) sin" dedy "fo —sin® xsin = a teed 7 dy = ne (- pre (, can) a2 + teosne - n) ines cos = ( 1)". Substitution of Eq, (2.220) into Eq. (2.229) leads to sa)? | amp? CD ad (1 r]) ea Anan — [Cou je? | ej] ae 1 EC o]) eas Substituting Eqs. (2.227) and (2.228) into Bq, (2.226) gives the complete solution as 4V, A sin M4 sinh vey ate + DD hers an! vy 35 where Ary is in Bq.(2.23). Mt mor cRcrRESSLLC28 Practical Applications ‘The scattering of EM waves by a dielectric sphere, known as the Mie scatter problem due to its frst investigator in 1908, is an important problem whose analyti solution is usually refered to i assessing some numerical computations, Thouelh the analysis of the problem is more rigorous, the procedure is similar to that of Exumple2.5, where scattering due toa conducting cylinder was treated, Our treatment here will be brief; for an in-depth treatment, consult Stratton (10). % 2.8.1 Scattering by Consider a dielectric sphere illuminated by a plane wave propagating in the 2 Where ¢» and dy ate expansion coefticients, fy is the propagation constant in the sphere. The functions Mf” and 4! in Eq. (2.240) are obtained by replacing zn(kr) in Eq. (2.237) by jy): i the only solution inthis case since the field must be finite at he vgn, the wenic uf the spt The unknown expansion coefficients dp, fy, ¢n, ard dy are determined by letting the helds satisty the boundary conditions, namely, the continuity ot the tangential components ofthe total electric and magnetic fields tthe surface ofthe sphere. Thus atr= 2.2410) (2.2416) mor cRcrRESSLLCThis is equivalent to bh + bh = 85. (2242a) Fy + Bh =, 2.242) Hi + Hj = Hj, (2.2420) H+ Hs = Hs, 22824) Substituting Eqs. (2.235), (2.238), and (2,240) into Bq. (2.242), multiplying the re- sulting equations by cos oF sing and integrating over <= < 2m, and then aPh dP, ‘multiplying by or — and integrating over 0 < @ = 2, we obtain Unkka) + ay’ (ka) = ey jn (kr) (2.243) 141 [Rajp(kaY’ + ayes [rah eka] = enpt Kkiajy Kia] (2.243) 1 nla) + bypeahiy (ka) = dyprjn (hx) (2.2436) Kkajp(kary + bok [ah (ka)] = dykrGyaip Gyan (2.2434) Solving Eqs. (2.243a) and (2.243b) gives ay and cy, while solving Eqs. (2.243e) and (2.2434) gives hy and dy. Ths, for 1 = to = tts ma ftp) — fnfelmajn (may 2 aa ieaa naa ay = Ae nina aa: Caan ule (a)Y = HY (ena men} gm tree! — ir ee 224%) 1? (adlmeajghmee)|! — me? jy mene )Y ie o = *$_,— (2.2440) AP eodLmecin mee) — jn reece? (CY ena cana (2.2440) IPO lmantmany — me janet OT Where = ka = 24/h and m = ky /kis te refractive inde ofthe dielectric, which may be real or complex depending on whether the dielectric is lossless or lossy. The primes athe square brackets indcae diferenation with respect the argument Of the Bese function ine te brackets, ie [xza x? = geczs)] To obtain Eqs. (2.244c) and (2.244d), we have made use of the Wronskian relationship. Jnl) Poo] = hx) Exjn(a] = —I/x (2.245) IF the dielectric is lossy and its surrounding medium is free space, KP = emo (we) — jo), P= eee (2.246) mor cRcrRESSLLCso that the (complex) refractive index m becomes Em! — jm" 245 Jae ij (2.247) = Vee yen ‘The problem of scattering by a conducting sphere can be obtained as a special ‘case of the problem considered above, Since the EM fields must vanish inside the conducting sphere, the right-hand sides of Eqs. (2.242a), (2.242), (2.243a), and (2.243d) must be equal to zet0 50 that (¢y = in(a) ne lein(aol [an oy (2.248a) (2.2486) ‘Thus we have completed the Mie solution; the field at any point inside or outside the sphere can now be determined. We will now apply the solution to problems of practical interest. 2.8.2 Scattering Cross Sections Often seattored radiation is mast conveniently meacined by the scattering crass section Qc, (in meter”) which may be defined as the ratio of the total energy scattered. per second W, to the energy density P of the incident wave, i, Beas (2.249) ‘The energy density of the incident wave is given by foip fe (2.250) m2 VE ‘The scattered energy from the sphere is weebne[™ [ [rong rane] Pounds here the star sign denotes complex conjugation and field components are evaluated at far field (r >> a). By using the asymptotic expressions for spherical Bessel functions, we can write the resulting field components as Ege!*? cos S28) (2.251a) Ege!* sin 81(8) 2.251b) mor cRcrRESSLLC‘where the amplitude functions 5; (6) and S3(@) are given by [11] eo ntl an ph, dP, (cos8) 50) = DBE (Be ricosey + EE) rasray ot Onell ae dP (cos) 20) = a 1 {CO3 iprnerenntrir (2,252 ae Dawe (ano ia do meee Ts, 1 jak (iSO + [S0)!* ‘This is evaluated with the help of the identities (10) i (ee ae “ae Pip. )s ade nom and We obtain 2 S7Qn +1) (lon!? + nl?) 0.253) 1 yal ‘Substituting Eqs. (2.250) and (2.25: to Eq. (2.249), the scattering cross section is. found to be Zen 1) (lanl? + Ibo?) 2.254) Ques Sinniay, dhe cross secrion for extinction Qow (in meter) is obined [11] from the amplitude functions for @ = 0,1. 4 eu = FF RESO) 2 pe Sen +1) thy) 2288) Leu = mor cRcrRESSLLCwhere 1S FLO Nah) 02% In obtaining Eg, (2.256), we have made use of Pe ap} a n(n + 1)/2 Ifthe sphere is absorbing, the absorption cross section Qt (in meter?) i obtained from ans = eu ~ Ques 2257) since the energy removed is partly scattered and parly absorbed. ‘A useful, measurable quantity in radar communications is the radar cross section or back-scattering cross section a of a scattering obstacle. It is a lump measure of the efficiency ofthe obstacle in scattering radiation back tothe source (@ = 180°) Wis defined in terms of the ir ze eaten eld as 2.258) From Fg. 2.251), 1G) + [Sym] But Ht Sir) = Site) = "n+ 1) ay — 2 ial where we have used ri an} 5 > -Fa| = SE] - Comes or Thos om aie + Dan bn) (2.259) ‘Similarly, we may determine the forward-scattering cross section (0 = 0") as of =F [ISP + 820?) ‘Substituting Eq, (2.256) into this yields ent Da +h et So mor cRcrRESSLLC2.9 Attenuation Due to Raindrops ‘The rapid growth in demand for additional communication capacity has put pressure ‘on engineers to develop microwave systems operating at higher frequencies. Tt tums ‘out, however, dat at fleyuencies above 10 GH attenuation caused by atmospheric particles can reduce the reliability and performance of radar and space communication Tinks. Such particles include oxygen, ive crystals, rain, fox, and snow, Prediction of the effect of these precipitates on the performance of a system becomes important. In this final subsection, we will examine attenuation and phase shift of an EM wave propagating through rain drops. We will assume that raindrops are spherical so that Mie rigorous solution can be applied. ‘This assumption is valid ifthe rate intensity 1S Jow. For high rain intensity, an oblate spheroidal model would be more realistic (12) “The magnitude of an EM wave traveling through a homogeneous medium (with N identical spherical particles per unit volume) in a distance £ is given by e~”*, where 1y is the attenuation coefficient given by [11] =e a2 y- Eres e261) “Thus the wave is attenuated by 1 A= wiog) 7 A=4343/€ — (indB) The attenuation per length (in B/m) is A=433y BN 4.343°X re s(0) 2.262) ‘Similarly, it ean be shown [11] that the phase shift of the EM wave caused by the ‘medium is ZN m5) Gace ims ndeyy e209) mor cRcrRESSLLCTable 2.9 Laws and Parsons Drop-size Distributions for Various Rain Rates ‘Rain Rate mmvhour) Dep 025 125 25 S$ 125 25 $0 100 150 diameter Percent of total volume (em) 005280 109 73 47 26 17 12 10 10 01 301 371 278 203 11S 76 Sd do dd O15 182-313 32K 310 24S Ik4 125 BR TH 02-30, 135 190 22 254 239 199 139 117 025° 07 49 79 118 173 199 29 171 139) 03 15333 57 IW 128 136 184 197 035 06 11 25 43 82 109 150 161 os. 02 06 10 23 33 67 90 as 02 05 12 21 33 58 os 03 06 11 18 30 oss 02 05 1 17 06 03 05 10 06s 02 07 o7 ‘To relate attenuation and phase shift to a realistic rainfall rather than identical ‘drops assumed 30 far, itis necessary to know the drop-size distribution for a given rate intensity. Representative distributions were obtained by Laws and Parsons (13] as shown in Tuble 29. To evaluate the effect of the drop-size distribution, suppose for a particular rain rate R, p isthe percent of the total volume of water reaching the ‘ground (as in Table 2.9), which consists of drops whose diameters fall in the interval ‘centered in D em (D = 2a), the number of drops in that interval is given by Nc = pN(D) (2.264) “The total attenaton and phase shift ver the entre volume become A= 0.463% 10°F pND)RES BAM) —_ 2.2685) = WL pnwrim seo) ew 2.266) ‘here isthe wavelength in em and V(D) is the number of raindrops with equivo- Jumic diameter D perem. The summations are taken over all drop sizes. Tn order to relate the attenuation and phase shift tothe rain intensity measured in rain rate & (in moun), itis necessary to havea relationship between Wand R. The telationship ‘bined by Best [13], shown in Table 2.10, involves the terminal velocity w (in m/s)of the rain drops, ie, R=u-N- (volume ofadiop) dna? = un Gary R=6rNuD> 10° (mmfhry Thus ND 10-5 2.2671 = ude enn Substituting a2 p AaAn5RY ARES am (2.268) op , 2. ® aipilms) (ew 2.269) where M(D) is in per em’, D and 2 are in em, w is in mjs, p is in percent, and S(O) is the complex forward-scattering amplitude defined in Eq. (2.256). The complex refractive index of raindrops [14] at 20°C required in calculating attenuation and phase shift is shown in Table 2.11. ‘Table 2,10 Raindrop Terminal Velocity Radius (Gm) Velorty nls) 0.025 0s 007s 0.10 A125 os 0175 835 020 87 02: 90 025 92 0275 935 030 95 0325 96 Example 2.13 For ice spheics, plot the normalized back-scateing cross section, op/tu?, as a function ofthe normalized circumference, « = 2ra/A. Assume that the refractiveTable 2.11 Refractive Index of Water Retauve index On 8.960 = 70.1713, 8.936 — 0.2172 8.952 — 70.2648 8.933 — 70.4105, 8.915 — 70.5078 8.858 — 70.7471 8.780 — 70.971 8574 — F130 7.884 — 72.184 7.148 — 2.614 6.614 = 72.780 5.581 = j2.848 40 4.886 — 2.725, 00 4.052 — 2.395 80 3.581 — 72.100 282 — 1.864 160, 2.820 = 1.382 200 2.668 ~ j1.174 300, 2.481 — 70.8466 20°C index of ice is independent of wavelength, making the normalized cross section for ice applicable over the entire microwave region. Take me = 1.78 ~ 72.4 x 107 at oe. Solution From Bq. 2.289), w 2 ~ z S21 On + Cay = by Since w = ka, the normalized back-scattering cross section is Sewers nie ta] eat Using this expression in conjunction with Eq, (2.244), the subroutine SCATTERING to determine oy/sra? for 0.2 Goce 30, CONTENVE. 063 Br ka4aencosumar fern PE fan snan CH) SSINDH PTE ort Pena ® RD AT, ora wate 40) it, Sora 40 FOMATCSE, atewuuTrON =),512.6,41, 49/7 st, Sore PHASE SHIFT =°[F12.6 (11, DEQ/RO 5 core 50 17 or? Sr0P bore ED) G00 © svensessvosessvensosssesssseensenssesesessenes Goose “Using nIE'S. SOLUTION Soot © «THIS SUBROUTINE CALCiLATES THE SCATTERING COEFFIEHTS boos © AND THE FORWARD ScaTTeRtea FUNCTION Goo? SUBROUTINE SCATTERIHG(,m, 80,58) oro CCORPLEEe 80, 3c, SUM, SUM ote {CORPLEK8 3H(O:26) ,3MD(:20) RCO: 20 ,¥DC0:20) faa Srarats acaran} nia. 90) ois DiMEaSt0H (20) (8(20).c(20) ,0(20) ota ore Jorr0"0,1.0) cour 5m = (6:0,0.0) 6020 0022 i ons bo to #1 AMAL fans ean wesai Creat, 19) 028, one near ADS eae mnds 2 3Gp« Da 028 125 dRansabOD = aaeIMDD 029 hap = ayaa ase Bit? Seadeinncn — (ee2)-2ncd-206 ost Bz = EEDeIMDUD > (aca) sINDeRBOD 032 yo B1/82 fave Pa) = weyers) 034 DGD = “"icenr Crema) oosg bese © CALCULATE THE FORWARD SCATTEATAG AMPLTTUDE FUNCTION &(0) 38 Ep Roem 4.0 040 SOME = SUMB + Fe(Cr1.DeeIDeC ACD ~ ott = 10CONTTHUE (Cont) FORTRAN program for Examples 2.13 and 2.14 (Continued. mor cRcrRESSLLCoot So = SUN/2.0. + FORMARD SCATTERING ANPLITUDE FUNCTION foes SBe(CABS( SUMB )/x)+02" "NORMALIZED RADAR CAOSS-SECTION ote ru boas BD 00a ¢"""“SuaRDUrIAE FoR SPHERICAL HANKEL FUNCTIONS W AKD 3D 008 008 SUBROUTINE HANRELCK,¥,8,8D) 2008 IMPLICIT REMLeBUA-d.1,0°2) ,cONPLEeBLHD ‘oor DIMEASLOW 3(0:20) 45(0:20);¥(0:20) (0:20) 008 BIMEASLON 34(0-205,JHDCO 25) ,8(0:26),HDCO'20) 010 PIR = 3.141592054 oot? GULL BRSSEL CE 3.20) 12 SAL BESSELACE iim, 29D) bors Ws oFLoWrG) oot Yap” C betscreri=vesG) - snap >/osracveetE) Sots BOB 2 G30) te anna OF aa. cots YoU) = € DC0S(VePiED®IDCN) ~ gHD(H) /OSIHCY=PTED oot? ED) > ewpixe Joa, - 70M > 0002" SuMRGUTINE FoR SPEERLCAL BESSEL FONCTION J AND JD fons @ «GF REAL ARGUWERT (SERIES EXPANSION METEDD) doa 005 SUBROUTINE BESSEL(X,1,3,.0) ane Twpu tert Beat #R( ton 1-027) 007 DIMENStON 1€0°20),36(6:20) 008 ote PIE = 3 tatsonesa ont V= DRLOMTOD + 0.5 013 sunt = 0.0 ota SUR = 0:0) 0016 ‘ChE cuMACPH GHD oir CLL FACTORIAL FD Soto Be Oa iyweny stds ede C09 268) 9 kee (UBER. 6))/ (GHEE 019 Sums soni? 020 Br aecr + FLONT(2N) + 0.8) boa eee bons TrCBS(A) oF TOL) Go To 20 boas = Dsqur(ere/2.0) 026 Jays gesont oor tosseeseesesesaserssseosssses oes «OF REAL ARQURENT BUT NEGATIVE ORDER (SERIES EXPANSION) oon Figure 2.16 (Cont.) FORTRAN program for Examples 2.13 and 2.14 (Continued). mor cRcrRESSLLCome sae seats 3.9.2) go Eee NEE Bt ao BE Sonaaten «1.2 oe ‘ oat aupyeteeCyea 9/0 See tas SB ruben ray oo 1010 = ae Se WSN a mae Soh g°° Suan sunt ate nets oe Sestalae eavase eaters tl Ss B= oem eon ees Be RET SEER so Se okaeaas eo cee se oa oe SB BG Trae +o. SB HU isto.eraoo ao 1010 sa Eat SH BEER ewys 0 Sat es ae Sh $77 ttn vi Sn vai oS Figere216 (Cont) FORTRAN program for Examples 2. 3 and 2.14 (Continued). mor cRcrRESSLLC2008 ‘unnouryae sana ¢y 0) oo DIDLICIY REALOBC4H, 0-2) toe Te(v-0"8) 10,20,20 G08 10 ews 08. Sora cee cCegrdeen)sBEgaT ODE) oF H/PH2 ona ier cos 20 PSV OS Sete Chui Faron (a, F4) cota (CALE FACTORIAL 3. F42 oro ha rrzepeghe (PTE5/(Fn+(2 2082)? oxo aera oat cH Gooa¢ "**"SuaRQUTINE' FoR Factontae OF ¥y tees ¥t Boos ‘suuaouriae FAcTORTALL.F) oe aie 007 Feo ee8 Trcu'kg.0) 0 70 20 008 pote tata Gero 10 Pee FeOeLOATCED Sess 25 Rerun oor 0 Figure 2.16 (Cont.) FORTRAN program for Examples 2.13 and 2.14. find j,(2). Subroutine HANKEL employs Eq, (2.162) 10 find y,(x), which involves calling subroutine IBESSELN to calculate j-»(x). The derivative of Bessel-Riccati function [xz,(x)] is obtained from (see Prob. 2.14) be Goy’ Ney) + Ey400) where 29 18 jy j-ny 9 OF fin(x). Subroutine GAMMA calculates M0 ++ 1/2) us- Ing Eq, 2.165), while subroutine FACTORIAL determines n!. All computations were done in double precision arithmetic, although it was observed that using single precision would only alter results slightly. ‘Typical results for 11 GHz are tabulated in Table 2.12. A graph of attenuation ws, ain rte is portrayed in Fie. 2.17. The plot shows that attention increases ‘vith rain rate and conforms with the comman rule of thumb. We must noe that the tndelying assumption of spherical raindrops renders the result as onlya fist order approximation ofthe practical rainfall eiaion,‘Table 2.12 Attenuation and Phase Shift at 11 GHz Rain rate (rumhe) Attenuation QB/an) Phase shift (degen) 025 2.56 x 10 0419 1.25 1702x1073 1.655 S 4.072 x 1073 3.040 5.0 9878x103 5.601 25 03155 12,58 5 0.7513 23:19 50 1.740 42.78 100 3947 1899 150 6.189 112.16 A Gt GHz Gir 40 0 1 10 ko 160 R Figure 2.17 Attenuation vs. rain rate, 2.10 Concluding Remarks We have reviewed analytic methods for solving partial differential equations. An- lytic solutions are of major interest as test models for comparison with numerical techniques. The emphasis has been on the method of separation of variables, the most powerful analytic method. Foran execllent, more in-depth exposition of this method, consult Myint-U [5]. In the course of applying the method of separation of variables, we have encountered suine nialematival fuuctivns suck as Bessel functions and Ley cendre polynomials. For a thorough treatment of these functions and their properties, ell [7] and Johnson and Johnson [8] are recommended. The mathematical hand- book by Abramowitz and Stegun [15] provides tabulated values of these functions mor cRcrRESSLLCfor specific orders and arguments. A few useful texts on the topics covered in this, chapter are also listed in the references, ‘Asan example of rea life problems, we have applied the analytical techniques de- veloped in this chapter to the problem of attenuation of microwaves due to spherical raindrops. Spherical models have also been used to assess the absorption charac- teristics of the human skull exposed to EM plane waves (16]-20] (see Probs. 2.46 102.49). ‘We conclude this chapter by remarking thatthe most satisfactory solution ofa field problem is an exact analytical one. In many practical situations, no solution can be ‘obtained by the analytical methods now available, and one must therefore resort 10 ‘numerical approximation, graphical or experimental solutions. (Experimental solu- tions are usually very expensive, while praphical solutions are not so accurate). The ‘remainder of this book will be devoted to a study of the numerical methods commonly used in EM. References [1] WJ. Gibbs, Conformal Transformation in Electrical Engineering. London: ‘Chapman & Hall, 1958, [2] N. Morita eta, Integral Equation Methods for Electromagnetics. Boston, MA: Artech House, 1990. [3] ILD, Weinberger, A First Cowse in Partial Differential Equations, New York John Wiley, 1965, Chap. IV, pp. 63-116. [4] RD. Kersten, Engineering Differential Systems. New York: McGraw-Hill, 1969, Chap. 5. pp. 66-106 [5] T. Myint-U, Partial Differential Equations of Mathematical Physics. New York North-Holland, 1980. IG] GN. Watson, Thevry uf Bessel Functions. Loudon: Cambridge University Press, 1966, [7] WwW. Bell, Special Functions for Scientists and Engineers. London: D. Van Nostrand, 1968, [8] DE. Johnson and JR. Johnson, Mathematical Methods in Engineering and Physics. Englewood Cliffs, NJ: Prentice-Hall, 1982. [9] K. Asano, “Electrostatic potential and fick in a cylindrical tank containing charged liquid,” Proc. IEEE, vol. 124, no, 12, Dec. 1977, pp. 1277-1281 mor cRcrRESSLLC[10] A. Stratton, Electromagnetic Theory. New York: McGraw-Hill, 1941, pp. 394-421, 563-573, [11] H.C. Van de Hulst, Ligh Scattering of Small Particles. New York: John Wiley, 1957, pp. 28-37, 114-136, 284. £121 J. Morrison and MJ. Cross, “Scattering of a plane electromagnetic wave by axisymmetric raindrops,” Bell Syst. Tech. J, vol. 53, no. 6, July-Aug. 1974, pp. 955-1019. [13] DEE. Setzer, “Computed transmission through rain at microwave and visible trequencies,” Bell Syst. Tech, J., vol. 49, no. 8, Oct. 1970, pp. 1873-1892. [11] MONO. Sadiku, “Refractive index of snow at microwave frequenci Optics, vol. 24, no. 4, Feb. 1985, pp. 572 [15] M, Abramowitz and LA. Stegun, Handbook of Mathematical Functions. New ‘York: Dover, 1965. [16] CH. Durney, “Electromagnetic dosimetry for models of humans and animals a review of theoretical and numerical techniques,” Prac. ILL. Vol. 08, 90.1, Jan. 1980, pp. 33-40, [17] M.A. Morgan, “Finite element calculation of microwave absorption by the cranial structure,” IEBE Trans, Bio. Engr, vol. BME 28, no. 10, Oct. 1981, pp. 687-695, (18) J.W. Hand, “Microwave heating pattern Biol., vol. 22, no. 5, 1977, pp. 981-987. [19] W-T. Joines and RJ. Spiegel, “Resonance absorption of microwaves by the ‘human skull,” ZEEE Trans. Bio. Engr, vol. BME-21, Jan. 1974, pp. 40-48. in simple tissue models,” Phys. Med. 20) CM. Weil, “Absorption characteristics of multilayered sphere madels expased to UHF/microwave radiation,” IEEE Trans. Bio. Engr., vol. BME-22, no. 6, Now. 1975, pp. 468176, Problems 2.1 Consider the PDE aby, FIV yy Ty Hdd, Pedy + SH =0 here the coeflicients u, 2, €,d,e, and f are in general functions of « and y, ‘Under what conditions is the PDE separable? mor cRcrRESSLLC0 © w Figure 2.18 For problem 2.2. 2.2 Determine the distribution of electrostatic potential inside the conducting rect- angular boxes with cross sections shown in Fig, 2.18. 23. The cross-sections of the cylindrical systems that extend to infinity in the z= direction are shown in Fig. 2.19. The potentials on the boundaries are as shown, For cach system, find the potential distibution. 244 Find the solution 1/ of (a) Laplace equation VU=0 O0 UON=01>0, UAN=1,t>0 Ua.o=0, OSaEt (©) Wave equation PU Ug OSrstt>0 UO) =0=UC,), 120 u@.0 Uex,0) 2.5. Find the solution © of: (2) Laplace equationvo vo v0 I 0 a Vevy x ® we veo v0 “| | 2 0 : ‘igure 2.19 For problem (b) Laplace equation Vo=0 O0, — &(p, 6,0) = p*c0s29, 05 @ < 2x mor cRcrRESSLLC2.6. Solve the PDE ao Pe Satter a0 O 00) = Bgr cos + Vy Find V and E everywhere. Determine the maximum value of the field strength. 2.28 Ina sphere of radius a, obtain the solution V(r, 8) of Laplace's equation VvGe=0, rsa subject 10 V(a.8) = 3c0s 6 + 30038 +1 2.26 Determine the solution to Laplace's equation wt =0 outside a sphere r > a subject to the boundary condition a FeV (a0) = e058 + 30530 Find the potential distribution inside and outside a dielectric sphere of radius 1a placed in a uniform electte field Ey. Hint: The problem to be solved is VV = 0 suibject to av _ ava a a Eyreos6 as +00 onr =a, y=V2 nr =a, 2.28 (a) Derive the recurrence relation of the associated Legendre polynomials = EE) = non D mim = DIRE) ents (6) Using the recurence relation on the formula for Pind P2, PB, P2. and PZ Expand V = cos29sin?@ in terms of the spherical harmonics Py"(cos#) in mg and P;"(cos#) cosmd. 2.30 In the prolate spheroidal coordinates (£,n, the equation Vo+RO=0 assumes the form a 167, 77,» 36 #le-92)-4[e-9) 3 ]+[ 1 ]opeeee- p)o=0‘Show that the separated equations are dfs. tt ale) Z]-[& 4 f(y) 22) faye 2 ale) el Pere wee anc re separation ont 231 soley, 2209 ita = (a) fe. yz) 2.49 Soke hsnkomogersone PDE 20 156 at ap pap ar snips ot coaiens (0,9) = 0,810, 0) = 0.04640) Faeyr Rotana bee wet ne andl (b) fx ys 5) = si 2.33 Infinitely long metal box has a rectangular cross section shown in Fig 2.22. If the box is filled with charge y= 2ox/a, find V inside the box. veo Figure 2.22 For problem 2.33. 2.34 In Section 2.7.2, find K, and Ke, the electric field intensities in gas and lqui respectively. 2.38 Consider the potential problem shown in Pig. 2.23. The potentials at x = 0, =a, and y =0 sides are zero while the potential at y = b side is Vp. ‘Using the series expansion technique similar to that used in Section 2.7.2, find the potential distribution inthe solution region.v=0-| Figure 2.23 Potential system for problem 2.35. 2.36 Consider a grounded rectangular pipe with the cross section shown in Fig. 2.24. Assuming that the pipe Is partially filed with hydrocarbons with charge density Po, apply the same series expansion technique used in Section 2.7.2 to find the potential distribution in the pipe. os % Figure 224 Earthed reetangular pipe partially filed with charged liquid—for problem 2.36. 37. Write program to generate associated Legendre polynomial, with x = cos 0,5. You may use either series expansion ot recurrence relations. Take 0 < n= 15,0 =m 0 (3.184) and initial condition O(x,0) (3.180) Sol ‘Thisproblem may be regarded as amathematical model ofthe temperature distribution inarodof length £ = 1m with its end in contacts with ee blocks (or held at 0°C) and the rod intially at 100°C, With that physical interpretation, our problem is finding the intemal temperature as fonction of position and time. We will solve this problem using both explicit and implicit methods. (a) Explicit Method For easy hand calculations, let us choose Ax = 0.1, 7 = 1/2 so that As rary 0S XTable 3.2 Finite Difference Approximation tothe Parabolic Equation: So HGR Feo Method Alger Molecule 1. First order (Euler) ots explicable for kan? S03 2. Crk Niholion 5. Leaping x oof! 4. Detore Pranks = explicit, conditionally sabe since k = 1. We need the solution for only 0 << x < 0.5 due to the fact that the problem is symmetric with respect to x = 0.5, First we calculate the initial and boundary values using Eq, (3.18). These values of @ at the fixed nodes are shown in Table 3.3 for x = 0,x = 1, and ¢ = 0. Notice that the values of (0.0) and ‘C1, 0) are taken as the average of O and 100, We now calculate at the free nodes using Bq, (3.14) or the molecule of Fig, 3.400). The result s shown in Table 3.3. The analytic solution to Eg, (3.17) subject to Eg. (3.18) is Oy sinnerexp(-n2et), na 2k+1 = Comparison of the explicit finite difference solution with the analytic solution at 14 is shown in Table 3.4. The table shows that the finite difference solution isreasonably accurate, Greater accuracy can be achieved by choosing smaller values of Ar and Ar ‘Table 3.3. Result for Example 3.1 x 0 Ot 03 04 05. 06. 10 i 0 50 100 100 100-100-100 50 0.003 “0 100 100 100100100, 0 001 0 875 100 100 100100 0 0.0150 75 93.75 100 100 100 0 0.02 0 68.75 87.5 9687 100 96.87 0 0.028 0 625 S281 93.75 9687 93.75 0 0 0 58.59 78.21 89.84 93.75 89.84 O10 14.66 27.92 3839 45.18 4744 45.18 Table 3.4 Comparison of Explicit Finite Difference Solution with Analytic Solution; for Example 3.1 1 Finitedifference Analytic solution Pereentage solutionatx=0.4 atx = 0.4 err (0.005 100 99.99 O01 0.01 100 99.53 0.47 0.15 100 O78 2.2 0.02 96.87 9518 18. owas 9379 visi 2.0 0.03 89.84 88320017 0.035 85.94 3461 16 oo 8203 8088 14 0.10 18 13 ot (b) Implicit Method Letus choose Ax = 0.2, r = I sothat At = 0.04, The values of @ at the fixed nodes are calculated as in part (a) (See Table 3.3). For the free nodes, we apply Eq. (3.16) ‘or the molecule of Fiz. 3.5(b). If we denote @¢i, j + 1) by (i = 1.2.3.4). the values of & forthe first time step (Fig. 3.6) can be obtained by solving the following simultaneous equations =0+ 40) =o. =; $42 +05 2 + dbs — oy =05 4404-0We obtain 58.13, 8254, sy —72, Oy 55.5 at r= 0.04. Using simultaneous equations for ¢ ese values of ©, we apply Eq. (3.16) to obtain another set of 0.08 as 0 +40) =) = 0482.54 WO) + Ady ~ 2 = R134 72 = 02 $405 ~ 4 = 82.544 -03 4404-0 = 7240 which results in = 644, = 55.23, ‘This procedure can be programmed and accuracy can be increased by choosing more points for each time step. 04 06k Figure 3.6 For Example 3.1, part (b).. 3.4 Finite Differencing of Hyperbolic PDEs ‘The simplest hyperbolic partial differential equation is the wave equation of the form aah ant OF G.19)‘where u isthe speed of the wave. An equivalent finite difference formula is ket ha 206, NE HE=A I) _ HE J+ N= 200, N+ HE, an? ane wherex =iAx, Ati, j =0,1,2,.... This equation can be written as OG J+ = 2 —NOE, JP FHOTFH1 N+ O0-1,Ml—O@j—D 320) where @(i, j) is an approximation to ® (x. f) and r is the “aspect ratio” given by uae? ra( 321) (ss) Equation (3.20) isan explicit formula forthe wave equation. The corresponding com- putational molecule is shown in Fi. 3.7(a), For the solution algorithm in Eg. (3.20) © iat © iat a 2) q i) oc i ag i Ht 7 i HI ¢ it 9 % Figure 3.7 Computational molecule for wave equation: (a) for arbitrary < 1,(b)forr = 1. to be stable, the aspect ratio r < 1, as will be shown in Example 3.5. If we choose 1 = 1,Eg. (3.20) becomes OG 4D SOU FL +00 -1,)- 05-1 G2 with the computational molecule in Fig, 3.7(b). Unlike the single-step schemes of Eqs. (3.13) and (3.15), the worstep schemes of Eys. (3.20) and (3.22) seguite dat the values of @ at times j and j — 1 be known to yet ® at time j + 1. Thus, we must derive & separate algorithm (0 “start” the solution of Ey, (3.20) or (3.22), that is, we must compute (7, 1) and (7,2), To do this, we utilize the preseribed initial ‘condition. For example, suppose the initial condition om the PDE in Eq. (3.19) Is, a aWe use the backward-iference formula ame, 0) G1) 0G, =1) an 2At 2 26D = 46,1) eax Subatimting Fe 73) inta Fg (320) and ting j = Ge at = 0), we got OG, 1) = 20 = OG, 0) + FOG = 1,0) + OG +1,0)] — 6,1) or OU. 0-06.04 200-19 400+101 a2H Using the starting formula in Eq. (3.24) together with the prescribed boundary and initial conditions, the value of & at any yuid point (i, j) can be obtained dveetly fiom Eq. (3.20), There are implicit methods for solving hyperbolic PDEs just as we have implicit, ‘methods for parabolic PDEs. However, for hyperbolic PDEs, implicit methods result in an infinite number of simultaneous equations to be solved and therefore cannot be used without making some simplifying assumptions. Interested readers are referred to Smith (5] or Ferziger {6}. Example 3.2 Solva the wave equation, Seb, Oxx ]BP) mine a and g(x, y) = —x(y = 1) = 10-%/eo. For the finite difference solution. it can be shown that ina rectangular region, the ‘optimum over-relaxation factor is given by the sinaller root of the quadratic equa- tion (10) Pe? — 160 +16 =0 where 1 = coster/N.) + c0s(7/.%y) and Ny and NV, are the number of intervals along x and y axes, respectively. Hence — Vea Te We try three eases of Ny = Ny = 4, 12 and 20 so that Ax and 1/20, respectively. Also we set 4, 1/12, Ps _ a= 10% € 107/36 Figure 3.12 presents the FORTRAN code for the solution of this problem. The potentials atthe free nodes for the different eases of ft are shown in Table 3.6. Notice that as the mesh size h reduces, the solution becomes more accurate, but it takes more iterations for the same tolerance. gay) —36rx(y =D000: 008 om 2008 10 012 ors Sore one 0036 oar ‘9049 Seat oes 080 ost 082 ose os? 058 Bose igure 3.12 Conseseseaeessen 20 ° 50 0 Frere pirreneace SOLvilon OF Porsson's nwarson weet Wy = USING THE METHOD OF SUCCESSIVE OVER-RELATATION n 0. OF INTERVALS ALONG Y-axis 7 NO, OF INTERVALS ALONG YoAKZS NCL3) | POPENTIAL AT GRED POINT Gust) = B6(Z) BOER AS Ost cc RE, 3 = Oy, pimasion ¥(0:20,0:30) Sita PTkys.19309/ DATA A B/10,1-0/ BATA Vii ¥2 ¥3,¥4/0.0,10.0,20.0, 19.07 SPECEFY SOUNDARY YALUES AND NECESSARY PARAMETERS FS iuaran Do 10 T=4 aR-1 noone Tet, wet Wero-¥8 Wo,a var, continue V(Ht,o)=(v1 + ¥2)/2.0 Mocs + 149/20 Vora stv2 + "¥39/3-0 rnp Tha OPTIMO OVER-RELAKATION FACTOR 'T = coscrin/an) + cas (PER/aY WE C50. SyRrt6d 0 = 16.0¢T¥T)3/(THTD werTe(6,40) EoumnzCit sk racron om ant 6.9 Do ToT stares 2 getvoaie) Bo"7o se tart Y= WeFLATOD) G = =Se,oertRene(y - 1.0) BE weeCHrst a) 4 vist, o9 + verse) + ¥a,s-1) 1 Mower sd) ~ ete) mun 2 aN IAs (a) Wa) ek covitnve nt 110.0) FORTRAN code for Example 3.4 (Continued).0061 oes e Figure 3.12 EM = aMTH/FLOATORCONY) SnLVTION BAS. CONVERGED oooen = vouun + t 07050 Hise WITE(6 20) FORMAT(GX, ‘SOLUTION ODES NOT CONVERGE IH 100 ITERATIONS?) 0 10 100 soit SOLUTION gis convERGKD suite ce190) HCOUAT FOUUT(Gx, "SOLUTION CONVERGES 14 ,24,12,21,"2TERATIONS?,/) 130 T= isnt DD 4203 = ila-t FORMATCAE, *1=",15,21, conTEHvE 33,2197 9,612.6, CALCULATE Tae Exact SoLUTZON {o1sSUN/S AgUATIUN WAH MURUGEAEUUS BOUNDARY GONDETIONS SOLVED BY SERIES. EXPANSION bo 150.1 =1,%-1 = bertoaT(D) Do tg03= tay-1 r | TAKE ONLY 10 TRANS OF THE SERES DO 130.8 = 1,20 FE = FLOATS Fhciukt 2" Ghevssyaseer + ceueptsyayee2 Pacron2 = ("C-1-0)es(hen) )4144_Oeheb/(PLeFNOFD, FACTORS = 1.0 ~" ("0 = Cr O)eei/b Phere = rhétunzsricraes/FicroRt SOM = SUM + FACTOReSINCERGP I+4/1)6SINCEN®PIEAY/B) vie S00 [LAPLACE’S EQUATION WITH LABONOGENEQUS BOUNDARY. coND:TIOHS, SOLVED USING THE METHOD OF SRPARATION OF ¥ARIASLES cisg.oe/rr8 eet oca/ele he8!00na/Ple (Chet ovva/PTe Dovido K=1,10 | TARE OLY 10 TEAMS GF THE SERIES ahr q AScanesinacune?iten/®) TERAE=CIeh1#82/43) (Cont.) FORTRAN code for Example 3.4 (Continued).eartecsepto02/03 EESIinGdehtie (e510 /0) ESqaNeSIWa (ANePTE*B/A) ‘Temndecaeriok2/83 voyp sve eu Corio urpur XACT RESULTS bo 36035 tai Werte Ce,110)'T,3,¥C1,29 o1a? Ea. Figure 3.12 (Cont) FORTRAN code for Example 3.4. Table 3. Successive Over-relaxation Solution of Example 3.4 haa h=12 h=1720 Node ope = LATL ope = 1.729 ope = 1.729 Exact Solution Biterations 26 iterations _ 43 iterations = =A247 = 3.409 b =1.703 1,982 c 4.305 4.279 a 0.0393 -0,0961 e 3.012 2.928 t 9.368 91556 8 3.044 2.921 b ell 6.072 i 11.04 11.42 3.6 Accuracy and Stability of FD Solutions ‘The question of accuracy and stability of numerical methods is extremely important if our solution is to be reliable and useful. Accuracy has to do with the closeness Of the approximate solution to exact solutions (assuming they exist). Stability is therequirement that the scheme does not increase the magnitude of the solution with inerease in time, ‘There are three sources of errors that are nearly navi ‘of physical problems [8]: + modeling errors, + truncation (or discretization) errors, and + roundoff errors. ‘Each of these error types will affect accuracy and therefore degrade the solution, ‘The modeling errors are due to several assumptions made in arriving atthe mathe- ‘matical model, For example, @ nonlinear system may be represented by a linear PDE. ‘Truncation errors arise from the fact that in numerical analysis, we can deal only ‘with a finite number of terms from processes which are usually deseribed by infinite series. For example, in deriving finite difference schemes, some higher-order terms in the Taylor series expansion were neglected, thereby introducing truncation error. ‘Truncation errors may be reduced by using finer meshes, that i, by reducing the mesh, size h and time increment At. Alternatively, truncation errors may be reduced by using a large number of terms inthe series expansion of derivatives, that is, by using higher-order approximations. However, care must be exercised in applying higher- ‘order approximations. Instability may result if we apply a difference equation of an ‘onder higher than the PDE being examined. These higher-order difference equations. may introduce “spurious solutions.” Roundoff errors refleet the fact that computations can be done only with a finite precision on a computer. This unavoidable source of errors is due to the limited size ‘of registers in the arithmetic unit of the computer. Roundoff errors can be minimized ‘by the use of double-precision arithmetic, The only way to avoid roundoff errors ‘completely it to code all operations using integer arithmetic, This is hardly possible in most practical situations. Although it has been noted that reducing the mesh size / will increase accuracy. itis not possible to indefinitely reduce h. Decreasing the truncation error by using a finer mesh may result in increasing the roundoff error due to the increased number of arithmetic operations. A pointisreached where the minimum total error occurs forany particular algorithm using any given word length (9). This is illustrated in Fig. 3.13 ‘The concem about accuracy leads us to question whether the finite difference solution ‘can grow unbounded, a property termed the instability ofthe difference scheme. A numerical algorithm is said to be stable i'a small error at any stage predluces a smaller ‘cumulative error. Itis unstable otherwise. The consequence of instability (producing, tunbonded solution) is disastrous. To determine whether a finite difference scheme is stable, we define an error, <", which oceurs at time step 1, assuming that there is one independent variable, We define the amplification of this error at time step n + 1 as eth a get G37rouni-ott “SS 4? aseriation mesic ure 3.13 Error as a function of the mesh size. where g is known as the amplification factor. In more complex situations, we have {wo or more independent variables, and Eq, (3.37) becomes tel"? = [)ler” (3.38) ‘where [G] is the amplification matrix. For the stability ofthe difference scheme, itis, required that Eq. (3.37) satisfy vst] or list 3.39) Pror the ease in Bg, (3.38), licnst 3.398) One useful and simple method of finding a stability criterion fora difference scheme is to construct a Fourier analysis ofthe difference equation and thereby derive the amplification factor. Weillustate this technique, owas von Newnannsmethod (4, 5,7, 10], by considering the explicit scheme of Eq. (3.13): Or = 2 OF +r (Of, +974) G40) where + — Ar/k(Ax)2, We have changed our usual notation so that we can use j= Tin the Fourier series Lette solton be op = Panne, Orsi Gata‘where is the wave number, Since the differential equation (3.10) approximated by Eq. (3.13) is linear, we need consider only a Fourier mode, ie. OF = A@el* Galby Substituting Eg. 3416) into Eq. 40) gives Anette 1 an Atel sr (et e-Ht) alt ate "[1 —2r + 2coskx] G42) Hence the amplification factor is an Sr = 1a r+ 2eoskx kx = 4rsin? = 1 4rsin? 3.43) In order to satisfy Eg. (3.394), ‘Since this condition must hold for every wave number k, we take the maximum value of the sine function so that 1-42 = or Of course, r = 0 implies A 3.44) Example 3.5 For the finite difference scheme of Eq, (3.20), use the von Neumann approach to determine the stability condition. Solution We assume a tril solution ofthe form Of = ArelttxSubstituting this into Eg. 3.20) results in API a = pate er (elt per) ater = grote AM 2 AN =r) + 2r coskx] — AN G45) Interms of ¢ = A"*"/A" Eq (3.45) becomes, 2p 1-0 2.46) where p = 1 — 2r sin? 4. The quadratic equation (3.46) has solutions p+[P 1], = e-[e-a]"” 8 For |gj| <1, where whereh=Ax— Ay. Ul 3.7. Practical Applications I — Guided Structures ‘The finite difference method has been applied successfully to solve many EM- related problems. Besides those simple examples we fave considered earlier in this chapter, the method has been applied to diverse problems [1] including: + Transmission-line problems (12}-{21], + Microwave circuit [27-130], + EM penetration and scattering problems (31, 32], + EM exploration of minerals (34), ant EM enerzy deposition in human bodies [35. 36] It is practically impossible to cover all those applications within the limited scope fF this text. Ta this section, we uusides the telalively eanier problents of ats sion lines and waveguides while the problems of penetration and scattering of EM ‘waves will be teated in the next section. Other applications utilize basically similar techniques.3.7.1. ‘Transmission Lines ‘The finite difference techniques are suited for computing the characteristic impedance, phase velocity, and attenuation of several transmission lines—poly gona lines, shielded strip Hines, coupled strip lines, microstrip lines, coaxial lines, and reet~ angular lines (12]-[19]. The knowledge of the basic parameters of these lines is of paramount importance in the design of microwave circuits, For concreteness, consider the microstrip line shown in Fi 3. 14(a). The geometry in Fig, 3.14(@) Is deliberately selected 10 be able to illustrate how one accounts for discrete inhomogeneities (i.e, homogeneous media separated by interfaces) and lines ‘of symmetry using a finite difference technique. The techniques presented are equally applicable fo other lines. Assuming that the mode is TEM, having components of neither E nor Hi fields in the direction of propagation, the fields obey Laplace's ‘equation over the line cross section. The TEM mode assumption provides good approximations if the line dimensions are much smaller than halt a wavelength, which means that the operating frequency is far below cutoff frequency forall higher order ‘modes [16]. Also owing to biaxial symmetry about the two axes only one quarter ot the cross section need to be considered as showin in Fig, 3.140). ov (a) Shielded double strip line with partial dielectric supports (b) problem in (a) simplified by making full use of symmetry.‘The finite difference approximation of Laplace’s equation, derived in Eq. (3.31), namely, 0, has been 1 VED= FEF DAVE-LDHVETENEVE I=) GAD, For the sake of conciseness, let us denote Vom VG MaVGG+) va-1.)) 3.48) V=VG5- Y= VO+1 so that Eq, (3.47) becomes 7 = 4M tv vt vl 3.49) ‘with the computation molecule shown in Fig, 3.13, Equation (3.49) is the general formula to be applied to all free nodes in the free space and diclectric region of Fig. 3.1400). as Computation molecule for Laplace’s equation, On the dielectric boundary, the boundary condition, Diy = Dy. 50) must be imposed. We recall that this condition is based on Gauss’s law for the electric field, ie ene Gs)since no free charge is deliberately placed on the dielectric boundary. Substituting E=-VV ing. G51) gives on fev as ; where 2V /an denotes the derivative of V normal tothe contour, Applying Eq, (3.52) to the interface in Fig. 3.16 yields _ = Ve, Wa Voy Va Voy 0 pS Vs— Wo Va—Vayh Ws Vo) h pote, Moh, toe Figure 116 Interface between media of dielectric permittivities ¢, and c2. Rearranging the terms, fata) Zea ta=ah+ear+ 2S Wet vy) a Tt athe t Soe intin 3.53 matey ere tE B59) ‘This ise ite atTerence equivalent ofthe boundary condition in Bq. (3.50). Notice ‘thatthe discrete inhomogeneity does not affect points 2 and 4 on the boundary but affects points 1 and 3 in proportion to their corresponding permittivities, Also note that when €) = ©, Ba, G53) reduces to Eg. 3.49). ‘On the line of symmetry, we impose the condition av or (3.54)‘This implies that on the line of symmetry along the y-axis, (x = 0 ori = 0) m7 (Vy = Va)/h or V3 / so that Bq. (3.49) becomes. Li + vs +2¥4) (3558) VOR tive. F+04+V07-D42V0, 9) (3.556) av On the line of synunetry along the x-axis (9 cor Vs = Vj so that Vi — Vat =0 1 qoute val (3.564) VGi,0) =F EVE, + VG -1,0+VE41,0) 3.560) 4 ‘The computation molecules for Eqs. (3.55) and (3.56) are displayed in Fig. 3.17, ‘gure 3.17 ‘Computation molecule used for satisfying symmetry conditions: (a) 9V/2 — 0, oo) avjay By setting the potential at the fixed nodes equal to their prescribed values and applying Eqs. (3.49), (3.53), (3.55), and (3.56) to the free nodes according to the ‘band matrix or iterative methods discussed in Section 3.5, the potential at the free nodes can be determined. Once this is accomplished, the quantities of interest can be calculated,‘The characteristic impedance Z, and phase velocity w of the line are defined as (3.578) (3.57) where L and C ate the inductance and capacitance per unit length, respectively. If the dielectric medium is nonmagnetic (1 = j1y), the charactoristic impedance Za. and phase velocity up with the dielectric removed (ie. the line is air-filled) are given by (3.584) (3.586) where Cp is the capacitance per unit ength without the dielectric. Combining Eqs. (3.57) and (3.58) yields 1a - tk oo vem (Se 2 5% e° ean (3399 = 3 108 mf, the speed of igh in free space, and eis the effective dicletic constant. Thus to find Zp and «for an inhomogeneous medium requires caleulating the capacitance per unit length of the structure, with and without the dliclotic substrate Tf Vis the potential difference between the inner and the outer conductors, 42 =i2 3 c-% 3.60) so thatthe problems reduced to finding the charge per nit length Q. (The factor 4 nceded since we are working on only one quarter ofthe cross sction) To find swe apply Gausss law toa closed path {enclosing the innerconduetor, We may select {asthe rectangular path between two adjacent rectangles as shovin in Fi, 3.18 a= fora feta (este)are(! Gon)Figure 3.18. ‘The rectangular path ¢ used in calculating charge enclosed. Since Ax = Ay =h, = (Vp +e Vu + eV HF eVG ++) — (CV +2 + eVE ++ 2 = co [Sei¥ for nodes i on outer rectangle GHIMP wih comer (ch as not counted] =co[Soen¥sforndes ion imer rectangle KIN with comer ech as coumted wice], 6 where Vj and ¢,; are the potential and dielectric constant atthe ith node. If is on the diclewe interface rr = (1 + 62)/2. Also iis onthe fine of symmetry, we Use V;/2 instead of V; to avoid including V; twice in Eq. (3.60), where factor 4 is applied, We also find Cy =1Q./ Vu 2.63) where 0..is obtained by removing the dielectric. finding V; atthe free nodes and then using Eq. (3.62) with ér1 = Latall nodes. Once Q and Q, are calculated, we obtain C and Ce from Eqs. (3.60) and (3.63) and Z. and w from Eg, (3.59), An outline of the procedure is given below: (1) Calculate V (with the dielectric space replaced by free space) 8.53), B55), and (3.56). (2) Determine Q using Eg. (3.62). Bas. 3.49),= () Find Cy = > ) it ie i 4 (4) Repeat steps (1) and (2) (withthe dieleetie space) and find C= +> 6 =3x 108 ms VCC (S) Finally. caleutate Z5 = ‘The attenvation of the line can be calculated by following similar procedure outlined in{14,20,21]. The procedure for handling boundariesat infinity and that for boundary singularities in finite difference analysis are discussed in [37, 38]. 3.7.2 Waveguides ‘The solution of waveguide problems is well suited for finite difference schemes because the solution region is closed. This amounts to solving the Helmholtz or wave ‘equation VO+ Ko =0 Bo where ® = E, for TM modes or & = H, for TE modes, while & is the wave number given by (3.65) ‘The permittivity € of the dielectric medium can be real for a lossless medium or ‘complex for a lossy medium, We consider all fields to vary with time and axial distance as exp j (at — 82). In the eigenvalue problem of Eg. (3.64). both k and & are to be determined, The cutoff wavelength is 2. = 2sr/ke. For each value of the ‘eof? wave number Ke, there is a solution forthe eigenfunction &,, which represents the field configuration of a propagating mode. "To apply the finite difference method, we discretize the cross section of the wave- ‘guide by a suitable square mesh. Applying Ea. (3.29) to Eq. (2.64) gives, +1) B.66) where Ax = Ay = kris the mesh size. Equation (3.66) applies to all the free or interior nodes. At the boundary points, we apply Dirichlet condition ( = 0) for the ‘TM modes and Neumann condition (3/2 = 0) for the TE modes. This implies that at point A in Fig, 3.19, for example, O4=0 (3.67) for TM modes. At point A, 44/on = 0 implies that p = +p so that By, (3.64) becomes Op + Oe +2p—(4— WENO, =O (3.68)for TE modes. By applying Eg. (3.66) and either Eq, (3.67) or (3.68) to all mesh points in the waveguide cross section, we obtain m simultaneous equations involving the m unknowns (1, ®2,... ,®,). These simultaneous equations may be conveniently B BE aol c Figure 3.19 te difference mesh for a waveguide, cast ito the matrix equation (A-Ao=0 3.69) or AO = 20 (3.696) ‘where A is an m x m band matrix of known integer elements, isan identity matrix, © = (@), Op, ... , By) 18 me eigenvector, and ( sa ) 6.70) is the eigenvalue. There are several ways of determining 4 and the corresponding ®. We consider 1wo of these options. The first option is the direct method. Equation (3.69) can be satisfied only if the determinant of (A ~ 21) vanishes, Le., kin lana ‘This results ina polynomial in, which can be solved [29] forthe various cigenvales 2. For each 2, we obtain the corresponding from Eq, (3.66). This method requires storing the relevant matrix clements and docs not take advantage of the fact that matrix Ais sparse. In favor of the method i the fact that a computer subroutine usually exists (Goce [40] o: Appendix D.A) that sulves the cingenvalue probleus in Dy, (3.78) ain that determines all the eigenvalues of the matrix. These eigenvalues give the dominant and, higher modes of the waveguide, although accuracy deteriorates rapidly with mode number‘The second option isthe iterative method. In this ease, the matrix elements are usually generated rather than stored, We begin with ©; = @: = Vand ‘a puessed value fork, The field !"" ar the (th node inthe (-+ 1th eration is ‘obtained from its known value in the kh iteration using oR, OG, = OG, a (3.72) GD MODE 5 c where «is the aoceleration factor, 1 < c <2, and Ryy is the residual atthe (i, th node given by Ry = OE J FDFOG, j-D FOUL) +00 -1,)-(4-#E) O09 (73) After three or four scans of the complete mesh using Eq. (3.73), the value of 2 = 242 should be updated using Raleigh formula ovtods Pes 3.74) ‘The finite difference equivalent of Eg, (3.74) is Sia D jas OU MOH. HOUND JOGA. D—$E)1 i 7 where @s are the latest field values after three or four scans of the mesh and the summation iscarred out overall pointsin the mesh. The new valuc of k obtained from Eq. (3.75) is now used in applying Eq, (3.72) over the mesh for another three or four times to give more accurate eld valves which are again substituted into Eg, (3.75) 10 update k. This process is continued until the difference between consecutive values of kis within a specified acceptable tolerance. If the first option is tobe applicd, matrix A must fist be found. To obtain matrix ‘Ais not easy. Assuming’TM modes, one way of calculating A is to number the fre nodes from left to right, bottom to top, starting from the left-hand corner as shown. typically in Fig, 320, If there ae », and n, divisions along the x and y directions, the number of tree nodes is ng = 0x1 (ny— 1) (G76) Bach free node must be assigned two sets of numbers, one to correspond to m in q_ and the other to correspond 10 (fj) in (7. j). An array NLG. j) = m, atte = Ly J = 1)2sese sly — Lis easily developed to relate the two numbering schemes. To determine the value of element Ayn, we seaich VE(i, J) 10 find (iy, jn) ald (in, Jn), Which ae the values of (i, 7) corresponding to nodes mand n, respectively. With these ideas, we obtain 4 oman Heim Sin i = in td Ang = fot ie ine In Jd orp HL im Sin tle in =n <1 im Sind, im =e 0, otherwise ujaxfie_fis_fs lis pnp Figure 3.20 1g node numbering schemes for ny ny =e Example 3.6 Calculate Zo for the n Solution This problem is representative of the various types of problems that can be solved using the concepts developed in Section 37.1. The computer program in Fig. 3.21 was developed based on the five-step procedure outlined above. By specifying the Step size hand the namber of trations, the program fst sets the potential tal nodes equal to zero. The potential onthe outer conductor is set equal o2er0, while that on the iner conductor is set to 100 volts so that Vs = 100. The program finds C, when the dielectric slab is removed and C when the slab sin place and finally determines 25. Por acelected h, the number of iterations mun he large enogh and feat than the numberof divisions along ory direction. Table 37 shows sme typical resus,008 002 008 008 ed 012 e013 sors sot fore 020 6: one 028 027 cone, ost 0038 Boas 2036 0038, 0039 oat oa ots oar 080 081 083 Figure 3.21 ‘USING THE PLATTE DIFFERENCE METHOD € THIS PROGRAN CALCULATES THE CHLAACTERISTIC|IHPEDANCE © OF A MIcKOSTRIP. LW DIMENSIOR V(-1:200,-1:200), $¥@2),9¢2) Dura 4,8,0,4/2 5,2'5,0.8.1'0/ 1 Lime aa DETK BR, E050/2 38,8. 886=12, 3.08007 £ = 0.028 P8060 + wn, oF rrERaTrONS: aay aH BD = bya ek WO = 100.0 CALCULATE CHARGE WITH AKD WITHOUT DIELECTRIC Bee” Ea © ElleeRR ¢ bo 10 10,4 Wa3y 60 10 GaNTINVE © Ser POTENTIAL OW INHER CONDUCTOR (FIXED nODES) EqUAL 0 9 20 1-0, 44 va,ap) =" © CALOULATE POTENTIAL AT FREE HODES BY/(2.00¢81 + 2D) B2/Q.oscEL SD) Do 40 Eoo;na-t Do 40 Jm0,51-1 TRC ED FQ'HTD AWD (7-1 HB) ) 60 10.40 TFG 0-90). Tae (¢ tnose BOUNDARY CONDITION AT THE TATERF vera) = a age(¥Cres a) ¢ Veted a) = 1 Pieverysei) ¢ P20¥CL 3-2) co 10 40 TFCL59.0) Tee ¢ raose SYMRETRY CINDETION ALOK Y-A1I8 VD CSGVIE DT VEIN) + ¥O,E2 2/80 Goo FPG.69.0) rar © rmpase SYMRETEY CINDTTION ALONG X-ANIS VOyD » (VI 3) VT1 3) + 2.00060 341) 9/40 ‘Computer program for Example 3.6 (Continued).Gees 30 HEtad) aeveaet apevtrrt nee eH 9/40 06s $0 CONTINUE oes sore fio eee na ors soot > GY aD /2 (0072 SUM POTENTIAL ON THTER AND OUTER LoOPS oreo) tt can an i a ore SUK = SUM + EIeV(0,20UT)/2.0 + SYMMETRY POLKT 80 sum + Ezevcaour,s) oer Som + (hiseayevGlaur,/2.0 foes 70 (CONTINUE i ee SUM= SIM + Ezey(ZOUT,0)/2.0 + SYMMETRY POrHT ose Teteeg.2) eva) > sum 86 ove 1" "FoR runER LooP oe? sour “1 088 SOm = SUN + 2.0¢ESeVCIOUT, 20UTD + CORRER POUT 090 som boot ape ayes) aveay > 092 i oss 90 GOTIMUE Ooo FINALLY, CALCULATE 20 foe bose €or pgsgesayee 588 GS EOGAD 2098 20 2 2.0/€ cesgnrcooecs) > oi Parit #3 aT 20 102 stor (Cont.) Computer program for Example 3.6. 38 Practical Applications I — Wave Scattering (FDTD) ‘The finite-difference time-domain (FDTD) formulation of EM field problems is a convenient tool for solving scattering problems. The FDTD method, first introduced, by Yee [42] in 1966 and later developed by Taflove and others (31, 32, 35], [43]-146], is direct solution of Maxwell's time-dependent curl equat the irradiation of the scatterer as an initial value problem, Our discussion on the FD TD method will eover: wns. The scheme treats + Yoe's finite difference algorithm, + accuracy and stability,Table 3.7 Characteristic Tnnpedance of a Mierostip Line fon umple 3.6 7__ Number Zo 025 700 49.05 on 300 37.83 0.05 500 65.82 0.05 700 63.10 0.05 1000 61.53 ‘Other method (41): Zy = 62.50 + absorbing boundary conditions, + initial fields, and + programming aspects, Some iadel examples with TORTRAN voles will be provided 1 ilustiate de method. 3.8.1. Yee’s Finite Difference Algorithm In an isotropic medium, Maxwell's equations can be written as VxE=- (3.78a) VxH=ok+e 3.780) ‘The vector Eg. (3.78) represents a system of six scalar equations, which can be expressed in rectangular coordinate system as: ay aes, a ay ). am 8.790) G.190) 3.790) GB.) (19)Following Yee’s notation, we define a grid point in the solution region as Gj.) = (FAx, Jy, RAD G80) and any funetion of space and time as FP j= FOB, j8, 48,0) G81 where 5 ‘Ay = Az is the space increment, Av is the time increment, while i, j, k,and ate integers. Using central finite difference approximation for space and time derivatives that are second-order accurate, BPUG iB) _ PME + 1/2.) — PMG ~ 1/2, 5.8) > seo = +0(0) as BRM IR)_ POE jk) PIPE, J.) 2 SRG PAGS PEL 2 3 ~ 7 +0(ar) 3s» In applying Eq, (3.82) t0 all the space derivatives in Eq, (3.79), Yee positions the ‘components of E and H about a unit cell of the lattice as shown in Fig. 3.22, To Positions of the field components in a unit cell of the Yee’s latti incorporate Eq, (3.83), the components of E and H are evaluated at alternate half time steps. Thus we obtain the explicit finite difference approximation of Ey. (3.79) asMe pte a1) = HE PU) EEE 2) ar adi § +12, 4+ 1/295

Potrebbero piacerti anche

• Fear: Trump in the White House

  

  Da Everand

  Fear: Trump in the White House

  Bob Woodward

  Valutazione: 3.5 su 5 stelle

  3.5/5 (738)
• A Man Called Ove: A Novel

  

  Da Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Valutazione: 4.5 su 5 stelle

  4.5/5 (4609)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  Da Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Valutazione: 3.5 su 5 stelle

  3.5/5 (231)
• • Riviste
  • Podcast
  • Spartito
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  Da Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Valutazione: 4.5 su 5 stelle

  4.5/5 (120)
• Grit: The Power of Passion and Perseverance

  

  Da Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Valutazione: 4 su 5 stelle

  4/5 (588)
• Yes Please

  

  Da Everand

  Yes Please

  Amy Poehler

  Valutazione: 4 su 5 stelle

  4/5 (1891)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  Da Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Valutazione: 4.5 su 5 stelle

  4.5/5 (266)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  Da Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Valutazione: 3.5 su 5 stelle

  3.5/5 (399)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  Da Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Valutazione: 4.5 su 5 stelle

  4.5/5 (838)
• Shoe Dog: A Memoir by the Creator of Nike

  

  Da Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Valutazione: 4.5 su 5 stelle

  4.5/5 (537)
• The Emperor of All Maladies: A Biography of Cancer

  

  Da Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Valutazione: 4.5 su 5 stelle

  4.5/5 (271)
• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  Da Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Valutazione: 4 su 5 stelle

  4/5 (5794)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  Da Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Valutazione: 3.5 su 5 stelle

  3.5/5 (2259)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  Da Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Valutazione: 4.5 su 5 stelle

  4.5/5 (344)
• Principles: Life and Work

  

  Da Everand

  Principles: Life and Work

  Ray Dalio

  Valutazione: 4 su 5 stelle

  4/5 (599)
• Rise of ISIS: A Threat We Can't Ignore

  

  Da Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Valutazione: 3.5 su 5 stelle

  3.5/5 (137)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  Da Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Valutazione: 4.5 su 5 stelle

  4.5/5 (234)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  Da Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brené Brown

  Valutazione: 4 su 5 stelle

  4/5 (1090)
• John Adams

  

  Da Everand

  John Adams

  David McCullough

  Valutazione: 4.5 su 5 stelle

  4.5/5 (2409)
• The Glass Castle: A Memoir

  

  Da Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Valutazione: 4.5 su 5 stelle

  4.5/5 (1712)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  Da Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Valutazione: 4 su 5 stelle

  4/5 (895)
• Her Body and Other Parties: Stories

  

  Da Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Valutazione: 4 su 5 stelle

  4/5 (821)
• Sing, Unburied, Sing: A Novel

  

  Da Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Valutazione: 4 su 5 stelle

  4/5 (1103)
• Angela's Ashes: A Memoir

  

  Da Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Valutazione: 4.5 su 5 stelle

  4.5/5 (440)
• Wolf Hall: A Novel

  

  Da Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Valutazione: 4 su 5 stelle

  4/5 (3811)
• A Tree Grows in Brooklyn

  

  Da Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Valutazione: 4.5 su 5 stelle

  4.5/5 (1929)
• The Woman in Cabin 10

  

  Da Everand

  The Woman in Cabin 10

  Ruth Ware

  Valutazione: 3.5 su 5 stelle

  3.5/5 (2322)
• The Light Between Oceans: A Novel

  

  Da Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Valutazione: 4.5 su 5 stelle

  4.5/5 (789)
• The Constant Gardener: A Novel

  

  Da Everand

  The Constant Gardener: A Novel

  John le Carré

  Valutazione: 3.5 su 5 stelle

  3.5/5 (104)
• The Perks of Being a Wallflower

  

  Da Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Valutazione: 4.5 su 5 stelle

  4.5/5 (2101)
• How To Paint Eyes by Sharon Kinzie (1989)

  

  Documento33 pagine

  How To Paint Eyes by Sharon Kinzie (1989)

  emikanter

  50% (2)
• The Art of Racing in the Rain: A Novel

  

  Da Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Valutazione: 4 su 5 stelle

  4/5 (4200)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  Da Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Valutazione: 4.5 su 5 stelle

  4.5/5 (474)
• The Outsider: A Novel

  

  Da Everand

  The Outsider: A Novel

  Stephen King

  Valutazione: 4 su 5 stelle

  4/5 (1839)
• The Unwinding: An Inner History of the New America

  

  Da Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Valutazione: 4 su 5 stelle

  4/5 (45)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  Da Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Valutazione: 4 su 5 stelle

  4/5 (98)
• On Fire: The (Burning) Case for a Green New Deal

  

  Da Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Valutazione: 4 su 5 stelle

  4/5 (73)
• 30GX Wiring 533-084 30HX Chiller Carrier

  

  Documento34 pagine

  30GX Wiring 533-084 30HX Chiller Carrier

  yamamoto_san

  100% (5)
• TQ Etm PDF

  

  Documento265 pagine

  TQ Etm PDF

  ́Vu Nguyen

  Nessuna valutazione finora
• Little Women

  

  Da Everand

  Little Women

  Louisa May Alcott

  Valutazione: 4 su 5 stelle

  4/5 (104)
• Brooklyn: A Novel

  

  Da Everand

  Brooklyn: A Novel

  Colm Tóibín

  Valutazione: 3.5 su 5 stelle

  3.5/5 (1937)
• Manhattan Beach: A Novel

  

  Da Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Valutazione: 3.5 su 5 stelle

  3.5/5 (792)
• Bad Feminist: Essays

  

  Da Everand

  Bad Feminist: Essays

  Roxane Gay

  Valutazione: 4 su 5 stelle

  4/5 (1015)
• Steve Jobs

  

  Da Everand

  Steve Jobs

  Walter Isaacson

  Valutazione: 4.5 su 5 stelle

  4.5/5 (806)
• Biophysics Basics in Electrotherapy

  

  Documento31 pagine

  Biophysics Basics in Electrotherapy

  rabeya

  Nessuna valutazione finora
• Conductor Impedance Data (Cable Impedance)

  

  Documento2 pagine

  Conductor Impedance Data (Cable Impedance)

  NOELGREGORIO

  100% (2)
• Electronics Reference - Digital Logic Circuits

  

  Documento23 pagine

  Electronics Reference - Digital Logic Circuits

  Amanuel Abrha

  Nessuna valutazione finora
• Matlab Signal Processing Examples

  

  Documento20 pagine

  Matlab Signal Processing Examples

  Kunchala Anil

  Nessuna valutazione finora
• EE5403 Lecture 2 Introduction To RF PA 20070930

  

  Documento30 pagine

  EE5403 Lecture 2 Introduction To RF PA 20070930

  Amanuel Abrha

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Bat Coda

  

  Documento1 pagina

  Bat Coda

  Haftamu Tekle

  Nessuna valutazione finora
• Hms Getye PDF

  

  Documento42 pagine

  Hms Getye PDF

  Amanuel Abrha

  Nessuna valutazione finora
• OO and System Design

  

  Documento40 pagine

  OO and System Design

  Amanuel Abrha

  Nessuna valutazione finora
• Digital Control of Switching Power Supply

  

  Documento11 pagine

  Digital Control of Switching Power Supply

  illuminatissm

  100% (1)
• A+ Operating System

  

  Documento58 pagine

  A+ Operating System

  Amanuel Abrha

  Nessuna valutazione finora
• Flyback Design and Implementation

  

  Documento19 pagine

  Flyback Design and Implementation

  Siyabonga Magagula

  Nessuna valutazione finora
• Week001-Electric Charge, Coulomb's Law, Electric Fields, and Electric Flux pUb7UF

  

  Documento17 pagine

  Week001-Electric Charge, Coulomb's Law, Electric Fields, and Electric Flux pUb7UF

  Benjonit Capulong

  Nessuna valutazione finora
• Winding Insulation

  

  Documento42 pagine

  Winding Insulation

  mwasaha

  100% (1)
• 31R17021119 Pdf-Merged

  

  Documento3 pagine

  31R17021119 Pdf-Merged

  ramesh

  Nessuna valutazione finora
• AppliedMathematicsandPhysics CK407, CK408

  

  Documento2 pagine

  AppliedMathematicsandPhysics CK407, CK408

  Yes Zam

  Nessuna valutazione finora
• TLRF 2m PDF

  

  Documento25 pagine

  TLRF 2m PDF

  Deepthi SL Creations

  Nessuna valutazione finora
• SFF1001G THRU SFF1008G: Isolation 10.0 AMPS. Glass Passivated Super Fast Rectifiers

  

  Documento2 pagine

  SFF1001G THRU SFF1008G: Isolation 10.0 AMPS. Glass Passivated Super Fast Rectifiers

  Samvel Khachatryan

  Nessuna valutazione finora
• ACES Journal March 2021 Paper 1

  

  Documento6 pagine

  ACES Journal March 2021 Paper 1

  Mohammad Yahya Azab

  Nessuna valutazione finora
• Art - No. 103826: Voltage Tester "Profi Lcdplus Ii"

  

  Documento2 pagine

  Art - No. 103826: Voltage Tester "Profi Lcdplus Ii"

  Mihai Butnaru-Palade

  Nessuna valutazione finora
• Mass, Weight and Gravitational Field - Notes

  

  Documento2 pagine

  Mass, Weight and Gravitational Field - Notes

  aasraj

  Nessuna valutazione finora
• General Physics Exercises

  

  Documento49 pagine

  General Physics Exercises

  alo

  Nessuna valutazione finora
• Solution Manual To Chapter 03 PDF

  

  Documento2 pagine

  Solution Manual To Chapter 03 PDF

  anjana

  Nessuna valutazione finora
• DC Motor Calculations

  

  Documento18 pagine

  DC Motor Calculations

  Harshana Randheera

  Nessuna valutazione finora
• Antenna and Wave Propagation: by T.MARTHANDAN.M.E.,M.B.A

  

  Documento22 pagine

  Antenna and Wave Propagation: by T.MARTHANDAN.M.E.,M.B.A

  MARTHANDAN

  Nessuna valutazione finora
• Krohne Optiflux 4000 Manual

  

  Documento36 pagine

  Krohne Optiflux 4000 Manual

  Luis Santos

  Nessuna valutazione finora
• ED099507

  

  Documento136 pagine

  ED099507

  Josue

  Nessuna valutazione finora
• Generator S637UUT 206J

  

  Documento52 pagine

  Generator S637UUT 206J

  Mantto Central

  Nessuna valutazione finora
• Key Features: P1200/Nd Series

  

  Documento2 pagine

  Key Features: P1200/Nd Series

  Juan Carlos Avila Gonzalez

  Nessuna valutazione finora
• S.2 Integrated Science Worksheet (Chapter 8.1 - 8.4)

  

  Documento12 pagine

  S.2 Integrated Science Worksheet (Chapter 8.1 - 8.4)

  elephantn63

  Nessuna valutazione finora
• XZ1 AL Eng - Rev4 PDF

  

  Documento5 pagine

  XZ1 AL Eng - Rev4 PDF

  jarias-solarpack

  Nessuna valutazione finora
• Magnetism and Electromagnetism

  

  Documento19 pagine

  Magnetism and Electromagnetism

  Ryan A. Ramos

  Nessuna valutazione finora
• Lec Phir11.01

  

  Documento16 pagine

  Lec Phir11.01

  R Tharunish

  Nessuna valutazione finora
• Science, Technology and Society: (GEED 10083)

  

  Documento23 pagine

  Science, Technology and Society: (GEED 10083)

  Xivaughn Sebastian

  Nessuna valutazione finora
• Pair Production by Schwinger and Breitwheeler Processes in Bifrequent Fields

  

  Documento12 pagine

  Pair Production by Schwinger and Breitwheeler Processes in Bifrequent Fields

  zeyadwaeli500

  Nessuna valutazione finora
• D-Tect 3: Motion Detector

  

  Documento2 pagine

  D-Tect 3: Motion Detector

  Adrian Deoanca

  Nessuna valutazione finora