Download Integral equations for real-life multiscale electromagnetic problems (electromagnetic waves

Page 1


https://ebookmass.com/product/integral-equations-for-real-

Instant digital products (PDF, ePub, MOBI) ready for you

Download now and discover formats that fit your needs...

Techniques of Functional Analysis for Differential and Integral Equations 1st Edition Paul Sacks

https://ebookmass.com/product/techniques-of-functional-analysis-fordifferential-and-integral-equations-1st-edition-paul-sacks/

ebookmass.com

Therapy in the Real World: Effective Treatments for Challenging Problems

https://ebookmass.com/product/therapy-in-the-real-world-effectivetreatments-for-challenging-problems/

ebookmass.com

eTextbook 978-1305965799 Differential Equations with Boundary-Value Problems

https://ebookmass.com/product/etextbook-978-1305965799-differentialequations-with-boundary-value-problems/

ebookmass.com

Divine Style: Walt Whitman and the King James Bible. 1st Edition F. W. Dobbs-Allsopp.

https://ebookmass.com/product/divine-style-walt-whitman-and-the-kingjames-bible-1st-edition-f-w-dobbs-allsopp/

ebookmass.com

Silas: Club Sin Book 4 Jasinda Wilder

https://ebookmass.com/product/silas-club-sin-book-4-jasinda-wilder/

ebookmass.com

Prescott’s Microbiology 10th Edition – Ebook PDF Version

https://ebookmass.com/product/prescotts-microbiology-10th-editionebook-pdf-version/

ebookmass.com

The Global Circulation of Chinese Materia Medica, 1700–1949 Di Lu

https://ebookmass.com/product/the-global-circulation-of-chinesemateria-medica-1700-1949-di-lu/

ebookmass.com

In and Out of Sight: Modernist Writing and the Photographic Unseen Alix Beeston

https://ebookmass.com/product/in-and-out-of-sight-modernist-writingand-the-photographic-unseen-alix-beeston/

ebookmass.com

(eBook PDF) Management, 7th Asia-Pacific Edition by John

https://ebookmass.com/product/ebook-pdf-management-7th-asia-pacificedition-by-john-r-schermerhorn/

ebookmass.com

Programming for Game Design: A Hands-On Guide with Godot 1st Edition Wallace Wang

https://ebookmass.com/product/programming-for-game-design-a-hands-onguide-with-godot-1st-edition-wallace-wang/

ebookmass.com

IntegralEquationsfor Real-LifeMultiscale Electromagnetic Problems

AndrewF.Peterson,PhD-SeriesEditor

Thevolumesinthisserieswillencompassthedevelopmentandapplicationofnumerical techniquestoelectricalandelectronicsystems,includingthemodellingofelectromagnetic phenomenaoverallfrequencyrangesandcloselyrelatedtechniquesforacousticandoptical analysis.Thescopeincludestheuseofcomputationforengineeringdesignandoptimization, aswellastheapplicationofcommercialmodellingtoolstopracticalproblems.Theserieswill includetitlesforseniorundergraduateandpostgraduateeducation,researchmonographsfor reference,andpractitionerguidesandhandbooks.

TitlesintheSeries

K.Warnick,“NumericalMethodsforEngineering,”2010. W.Yu,X.YangandW.Li,“VALU,AVXandGPUAccelerationTechniquesforParallelFDTD Methods,”2014.

A.Z.Elsherbeni,P.NayeriandC.J.Reddy,“AntennaAnalysisandDesignUsingFEKOElectromagneticSimulationSoftware,”2014.

A.Z.ElsherbeniandV.Demir,“TheFinite-DifferenceTime-DomainMethodinElectromagneticswithMATLAB®Simulations,2ndEdition,”2015.

M.Bakr,A.Z.ElsherbeniandV.Demir,“AdjointSensitivityAnalysisofHighFrequencyStructureswithMATLAB®,”2017.

O.Ergul,“NewTrendsinComputationalElectromagnetics,”2019.

D.Werner,“NanoantennasandPlasmonics:Modelling,designandfabrication,”2020. K.KobayashiandP.D.Smith,“AdvancesinMathematicalMethodsforElectromagnetics,” 2020

V.Lancellotti,“AdvancedTheoreticalandNumericalElectromagnetics,Volume1:Static, stationaryandtime-varyingfields,”2021.

V.Lancellotti,“AdvancedTheoreticalandNumericalElectromagnetics,Volume2:Field representationsandthemethodofmoments,”2021.

S.Roy,“UncertaintyQuantificationofElectromagneticDevices,Circuits,andSystems,” 2021

A.Baghai-Wadji“MathematicalQuantumPhysicsforEngineersandTechnologists,Volume1:Fundamentals,”2023.

IntegralEquationsfor Real-LifeMultiscale Electromagnetic Problems

TheInstitutionofEngineeringandTechnology

PublishedbySciTechPublishing,animprintofTheInstitutionofEngineeringand Technology,London,UnitedKingdom

TheInstitutionofEngineeringandTechnologyisregisteredasaCharityinEngland&Wales (no.211014)and Scotland(no.SC038698).

©TheInstitutionofEngineeringandTechnology2024 Firstpublished2023

ThispublicationiscopyrightundertheBerneConventionandtheUniversalCopyright Convention.Allrightsreserved.Apartfromanyfairdealingforthepurposesofresearch orprivatestudy,orcriticismorreview,aspermittedundertheCopyright,Designsand PatentsAct1988,thispublicationmaybereproduced,storedortransmitted,inany formorbyanymeans,onlywiththepriorpermissioninwritingofthepublishers,orin thecaseofreprographicreproductioninaccordancewiththetermsoflicencesissued bytheCopyrightLicensingAgency.Enquiriesconcerningreproductionoutsidethose termsshouldbesenttothepublisherattheundermentionedaddress:

TheInstitutionofEngineeringandTechnology FuturesPlace KingsWay,Stevenage Hertfordshire,SG12UA,UnitedKingdom

www.theiet.org

Whiletheauthorsandpublisherbelievethattheinformationandguidancegiveninthis workarecorrect,allpartiesmustrelyupontheirownskillandjudgementwhenmaking useofthem.Neithertheauthorsnorpublisherassumesanyliabilitytoanyoneforany lossordamagecausedbyanyerrororomissioninthework,whethersuchanerroror omissionistheresultofnegligenceoranyothercause.Anyandallsuchliability isdisclaimed.

Themoralrightsoftheauthorstobeidentifiedasauthorsofthisworkhavebeen assertedbytheminaccordancewiththeCopyright,DesignsandPatentsAct1988.

BritishLibraryCataloguinginPublicationData

AcataloguerecordforthisproductisavailablefromtheBritishLibrary

ISBN978-1-83953-476-8(hardback)

ISBN978-1-83953-477-5(PDF)

TypesetinIndiabyMPSLimited PrintedintheUKbyCPIGroup(UK)Ltd,Eastbourne

CoverImage:Pobytov/DigitalVisionVectorsviaGettyImages

Abouttheeditorsxi

1Introduction1

FrancescaVipianaandZhenPeng References3

2Surfaceintegralequationformulations5 DonaldR.WiltonandWilliamA.Johnson

2.1Maxwell’sequations5

2.1.1IntegralformofMaxwell’sequations5

2.1.2PointordifferentialformofMaxwell’sequations7

2.1.3BoundaryformofMaxwell’sequations7

2.1.4TheHelmholtzequationsandpotentialrepresentations9

2.1.5Farfieldsandfarpotentials13

2.1.6Thedualityprinciple14

2.1.7Uniquenesstheorem15

2.2Equivalenceprinciples18

2.2.1Thevolumetricequivalenceprinciple18

2.2.2Thesurfaceequivalenceprinciple18

2.3Boundaryfieldrepresentations21

2.3.1TheCalderónidentities28

2.4TheLorentzreciprocitytheorem29

2.5Surfaceintegralequationformulationsandsolutionsbymoment methods31

2.5.1Surfacerepresentationbytriangulation31

2.5.2Definingelectromagneticquantitiesonamesh36

2.5.3Theelectricfieldintegralequation(EFIE)38

2.5.4Fillandassemblyofelementandsystemmatricesand columnexcitationvectors42

2.5.5Themagneticfieldintegralequation(MFIE)48

2.5.6ConductingsheetsandtheEFIEandMFIE52

2.5.7InternalresonancesandtheCFIE54

2.5.8Integralequationformulationsfordielectrics55

2.6Surfaceintegralequationchallenges58

2.6.1Vectornorms,matrixnorms,andconditionnumber58

2.6.2TheEFIEand L operator61

Integralequationsforreal-lifemultiscaleelectromagneticproblems

2.6.3TheMFIEand K operator64

2.6.4Mixedoperatorintegralequations70 References70

3Kernel-basedfastfactorizationtechniques75 ÖzgürErgül,BahramKhalichiandVakurB.Ertürk

3.1Introduction75

3.2Multilevelfastmultipolealgorithm76

3.2.1ConventionalMLFMAbasedonplanewaves77

3.2.2Low-frequencyandbroadbandMLFMAimplementations83

3.3Large-scalesimulationsandparallelcomputing86

3.4Materialmodeling89

3.4.1MaterialsimulationswiththeconventionalMLFMA90

3.4.2Simulationsofplasmonicstructures98

3.4.3Simulationsofnear-zero-index(NZI)structures101

3.5Problemswithdensediscretizations103

3.6Problemswithnon-uniformdiscretizations108

3.7Conclusionsandnewtrends111 Acknowledgments113 References113

4Kernel-independentfastfactorizationmethodsformultiscale electromagneticproblems125 MengmengLi,PaolaPirinoli,FrancescaVipianaandGiuseppeVecchi

4.1Introduction125

4.2Adaptivecrossapproximation(ACA)method126

4.3Multilevelmatrixcompressionmethodformultiscaleproblems128

4.3.1Backgroundandtheory128

4.3.2Accuracyvalidation130

4.3.3Computationalcomplexityanalysis130

4.3.4Numericalevaluationoftheinducedfieldsinareal-life aircraft131

4.4Nestedequivalencesourceapproximationforlow-frequency multiscaleproblems134

4.4.1Equivalentsourcedistributionsforfieldrepresentation134

4.4.2FieldrepresentationviaequivalentRWGbasisfunctions135

4.4.3Single-levelnestedmatrixcompressionapproximation algorithm135

4.4.4MultilevelNESA137

4.4.5Matrix–vectorproductandcomputationcomplexity140

4.4.6Numericalresults142

4.5Widebandnestedequivalencesourceapproximationformultiscale problems150

4.5.1Far-fieldfactorizationadmissibilityconditions151

4.5.2High-frequency-nestedapproximationindirections153

4.5.3MultilevelWNESA155

4.5.4MVPandcomputationcomplexity157

4.5.5Numericalresults160

4.6Mixed-formnestedequivalencesourceapproximationformultiscaleproblems163

4.6.1Multiscalesamplingforskeletons164

4.6.2Mixed-formwideband-nestedapproximation165

4.6.3Numericalresults167

4.7Conclusionandprospect172 Acknowledgments173 References173

5Domaindecompositionmethod(DDM)179

VíctorMartín,Hong-WeiGao,DiegoM.Solís,JoséM.Taboada, andZhenPeng

5.1DiscontinuousGalerkinDDmethodforPECobjects180

5.1.1IntroductiontodiscontinuousGalerkinmethod180

5.1.2SIEformulation181

5.1.3Domainpartitioningandbasisfunctionspace182

5.1.4Interiorpenaltyformulation184

5.1.5Matrixequationandpreconditioner186

5.1.6Iterativesolutionofpreconditionedmatrixequation187

5.1.7Numericalexperiments188

5.2DGDDmethodforpenetrableobjects194

5.2.1DG-DDM-SIEforhomogeneousobjects194

5.2.2DG-DDM-SIEforpiecewisehomogeneousobjects201

5.3Tear-and-interconnectDDM211

5.3.1Preconditionerformulation211

5.3.2Anoteonparallelization213

5.3.3Numericalexamples213 References223

6Multi-resolutionpreconditioner231

FrancescaVipiana,VictorF.MartinandJoseM.Taboada

6.1Preliminaries231

6.1.1Introductionandscope231

6.1.2Basisfunctions232

6.1.3MoMlinearsystem235

6.1.4Multi-resolutionstrategy236

6.2Basisfunctionsgeneration236

6.2.1Generalizedbasisfunctions238

6.2.2Multi-resolutionbasisfunctions245

6.2.3PECgroundplanehandling250

viii Integralequationsforreal-lifemultiscaleelectromagneticproblems

6.2.4Basisforelectricalsizesbeyondtheresonanceregion251

6.2.5Algorithmflowchartandcomputationalcomplexity251

6.3Generationofahierarchicalfamilyofmeshes253

6.3.1Cellsgroupingstrategy253

6.3.2Cellsrankingandaggregation257

6.3.3Cellsgroupingrefinement259

6.3.4Maximumcellsizegroupinglimiting260

6.3.5Computationalcomplexity261

6.4ApplicationtoMoM261

6.4.1Change-of-basismatrixmemoryallocation261

6.4.2Directsolution262

6.4.3Applicationtoiterativesolvers263

6.4.4Applicationtoelectricallylargemulti-scalestructures264

6.4.5Low-frequencymatrixentriesevaluation266

6.5Numericalresults268

6.5.1FerrariTestarossatestcase269

6.5.2Realisticvesseltestcase272

6.6Conclusionandperspectives273 Acknowledgments274 References274

7Calderónpreconditionersforelectromagneticintegralequations277 AdrienMerlini,SimonB.Adrian,AlexandreDély, andFrancescoP.Andriulli

7.1Introduction277

7.2Backgroundandnotations279

7.3Calderónidentities280

7.4Discretization282

7.5ElectricfieldIE284

7.5.1Theoriginalequation284

7.5.2Thepreconditionedequation288

7.6CombinedfieldIE292

7.6.1Theoriginalequation292

7.6.2Thepreconditionedequation292

7.7PMCHWT294

7.7.1Theoriginalequation294

7.7.2Thepreconditionedequation296

7.7.3Differentsolutionstrategies298

7.8Conclusions301 References301

8Decoupledpotentialintegralequation307 FelipeVicoandMiguelFerrando-Bataller

8.1Scatteringproblemandboundaryconditions307

8.2Low-frequencylimitboundaryvalueproblems309

8.3Stabilizingconditions314

8.4DecoupledpotentialsanddifferentLorenzgaugefixings316

8.5Incomingpotentialsinalow-frequencystableLorenzgauge319

8.6Decoupledpotentialboundaryvalueproblems322

8.7Second-kindintegralequation326

8.8Discretizationofanintegralequationofthesecondkind329

8.8.1High-orderaccurateself-interactionintegral340

8.9Nearinteractionquadrature346

AppendixA:Differentialgeometryofsurfaces346

AppendixB:Numericalintegrationandinterpolationin1D353

AppendixC:Numericalintegrationandinterpolationin2D355

AppendixD:GeneralizedGaussianquadratureforarbitrary non-smoothfunctions361

AppendixE:Functionspaces364 References365

9Conclusionandperspectives369 ZhenPengandFrancescaVipiana References371 Index375

This page intentionally left blank

Abouttheeditors

FrancescaVipiana isafullprofessorintheDepartmentofElectronicsand TelecommunicationsatPolitecnicodiTorino(POLITO),Italy.

Hermainresearchactivitiesconcernnumericaltechniquesbasedonintegral equationsandmethodofmomentapproaches,withafocusonmultiresolutionand hierarchicalschemes,domaindecomposition,preconditioningandfastsolutionmethods,andadvancedquadratureintegrationschemes.Moreover,herresearchinterests includethemodeling,design,realization,andtestingofmicrowaveimagingsystems formedicalandindustrialapplications.

Currently,Prof.Vipianacoordinates“THERAD–MicrowaveTheranosticsfor Alzheimer’sDisease,”researchprojectfundedbythe“CompagniadiSanPaolo”bank foundation,and“INSIGHT–Aninnovativemicrowavesensingsystemfortheevaluationandmonitoringoffoodqualityandsafety,”jointresearchprojectwithinthe ExecutiveProgramofScientificandTechnologicalCooperationbetweenItalyand China,fundedbytheNationalNaturalScienceFoundationofChina(NSFC)and theItalianMinistryofForeignAffairsandInternationalCooperation.Moreover,she istheprincipalinvestigator,forthePOLITOresearchunit,inthenationalproject “BEST-Food,BroadbandElectromagneticSensingTechnologiesforFoodquality andsecurityassessment,”andintheMarieSkłodowska-CurieDoctoralNetwork “GENIUS–Glide-symmetricmEtamaterialsforiNnovativeradIo-frequencycommUnicationandSensing,”fundedbytheEuropeanUnion’sHorizonEuropeProgramme andbytheUKResearchandInnovation.

Shehas20yearsfull-timeequivalentresearchexperience.Since2007,shehas beenaninstructorattheEuropeanSchoolofAntennas(ESoA)coursesand,since 2008,theteachingprofessorofthecourse,“AdvancedComputationalEMforAntenna Analysis”atthePOLITODoctoralSchoolwheresheispartofthePhDadvisors board.Prof.VipianareceivedtheLotShafaiMid-CareerDistinguishedAwardfrom theIEEEAntennasandPropagationSocietyin2017andsheisanassociateeditor of IEEETransactionsonAntennasandPropagation andofthe IEEEAntennasand PropagationMagazine,where,in2020,shewasalsoaguesteditorforthespecialissue “ElectromagneticImagingandSensingforFoodQualityandSafetyAssessment.”

ZhenPeng isanassociateprofessorintheDepartmentofElectricalandComputer EngineeringattheUniversityofIllinoisatUrbana-Champaign,USA.Hisresearch interestsincludeclassicalelectromagnetismwithscalablealgorithms;statisticalelectromagneticsforcomplexenvironments,forexample,physics-orientedstatistical waveanalysisintegratingorderandchaos,andelectromagneticinformationtheory

xii Integralequationsforreal-lifemultiscaleelectromagneticproblems forwirelesscommunication;quantumelectromagnetics;andmeasurementandcontrolofuncertaintiesinchaoticreverberationchambers.HewasaguesteditorofIEEE TransactionsonComponents,PackagingandManufacturingTechnology in2023,and anassociateeditorofIEEE TransactionsonMicrowaveTheoryandTechniques from 2018to2020.HehaswonseveralbestpaperawardsincludingBestElectromagnetics PaperAwardatthe16thEuropeanConferenceonAntennasandPropagationin2022, theEPEPSBestPaperAwardatthe30thConferenceonElectricalPerformanceof ElectronicPackagingandSystems,theIEEEEMCSymposiumBestPaperAward atthe2019IEEEInternationalSymposiumonElectromagneticCompatibility,Signal&PowerIntegrity,the2018BestTransactionPaperAward-IEEETransactions onComponents,PackagingandManufacturingTechnology,the2014IEEEAntenna andPropagationSergeiA.SchelkunoffTransactionsPrizePaperAward.Hewasa recipientoftheNationalScienceFoundationCAREERAward(ENG/ECCS/CCSS) in2018.

Chapter1 Introduction

FrancescaVipiana1 andZhenPeng2

Inthecontextofcomputationalelectromagnetics(CEM),surfaceintegralequation (SIE)techniquesbasedonthemethodofmoments(MoM)[1]offerapotenttoolthat hasbecomeessentialforsimulatingandengineeringadiverserangeofapplications. Theseapplicationsencompassadvancedantennadesign[2,3],radarcross-section (RCS)[4],stealthtechnologies[5],electromagneticcompatibilityandinterference (EMC/EMI)[6],andnanoscienceapplications[7],amongothers.SIEmethodsare particularlyattractivewhendealingwithlarge-scaleradiationandscatteringissues. Unlikevolumetricapproachesthatrequirethecharacterizationofthree-dimensional (3D)structuresandembeddingspace,SIEmethodsnecessitatetheparameterization oftwo-dimensional(2D)boundarysurfacesonly.Althoughtheyresultindenseand extensivematrixsystemsforlarge-scaleproblems,theutilizationofiterativefast solvers,suchasthemultilevelfastmultipolealgorithm(MLFMA)[8,9],enables efficientresolutionofsuchproblems.

Thescopeofthisbook“IntegralEquationsforReal-LifeMultiscaleElectromagneticProblems”istocollectanddescribethemainrecentavailableapproachesforthe numericalsolutionofSIEstoanalyzereal-lifemultiscaleelectromagneticproblems. InCEM,formulationsbasedonSIEsarecurrentlythemostusedfortheanalysis ofelectricallylargeandcomplexstructures.Still,itisessentialtohaveavailable state-of-the-arttechniquestosolvetheminanefficientandaccurateway.

Thebookisorganizedintosevenscientificchapters,completedwiththe “Introduction”and“Conclusionandperspectives”chapters.

Chapter2“Surfaceintegralequationformulations,”authoredbyDonaldR. WiltonandWilliamA.Johnson,encompassesaconciseoverviewofessentialconcepts requiredtocomprehend,formulate,andcomputationallyaddressSIEsencounteredin thefieldofelectromagnetics.Utilizingthisknowledge,theprevalentintegralequationsemployedintime-harmonicproblemsareestablished,involvinglinear,piecewise homogeneous,andisotropicmaterials.Then,numericalmethodsemployedtosolve

1 WavisionResearchGroup,DepartmentofElectronicsandTelecommunications,PolitecnicodiTorino, Italy

2 ElectromagneticsLabandCenterforComputationalElectromagnetics,DepartmentofElectricaland ComputerEngineeringUniversityofIllinoisatUrbana—Champaign,USA

Integralequationsforreal-lifemultiscaleelectromagneticproblems theseintegralequationsarepresented,includingtechniquesforaccuratelyevaluating thesingularandnear-singularintegralsthatemergeintheprocess.

Thefollowingtwochapters,Chapter3“Kernel-basedfastfactorizationtechniques,”authoredbyÖzgürErgül,BahramKhalichi,andVakurB.Ertürk,and Chapter4“Kernel-independentfastfactorizationmethodsformultiscaleelectromagneticproblems,”authoredbyMengmengLi etal.,arebothdedicatedtofast factorizationtechniquesforanefficientandaccuratesolutiontotheelectromagnetic problem.Chapter3describeskernel-basedmethods,wheretheprimaryemphasislies ontheunderlyingkerneloftheproblem,adjustingitsutilizationtoeffectivelyhandleelectromagneticinteractionsinmoreefficientmannerswhilemaintainingaccurate numericalperformances.Inparticular,itfocusesonthemultilevelfastmultipolealgorithm(MLFMA),analyzingallitspropertiesandpossibleimplementationstoobtain accurate,efficient,andstablesolutionstomulti-scaleproblems.Instead,Chapter4 describeskernel-independenttechniquesthatareentirelyalgebraicandtakeadvantageoftherank-deficientnatureofMoMcouplingmatrixblocks,generatedbytwo distinctgroupsofbasicfunctionsthatarewellseparatedinspace.Byemployinglowrankfactorizationmethods,theMoMmatrixcanbeapproximated,enablingswift evaluationsofmatrix–vectorproductsiniterativesolutionsorrapiddirectsolvers.

Chapter5,entitled“Domaindecompositionmethod”andauthoredbyVíctor Martín,Hong-WeiGao,DiegoM.Solís,JoséM.Taboada,andZhenPeng,focuses ontheapplicationofdomaindecomposition(DD)methodsinsolvingtime-harmonic electromagneticwaveproblemsbasedonSIE.Thesemethodsarehighlydesirable duetotheircapacitytoyieldefficientandeffectivepreconditionediterativesolution algorithms,andtotheirinherentlyparallelnaturethatmakesthemparticularlyattractive,accordingtothecurrenttrendsincomputerarchitecture.Thechapterpresents twoclassesofDDmethods.OneclassutilizesthelatestdevelopmentsinthesurfacebaseddiscontinuousGalerkin(DG)formulationwherethecontinuityofcurrentsat domainboundariesisdirectlyenforcedbyemployinganinteriorpenaltyDGformulation.Instead,theotherclassofDDmethodsfollowsthe“tear-and-interconnect” approach,wheretransmissionconditionsareimposedalongthetearingcontours betweensubdomains.

Thenexttwochapters,Chapter6“Multi-resolutionpreconditioning,”authored byFrancescaVipiana,VíctorMartínandJoséM.Taboada,andChapter7“Calderón preconditionersforelectromagneticintegralequations,”authoredbyAdrienMerlini, SimonB.Adrian,AlexandreDély,andFrancescoP.Andriulli,arebothdevotedtopreconditioningtechniquesappliedtotheMoMmatrixtoimproveitsconditioningandso enablingafasterconvergenceoftheusediterativesolutionalgorithm.Chapter6aims toprovideallthetheoreticalandpracticalknowledgeforaproficientimplementation ofthemulti-resolution(MR)preconditionerintheelectromagneticanalysisofperfect electricconductor(PEC)structureswitharbitrary3Dshapesviaboththeelectricfield integralequation(EFIE)andthecombinedfieldintegralequation(CFIE).TheobjectiveofChapter7istoofferabroadcomprehensionoftheunderlyingmechanisms ofCalderónpreconditioning,presentinganoverviewofitsdiverseapplicationsto commonlyusedelectromagneticformulations.Whilethechapteracknowledgesthe

existenceofintricatemathematicaldevelopments,itprimarilyfocusesonproviding referencestodetailedanalysesratherthandelvingintothosecomplexitiesextensively.

Finally,Chapter8,entitled“Thedecoupledpotentialintegralequation”and authoredbyFelipeVicoandMiguelFerrando-Bataller,exploresanexperimental approachknownasthedecoupledpotentialintegralequation(DPIE).Theobjective ofthisformulationistodevelopamethodthatexhibitsrobustnessacrossallfrequencies,withaspecificfocusonlowfrequencieswhendealingwithmultipleconnected geometries.

References

[1]HarringtonRF. FieldComputationbyMomentMethod.Piscataway,NJ:IEEE Press;1993.

[2]WangX,PengZ,LimKH, etal.Multisolverdomaindecompositionmethod formodelingEMCeffectsofmultipleantennasonalargeairplatform. IEEE TransactionsonElectromagneticCompatibility.2012;54(2):375–388.

[3]HesfordAJandChewWC.Onpreconditioningandtheeigensystemsofelectromagneticradiationproblems. IEEETransactionsonAntennasandPropagation. 2008;56(8):2413–2420.

[4]BlancaIGT,RodríguezJL,ObelleiroF, etal.Experienceonradarcross sectionreductionofawarship. MicrowaveandOpticalTechnologyLetters. 2014;56(10):2270–2273.

[5]PengZ,LimKH,andLeeJF.Nonconformaldomaindecompositionmethods forsolvinglargemultiscaleelectromagneticscatteringproblems. Proceedings oftheIEEE.2013;101(2):298–319.

[6]SolísDM,MartínVF,AraújoMG, etal.AccurateEMCengineeringonrealistic platformsusinganintegralequationdomaindecompositionapproach. IEEE TransactionsonAntennasandPropagation.2020;68(4):3002–3015.

[7]ObelleiroF,TaboadaJM,SolísDM, etal.Directiveantennananocouplerto plasmonicgapwaveguides. OpticsLetters.2013;38(10):1630–1632.

[8]SongJMandChewWC.Multilevelfast-multipolealgorithmforsolvingcombinedfieldintegralequationsofelectromagneticscattering. Microwaveand OpticalTechnologyLetters.1995;10(1):14–19.

[9]TaboadaJM,AraujoMG,BertoloJM, etal.MLFMA-FFTparallelalgorithm forthesolutionoflarge-scaleproblemsinelectromagnetics(InvitedPaper). ProgressinElectromagneticsResearch.2010;105:15–30.

This page intentionally left blank

Chapter2

Surfaceintegralequationformulations

Thischapterincludesabriefreviewoffundamentalmaterialneededforunderstanding,formulating,andnumericallysolvingsurfaceintegralequationsappearingin electromagnetics.Usingthismaterial,wethendevelopthemostcommonintegral equationsfortime-harmonicproblemsinvolvinglinear,piecewisehomogeneous,and isotropicmaterials.Methodsfornumericallysolvingtheintegralequationsaredevelopedanddiscussed,includingapproachesfornumericallyevaluatingthesingularand near-singularintegralsthatarise.

2.1Maxwell’sequations

TheMaxwellequationsareasetoffourlaws:Faraday’slaw,Ampere’slaw,andthe electricandmagneticformsofGauss’slaw.Eachofthesetoffourequationsmay, inturn,bewritteninthefollowingthreedifferentforms:integral,differential,and boundaryforms.Ithasbeenargued[1]thattheintegralformsofMaxwell’sequations arethemostfundamentalinthesensethatallotherformsderivefromthem.Wedeal hereonlywithtime-harmonicproblemsandassumethatallsourceandfieldquantities varyas e j ω t ,effectivelyreplacingtheoperator ∂/∂ t by j ω inthetime-domainforms ofMaxwell’sequations[2].

2.1.1IntegralformofMaxwell’sequations

TheintegralformofMaxwell’sequationsforexp(j ω t )timedependenceis

1 DepartmentofElectricalandComputerEngineering,UniversityofHouston,USA

2 Consultant,JemezSprings,NM,USA

Integralequationsforreal-lifemultiscaleelectromagneticproblems

where S isan open surfacewith closed boundary C .Ontheotherhand,if S isa closed surface,enclosingavolume V ,wecanalsowrite

Theelectricandmagneticfieldstrengthquantities,(E , H ),arerelatedtothecorrespondingfluxdensityquantities(D, B)viathe constitutiveequations involvingthe localpermittivityandpermeability,(ε , μ),respectively:

Intheproblemsconsideredhere,weassumethatthematerialparameters(ε , μ)are linear,piecewisehomogeneous,andisotropic.InthefirsttwoMaxwell’sequations, asshowninFigure2.1, S hasaunitnormal ˆ n andtheclosedcurve C hasaunit tangent ˆ chosensuchthattheunitvector ˆ u = ˆ ׈n isbothnormalto C andpoints awayfrom S .Inthelasttwoequations, S isa closed surfacewithunitnormal ˆ n,the boundaryofavolume V .Volumetricelectricandmagneticsourcecurrents J and M , withunits[A /m 2 ]and[V /m 2 ]appearin(2.1)and(2.2),respectively.Whilethere isnoexperimentalevidencefortheexistenceofmagneticmonopolesorcurrents, magneticcurrentsandchargesnotonlygiveanelegantsymmetrytoMaxwell’sequationsbutalsoprovideflexibleandmathematicallyconvenientmeansforrepresenting electromagneticfields.Magneticandelectriccurrentsmerelycomprisemagneticand electricchargesinmotion;thevolumechargedensitiesofthosechargeswedefineas m [Wb/m 3 ]and q [C/m 3 ],respectively.Theconservationofchargeprinciplestates thatineveryboundedregionofspace,electricchargeisconserved;forbothconvenienceandmathematicalsymmetries,weassumethatthissameconservationlawalso holdsformagneticcharges.Thus,thetotalchargeforeitherchargetypechangesina regiononlyaschargescrosstheregion’sboundaries(i.e.,enterorleavetheregion). Therate(in[C/s]),forexample,atwhichelectricchargesdecreaseinaregionmust

Figure2.1ThesurfaceSwithunitnormal ˆ n isboundedbythecurve ∂ S = C.The unitvectors ˆ and ˆ u lieinthetangentplaneofS; ˆ isalsotangenttoC while ˆ u isnormaltoCandpointsawayfromS.Thethreeunitvectors satisfy ˆ u × ˆ =ˆn andformamutuallyorthogonalright-handedtriad alongC.

equalthenetelectriccurrentflux(in[C/s])exitingtheregion’sboundaries.Magnetic chargesandcurrentsaresimilarlyrelated;bothresultsaresuccinctlysummarizedby the continuityequations,

where,fortime-harmonicquantities, j ω playstheroleof ∂/∂ t ,withbothsidesof (2.7)and(2.8)representingratesofdecreaseofthetotalchargein V .Thoughthecontinuityequationsfollowfromphysics,independentofMaxwell’sequations,thelatter areconsistentwiththem.Forexample,ifwereplacethefirstsurfaceintegralin(2.2) withthe closed boundarysurfaceof(2.3),thenthecontourintegralof(2.2)vanishes (i.e.,surface S nolongerhasaboundarycontour C ),and(2.7)follows.Similarly, (2.8)followsfromapplyingboth(2.1)and(2.4)toacommonclosedsurface S

2.1.2PointordifferentialformofMaxwell’sequations

EachMaxwellequationindifferentialorso-calledpointformcorrespondstoanequationofthesamenameinintegralform.Moreover,eachdifferentialformequation maybederivedfromalimitingprocedureofitscorrespondingintegralform.For instance,oneusuallyappliesStokestheoremtothetwocurlequations,andthedivergencetheoremtothetwoscalarequations,shrinkingeachtoadifferentialsurfaceor volume,respectively,andfinallyobtaining

Theintegralformofthecontinuityequationscanbesimilarlytreated,resultingin

NotethattakingthedivergenceofthefirsttwoMaxwellequations,usingtheidentity ∇ ∇ × A = 0,theconstitutiveequationsandthecontinuityequations,oneobtains thelasttwoMaxwellequations,provided ω = 0 Forthisreason,whendealingwith time-harmonicelectromagneticsproblems,ifchargesandcurrentsareassumedtobe relatedviathecontinuityequations,oneneedsonlysatisfyFaraday’sandAmpere’s laws,withtheGausslawsautomaticallyfollowing.

2.1.3BoundaryformofMaxwell’sequations

TheboundaryformsofMaxwell’sequationsarealsoobtainedfromtheintegralforms. Forthefirstpair,Figure2.2,ashortlength,infinitesimalheight(h << a)rectangular

Rectangular path for deriving the boundary form of Ampere's law. (a)(b) Pillbox surface for the boundary form of Gauss's law.

Figure2.2GeometriesusedtoderiveboundaryformsofMaxwell’sequations pathwhoselongestsidesareparalleltoandonoppositesidesofamaterialboundary relatethefieldintensitycomponents(E , H )onoppositesidestoany surfacecurrent flux(Js , Ms )throughtherectanglethatmightbepresentontheboundary;fluxintegral contributionsfromfinitevolumetriccurrents(J , M )orthefluxfields(B, D)donot contributetothelimitastheareaenclosedbythepathvanishes,leavingonlysurface currentfluxcontributions.Thisresultsintheequations(2.15)and(2.16)belowwhere ˆ n isnormaltotheinterfaceandpointsfromthe tothe + regionassociatedwitheach sideoftheinterface.Forthesecondpair,twosuchrectangularpathsperpendicularto oneanotherarethecross-sectionsofacylindricalpillboxofinfinitesimalheightsides andradius a (h << a)whosecircularendfacesareparalleltoandonoppositesides oftheinterfaceofthetwomedia.IntegratingthetwoGausslawsoverthevolume enclosedcapturesanysurfacechargedensitycontributionsfromtheinterface;volume chargeandcylindricalfacecontributionsvanishinthelimitastheheightvanishes, whereasthearea(π a2 )commontothetwofluxfaceandchargedensityintegrals cancels,resultingin(2.17)and(2.18)below.

s = (E + E ) ׈n,(2.15) Js =ˆn × (H + H ),(2.16) qs =ˆn (D+ D ),(2.17) ms =ˆn · (B+ B ),(2.18)

where Ms and Js aremagneticandelectricsurfacecurrentdensitiesand qs and ms are electricandmagneticsurfacecurrentdensities,respectively.Similarly,theboundary formofthecontinuityequationsforvolumetriccurrents J and M ataninterfaceis obtainedas

ˆ n (J + J ) + ∇ s Js =−j ω qs ,(2.19)

ˆ n · (M + M ) + ∇ s · Ms =−j ω ms ,(2.20)

where ∇s · isthesurfacedivergenceoperator.Weremindthereaderthatthetwoline integralsontheLHSoftheintegralformsofFaraday’sandAmpere’slawshavethe units(V)and(A),respectively,asdothemagneticandelectricline(filament)currents K and I passingthroughanenclosedpath;hence,theunitsof surface currentdensities Ms and Js mustbe[V /m]and[A /m],while volumetric currentdensityunitsfor M and J are[V /m 2 ]and[A /m 2 ].Notealsothatadifferentialvolumetricintegralover afilamentcurrentproducesa dipolemoment result, Id ,where d isthedifferential lengthofthefilamentthatfallsinsidethedifferentialvolume.Thus,inthesense thattheydonotradiate,pointsourcesofcurrentareessentiallynon-existentbecause aradiatingcurrentelementmusthaveatleastsomenon-vanishinglength, d > 0. Sinceadipole’sorientationisalsoimportant,itsmomentisoftenexpressedinvector formas Id or Id .

2.1.4TheHelmholtzequationsandpotentialrepresentations

Wehavealreadynotedthatif ω = 0,thetwoGausslawsareimpliedbythedifferential formsofFaraday’sandAmpere’slawsplusthecontinuityequations.Togetherwith theirbehavioratinfinity,thetwocurlequationsshouldbesufficienttodetermine (E , H )foragivenpairofcurrentsources(J , M ).Onestepinthisdirectionisto eliminateeither E or H betweenthetwo.Forexample,theeliminationofthemagnetic fieldiseasilyaccomplishedbytakingthecurlofbothsidesofFaraday’slaw,(2.9),and substitutingfromAmpere’slaw,(2.10),toeliminatethemagneticfield.Thisyields theso-calledvectorHelmholtzequationfortheelectricfield:

where k = ω √με = 2π/λ isthe wavenumber andwhere λ isthewavelengthofa wavepropagatinginthematerialifboth μ and ε arebothreal(i.e.,themedium islossless).Applyingthesameprocedure,butreversingtherolesofFaraday’sand Ampere’slaws,andeliminatingtheelectricfieldleadstothemagneticformofthe vectorHelmholtzequation:

Unfortunately,itstillremainsdifficulttofindgeneralsolutionsof(2.21)and(2.22). Itturnsoutthat potential representationsofthefieldsarebettersuitedtofinding solutionssincetheycanbechosentoautomaticallysatisfyconditionssuchasthe vanishingofacurlordivergence.Tothisend,wefirstnotethatsinceMaxwell’s equationsarelinear,thesourcepair(J , M )canbepartitionedintoasumofsimpler sourcepairsetsas(J , 0) + (0, M ),eachofwhichinvolvesbuta single currentspecies (electricormagnetic,respectively).Onceweobtainthefieldsduetoeachsingle currentspeciesactingalone,weobtainsolutionsforbothpresentbysuperposingthe resultsforeachspecies.Forexample,forelectriccurrentsonly,weset M = 0 and m = 0in(2.9)and(2.12),respectively,anddeterminethefields(E J , H J ),withthe superscriptindicatingthesourcetype.Forthiscase,andunderthemildhypothesisof derivativecontinuity[3],themagneticformofGauss’slaw, ∇ ∇· BJ = 0,impliesthat BJ =∇ ∇ ∇× A (andviceversa)foravector A wecallthe magneticvectorpotential since

Integralequationsforreal-lifemultiscaleelectromagneticproblems

weuseittodirectlydeterminemagneticfieldquantities, H J = 1 μ BJ = 1 μ ∇ ∇ ∇× A. A is notuniquelyspecified,however,sincebothitscurlanddivergencecanbespecified independently,accordingtotheHelmholtzdecompositiontheorem[4].Laterwewill alsospecify ∇ ∇ ∇· A soastosimplifythedefiningequationfor A;forthemoment, however,wesubstitutethisrepresentationof H into(2.9),andrearrangeto ∇ ∇ ∇× (E J + j ω A) = 0.Theidentity ∇ ∇ ∇× (∇ ∇ ∇ ) = 0 impliesthatif E J hastheform E J = j ω A −∇ ∇ ∇ ,itwouldautomaticallysatisfythiszerocurlcondition;butitcanalsobe shownthatunderrathermildrestrictions,every E J canalwaysbesorepresented[3]. Thenewlyintroducedfunction iscalledthe electricscalarpotential andthenegative signischosentoagreewiththefamiliarelectrostatic(i.e., ω → 0)representationfor E J involvingscalarpotentialonly.Insummary,wehavenowdevelopedthepotential fieldrepresentation(E J , H J ) = ( j ω A −∇ ∇ ∇ , 1 μ ∇ ∇ ∇× A)forthefieldsduetoelectric currentsources J actingalone.Similarly,formagneticcurrents M actingalone,the potentialfieldrepresentationis(E M , H M ) = ( 1 ε ∇ ∇ ∇× F , j ω F −∇ ∇ ∇ ),involving the electricvectorpotential F and magneticscalarpotential .Superposingthese results,weobtainfinallytheso-called mixedpotential representationof(E , H ) = (E J + E M , H J + H M ):

Toobtainwaveequationsforthepotentials,itisconvenienttoreturntothepotentialfieldrepresentationsgeneratedbyelectricormagneticsourcespecies,respectively, asfollows:

M = 0 : E J =−j ω A −∇ ∇

Substitutingthepotentialrepresentation(2.25)for(E J , H J )intoAmpere’slaw,using theidentity ∇ ∇ ∇× (∇ ∇× A) =∇ ∇ (∇ ∇ ∇·

2 A,andrearranging,weobtain

Similarly,substituting(E M , H M ),(2.26),intoFaraday’slaw,wefind

Despitestrongsimilaritiesbetweentheequationpair(2.21)and(2.22)andthepair (2.27)and(2.28),thelatterprovideanimportantopportunityforsimplificationsince thedivergencesof A and F remainyettobespecified.Ofmanypossibleso-called gauge choiceshere,theso-calledLorenzgauges,areobviousones:

sothat(2.27)and(2.28)immediatelysimplifyto ∇ 2 A + k 2 A =−μJ ,(2.31)

2 F + k 2 F =−ε M ,(2.32)

respectively.Equationsforthescalarpotentials and areeasilyobtainedbysubstituting(2.25)and(2.26)intothetwoGausslaws,(2.11)and(2.12),respectively, yielding

and

Additionaladvantagesofthevector-valuedequations(2.31)and(2.32)arethatthe operator ∇ 2 ,whenappliedtotherectangularcomponentsofthepotentialvectors, reducestothescalarLaplacianoperatorsothatalltherectangularcomponentsof (2.31)and(2.32)plusthetwoscalarpotentials(2.33)and(2.34)satisfythesame genericscalarwaveequation,

2 ψ + k 2 ψ =−f (r ),(2.35) forsomepotential ψ withscalarvolumesourcedensity f (r ).Wenotethesourceterm f (r )canbewrittenasasuperpositionorconvolutionintegral

f (r ) = V f (r )δ (r r ) dV ,(2.36)

where V includesatleastthe support of f (r ),supp (f (r )),thatis,thesmallestclosed regionoutsideofwhichthefunction f (r )vanishesidentically.Thus,thesetsupp (f (r )) iscompletelycontainedwithin V .Thethree-dimensionaldeltafunction δ (r r ) = δ (x x )δ (y y )δ (z z )representsa pointsourceofunitdensityat r ;noteits unitsare[1/m 3 ].Wewishtoexpress ψ asasource-weightedsumoverpointsource solutions,andtodosowefirstintroduceascalarfunction G (r , r ),asolutionofthe point-sourceequation

∇ 2 G + k 2 G =−δ (r r ),(2.37)

where G (r , r )representsagenericpotentialfieldat r duetoaunitdensitysource at r .Sincea(scalar)pointsourceisbothnon-directionaland(2.37)iscoordinate systemindependent,weshouldalsoexpectthesameof G ,i.e.,thatitdependsonly ontheradialseparation R =|r r | betweentheobservationpoint r andsourcepoint r .Wealsonotethatinalocalsphericalcoordinatesystemwithanoriginat r , ∇ 2 G = 1 R2 d dR R2 dG dR ,fromwhichoneeasilyverifiesthat

G (r , r ) = e jkR 4π R (2.38) satisfies(2.37)everywhereexceptpossiblyat R = 0(sincetheexpressionfor ∇ 2 involvesdivisionbyzerothere).Toalsoverifytheunitvolumeintegralbehaviorof the3Ddeltafunctionat R = 0,wemustexaminetheequalityofvolumeintegralsover

Integralequationsforreal-lifemultiscaleelectromagneticproblems termsonbothsidesof(2.37).Thenon-directionalnatureoftheproblemalsosuggests integratingoveravolumetrictestsphere, Va ,ofradius a centeredat R = 0,noting that ∇ 2 = ∇ · ∇ ,andevaluatingtheintegraloftheLaplaciantermviathedivergencetheorem.Indeed,onefinds Va (

independent ofthetestsphereradius a.

Finally,thelinearityofthescalarwaveequationsuggeststhatifanarbitrarysource densitydistributioncanbeexpressedasaweightedsuperpositionofpointsources, (2.36),thenthedistribution’spotentialcanbeexpressedasasimilarlyweighted superpositionofpotentialsduetopointsources:

Indeed,that(2.39)isasolutionof(2.35)iseasilyverifiedbydirectsubstitutionand useof(2.37)and(2.36).Weremarkthat,byimplication,theintegralin(2.39)must alsoincludeanycontributionsfromsurface,filament,orpointsourcespresent,which onewritesintheforms S G (r , r )fs (r ) dS , C G (r , r )f (r ) d ,or

respectively.

Using(2.23),(2.24),(2.31)–(2.35),and(2.39),wearenowabletowritethefields intermsofpotentialsasfollows[2]:

Beforeleavingthissection,wecallthereader’sattentiontothefollowingobservations concerning(2.40)–(2.45):

● Thoughtheintegrationdomainsandsourcesintheintegralsabovearevolumetric, theyimplicitlyincludesourceintegralsoversurfacesandfilaments,aswellas sumsoverpointsources.

● Notetheextenttowhichwerelyonlinearityandsuperposition:thefields(E , H ) arevectorcomponentsums,eachtermofwhichisasumofrealandimaginary parts,andconstructedasalinearsuperpositionofavectorpotential,thevectorvaluedgradientofascalarpotential,andthecurlofavectorpotential,i.e.,a

sumoverbothcurrentandchargeterms,respectively.Eachofthese,inturn,is aconvolutionintegral,i.e.,asumofapoint-sourceresponses G (r , r )weighted bysourcedensitiesoftheappropriatechargeorcurrentspeciesdefinedon V . Lateron,indiscretizingaproblem,wefurthersubdividethesourcedomainsinto collectionsofsimplersubdomains,eachwithsimplifiedlocalsourcerepresentations.Inthelattercase,however,thenewlevelofsumsistypicallyrepresented asmatrix–vectorproducts.

2.1.5Farfieldsandfarpotentials

Inthefarfield,itisconvenienttoexpressafieldobservationpoint r inspherical coordinatesas r = r (cos φ sin θ ˆ x + sin φ sin θ ˆ y + cos θ ˆ z )whoseoriginisat r = 0, assumednotfarfromorevenwithinthesourceregion.Atlargeradialdistances r =|r | fromallsourcepoints r ,theapproximation R =|r r |≈ r −ˆ r r where ˆ r = r /r , canbeusedinthelocallyvaryingphasefactorterminthenumeratorof G (r , r )while R ≈ r canbeusedinthedenominatortoapproximate G (r , r )as e jkr 4π r e jk r r inthefar vectorpotentials,yielding

Similarly,thefarscalarpotentialsare

Butsince

,wemaysimplyreplace

r inthefar field.CombinedwiththeLorenzgaugeconditions,itthenfollowsthatthegradients ofthescalarpotentialsareentirelyradiallydirectedandcompletelycancelradial componentsofthevectorpotentialcontributionstothefarfield;thus,inthefarfield, fields(E , H )becomecompletelytransversetotheobservationvector r andcanbe writtenas

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.