• Aucun résultat trouvé

Optimized domain decomposition methods for the spherical Laplacian

N/A
N/A
Protected

Academic year: 2022

Partager "Optimized domain decomposition methods for the spherical Laplacian"

Copied!
27
0
0

Texte intégral

(1)

Article

Reference

Optimized domain decomposition methods for the spherical Laplacian

LOISEL, Sébastien, et al.

LOISEL, Sébastien, et al . Optimized domain decomposition methods for the spherical Laplacian. SIAM Journal on Numerical Analysis , 2010, vol. 48, no. 2, p. 524–551

DOI : 10.1137/080727014

Available at:

http://archive-ouverte.unige.ch/unige:6259

Disclaimer: layout of this document may differ from the published version.

(2)

Laplaian

S. Loisel

J. Cté

M. J.Gander

L.Laayouni

A. Qaddouri

September 21, 2009

Abstrat

The Shwarz iteration deomposes a boundary value problem overa large domain

Ω

into smaller

subproblems by iteratively solving Dirihlet problems on a over

Ω 1 , ..., Ω 𝑝

of

Ω

. In this paper, we

disussalternatetransmissiononditionsthatleadtofasteronvergenefortheLaplaianonthesphere

Ω

. We look at Robin onditions, seond order tangential onditions and a disretized version of an

optimalbutnonloaloperator.

1 Introdution

AttheheartofnumerialweatherpreditionalgorithmslieaLaplaeandpositivedeniteHelmholtzprob-

lemsonthesphere[35℄. Reently,therehasbeeninterestinusingniteelements[5℄anddomaindeomposi-

tionmethods[4,26℄. TheShwarziteration[23,24,25℄anditsoptimizedvariantshavebeenverysuessful

andthesubjetofmuhreentresearh. WenowoutlinethehistoryofOptimizedShwarzMethods(OSM)

andreferto[14℄forfurtherdetails,aswellasaompletederivationforHelmholtzandLaplaeproblems in

theplane.

The optimizedShwarzmethod was introduedin [2,31, 32℄under various names. SinetheOSMhas

oneormorefreeparameters,whihneedtobehosenarefullytoguaranteethebestonvergenerate,there

followedaneorttondthebestpossibleparameterhoies[13,21,20℄. Despitethesepositivedevelopments,

aproofofonvergeneforageneralsituationhasprovenelusive[22,27℄. Onewayofobtainingonvergene

istodenearelaxationofthemethod[7℄.

TheusualmethodforoptimizingthefreeparametersistotakeaFouriertransformofthepartialdier-

entialequation, obtaininganexpliitreurrenerelationfortheiteration. This works onlyinspeialases

(e.g., thedomain is aretangle and the dierentialoperator is the Laplaianwith homogeneousDirihlet

onditions). In addition to theLaplae and Helmholtzproblems [19℄, the method an beused for various

otheranonialproblems(f. [9,18,17,29,28℄foronvetion-diusionproblems,[16℄forthewaveequation,

[8℄ for uid dynamis, [33℄ for the shallow water equation). The urrent paper is a ontinuation of this

researh.

OurpaperprovidesanOSMforthespherialLaplaian,adierentialoperatorwhihhadpreviouslynot

beenanalyzed,usingFourieranalysis. However,webrieymentionthattherehasbeenmuhreentresearh

onvariousother aspetsoftheOSM.Forinstane, domainswithornersan giverisetosingularities,and

this isanalyzed in [3℄. Inasimilar vein, dierentialoperators with disontinuousoeientsoften leadto

ill-onditionedproblems, and anOSMforthis situation isprovidedin [12℄. A ompletelydierent issueis

to phrasethe OSMin analgebrai way. In atypialappliation, one hasaode to omputethe stiness

matrix

𝐴

,butonewouldideallyprefernottohavetorewritetheentireodeinordertoimplementanOSM.

Inthatvein,[34℄providesaframeworkforexpressingtheOSMin thelanguageofmatries.

Setion de Mathématiques, Université de Genéve, CP 64, CH-1211 Geneva, Switzerland (sloiselgmail.om,

gandermath.unige.h).

Reherheenprévisionnumérique,EnvironnementCanada,2121Route Transanadienne,Dorval, Québe,Canada H9P

1J3(jean.otee.g.a, abdessamad.qaddourie. g. a)

Shool of Siene and Engineering Al AkhawaynUniversity Avenue Hassan II, 53000 P.O. Box 2165, Ifrane, Moroo

(L.Laayouniaui.ma)

(3)

deompositionmethods areoftennaturallyparallel,andheneanbeusedonlargelustersandsuperom-

puters. Thealgorithmthat runsoneahproessorisasmall,loalversionoftheglobalproblem,soproven

sequentialodesan beadapted and beome loal solverswhih will thenrun in parallel onsuh lusters.

Insuhasenario,animportantquestionisthat ofsaling.

In the simplest saling senario, one xes the number

𝑝

of subdomains, and one piks subdomains

Ω 1 , . . . , Ω 𝑝

insuhawaythattheoverlapbetweenadjaentsubdomainsisexatlyoneelementthik. Inthis

situation,itiswellknown[36℄thatthe(1-level)Shwarzmethodhasaonvergenefatorof

𝜌 𝐶𝑆 = 1 − 𝑂(ℎ)

.

Inotherwords,ateahstepoftheiterativemethod,theerrorismultipliedbytheoeient

𝜌 𝐶𝑆 < 1

. Beause

this oeientdepends on

, and tends to 1when

tendsto zero,wesee that moreand moreiterations

willbeneededtoreahagiventoleraneaswemake

smaller. It ispreferableto ndanalgorithmwhose

onvergenefatoreitherdoesnotapproah

1

,oratleastapproahes

1

nottooquiklywhen

tendstozero.

In this ontext, themain advantageof theOptimized Shwarzmethods is that they havemuh better

onvergenefatorsthanthelassialShwarzmethod. Forexample,theone-sidedoverlappingOO0method

hasaonvergenefatorof

𝜌 𝑂𝑂0 = 1 − 𝑂(ℎ 1/3 )

,andtheoverlappingOO2methodhasaonvergenefator of

𝜌 𝑂𝑂2 = 1 − 𝑂(ℎ 1/5 )

.

Inadditiontooverlappingmethods,Lions[25℄proposedtouseRobintransmissiononditionsalongthe

interfae to obtain nonoverlappingdomain deomposition methods. There is no nonoverlapping lassial

Shwarzmethodwithwhihtoompare,buttheOO0andOO2methodsanbeusedinthenonoverlapping

ontext. Theonvergenefators are

1 − 𝑂(ℎ 1/2 )

and

1 − 𝑂(ℎ 1/4 )

, forthenonoverlappingOO0 andOO2 methods,respetively.

Although ouranalysis islimitedtolatitudinalsubdomains,wewillseein Setion5thatweobtainvery

good performane,evenwhensubdomainshavemorevariedshapes.

Wenotethat,ifthenumberofsubdomains

𝑝

inreasesas

tendstozero,itisneessarytodesigna2-level

algorithm to maintaingood performane. Ina 2-levelalgorithm, detailed informationsfrom neighbors are

ombinedwithlesspreise informationstoredonalow-resolutionoarsegrid. Thedesignand analysisof

anOptimizedShwarzMethodwithaoarsegridorretionisahallengingproblemwhihwillbeanalyzed

inanupomingpaper[10℄.

1.1 Our ontributions

Inthispaper,weintrodueimprovedtransmissionoperators fortheLaplaeproblemonthe unitsphere

Ω

in

3

. We modify the Shwarz iteration by replaing theDirihlet onditions onthe interfaes by Robin

or seondordertangentialonditions. Thisallowsusto signiantlyimprovetheonvergenefatorofthe

iteration. Toanalyzetheonvergenerateandoptimizetheoeients,weusetheFouriertransformonthe

sphere. Inadditionto theusual Robinand seondordertangentialonditions,wealsodisusstheuseofa

nonloaloperatoralongtheinterfaes. ThisnonloaloperatorisrelatedtothesquarerootoftheLaplaian

oftheinterfaesandwasrstintroduedfornumerialalulationsin[4,26℄. Whiletheontinuousanalysis

showsthat theseoperators resultin aniterationthatonvergesin twosteps,in pratiewedonotseethis

exatbehavior. However,theresultingiterationseemstoonvergeataveryfastratethatisindependentof

themeshsize

orthethiknessoftheoverlap

𝐿

.

Wehighlightfour ontributionsofthispaper. First,wehaveaninnovativeuseoftheenvelopetheorem

toestablishanequiosillationproperty. Seond,wehighlightourontinuousanddisreteanalysesandtheir

omparison. Third,ouranalysistakesplaeonthesphere,whihisidealforweatherandlimatesimulation.

Fourth, we provide a nite element method whih is appropriate for spherial disretizations. Sine the

spherialLaplaianissingular,thehoieoftheniteelementspaeisimportantandapoorhoieleadsto

divergentiterationsandpoorlyonditionedsystems. Weshowthatwiththeproperhoieofbasisfuntions,

theonvergenefatoroftheoptimizedShwarzmethodsareunaeted.

This paperis organizedas follows. InSetion 2, weintrodue ourmodel problem andstate our main

results. InSetion 3, we review the Laplae operator on the sphereand reall theShwarz iteration and

its onvergene estimates, previously published in [4℄; we also give a new semidisrete estimate whih is

substantiallysimilartotheontinuousone. Insetion4,weperformtheFourieranalysisofthenewOptimized

ShwarzMethods. Insetion5,wepresentnumerialresultsthat agreewiththetheoretialpreditions.

(4)

2

1

a b

2

Ω 1

a

3

2

b 2

a 3 b 1

Figure1: Latitudinal domaindeomposition. Left: twosubdomains;right: multiplesubdomains.

2 OSM on the sphere

Considerthe followingiteration. Let

𝑏 < 𝑎

. Begin with random andidate solutions

𝑢 0

and

𝑣 0

. Dene

𝑢 𝑘+1

and

𝑣 𝑘+1

iterativelyby:

 

 

Δ𝑢 𝑘+1 = 𝑓

in

Ω 1 = { (𝜑, 𝜃) ∣ 0 ≤ 𝜑 < 𝑎 } , 𝑢 𝑘+1 (𝑎, 𝜃) = 𝑣 𝑘 (𝑎, 𝜃) 𝜃 ∈ [0, 2𝜋),

Δ𝑣 𝑘+1 = 𝑓

in

Ω 2 = { (𝜑, 𝜃) ∣ 𝑏 < 𝜑 ≤ 𝜋 } , 𝑣 𝑘+1 (𝑏, 𝜃) = 𝑢 𝑘 (𝑏, 𝜃) 𝜃 ∈ [0, 2𝜋);

(1)

(see gure 1.) This is the Shwarz iteration for two latitudinal subdomains. The idea of the Optimized

Shwarz Methods is to replae the Dirihlet onditions on the interfae by Robin or other transmission

onditions. Themodiediterationis:

 

 

Δ𝑢 𝑘+1 = 𝑓

in

Ω 1

((𝜓 + ∂𝜑 )𝑢 𝑘+1 )(𝑎, 𝜃) = ((𝜓 + ∂𝜑 )𝑣 𝑘 )(𝑎, 𝜃) 𝜃 ∈ [0, 2𝜋),

Δ𝑣 𝑘+1 = 𝑓

in

Ω 2

((𝜉 + ∂𝜑 )𝑣 𝑘+1 )(𝑏, 𝜃) = ((𝜉 + ∂𝜑 )𝑢 𝑘 )(𝑏, 𝜃) 𝜃 ∈ [0, 2𝜋);

(2)

where

𝜓

and

𝜉

are either real oeients or trae operators, and

Ω 1 , Ω 2

are as previously dened. If

𝜓

and

𝜉

aredierentialtraeoperatorsoforder

𝑘

,wesaythattheoptimizedShwarzmethodisOptimizedof

Order

𝑘

, orOOk[15℄. Choiesinlude:

1.

(𝜓𝑤)(𝜃) = 𝑐𝑤(𝜃)

where

𝑐

isarealoeient. Thisresultsin aRobinor OO0transmissionondition.

2.

(𝜓𝑤)(𝜃) = 𝑐𝑤(𝜃)+𝑑𝑤 ′′ (𝜃)

,where

𝑐

and

𝑑

arerealoeients. Thisresultsinaseondordertangential, or OO2transmissionondition.

3. Anonloalhoieof

𝜓

leadingtoaniterationthatonvergesintwosteps.

Wewill showin Setion 3thattheonvergenefator depends onthethiknessof theoverlap

𝐿

, given

by

𝐿 := − 2 log 1 + cos 𝑎 sin 𝑎

sin 𝑏 1 + cos 𝑏 ≥ 0.

OurmainresultforRobin(orOO0)transmissiononditionsisas follows.

Theorem2.1(OptimizedRobinorOO0oeients). Considerthe iteration(2)with

(𝜓𝑤)(𝜃) = 𝑐𝑤(𝜃)

and

(𝜉𝑤)(𝜃) = 𝑑𝑤(𝜃)

. Set

𝑥 = − 𝑑 sin 𝑏

,

𝑦 = 𝑐 sin 𝑎

. Let

ℎ = 1/𝑁

bethe grid size.

The one-sidedonditionis

𝑥 = 𝑦

. Withthisondition,andfurtherassumingthat

𝐿 < 4/3

and

𝐿 > ℎ 𝛼

for some

𝛼 < 3/2

, there is a unique optimized hoie of

𝑥

leading to the best possible onvergene

fator. Using

𝑥 = 𝑦 = ˆ 𝑥 1 = 𝐿 1 3

, theonvergenefatorevery other stepis

𝜌 = 1 − 4𝐿 1 3 + 𝑂(𝐿 2 3 )

.

If

𝑥 ∕ = 𝑦

is allowed, the optimizedhoie

(𝑐, 𝑑)

is two-sided. Assumethat

𝐿 > ℎ 𝛼

,

𝛼 < 5/4

. Thereis

a unique

(𝑐, 𝑑)

leading tothe bestonvergenefator. Using

𝑥 = 8 1 5 𝐿 1 5

and

𝑦 = 8

2 5

2 𝐿 3 5

leads toa

onvergenefator of

𝜌 = 1 − 2 ⋅ 2 3 5 𝐿 1 5 + 𝑂(𝐿 2 5 )

everyother step.

(5)

Theonvergenefator

𝜌 < 1

multipliesthe

𝐿 2

errorofthetraeoneahsubdomaineveryotherstep. We

seethatusingasymmetri Robinonditionsleadstoanimprovementonthenumberofiterationsrequired

for a given tolerane, from

𝑂(𝐿 1 3 )

to

𝑂(𝐿 1 5 )

. The one-sided estimate however is easier to use with

𝑝

subdomains,andsowewouldbeinterestedinahievinga

𝑂(𝐿 1 5 )

algorithmusingonlyone-sidedestimates.

Thisispossibleifweuseseondordertangentialtransmissiononditions.

Theorem 2.2 (One-sidedseond order tangential,or OO2,optimized oeients). Consider the iteration

(2)with

(𝜓𝑤)(𝜃) = 𝑐𝑤(𝜃) + 𝑑𝑤 ′′ (𝜃)

,

(𝜉𝑤)(𝜃) = ˜ 𝑐𝑤(𝜃) + ˜ 𝑑𝑤 ′′ (𝜃)

. Let

𝑠 = 𝑐 sin 𝑎 = − 𝑐 ˜ sin 𝑏

and

𝑡 = 𝑑 sin 𝑎 =

− 𝑑 ˜ sin 𝑏

, whih is the one-sided assumption. Assumethat

𝐿 < ℎ 𝛼

,

𝛼 < 5/4

. There isa uniquehoie of

𝑠, 𝑡

leadingtothebestpossibleonvergenefator. Using

𝑠 = 1 2 2 3 5 𝐿 1 5

and

𝑡 = 1 2 2 1 5 𝐿 3 5

leadstoaonvergene

fatorof

𝜌 = 1 − 4 ⋅ 2 2 5 𝐿 1 5 + 𝑂(𝐿 2 5 )

everyother step.

Theonvergerateestimateshold evenif

𝐿

issmallerthan

𝛼

(e.g., whenthegridishighlyanisotropi, with

𝛼 = 3/2

or

5/4

,asabove),butthentheoeientsgivenmaynolongerbeoptimized. Anextremease

iswhen

𝐿 = 0

,anonoverlappingase,andweanalyzethisaseinSetion3.1. Themain aseis

𝐿 = 𝑂(ℎ)

andishandledbytheTheorems2.1and2.2.

Our numerialexamplesarebased ontwodierentodes. Many odesin meteorologyuse thedisrete

Fouriertransformin thelongitudevariable

𝜃

toobtainasemispetral solver. This solvermakestheimple- mentationofourthreetypesoftransmissiononditionsequallyeasy,forlatitudinalinterfaes. However,suh

solversdonoteasilyhandlearbitrarilyshapedsubdomains,andthegridshavesingularitiesat thepoles.

Ourseond odeisaniteelementsolveronthesphere

Ω

. Thevariationalformulationofthespherial Laplaianleadsto integralsonthesphere

Ω

,and sowemust hoose pieewise

𝐶 1

basis funtionsfor some

niteelementspaeonthesphere

Ω

. Weuseastandardprojetiontehniquetogeneratespherialelements

frompieewiselinearelementsonapolyhedralapproximationof thesphere

Ω

.

3 The Laplae operator on the sphere

Ω

WetaketheLaplaeoperatorin

3

,givenby

Δ𝑢 = 𝑢 𝑥𝑥 + 𝑢 𝑦𝑦 + 𝑢 𝑧𝑧 ,

rephraseitinspherialoordinatesandset

∂𝑢

∂𝑟 = 0

to obtain

Δ𝑢 = 1

sin 2 𝜑

2 𝑢

∂𝜃 2 + 1 sin 𝜑

∂𝜑 (

sin 𝜑 ∂𝑢

∂𝜑 )

,

where

𝜑 ∈ [0, 𝜋]

istheolatitudeand

𝜃 ∈ [ − 𝜋, 𝜋]

thelongitude.

3.1 The solution of the Laplae problem

WetakeaFouriertransformin

𝜃

butnotin

𝜑

;this letsusanalyzedomaindeompositionswithlatitudinal boundaries. TheLaplaianbeomes

Δˆ ˆ 𝑢(𝜑, 𝑚) = − 𝑚 2

sin 2 𝜑 𝑢(𝜑, 𝑚) + ˆ 1 sin 𝜑

∂𝜑 (

sin 𝜑 ∂ 𝑢(𝜑, 𝑚) ˆ

∂𝜑 )

, 𝜑 ∈ [0, 𝜋], 𝑚 ∈ ℤ.

(3)

Forboundaryonditions,theperiodiityin

𝜃

istakenareofbytheFourierdeomposition.Thepolesimpose that

𝑢(0, 𝜃)

and

𝑢(𝜋, 𝜃)

donotvaryin

𝜃

. For

𝑚 ∕ = 0

thisisequivalentto

ˆ

𝑢(0, 𝑚) = 𝑢(𝜋, 𝑚) = 0, 𝑚 ˆ ∈ ℤ, 𝑚 ∕ = 0.

(4)

For

𝑚 = 0

,therelation

𝑢 𝜑 (0, 𝜃) = − 𝑢 𝜑 (0, 𝜃 + 𝜋)

leadsto

∫ 2𝜋

0 𝑢 𝜑 (0, 𝜃) 𝑑𝜃 = − ∫ 2𝜋

0 𝑢 𝜑 (0, 𝜃) 𝑑𝜃

,i.e.,

ˆ

𝑢 𝜑 (0, 0) = 𝑢 ˆ 𝜑 (𝜋, 0) = 0.

(5)

If

𝑢

is asolution of

Δ𝑢 = 𝑓

then so is

𝑢 + 𝑐

(

𝑐 ∈ ℝ

), henethe ODE for

𝑚 = 0

is determined up to an

additiveonstant.

(6)

With

𝑚 ∕ = 0

xed,thetwoindependentsolutionsof

Δ𝑢 = 0

are

𝑔 ± (𝜑, 𝑚) =

( sin(𝜑) cos(𝜑) + 1

) ±∣ 𝑚 ∣

= 𝑒 ±∣ 𝑚 log cos(𝜑)+1 sin(𝜑) , 𝑚 ∈ ℤ ∖ { 0 } .

For

𝑚 = 0

thetwoindependentsolutionsgive

ˆ

𝑢(𝜑, 0) = 𝐶 1 + 𝐶 2 log 1 − cos 𝜑 sin 𝜑 .

Ifweinsist,forinstane, that

𝑢 ∈ 𝐻 1 (Ω)

,weeliminatethelogarithmitermandweobtainthat

𝑢(𝜑, ˆ 0)

isa

onstantin

𝜑

.

All theeigenvaluesof

Δ

areoftheformof

− 𝑛(𝑛 + 1)

for

𝑛 = 0, 1, ...

;in partiular,theyarenon-positive (and

Δ

isnegativesemi-denite).

3.2 The Shwarz iteration for

Δ

with two latitudinal subdomains

ConsidertheShwarziteration(1). Weareinterestedin studyingtheerrorterms

𝑢 0 − 𝑢

and

𝑣 0 − 𝑢

where

Δ𝑢 = 𝑓

. Observethat

𝑢 𝑘 − 𝑢

and

𝑣 𝑘 − 𝑢

arelassialShwarziterates asperequation(1),butwith

𝑓 = 0

.

For this reason,it sues to analyzethe lassialShwarziteration with

𝑓 = 0

, whih weassume forthe

remainderofouranalysis.

UsingtheFouriertransformin

𝜃

,weanwrite

𝑢 ˆ 𝑘+2 (𝑏, 𝑚)

expliitlyin termsof

𝑢 ˆ 𝑘 (𝑏, 𝑚)

. Thisallowsus

toobtainaonvergenefatorestimate,whihwereallfrom [4℄.

Lemma3.1. The Shwarziteration onthesphere

Ω

partitionedalong twolatitudes

𝑏 < 𝑎

onverges(exept

for the onstantterm). The onvergene fator

∣ 𝑢 ˆ 𝑘+2 (𝑏, 𝑚)/ˆ 𝑢 𝑘 (𝑏, 𝑚) ∣

is

𝐶(𝑚) =

( sin(𝑏) cos(𝑏) + 1

) 2 ∣ 𝑚 ∣ (

sin(𝑎) cos(𝑎) + 1

) − 2 ∣ 𝑚 ∣

< 1.

(6)

This onvergenefatordependsonthefrequeny

𝑚

of

𝑢 𝑘

onthelatitude

𝑏

.

AnanalysisthatislosertothenumerialalgorithmwouldbetoreplaetheontinuousFouriertransform

in

𝜃

byadisreteone.

Lemma3.2 (Semidisreteanalysis.). The Laplaian disretizedin

𝜃

with

𝑛

sample points:

Δ 𝑛 𝑢 = 𝑛 2 4𝜋 2 sin 2 𝜑

( 𝑢

(

𝜑, 𝑗 + 1 2𝜋𝑛

)

− 2𝑢(𝜑, 𝑗) + 𝑢 (

𝜑, 𝑗 − 1 2𝜋𝑛

))

+ cot 𝜑𝑢 𝜑 + 𝑢 𝜑𝜑

(7)

leads toaShwarz iterationwhose onvergenefatoris

( sin(𝑏) cos(𝑏) + 1

) 2 ∣ 𝑚 ˜ ∣ (

sin(𝑎) cos(𝑎) + 1

) − 2 ∣ 𝑚 ˜ ∣

< 1

every twoiterations, where

˜

𝑚 2 = 𝑛 2

4𝜋 2 (1 − cos(2𝜋𝑘/𝑛))

for the

𝑘

thfrequeny.

Proof. WedoadisreteFouriertransformin

𝜃

andtheequationtosolvebeomes

𝐿 𝑛 𝑢(𝜑, 𝑘) = ˆ 𝑛 2

4𝜋 2 sin 2 𝜑 (cos(2𝜋𝑘/𝑛) − 1)ˆ 𝑢(𝜑, 𝑘) + cot 𝜑ˆ 𝑢 𝜑 (𝜑, 𝑘) + ˆ 𝑢 𝜑𝜑 (𝜑, 𝑘) = 0.

(8)

Bywriting

𝑚

as notedin thestatementof theLemma,wehavereduedthis problemto the previousone

andweanreuse Lemma3.1.

The two ontration onstants are verysimilar. For small valuesof

𝑚

(ignoring

𝑚 = 0

beause that

mode need notonvergeat all),the speedof onvergene is very poor. The overall onvergene fator as

measuredinthe

𝐿 2

normalongtheinterfaesisgivenby

sup

𝑚 ≥ 1

𝐶(𝑚) = 𝐶(1)

= 1 − 𝑂(𝑎 − 𝑏),

andsotheonvergenefatoroftheShwarziterationdeterioratesrapidlyas

𝑎 − 𝑏

beomessmall.

(7)

4 Optimized Shwarz iteration for

Δ

with latitudinal boundaries Inthissetion,weproveTheorems2.1and2.2. Thisrequiresmanytehnialresultsandsoweproeedwith

asequeneofsimplelemmas.

Toanalyze the onvergene fator, as we did in Setion 3, weassume, without lossof generality, that

𝑓 = 0

. Beause theLaplaianof aonstantis zero, we will seethat theerrorterms onvergeto onstant

funtions,whihisto say,theloalsolutionsonvergeupto anadditiveonstant.

Onekeytehniqueisto usethefat that,although

𝑢 0

and

𝑣 0

maybearbitraryin someSobolevspaes,

𝑢 1

and

𝑣 1

areharmoni in

Ω 1

and

Ω 2

, respetively. This keyobservationallowsus to nditerations that onvergefaststarting from the seond step. Thisobservation hasanimpaton thenumerialsimulations.

Often, the rstiterations

𝑢 1

and

𝑣 1

are notmuh better than the initial guesses

𝑢 0

and

𝑣 0

, andthey are

sometimesevenworse.

Lemma 4.1 (Nonloal operator). If, for eah

𝑚

,

𝜓(𝑚) = ˆ ∣ 𝑚 ∣ / sin 𝑎

and

𝜉(𝑚) = ˆ −∣ 𝑚 ∣ / sin 𝑏

,

Δ𝑢 1 = 0

in

Ω 1

and

Δ𝑣 1 = 0

in

Ω 2

, then

𝑢 2 = 0

and

𝑣 2 = 0

.

Proof. Let

𝑢 1 (𝜑, 𝜃) = ˆ 𝑢 1 (𝑎, 0)+ ∑

𝑘 ∕ =0 𝑢 ˆ 1 (𝑎, 𝑘)𝑔 + (𝜑, 𝑘)𝑒 𝑖𝑘𝜃

and

𝑣 1 (𝜑, 𝜃) = ˆ 𝑣 1 (𝑎, 0)+ ∑

𝑘 ∕ =0 𝑣 ˆ 1 (𝑎, 𝑘)𝑔 (𝜑, 𝑘)𝑒 𝑖𝑘𝜃

,

thentheboundaryvalueproblemforthenorthhemisphereyieldsforeah

𝑚 ∕ = 0

:

ˆ

𝑢 2 (𝑎, 𝑚) (

𝜓(𝑚) + ˆ ∣ 𝑚 ∣ sin 𝑎

)

= ˆ 𝑣 1 (𝑎, 𝑚) (

𝜓(𝑚) ˆ − ∣ 𝑚 ∣ sin 𝑎

)

= 0,

(9)

and so

𝑢 ˆ 2 (𝑎, 𝑚) = 0

. Likewise, we nd that

𝑣 ˆ 2 (𝑏, 𝑚) = 0

for eah

𝑚 ∕ = 0

. For

𝑚 = 0

, we have that

ˆ

𝑢 2 (𝑎, 𝑚) = ˆ 𝑣 1 (𝑎, 𝑚)

: the iteration does not aet the onstant mode

𝑚 = 0

. Sine the solution of the

Laplaeproblemisonlydeneduptoaonstant,theiterationhasonverged.

Theseondtangentialderivative

𝐷 2 𝜃

alonglatitude

𝜑

hasFouriertransform

− 𝑚 2

. Theoptimaloperators

are therefore the symmetri and positive denite square root of

− 𝐷 𝜃 2

, multiplied by

1/ sin 𝑎

or

1/ sin 𝑏

aordingto thelatitude. This normalizationonstantisrelatedtothelengthofthelatitudeat

𝜑 = 𝑎

and

𝜑 = 𝑏

,respetively,sinealatitude

𝜑

isairle ofradius

𝑟 = sin 𝜑

.

Next,westatetheonvergenefatoroftheoptimizedShwarziteration,if

𝜓

and

𝜉

arenotthesenonloal

operators.

Lemma4.2. If

Δ𝑢 1 = 0

in

Ω 1

then

ˆ

𝑢 3 (𝑏, 𝑚) = 𝑢 ˆ 1 (𝑏, 𝑚) 𝜉(𝑚) sin ˆ 𝑏 + ∣ 𝑚 ∣ 𝜉(𝑚) sin ˆ 𝑏 − ∣ 𝑚 ∣

𝜓(𝑚) sin ˆ 𝑎 − ∣ 𝑚 ∣ 𝜓(𝑚) sin ˆ 𝑎 + ∣ 𝑚 ∣

𝑔 (𝑚, 𝑎) 𝑔 (𝑚, 𝑏)

𝑔 + (𝑚, 𝑏) 𝑔 + (𝑚, 𝑎) .

Theproof isasimpleomputation usingequation(9). Wenowlookforgoodonvolutionkernels

𝜉

and

𝜓

so thatthetheonvergenefator

𝜅(𝜓, 𝜉, 𝑚, 𝑎, 𝑏) =

𝜅 1

z }| {

𝜉(𝑚) sin𝑏 ˆ + ∣𝑚∣

𝜉(𝑚) sin𝑏 ˆ − ∣𝑚∣

𝜅 2

z }| {

𝜓(𝑚) sin ˆ 𝑎 − ∣𝑚∣

𝜓(𝑚) sin ˆ 𝑎 + ∣𝑚∣

𝜅 3

z }| { 𝑔 − (𝑚, 𝑎) 𝑔 − (𝑚, 𝑏)

𝜅 4

z }| { 𝑔 + (𝑚, 𝑏)

𝑔 + (𝑚, 𝑎) ,

(10)

with

𝑚 ∈ ℤ ∖ { 0 }

,isassmallaspossible. Foragivenhoieof

𝜓 ˆ

and

𝜉 ˆ

,theonvergenefatorasmeasured

inthe

𝐿 2

normalongtheinterfae

𝜑 = 𝑏

will begivenby

sup 𝑚 ∈ℤ∖{ 0 } ∣ 𝜅(𝜓, 𝜉, 𝑚, 𝑎, 𝑏) ∣ .

Theonvergenefatorestimate(10)alsoleadstothefollowingresult:

Corollary 4.3. The iteration (2) is onvergent (modulo the onstant mode)if

𝜓

and

− 𝜉

are onvolution

operatorsthatarepositive denite,regardless ofoverlap.

Proof. Let

𝜓(𝑚) ˆ > 0

and

𝜉(𝑚) ˆ < 0

,andlet

𝜅 1 , 𝜅 2 , 𝜅 3 , 𝜅 4

beasin (10). Sine

sin 𝑎 > 0

and

sin 𝑏 > 0

,wesee

that

∣ 𝜅 1 ∣ < 1

and

∣ 𝜅 2 ∣ < 1

. Furthermore,

𝜅 3 𝜅 4 =

( sin(𝑏) cos(𝑏) + 1

) 2 ∣ 𝑚 ∣ (

sin(𝑎) cos(𝑎) + 1

) − 2 ∣ 𝑚 ∣

=

⎜ ⎜

⎜ ⎝

≤ 1

z }| { ( sin(𝑏)

sin(𝑎) )

≤ 1

z }| { ( cos(𝑎) + 1

cos(𝑏) + 1 )

⎟ ⎟

⎟ ⎠

2 ∣ 𝑚 ∣

≤ 1,

wherewehaveusedthat

𝑏 < 𝑎

. Hene,

∣ 𝜅(𝜓, 𝜉, 𝑚, 𝑎, 𝑏) ∣ < 1

forevery

𝑚 ∕ = 0

.

(8)

By omparison, in [27℄, it was shown that if

𝜉

and

𝜓

are Robin transmission onditionsfor a domain deomposition with relativelyuniform overlap,then the iterationonvergesso long as

𝜉

and

𝜓

are both

positivedenite.

Our analysis willproeed by onsideringthe ontrationonstant (10)forthevarious hoiesof trans-

missiononditions

𝜉

and

𝜓

. WeonsiderrsttheuseofRobintransmissiononditionsand proveTheorem 2.1. Seond,weproveTheorem2.2forseondordertangentialtransmissiononditions.

4.1 Robin or OO0 transmission onditions

Let

𝜓𝑢 = 𝑐𝑢

and

𝜉𝑢 = 𝑑𝑢

,then

𝜅 𝑂𝑂0 (𝜓, 𝜉, 𝑚, 𝑎, 𝑏) = 𝑑 sin 𝑏 + ∣ 𝑚 ∣ 𝑑 sin 𝑏 − ∣ 𝑚 ∣

𝑐 sin 𝑎 − ∣ 𝑚 ∣ 𝑐 sin 𝑎 + ∣ 𝑚 ∣

𝑔 (𝑚, 𝑎) 𝑔 (𝑚, 𝑏)

𝑔 + (𝑚, 𝑏) 𝑔 + (𝑚, 𝑎)

= 𝜅(𝑐, 𝑑, 𝑚, 𝑎, 𝑏).

For

𝑎, 𝑏

xed,wenowneedtoomputethehoiesof

𝑐, 𝑑

minimizing

𝜑(𝑐, 𝑑) = sup

𝑚 𝜅 𝑂𝑂0 (𝑐, 𝑑, 𝑚, 𝑎, 𝑏) 2 ,

where wehaveintrodued asquare to simplify somealulations; theatual ontration onstantwill be

givenby

√ 𝜑(𝑐, 𝑑)

. Tosimplifyournotation,itisonvenienttodoahangeofvariables. Weset

𝑥 = − 𝑑 sin 𝑏

,

𝑦 = 𝑐 sin 𝑎

and

𝐿 = − 2 log

( 1 + cos 𝑎 sin 𝑎

sin 𝑏 1 + cos 𝑏

)

≥ 0,

andrewritingthe

𝑔 ±

termsasexponentials,theoptimizationproblembeomes Minimize

𝜑(𝑥, 𝑦)

where

𝜑(𝑥, 𝑦) = sup

𝑚 ∈{ 1,2,... }

𝐹 (𝑥, 𝑦, 𝑚) 𝐹 (𝑥, 𝑦, 𝑚) =

𝑥 − 𝑚 𝑥 + 𝑚

𝑦 − 𝑚 𝑦 + 𝑚 𝑒 𝐿𝑚

2

(11)

= 𝐹 𝐿 (𝑥, 𝑦, 𝑚).

The onvergene fator every other step for partiular parameter hoies

𝑥

and

𝑦

is

𝜑(𝑥, 𝑦)

. The ase

𝐿 = 0

onlyours when

𝑎 = 𝑏

,in whih ase

𝜑(𝑥, 𝑦) = 1

forall

𝑥, 𝑦 ≥ 1

and thenwewill needtotakeinto

aountertaindisretizationparameterstoobtainaonvergenefatorestimate. Thisasewillbetreated

separately.

WenowusetwodierentanalysestoomputeoptimizedRobinparameters

𝑐

and

𝑑

. First,wemakethe

simplifyingassumption that

𝑥 = 𝑦

,allowingustooptimizeinsteadtheobjetivefuntion

max 𝑚 ∣ (𝑥 − 𝑚)/(𝑥 + 𝑚)𝑒 𝐿 2 𝑚2 .

This situation is interesting beause this an lead to an algorithm where all subdomains do the same

thing. Thisisthemethodwhihmostobviouslygeneralizesto manysubdomains. Seond,weonsiderthe

expression

𝐹 (𝑥, 𝑦, 𝑚)

withbothparameterssimultaneously. Thisisalsointeresting,beauseinthissituation, the iteration does something dierent on eah subdomain. We will see that, although this approah is

diulttogeneralizetomanysubdomains,itleadsto substantiallyimprovedonvergenefators.

4.2 OO0 one-sided analysis,

𝐿 > 0

.

Weanwrite

𝐹 𝐿 (𝑥, 𝑦, 𝑚) = 𝐹 𝐿 (𝑥, 𝑚)𝐹 𝐿 (𝑦, 𝑚)

(f. (11)),wherewedene

𝐹 𝐿 (𝑥, 𝑚) = 𝐹(𝑥, 𝑚) =

𝑥 − 𝑚 𝑥 + 𝑚 𝑒 𝐿 2 𝑚

2

.

(9)

Werstlookattheproblemofomputingtheoptimalparameter

𝑥 0

minimizing

𝜑 𝐿 (𝑥) = sup

𝑚 ∈{ 1,2,... }

𝐹 𝐿 (𝑥, 𝑚),

sine it is simpler. The solution

𝑥 0

minimizing

𝜑 𝐿 (𝑥)

leads to the hoies

𝑥 = 𝑦 = 𝑥 0

, whih minimizes

𝜑 𝐿 (𝑥, 𝑦 )

subjettothesymmetryonstraint

𝑥 = 𝑦

. Theonvergenefatorforapartiularparameterhoie

𝑥

is

𝜑(𝑥, 𝑥) = 𝜑 𝐿 (𝑥)

.

We now deal with the various spetra over whih we an optimize. Our ontinuous problem has a

frequenyvariable

𝑚 ∈ ℤ

,butsinethe

𝑚 = 0

frequenyneednotonverge,andsine

𝐹 𝐿 (𝑥, − 𝑚) = 𝐹 𝐿 (𝑥, 𝑚)

,

wehavealreadysimpliedourproblembyonsideringthefrequenies

𝑚 ∈ { 1, 2, . . . }

. Beauseofoureventual

disretization,wewillwanttoonsiderthefrequenydomain

{ 1, 2, ..., 𝑁 }

,where

𝑁

isthehighestfrequeny

thatan beresolvedontheinterfae,givenourgrid. However,beauseofthediulties inoptimizingover

disretesets, wealso onsider thepossibilityof allowing

𝑚

tovary in theintervals

[1, ∞ )

aswellas

[1, 𝑁]

.

For tehnialreasons,itisalsoonvenienttoinlude

𝑚 = ∞

inouralulations.

Let

2 < 𝑁 < ∞

anddene

𝜑 (𝑐) 𝐿 (𝑥) = sup

𝑚 ∈ [1, ∞ ]

𝐹 𝐿 (𝑥, 𝑚), (𝐼 (𝑐) = [1, ∞ ]), 𝜑 (𝑑) 𝐿 (𝑥) = sup

𝑚 ∈{ 1,2,... }

𝐹 𝐿 (𝑥, 𝑚), (𝐼 (𝑑) = { 1, 2, ..., ∞} ), 𝜑 (𝑐𝑏) 𝐿 (𝑥) = sup

𝑚 ∈ [1,𝑁]

𝐹 𝐿 (𝑥, 𝑚), (𝐼 (𝑐𝑏) = [1, 𝑁]), 𝜑 (𝑑𝑏) 𝐿 (𝑥) = sup

𝑚 ∈{ 1,2,...,𝑁 }

𝐹 𝐿 (𝑥, 𝑚), (𝐼 (𝑑𝑏) = { 1, 2, ..., 𝑁 } ).

Aordingtoourdenitions, forall

𝑥

,

𝜑 (𝑑) 𝐿 (𝑥) = 𝜑 𝐿 (𝑥)

. Theother funtionsaresimilarbut maximizeover

adierentset offrequenies.

One of the main toolsfor solvingmin-max problems is theenvelopetheorem, whih westate herefor

onveniene.

Theorem 4.4 (The EnvelopeTheorem). Let

𝑈

be open in

𝑛

and

𝜑 : 𝑈 → 𝑅

,

𝛾 ∈ ℝ 𝑛

,

∣∣ 𝛾 ∣∣ = 1

. For

𝑥 ∈ 𝑈

, denethe one-sided diretional derivative

𝐷 𝛾 𝜑(𝑥)

by

𝐷 𝛾 𝜑(𝑥) = lim

𝜖 ↓ 0 +

𝜑(𝑥 + 𝜖𝛾) − 𝜑(𝑥)

𝜖 ,

if the(one-sided)limit exists. Let

𝑉

beaompat metrispae. Let

𝐹 : 𝑈 × 𝑉 → ℝ

(𝑥, 𝑦) 7→ 𝐹 (𝑥, 𝑦)

beontinuousandassumethatthepartialgradient

𝐹 𝑥 (𝑥, 𝑦)

existsforall

𝑥, 𝑦 ∈ 𝑈 × 𝑉

andvariesontinuously in

(𝑥, 𝑦)

. Dene

𝜑(𝑥) = max

𝑦 ∈ 𝑉 𝐹(𝑥, 𝑦),

𝑌 (𝑥) = { 𝑦 ∈ 𝑉 ∣ 𝐹 (𝑥, 𝑦) = 𝜑(𝑥) } ∕ = ∅ .

Then,

1. Forall

𝑥 ∈ 𝑈

and

𝛾 ∈ ℝ 𝑛 , ∣∣ 𝛾 ∣∣ = 1

,

𝐷 𝛾 𝑢(𝑥)

existsandisgiven bythe formula

𝐷 𝛾 𝜑(𝑥) = max

𝑦 ∈ 𝑌 (𝑥)

∑ 𝑛

𝑖=1

𝛾 𝑖 𝐹 𝑥 𝑖 (𝑥, 𝑦).

(12)

2. Let

𝑥 ˜ ∈ 𝑈

. If

𝑌 (˜ 𝑥) = { 𝑦(˜ 𝑥) }

isasinglepointthen

𝐷 𝛾 𝜑(˜ 𝑥) =

∑ 𝑛

𝑖=1

𝛾 𝑖 𝐹 𝑥 𝑖 (˜ 𝑥, 𝑦)

(13)

(10)

and

𝜑

is(fully) dierentiableat

𝑥 ˜

. If furthermore

𝑌 (𝑥) = { 𝑦(𝑥) }

isasingleton for all

𝑥 ∈ 𝑂 ⊂ 𝑈

an

open set,then

𝜑

isontinuouslydierentiablein

𝑂

.

This immediatelyleadstoanequiosillationproperty.

Lemma4.5. For

𝑗 ∈ { 𝑐, 𝑑, 𝑐𝑏, 𝑑𝑏 }

, thereisaminimizer

𝑥 𝑗

of

𝜑 (𝑗) (𝑥)

,and

∣ 𝑀 (𝑗) (𝑥 𝑗 ) ∣ ≥ 2

, where

𝑀 (𝑗) (𝑥) = { 𝑚 ∈ 𝐼 (𝑗) ∣ 𝜑 (𝑗) (𝑥) = 𝐹 𝐿 (𝑥, 𝑚) }

.

Proof. Wemustplaeourselvesinthehypothesesoftheenvelopetheorem. Eahset

[1, ∞ ], { 1, 2, ..., ∞} , [1, 𝑁 ], { 1, 2, ..., 𝑁 }

isaompatmetrispae(for

𝐼 (𝑐)

and

𝐼 (𝑑)

onemayusethemetri

𝑑(𝑥, 𝑦) = ∣ 𝑥 1 − 𝑦 1

). Sinethefuntion

𝜑 (𝑗)

is a ontinuous funtion on the ompat metri spae

𝐼 (𝑗)

, it must have a minimum. The optimal

parameter

𝑥 𝑗

isin

[0, ∞ )

sowetake

𝑈 = ( − 1 2 , ∞ )

,andthehypothesesoftheenvelopetheoremaresatised.

Assumethat

𝑀 (𝑗) (𝑥 𝑗 ) = { 𝑚 ˜ }

isasingleton(thereisnoequiosillation). Then,bytheenvelopetheorem, thetwo-sidedderivative

𝐷 𝑥 𝜑 (𝑗) (𝑥 𝑗 )

existsandsine

𝜑 (𝑗) (𝑥 𝑗 )

isminimal,thederivativemustbezero. Using

formula(13),weobtain

0 = 𝐹 𝑥 (𝑥 𝑗 , 𝑚) = ˜ − 4 ˜ 𝑚 𝑚 ˜ − 𝑥 𝑗

( ˜ 𝑚 + 𝑥 𝑗 ) 3 𝑒 2𝐿 𝑚 ˜ .

Sine

𝑥 𝑗 , 𝑚 > ˜ 0

, itmust bethat

𝑚 ˜ = 𝑥 𝑗

. Hene,

𝜑 (𝑗) (𝑥 𝑗 ) = 𝐹 (𝑥 𝑗 , 𝑚) = 0 ˜

. Butwe knowthat

𝜑 (𝑗) (𝑥 𝑗 ) ≥ 𝐹 (𝑥, 𝑥 + 1) > 0

.

Nowweharaterizetheuniqueminimizerforthease

𝑗 = 𝑐

.

Lemma4.6. Thereisaunique

𝑥 𝑐

minimizing

𝜑 (𝑐) 𝐿 (𝑥)

anditisthe uniqueroot

𝜓(𝑥 𝑐 ) = 0

of

𝜓(𝑥) = 𝜓(𝑥, 𝐿) =

( 𝑚(𝑥) − 𝑥 𝑚(𝑥) + 𝑥

) 2

𝑒 𝐿𝑚(𝑥) − ( 1 − 𝑥

1 + 𝑥 ) 2

𝑒 𝐿 ,

where

𝑚(𝑥) =

√ 4𝑥 𝐿 + 𝑥 2

.

Proof. Beause any minimizer

𝑥 𝑐

is in

[1, ∞ )

, the onlyloal maxima of

𝐹 𝐿 (𝑥, 𝑚)

as

𝑚

runs in

[1, ∞ ]

are

at

𝑚 = 1

and

𝑚 = 𝑚(𝑥) = √

4𝑥

𝐿 + 𝑥 2

and so

𝑀 (𝑐) (𝑥 𝑐 ) ⊆ { 1, 𝑚(𝑥 𝑐 ) }

. By Lemma 4.5, we know that

#𝑀 (𝑐) (𝑥 𝑐 ) ≥ 2

so

𝑀 (𝑐) (𝑥 𝑐 ) = {

1,

√ 4𝑥 𝑐

𝐿 + 𝑥 2 𝑐 }

.

Hene,

𝜓(𝑥 𝑐 ) = 0

. Thisshowstheequiosillation

𝜓(𝑥) = 0

.

Wenowshowtheuniquenessofthesolutionoftheequiosillationproblem. Tothatend, let

𝑔(𝑥) = 𝐹 (𝑥, 𝑚(𝑥)),

𝑔 (𝑥) = 𝐹 𝑥 (𝑥, 𝑚(𝑥)) + 𝐹 𝑚 (𝑥, 𝑚(𝑥))𝑚 (𝑥)

= 𝐹 𝑥 (𝑥, 𝑚(𝑥))

= − 4𝑚(𝑥) 𝑚(𝑥) − 𝑥

(𝑚(𝑥) + 𝑥) 3 𝑒 𝐿𝑚(𝑥) < 0,

and

ℎ(𝑥) = 𝐹 (𝑥, 1), ℎ (𝑥) = − 4 1 − 𝑥

(1 + 𝑥) 3 𝑒 𝐿 > 0.

Hene,

𝜓 = 𝑔 − ℎ < 0

and

𝜓

maynothavemultiplezeros.

Thisresultgivessomemoreinformationforthease

𝑗 = 𝑑

(thatis,

𝐼 (𝑑) = { 1, 2, ..., ∞}

).Wemustassume

that

𝐿

isnottoobig,butourfousistheasewherethereislittleoverlapsothisisnotaproblem. Thespeial

point

𝑚(𝑥)

ontinuestoplayanimportantrole. Notiethat

𝑚 (𝑥) > 1

andthat

𝑚(1) = √

4 + 𝐿/ √

𝐿 > 1

so

that, forall

𝑥 > 1

,

𝑚(𝑥) > 𝑚(1) + (𝑥 − 1) > 𝑥

.

(11)

Lemma 4.7. Let

𝐿 < 4/3

and let

𝑥 𝑐

minimize

𝜑 (𝑐) 𝐿 (𝑥)

and

𝑥 𝑑

minimize

𝜑 (𝑑) 𝐿 (𝑥)

. Then

1 ∈ 𝑀 (𝑑) (𝑥 𝑑 ) ⊆ { 1, ⌈ 𝑚(𝑥 𝑑 ) ⌉ , ⌊ 𝑚(𝑥 𝑑 ) ⌋}

and

𝑥 𝑐 − 1 ≤ 𝑥 𝑑 ≤ 𝑥 𝑐

.

Proof. Beause

𝑚 = 𝑚(𝑥 𝑑 )

istheonlyloalmaximumof

𝐹 𝐿 (𝑥, 𝑚)

,weknowthat

𝑀 (𝑑) (𝑥 𝑑 ) ⊆ { 1, ⌈ 𝑚(𝑥 𝑑 ) ⌉ , ⌊ 𝑚(𝑥 𝑑 ) ⌋}

,

and we also knowfrom Lemma 4.5 that

#𝑀 (𝑑) (ˆ 𝑥) ≥ 2

. Assume that

𝑀 (𝑑) (ˆ 𝑥) = {⌈ 𝑚(ˆ 𝑥) ⌉ , ⌊ 𝑚(ˆ 𝑥) ⌋}

. The

hypothesis on

𝐿

ombinedwiththefat that

𝑥 𝑑 ≥ 1

guaranteesthat

𝑚(𝑥 𝑑 ) > 𝑥 𝑑 + 1

so

𝑥 𝑑 < ⌊ 𝑚(𝑥 𝑑 ) ⌋

. We

annowusetheone-sidedderivativeoftheEnvelopeTheorem,obtainingthat

𝐷 (1) 𝜑 (𝑑) (𝑥 𝑑 ) = max

𝑚 ∈{⌈ 𝑚(𝑥 𝑑 ) ⌉ , ⌊ 𝑚(𝑥 𝑑 ) ⌋} 𝐹 𝑥 (𝑥 𝑑 , 𝑚)

= max

𝑚 ∈{⌈ 𝑚(𝑥 𝑑 ) ⌉ , ⌊ 𝑚(𝑥 𝑑 ) ⌋} − 𝑚 𝑚 − 𝑥 𝑑

(𝑚 + 𝑥 𝑑 ) 3 𝑒 𝐿𝑚 < 0.

Therefore,

1 ∈ 𝑀 (𝑑) (ˆ 𝑥)

.

Clearly, wemusthave

𝜑 (𝑑) (𝑥 𝑑 ) ≤ 𝜑 (𝑐) (𝑥 𝑐 )

. If

𝑥 𝑑 > 𝑥 𝑐

then

𝜑 (𝑑) (𝑥 𝑑 ) ≥ 𝐹(𝑥 𝑑 , 1) = 𝜑 (𝑐) (𝑥 𝑑 ) > 𝜑 (𝑐) (𝑥 𝑐 )

,

henewemust have that

𝑥 𝑑 ≤ 𝑥 𝑐

. If

𝑥 𝑑 < 𝑥 𝑐 − 1

, then in viewof the estimate

𝑚 (𝑥) > 1

, wehave that

𝑚(𝑥 𝑑 ) < 𝑚(𝑥 𝑐 ) − 1

. Let

𝑚 0 = ⌊ 𝑚(𝑥 𝑐 ) ⌋

andndthe unique

𝑥 0 ∈ (1, 𝑚 0 )

suhthat

𝑚(𝑥 0 ) = 𝑚 0

. Wehave

that

𝑥 𝑑 ∈ (1, 𝑥 0 ) ⊂ (1, 𝑚 0 )

. For

𝑥 ∈ (1, 𝑚 0 )

,wehavethat

𝐹 𝑥 (𝑥, 𝑚 0 ) < 0

. Hene,

𝜑 (𝑑) (𝑥 𝑑 ) ≥ 𝐹 (𝑥 𝑑 , 𝑚 0 ) >

𝐹 (𝑥 0 , 𝑚 0 ) = 𝜑 (𝑐) (𝑥 0 ) > 𝜑 (𝑐) (𝑥 𝑐 )

.

As aresult of Lemma 4.7, we see that using thedisrete frequenyspetrum

{ 1, 2, ..., ∞}

or using the

ontinuousfrequenyspetrum

[1, ∞ ]

resultsinapproximatelythesameRobinparameter.

We would liketo givea formula for the optimized Robin parameter, but the equiosillation property

annotbesolvedexpliitly. However,wean solveitapproximatelyfor small

𝐿

. Thisfurther allowsus to

onsiderthespetrum

[1, 𝑁 ]

. If

𝑥 𝑐𝑏

minimizes

𝜑 (𝑐𝑏) (𝑥)

then

𝑀 (𝑐𝑏) (𝑥 𝑐𝑏 ) ⊆ { 1, 𝑁, 𝑚(𝑥 𝑐𝑏 ) }

. When

𝐿

isnottoo

smallomparedto

𝑁 1

,wenowshowthat

𝑀 (𝑐𝑏) (𝑥 𝑐𝑏 ) = { 1, 𝑚(𝑥 𝑐𝑏 ) }

. This impliesthat, forthosevaluesof

𝐿

and

𝑁

,theoptimizedparameterforthefrequenyspetrum

[1, 𝑁]

isthesameastheoneforthefrequeny

spetrum

[1, ∞ ]

.

Lemma4.8. The optimizedhoie

𝑥 𝑐

for

𝜑 (𝑐) (𝑥 𝑐 )

isasymptotito

𝐿 1 3

:

𝐿 lim → 0 + 𝑥 𝑐 𝐿 1 3 = 1.

If wehoose

𝑥 ˆ 𝑐 = 𝐿 1 3

asouroptimizedparameter, the onvergene fatorevery otherstepis

𝐹 (ˆ 𝑥 𝑐 , 1) = 1 − 4𝐿 1 3 + 𝑂(𝐿 2 3 ).

(14)

Let

𝛼 > − 3/2

. If

𝐿 = 𝐿(𝑁) > 𝑁 𝛼

then for all

𝑁

suiently large,

𝑚(𝑥 𝑐 ) < 𝑁

, and for suhvalues of

𝑁

and

𝐿

, wehave that

𝑥 𝑐 = 𝑥 𝑐𝑏

.

Proof. Wetrytosolve

𝐹 (𝑥, 𝑚(𝑥)) − 𝐹 (𝑥, 1) = 0

bywriting

𝑥

asapowerof

𝐿

. Let

𝑥 = 𝐶𝐿 𝛽

,thenweget

0 = 𝜓 =

( √

𝐶𝐿 𝛽 1 + 𝐶 2 𝐿 2𝛽 − 𝐶𝐿 𝛽

√ 𝐶𝐿 𝛽 1 + 𝐶 2 𝐿 2𝛽 + 𝐶𝐿 𝛽 ) 2

𝑒 𝐿

4𝐶𝐿 𝛽− 1 +𝐶 2 𝐿 2𝛽

( 1 − 𝐶𝐿 𝛽 1 + 𝐶𝐿 𝛽

) 2

𝑒 𝐿 ,

(15)

whih is dened for

𝐿 > 0

and

𝐶 > 0

. We want to take a series expansion for

𝜓

, and we nd that for

𝛽 = − 1/3

,

𝜓 = ( 4

𝐶 − 4 √ 𝐶

)

𝐿 3 1 + 𝑂(𝐿 2 3 ).

Wesee that

𝐺(𝑥) = 𝐹 (𝑥, 𝑚(𝑥)) − 𝐹(𝑥, 1) < 0

for suientlysmall

𝐿

ifwehoose

𝑥 = 𝐶𝐿 1 3

with

𝐶 > 1

,

and

𝐺(𝑥) > 0

forallsuientlysmall

𝐿

if

𝐶 < 1

. Hene,

𝑥 𝑐 𝐿 1 3 → 1

as

𝐿 → 0 +

.

Toshowthat

𝑥 𝑐 = 𝑥 𝑐𝑏

,itsuestoshowthat

𝑚(𝑥 𝑐 ) < 𝑁,

(16)

forsuientlysmall

𝐿

. Wesubstitute

𝑥 = 𝐿 1 3

andthen

𝐿 = 𝑁 𝛼

into

𝑚(𝑥)

obtaining

𝑚(𝑥) =

4𝑁 4 3 𝛼 + 𝑁 2 3 𝛼 .

Hene if

𝛼 > − 3/2

,

𝑚(𝑥)/𝑁 → 0

as

𝑁 → ∞

and so

𝑚(𝑥) < 𝑁

for large

𝑁

. Conversely, if

𝛼 < − 3/2

,

𝑚(𝑥)/𝑁 → ∞

and

𝑚(𝑥) > 𝑁

forlarge

𝑁

.

(12)

Thease

{ 1, 2, ..., 𝑁 }

followsimmediately.

Corollary 4.9. Let

𝛼 > − 2 3

. Let

𝐿 = 𝐿(𝑁 ) > 𝑂(𝑁 𝛼 )

,

𝐿 < 4 3

and let

𝑥 𝑐𝑏

minimize

𝜑 (𝑐𝑏) (𝑥)

. If

𝑥 𝑑𝑏

minimizes

𝜑 (𝑑𝑏) (𝑥)

thenfor all

𝑁

suiently large,

1 ∈ 𝑀 (𝑥 𝑑𝑏 ) ⊂ { 1, ⌊ 𝑚(𝑥 𝑐 ) ⌋ , ⌈ 𝑚(𝑥 𝑐 ) ⌉}

.

Proof. Say that

𝑥 𝑑

minimizes

𝜑 (𝑑) (𝑥)

. By lemma 4.7, we have that

𝑀 (𝑑) (𝑥 𝑑 ) ⊆ { 1, ⌈ 𝑚(𝑥 𝑐 ) ⌉ , ⌊ 𝑚(𝑥 𝑐 ) ⌋}

,

where

𝑥 𝑐

minimizes

𝜑 (𝑐) (𝑥)

. Lemma 4.8 gives

⌈ 𝑚(𝑥 𝑐 ) ⌉ < 𝑁

for large

𝑁

, hene

𝑀 (𝑑) (𝑥 𝑑 ) ⊂ { 1, ..., 𝑁 }

. If

𝑥 𝑑𝑏 > 𝑥 𝑑

then

𝐹 (𝑥 𝑑𝑏 , 1) > 𝐹 (𝑥 𝑑 , 1)

and

𝑥 𝑑𝑏

is not a minimizer. If

𝑥 𝑑𝑏 < 𝑥 𝑑

and if

1 ∕ = 𝑚 𝑑 ∈ 𝑀 (𝑑) (𝑥 𝑑 )

then

𝜑 (𝑑𝑏) (𝑥 𝑑𝑏 ) ≥ 𝐹(𝑥 𝑑𝑏 , 𝑚 𝑑 ) ≥ 𝐹 (𝑥 𝑑 , 𝑚 𝑑 ) = 𝜑 (𝑑) (𝑥 𝑑 )

and

𝑥 𝑑𝑏

is nota minimum. Hene

𝑥 𝑑 = 𝑥 𝑑𝑏

for all

suientlylarge

𝑁

.

Notethat theasymptotionvergenefator(14)isapproximatelyindependentofthespetrumweuse,

andsowemayusetheoptimizedparameter

𝑥 ˆ 𝑐 = 𝐿 1 3

in allsituations.

This ompletestheproofoftherstpartofTheorem2.1.

4.3 Two-parameter OO0 analysis,

𝐿 > 0

.

Returningtoformula(11),wenowomputetheoptimalparameters

𝑥

and

𝑦

withoutassumingthat

𝑥 = 𝑦

.

Calulationsinthissettingaremoreompliatedthanintheone-parameterase,so wetreatonlytheases

of thespetrum

[1, ∞ ]

and

[1, 𝑁]

. We an get arudimentary equiosillationpropertyout of theenvelope theorem.

Lemma4.10. Let

𝐿 > 0

and

𝐹 (𝑥, 𝑦, 𝑚) =

( 𝑥 − 𝑚 𝑥 + 𝑚

𝑦 − 𝑚 𝑦 + 𝑚 𝑒 𝐿𝑚

) 2

, 𝜑(𝑥, 𝑦) = sup

𝑚 ∈ [1, ∞ ]

𝐹 (𝑥, 𝑦, 𝑚),

𝑀 (𝑥, 𝑦) = { 𝑚 ∈ [1, ∞ ] ∣ 𝐹 (𝑥, 𝑦, 𝑚) = 𝜑 (𝑗) (𝑥, 𝑦) } .

Then

𝜑(𝑥, 𝑦)

hasa minimum

(𝑥 0 , 𝑦 0 )

. Inaddition,

#𝑀 (𝑗) (𝑥 0 , 𝑦 0 ) ≥ 2

and

min(𝑥 0 , 𝑦 0 ) ≥ 1

.

Proof. Note that

𝜑(1, 1) < 1

. Sine

𝜑(𝑥, 𝑦) ≥ 𝐹 (𝑥, 𝑦, 1)

and sine

𝐹 (𝑥, 𝑦, 1)

tends to

1

as

(𝑥, 𝑦)

tends to

, thereis an

𝑀 < ∞

suh that, if

(𝑥, 𝑦) ∈ / [1, 𝑀 ] × [1, 𝑀 ]

, then

𝜑(𝑥, 𝑦) > 𝜑(1, 1)

. Hene,

𝜑

must havea

minimumin

[1, 𝑀 ] × [1, 𝑀 ]

.

We now plae ourselves under the hypotheses of the Envelope Theorem. We have that

(𝑥 0 , 𝑦 0 ) ∈ ( 1 2 , ∞ ) 2 = 𝑈

andthehoie

𝑉 = [1, ∞ ]

doesthetrik.

If

𝑀 (𝑗) (𝑥 0 , 𝑦 0 ) = { 𝑚 }

isasingleton,then

𝜑(𝑥, 𝑦)

isdierentiableat

𝑥 0 , 𝑦 0

and

0 = ∇ 𝜑(𝑥 0 , 𝑦 0 )

= ( ∂𝐹

∂𝑥 (𝑥 0 , 𝑦 0 , 𝑚), ∂𝐹

∂𝑦 (𝑥, 𝑦, 𝑚) )

.

Itsuestolook atthepartialin

𝑥

,

0 = − 4𝑚 𝑚 − 𝑥 (𝑚 + 𝑥) 3

( 𝑚 − 𝑦 𝑚 + 𝑦

) 2

𝑒 2𝐿𝑚 ,

toonludethateither

𝑚 = 𝑥

or

𝑚 = 𝑦

. Ineitherase,

𝜑(𝑥, 𝑦) = 𝐹 (𝑥, 𝑦, 𝑚) = 0

,whih isabsurd.

In theone-sided ase,this result was suient beause theonly remainingpossibilitywas tohavetwo

pointsequiosillating. However,in thetwo-sidedase,weanequiosillatetwoorthree points,but lemma

4.10only saysthat there are at least twopointsequiosillating. Using thefat that

𝜑(𝑥, 𝑦)

hasone-sided

derivativeseverywhere,andthatforanydiretion

𝛾

thisone-sidedderivative

𝐷 𝛾 𝜑(𝑥, 𝑦)

mustbenon-negative at

(𝑥, 𝑦 )

wean obtainequiosillationof threepoints. First wemust desribetheshapeof

𝐹 (𝑥, 𝑦, 𝑚)

asa

funtionof

𝑚

.

(13)

Lemma 4.11. Let

1 ≤ 𝑥 ≤ 𝑦

(without loss of generality). The two loal maxima

𝑚 1 (𝑥, 𝑦) ≤ 𝑚 2 (𝑥, 𝑦)

of

𝐹 (𝑥, 𝑦, 𝑚)

inside

𝑚 > 1

aregiven by

𝑚 2 1 (𝑥, 𝑦) = 2𝑥 + 2𝑦 + 𝐿𝑥 2 + 𝐿𝑦 2 − √ 𝑝 𝑥,𝑦 (𝐿)

2𝐿 ,

𝑚 2 2 (𝑥, 𝑦) = 2𝑥 + 2𝑦 + 𝐿𝑥 2 + 𝐿𝑦 2 + √ 𝑝 𝑥,𝑦 (𝐿)

2𝐿 ,

𝑝 𝑥,𝑦 (𝐿) = (𝑥 2 − 𝑦 2 ) 2 𝐿 2 + 4(𝑥 3 − 𝑥𝑦 2 − 𝑥 2 𝑦 + 𝑦 3 )𝐿 +4(𝑥 + 𝑦) 2 > 0.

Furthermore, if

𝑥 < 𝑦

then

𝑚 1 (𝑥, 𝑦) ∈ (𝑥, 𝑦)

and

𝑚 2 ∈ (𝑦, ∞ )

. If

𝑥 = 𝑦

, then

𝑚 1 (𝑥, 𝑥) = 𝑥

and

𝑚 2 (𝑥, 𝑥) ∈ (𝑥, ∞ )

.

Theritialpoints

𝑚 1

and

𝑚 2

alsodependon

𝐿

. Whenthatdependenemustbemadeexpliit,wewrite

𝑚 1 (𝑥, 𝑦, 𝐿)

and

𝑚 2 (𝑥, 𝑦, 𝐿)

.

Proof. Wesetthederivative

∂𝐹/∂𝑚

tozero,obtaining

0 = ∂𝐹

∂𝑚 (𝑥, 𝑦, 𝑚)

= 2 𝑥 − 𝑚 (𝑥 + 𝑚) 3

𝑦 − 𝑚

(𝑦 + 𝑚) 3 𝑞 𝑥,𝑦 (𝑚)𝑒 2𝐿𝑚 , 𝑞 𝑥,𝑦 (𝑚) = − 𝐿𝑚 4 + (2𝑥 + 2𝑦 + 𝐿𝑥 2 + 𝐿𝑦 2 )𝑚 2

− (2𝑦 2 𝑥 − 2𝑥 2 𝑦 − 𝐿𝑥 2 𝑦 2 ).

The hoies

𝑚 = 𝑥

and

𝑚 = 𝑦

lead to

𝐹 (𝑥, 𝑦, 𝑚) = 0

so they are minima, so we restrit our attention

to

𝑞 𝑥,𝑦 (𝑚) = 0

. Bysubstituting

𝑚 2 = 𝑛

weobtaina quadratipolynomialin

𝑛

whose roots are

𝑚 2 1 (𝑥, 𝑦)

and

𝑚 2 2 (𝑥, 𝑦)

; we need to show that

1 ≤ 𝑚 1 ≤ 𝑚 2

. First to show that they are real, we show that

the disriminant

𝑝 𝑥,𝑦 (𝐿)

is non-negative. It sues to show that, as a quadrati polynomial over

𝐿

, its

oeientsarenon-negative. Theoeientof

𝐿 2

andtheonstantoeientaresquares andhenenon-

negative. Werewrite

𝑥 = 𝜆𝑦

with

𝜆 ∈ (0, 1]

andsubstituteintotheoeientof

𝐿

in

𝑝 𝑥,𝑦 (𝐿)

to obtainthe

expression

𝑦 3 (4𝜆 3 − 4𝜆 2 − 4𝜆 + 4) > 0

sine

𝑦 > 0

. Hene

𝑝 𝑥,𝑦 (𝐿) > 0

and

𝑚 1 < 𝑚 2

.

Assume that

𝑥 < 𝑦

are xed. Sine

𝐹(𝑥, 𝑦, 𝑥) = 0

and

𝐹(𝑥, 𝑦, 𝑦) = 0

but

𝐹 (𝑥, 𝑦, 𝑚) > 0

for any

𝑚 ∈ (𝑥, 𝑦)

, there is a loal maximum in

𝑚

in the interval

(𝑥, 𝑦)

. Moreover, sine

𝐹 (𝑥, 𝑦, 𝑦) = 0

and

lim 𝑚 →∞ 𝐹 (𝑥, 𝑦, 𝑚) = 0

, there must be another loal maximum in theinterval

(𝑦, ∞ )

. However,wehave

already shown that the onlyandidates for loal maxima are

𝑚 1

and

𝑚 2

, and sine

𝑚 1 ≤ 𝑚 2

it must be

that

𝑚 1 ∈ (𝑥, 𝑦)

and

𝑚 2 ∈ (𝑦, ∞ )

.

Lemma4.12. Let

(𝑥 0 , 𝑦 0 )

minimize

𝜑(𝑥, 𝑦)

. Then

𝑀 (𝑥, 𝑦) = { 1, 𝑚 1 (𝑥, 𝑦), 𝑚 2 (𝑥, 𝑦) }

with

𝑥 ∕ = 𝑦

.

Proof. Thederivativeof

𝜑

inthediretion

𝛾 = (𝛼, 𝛽)

isgivenby

𝐷 𝛾 𝜑(𝑥, 𝑦) = max

𝑚 ∈ 𝑀(𝑥,𝑦) 𝛼 ∂𝐹

∂𝑥 (𝑥, 𝑦) + 𝛽 ∂𝐹

∂𝑦 (𝑥, 𝑦)

= max

𝑚 ∈ 𝑀(𝑥,𝑦) (𝛼, 𝛽) ⋅ 𝐺(𝑚),

where

𝐺(𝑚) = 4𝑚𝑒 2𝐿𝑚 𝑥 − 𝑚 (𝑥 + 𝑚) 2

𝑦 − 𝑚 (𝑦 + 𝑚) 2

( 𝑦 − 𝑚 𝑥 + 𝑚 , 𝑥 − 𝑚

𝑦 + 𝑚 )

.

Inthease

𝑥 = 𝑦

,

𝑚 1

isatuallyaminimumso

𝑀 (𝑥, 𝑥) = { 1, 𝑚 2 (𝑥, 𝑥) }

andinpartiular

𝑥 > 1

. Thevetors

( 𝑥 − 1

𝑥 + 1 , 𝑥 − 1 𝑥 + 1

)

and

( 𝑥 − 𝑚 2

𝑥 + 𝑚 2 , 𝑥 − 𝑚 2

𝑥 + 𝑚 2

)

Références

Documents relatifs

Patients - Les patients admissibles à l’étude HOPE devaient être âgés de 55 ans et plus, avoir des antécédents de maladie cardiovasculaire (coronaropathie, accident

Whereas, the split coarse space solves take 104 iterations for the first (one-level) solve plus 14 iterations for building the coarse operator plus 4 × 25 iterations for the

In order to achieve this goal, we use algebraic methods developed in constructive algebra, D-modules (differential modules) theory, and symbolic computation such as the so-called

We guess that for an elliptic operator there are no interface conditions such that the additive Schwarz method converges in a finite number of steps in the case of a general

We derive an implicit solution on each subdomain from the optimized Schwarz method for the mesh BVP, and then introduce an interface iteration from the Robin transmission

Here, the case of the linear elasticity problem approximated by the finite element method in combination with non-overlapping domain decomposition method such as the finite

JJs, the weak link needs to be embedded into the system during the fabrication process and can hardly be tuned afterwards. The versatility and performance of Josephson devices

The total environmental impact of each process can be characterized by four main categories of impacts: Climate Change Human Health (CCHH), Human Toxicity (HT),