RESEARCH OUTPUTS / RÉSULTATS DE RECHERCHE
Author(s) - Auteur(s) :
Publication date - Date de publication :
Permanent link - Permalien :
Rights / License - Licence de droit d’auteur :
Bibliothèque Universitaire Moretus Plantin
Institutional Repository - Research Portal
Dépôt Institutionnel - Portail de la Recherche
researchportal.unamur.be
University of Namur
Ginzburg-Landau approximation for self-sustained oscillators weakly coupled on
complex directed graphs
Di Patti, Francesca ; Fanelli, Duccio; Miele, Filippo; Carletti, Timoteo
Published in:
Communication in Nonlinear Science and Numerical Simulation
DOI:
10.1016/j.cnsns.2017.08.012
Publication date:
2017
Document Version
Publisher's PDF, also known as Version of record
Link to publication
Citation for pulished version (HARVARD):
Di Patti, F, Fanelli, D, Miele, F & Carletti, T 2017, 'Ginzburg-Landau approximation for self-sustained oscillators
weakly coupled on complex directed graphs', Communication in Nonlinear Science and Numerical Simulation,
vol. 56, pp. 447-456. https://doi.org/10.1016/j.cnsns.2017.08.012
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal ? Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Commun Nonlinear Sci Numer Simulat 56 (2018) 447–456
ContentslistsavailableatScienceDirect
Commun
Nonlinear
Sci
Numer
Simulat
journalhomepage:www.elsevier.com/locate/cnsns
Research
paper
Ginzburg-Landau
approximation
for
self-sustained
oscillators
weakly
coupled
on
complex
directed
graphs
Francesca
Di
Patti
a,b,c,∗,
Duccio
Fanelli
a,b,c,
Filippo
Miele
d,
Timoteo
Carletti
e a Dipartimento di Fisica e Astronomia, Università degli Studi di Firenze, via G. Sansone 1, 50019 Sesto Fiorentino, Italia b INFN Sezione di Firenze, via G. Sansone 1, 50019 Sesto Fiorentino, Italiac CSDC, via G. Sansone 1, 50019 Sesto Fiorentino, Italia
d Spanish National Research Council (IDAEA-CSIC), E-08034 Barcelona, Spain
e Department of Mathematics and Namur Center for Complex Systems - naXys, University of Namur, rempart de la Vierge 8, B 50 0 0 Namur, Belgium
a
r
t
i
c
l
e
i
n
f
o
Article history:
Received 7 February 2017 Accepted 15 August 2017 Available online 16 August 2017 Keywords:
Reaction-diffusion model Complex Ginzburg–Landau equation Pattern formation
Synchronization
a
b
s
t
r
a
c
t
Anormalformapproximationfortheevolutionofareaction-diffusionsystemhostedona directedgraphisderived,inthevicinityofasupercriticalHopfbifurcation.Weakdiffusive couplingsareassumedtoholdbetweenadjacentnodes.Underthisworkingassumption, aComplexGinzburg–Landauequation(CGLE) isobtained, whose coefficientsdepend on theparametersofthemodelandthetopologicalcharacteristicsoftheunderlyingnetwork. TheCGLEenablesonetoprobethestabilityofthesynchronousoscillatingsolution,as dis-playedbythereaction-diffusionsystemabove Hopfbifurcation.Morespecifically, condi-tionscanbeworkedoutfortheonsetofthesymmetrybreakinginstabilitythateventually destroystheuniformoscillatorystate.NumericaltestsperformedfortheBrusselatormodel confirmthevalidityoftheproposedtheoreticalscheme.PatternsrecordedfortheCGLE re-semblecloselythoserecovereduponintegrationoftheoriginalBrussellatordynamics.
© 2017ElsevierB.V.Allrightsreserved.
1. Introduction
Manyreal-life phenomenacanbe ultimatelydescribedintermsofmutuallyinteractingentitieswhichcan occasionally giverise tocollectivebehaviors[1,2].The emergingpatternsmayplay importantfunctionalroles,asit isforinstancethe caseforbiochemicalprocesses[3]andecologicalapplications[4–6].Synchronyisamongthemoststrikingexampleof self-organizeddynamics[7].Itisencounteredinawidegalleryofnaturalsystems,thinktothebeatingoftheheart[8]andthe firingofthefirefly[9].Italsoplaysaroleofparamountimportanceforthecorrectfunctioningofman-designedtechnology, ase.g.inpowergrids[10,11].
Synchronizationofself-sustained oscillations isa particularlyrich field ofinvestigation. Mathematically, self-sustained oscillationscorrespond tostablelimitcycles inthestate spaceofanautonomouscontinuous-time dynamicalsystem. Os-cillatorscanbeembeddedincontinuum space,beingsubjecttodiffusivecouplings.Theensuingreaction-diffusionsystem displayssynchronousoscillations, which,underspecific conditions,prove robustto external perturbations. Eachoscillator canalternativelyoccupya node ofacomplexnetwork [12–15],a generalization thatopensup theperspectiveto tacklea largeplethoraofproblemsthatdealwithadiscreteandheterogeneoushostingsupport[15–17].
∗ Corresponding author at: Dipartimento di Fisica e Astronomia, Università degli Studi di Firenze, via G. Sansone 1, 50019 Sesto Fiorentino, Italia. E-mail address: [email protected] (F. Di Patti).
http://dx.doi.org/10.1016/j.cnsns.2017.08.012 1007-5704/© 2017 Elsevier B.V. All rights reserved.
Forcontinuous reaction-diffusionsystems nearthe supercritical Hopf bifurcation, one can obtain asimplified, normal formdescription ofthe dynamicsinterms ofan associatedGinzburg–Landau equation (CGLE) [7,18,19].Thisis a reduced picturewhichprovidesanaccuraterepresentationoftheoriginalmodel,whileallowingforanalyticalprogresstobemade. Fromamoregeneralperspective,itisalsoimportanttoclassifyminimal,thougheffectivedescriptiveframeworksthatcould guideinthesearch foruniversaltraitsthat happentobe sharedbymultispeciesmodels,besides theirintrinsicdegreeof inherentspecificity. Simplifiedschemes thatexemplifyan exactunderlyingdynamicscan beeffectivelyemployed toshed lightontheonsetofspatio-temporalchaosandpropagationofnonlinearwaves.Externallyimposednonhomogeneous per-turbationcan, forinstance,breakthe synchronyofthe oscillationsasdisplayedby the originalreaction-diffusion system, orequivalentlyits CGLEanalogue,somaterializinginacolorful densitypatternsthat sustainspatio-temporal propagation. The formal link betweencontinuousreaction diffusion-systemsandthe CGLE wasestablished inthe pioneeringwork by Kuramoto [18],exploitinga multiple-scale perturbative analysis.More recently theanalysis hasbeen extended by Nakao
[20] totherelevantsettingwherethereaction-diffusionsystemismadetoevolveonasymmetric,henceundirected, net-work. As remarked in [21], directionality mattersand can seed the emergence of non trivial collective dynamics which cannotmanifestwhenthescrutinizedsystemismadetoevolveonasymmetricdiscretesupport.Motivatedbythisfinding, weconsideredin[22]thedynamicsofareaction-diffusionsystemdefinedonadirectedgraphwhichdisplaysastablefixed pointandobtainedaneffectivedescriptionfortheevolutionmodetriggeredunstable,justabovethethresholdofcriticality. Theanalysisexploitsamultipletime-scaleanalysisandeventuallyyieldsaStuart–Landaufortheamplitudeoftheunstable mode,whosecomplexcoefficientsreflectthetopologyofthenetwork,thefactualdrivetotheinstability.
Inthispaperwe aimatfollowingthesimilar strategyof[22]toderive anapproximateequation fortheevolutionofa reactiondiffusionsystemonadirected(andbalanced)graph,inthevicinityofasupercriticalHopfbifurcation.Asamatter of fact we willassume weakthe strength of thecoupling that linksadjacent nodes.In doing so,we will generalize the workofNakao[20]totheinterestingsettingwhereasymmetryinthecouplingsneedstobeaccommodatedforand,atthe sametime, reformulate theclassical workof Kuramoto[18],ona discrete spatialbacking.To anticipateourfindings, and atvariancewiththeanalysisreportedin[22],we willfinallyobtain aCGLEasaminimal descriptionforthedynamicsof theself-sustainedoscillators coupledona complexandasymmetricgraph.TheobtainedCGLE enablesone toanalytically probethestabilityofthesynchronousuniformstate,asdisplayedbythereaction-diffusionsystem.Specifically,itallowsto determinetheparameters settingthatinstigatesa symmetrybreakingtransitiontonon-uniformpatterns.Numericaltests madefortheBrusselatormodel,hereassumedasareferencemodelforitspedagogicalinterests,willconfirmtheadequacy oftheproposedapproximatescheme.
2. Diffusiveoscillatorsonnetworks
We will here consider a generic two dimensional reaction-diffusion system and label with xj
(
t)
=(
φ
j,ψ
j)
T for j=1,...,N the two-dimensional real vector of the concentrations.The index j refers to the node of the network to which the selected componentsrefer to. N standsfor thesize of the network,i.e. thetotal number ofnodes. The onlyfurther assumptionthatweshallmakeistheexistenceofself-sustainedoscillationsforthesystemunderscrutinyandforthis rea-son wewillpoint toxj(t) asto theoscillators’variables.The dynamicsofthesystemishencedescribedbythefollowing differentialequation ˙ x j= F
(
x j,μ
)
+D N k=1jkx k (1)
wherethetwo-dimensionalnonlinearfunctionFspecifiesthereactionterms:itdependsonthelocalconcentrationsofthe speciesxjandon
μ
whichisavectorofarbitrarydimensionthatgathertogethertheparametersofthemodel.Thesecondtermrepresentsthediffusivecoupling:D=Kdiag
(
Dφ,Dψ)
denotesthediagonalmatrixofthediffusioncoefficients. Kisa constantparameter thatsetthestrengthofthecoupling.Aswewillmakeclearinthefollowingtheperturbative analysis thatweshalldevelop,holdforK<<1.InEq.(1),istheLaplacianmatrixwhoseelementsread
i j=Ai j−
δ
i jki.Herewefocusondirectednetworks, thustheadjacencymatrixAisnotsymmetric.Usingstandardnotations,Ai j=1ifalinkexists
that goesfromnodei tonodej.Otherwise,Ai j=0.ki isthenumberofoutgoingedges fromnode i,
δ
ijistheKronecker’sdelta.Weassumethatsystem(1)admitsahomogeneousequilibriumpointthatweherelabelx∗=
(
φ
∗,ψ
∗)
T.Thisrequestimpliesdealingwithabalancednetwork,namelyanetworkwheretheoutgoingandincomingconnectivitiesareequal. Weadditionallyrequirethatx∗undergoesaHopfbifurcationfor
μ
=μ
0.Accordingly,theJacobianmatrixassociatedto system(1)hasapairofimaginaryeigenvalues ± iω
0.Slightlyabovethesupercritical Hopfbifurcation,x∗becomesunsta-ble,thereaction-diffusionsystemadmitsan timedependent homogeneoussolution.Thisistheuniformstateobtainedby replicatingoneachnodeofthenetworkandincompletesynchrony,thelimitcycledisplayedbythesysteminitsa-spatial limit(K=0).Thespatialcoupling,sensitivetotinynonhomogeneities,whichconfigureasinjectedperturbation,can even-tually destabilized theuniform synchronous equilibrium. When diffusion is small(K=
2<<1), the method ofmultiple
timescales[18]constitutesaviablestrategytocharacterizethenonlinearevolutionoftheperturbationandhenceelaborate onthestabilityofthetime-dependentuniformperiodicsolution.
To this end, inspired by the analysis carried out in [20], we introduce small inhomogeneous perturbations,
δφ
j andδψ
j, to the uniform equilibriumpoint, namely(
φ
j,ψ
j)
=(
φ
∗,ψ
∗)
+(
δφ
j,δψ
j)
for j=1,...,N.We then substitute thisF. Di Patti et al. / Commun Nonlinear Sci Numer Simulat 56 (2018) 447–456 449
ansatzintoEq.(1).PerformingaTaylorexpansionoftheresultingsystemandpacking
δφ
jandδψ
jintothecolumnvectoruj=
δφ
j,δψ
jT,oneendsupwiththefollowingequationforthetimeevolutionofuj:∂
dtu j= Lu j+D N k=1jku k+Mu ju j+Nu ju ju j+... (2)
whereListheJacobianmatrixevaluatedatthesteadystatex∗.Adoptingthesamesymbolicnotationsof[18],Mujujand
Nujujujdenotetwo-dimensionalvectorswhosecomponentsrespectivelyread
Mu ju j l= 1 2! 2 m,n=1∂
2F l(
x ∗,μ
)
∂
xn∂
xm(
u j)
m(
u j)
n Nu ju ju l l= 1 3! 2 m,n,p=1∂
3F l(
x ∗,μ
)
∂
xn∂
xm∂
xp(
u j)
m(
u j)
n(
u j)
p forl=1,2.SettingthemodelabovethesupercriticalHopfbifurcation,translatesintoredefining
μ
bymeansofasmallparameteras
μ
=μ
0+2
μ
1,whereμ
1 isorderone.Tofollowthenonlinearevolutionoftheamplitudeofthemostunstablemode,itisappropriateto introduceaslowtimevariable
τ
=2t and,accordingly,torewritethetotalderivative withrespectto
theoriginaltimet: d dt −→
∂
∂
t +2
∂
∂τ
. (3)Asa key assumption,alreadyalluded tointhe precedingdiscussion,we setK=
2:inother words we supposethat the
impactofdiffusionisofthesameorderasthedeviationfromthebifurcationpoint.Moreover,nearcriticality,thematrixL
andtheoperatorsMandN maybeexpandedinpowersof
2 L = L 0+
2L 1+...
M= M0+
2M1+... N= N0+
2N1+...
(4)
Wefurtherassumethatucanbeexpressedasaseries,functionofbothtand
τ
:u j
(
t)
=u (j1)
(
t,τ
)
+2u (2)
j
(
t,τ
)
+... (5)InsertingEqs.(3)–(5)intoEq.(2)weget:
∂
∂
tI2+2
∂
∂τ
I2− L0−2L 1− ...
(
u (j1)+
2u (2) j +...
)
−2D N k=1
jk
(
u k(1)+
2u (k2)+...
)
=2M0u (j1)u ( 1) j +
3
(
2M 0u (j1)u ( 2) j +N0u (j1)u ( 1) j u ( 1) j)
+O(
4
)
Collectingtogethertermsofthesameorderin
returnsthefollowingsetofequations
∂
∂
tI2− L0u (ν)j = B (ν)j (6)
for
ν
=1,2,3....TheexistenceofanontrivialsolutionfortheaforementionedlinearsystemsisguaranteedbytheFredholm theorem[23].Asweshalldiscussinamoment,thesolvabilitycondition(seeAppendixA)isdirectlysatisfiedforν
=1andν
=2,whileitmustbeexplicitlyimposedforν
=3.Letusstart byfocusingon
ν
=1.Thecorrespondingrighthandsideofsystem(6)isB(j1)=0.Itiseasytoseethatthe analyticalsolutionisu (j1)
(
t,τ
)
=Wj(
τ
)
U 0eiω0t+c.c. (7)whereU0 therighteigenvectorofLcorresponding totheeigenvaluei
ω
0.The complexvariableWj(τ
)istheamplitudeoftheperturbation.Itsdynamicsissupposed tobe slow,i.e.ruledbytherescaled time
τ
.As weshallsee,by imposingthe solvabilityconditionforν
=3willreturnanequationforthetimeevolutionofWj(τ
).Movingto
ν
=2,wefindB(j2)=M0u(j1)u(1)
j andwetrytocastthesolutionofthelinearsystemintheform
u (j2)=W2
jV 2e2iω0t+
γ
u (j1)+|
Wj|
2V 0.Inserting this ansatz into (6) and grouping together terms that do not depend on t, one finds V0=−2L−10 M0U0U¯0
where the bar stands for the complex conjugate. In the same way, collecting terms proportional to e2iω0t yields V
2=
(
2iω
0I2− L0)
−1M0U0U0.The third constant term appearing in Eq. (6) is explicitly given by B(j3)=
L1−ddτI2 u(j1) +DkN=1jku(k1)+ 2M0u(j1)u( 2) j +N0u(j1)u( 1) j u( 1)
j .Byimposingthesolvabilitycondition,oneeventuallyobtainsthefollowingCGLequationfor
Wj(
τ
) d dτ
Wj(
τ
)
=σ
Wj− g|
Wj|
2W j+d N k=1jkWk (8) where
σ
=σ
Re+iσ
Im=(
U∗0)
†L1U0, g=gRe+igIm= −(
U∗0)
† 2M0V2U¯0 +2M0V0U0+3N0U0U0U¯0and d=dRe+idIm=
(
U∗0)
†DU0 arecomplexnumbers.
Usingthefollowingscaletransformation
τ
σ
Re →τ
dReσ
Rei j→
i j gRe
σ
Re Wj→Wj,introducingthreenewcoefficients c0=
σ
Imσ
Re c1= dIm dRe c2= gIm gReandmakingthefurthertransformationWj→Wjexp(ico
τ
),wefindthefollowingreducedCGLEd d
τ
Wj=Wj−(
1+ ic2)
|
Wj|
2W j+(
1+ic1)
kjkWk (9)
forthecomplexamplitudeWjofthej-thoscillator.
Eq.(9)displaysauniformlysynchronizedsolutiongivenby
Wj
(
τ
)
≡ WLC=e−ic2τ (10)for j=1,...,N.Inthisrespectitprovidesalocalapproximationforthedynamicsoftheoriginalreaction-diffusionsystem neartheHopfbifurcation,namelywhenself-organizedoscillationsrise.ThedynamicsoftheensemblemadeofNoscillators issensitivetothecouplingimposed,thislatterbeingencodedintheLaplacianoperator.Mutualinterferencesamong oscil-lators,asmediatedbytheweakdiffusivecouplinghereatplay,candisruptthecoherencyofthesynchronousstate,making thesystemtoevolvetowardsa differentattractor.Thestabilityofthesynchronizeddynamicscanbe assessedviaalinear stabilityanalysis whichis carriedout forCGLE, andwhoseconclusion can be readily translatedto theoriginal reaction-diffusion framework. Havingobtaineda universal description ofthe dynamics ofthe systemin termsof anormal mode representationallowsonetoobtaingeneralconditionsoftheonsetoftheinstability,whichprescindthespecificcontextof analysis.
Toinvestigatethe linearstabilityofsolution(10),we briefly review themethod describedin[20,24]andpoint tothe specificaspectsthatrelatetothedirectednatureofthecouplings,asdiscussedin[25].Noticethatinthislatterpaper,the CGLEisassumedasthereferencemodel,whilehereitisobtainedasalocalapproximationofagenericreaction-diffusion systemdefinedonadirectednetwork,underweakdiffusivecouplings.
Tocarryoutthelinearstabilitycalculation,wesubstitutethefollowingamplitudeandphaseperturbation[20,24]
Wj
(
τ
)
=WLC(
τ
)(
1+ρ
j(
τ
))
eiθj(τ )intoEq.(9)andwelinearizearound
ρ
j=0andθ
j=0obtaining ˙ρ
j ˙θ
j = −2 0 −2c2 0ρ
jθ
j + N k=1jk 1 −c1 c1 1
ρ
kθ
k . (11)Using theLaplacianeigenvaluesandeigenvectorsthat, bydefinition,satisfy
(α) =
(α)
(α) for
α
=1,...,N,theinho-mogeneousperturbations
ρ
jandθ
jcanbeexpandedasρ
jθ
j = N α=1ρ
(α)θ
(α) eλ(α)τ(α)j (12)
where
ρ
(α)andθ
(α)denotetheexpansioncoefficients, whileλ
(α) controlstheexponentialgrowthoftheα
thmode.Ifthereal part of
λ
(α) (λ
(α)Re ) is negative, perturbations shrink to zero,otherwise they grow exponentially. Plugging expansion
(12)intoEq.(11)yieldsthefollowingcondition
det
J α−λ
(α)I2=0 (13) with J α= −2+(α) −c1
(α) −2c2+c1
(α)
(α) .
F. Di Patti et al. / Commun Nonlinear Sci Numer Simulat 56 (2018) 447–456 451
Constrain(13)returnsanequation for
λ
(α) asa functionof(α) knownasthedispersion relation.Fora symmetric
undi-rectednetwork,astheoneexploredbyNakao[20,24],theeigenvalues
(α) oftheLaplacianoperatorsarerealandthis
ob-servationtranslatesinsimpleconditionfortheinstabilitytodevelop.Astraightforwardcalculation,allowsonetoconclude thatthe perturbationcan grow,andso destroythe synchronyofthesystem, provided 1+c1c2<0.When theunderlying
graphofcouplingisinsteaddirected,thespectrumoftheLaplacianoperatoriscomplexandtheinstabilitycansetinalso when1+c1c2>0.Forthisreason,thegeneralizedclassofinstabilitythatisencounteredwhenthesystemisconfinedona
asymmetricsupportisreferredtoastotopologydriven.Theconditionswhichresultin
λ
(α)Re >0whenthenetworkof cou-plingsisdirected,or,equivalently,when(α) iscomplex(
(α)=
(α)
Re +i
(α)Im),havebeenworkedoutin[25],buildingon
therecipeoutlinedin[21].Theconclusionof[25] areherebrieflyreviewedforthesakeofcompleteness.Indeed,
λ
Re(α)>0 if S2(
(α)Re)
S1(
(α)Re)
Im(α)
2 (14) where S2
(
αRe)
=C2,4(
(α)Re)
4− C2,3(
(α)Re)
3+C2,2(
Re(α))
2− C2,1Re(α) S1
(
αRe)
=C1,2(
αRe)
2− C1,1Re(α)+C1,0 with C2,4=1+c21 C2,3=4+2c1c2+2c21 C2,2=5+4c1c2+c21 C2,1=2+2c1c2 C1,2=c41+c21 C1,1=2c31c2+2c21 C1,0=c21
(
1+c22)
.Whencondition(14)ismetforsomeeigenvalue
(α),thelimitcycleloosesits stabilityandthesystemevolvestowardsa
differentattractors,whichisnonhomogeneousinspace.
Thenext sectionisaimedatchallengingthevalidityofEq.(9),asanapproximateequationforthenonlinearevolution ofan injectednon homogeneousperturbation. Toanticipateourfindings, wewill showthat theCGLE allows tocorrectly delineatethe region of parameters that yields stableandsynchronized self-sustained oscillators,and so accurately mark the transitionto the pattern forming domain. In both cases,i.e. when uniform oscillations or complexpatterns in den-sityare displayed,we obtain asatisfyingdegree ofcorrespondencebetweentheoriginal reaction-diffusionmodelandits correspondingCGLE.ToperformthenumericaltestswewillmakeuseoftheBrusselatormodel,asareferencecasestudy.
3. Numericalvalidationofthetheory:theBrusselatormodel
Tochallengethe analysisperformedinthe previoussection, we hereconsideraspecific reaction-diffusionsystem, the Brusselatormodel,whichallowsforself-sustainedoscillations.Thegoverningequationsread:
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
˙φ
j=A−(
B+1)
φ
j+φ
2jψ
j+Dφ N k=1jk
φ
k ˙ψ
j=Bφ
j−φ
2jψ
j+Dψ N k=1jk
ψ
k (15)where A, B stand for non negative constants parameters. This system admits a homogeneous steady state given by
(
φ
∗,ψ
∗)
=(
A,B/A)
.ByactingontheparameterB,whilekeepingA fixed,onecan induceaHopfbifurcation,whichopensuptheroutetoself-sustainedoscillations.ThebifurcationoccursforBc=1+A2,asexplainedintheannexedAppendixB.
Abovethebifurcationpoint,theBrusselatordynamicscanbeapproximatedbyaCGLE.Followingthederivationreportedin theAppendixB,whichbuildsontheseminalpaperbyKuramoto[18],oneendsupwiththecompactexpressionforc1and
c2: c1=−A Dφ− Dψ Dφ+Dψ c2= 4− 7A2+4A4 3A
(
2+A2)
. (16)The CGLE is therefore completely characterized andit can be readily employed to asses the stability of the uniform oscillatorystate,namelythespatiallyextendedconfigurationthatisobtainedasthesynchronousreplicaofNself-sustained
Fig. 1. Panel (a): eigenvalues of the Laplacian for a balanced Newman–Watts [26] network generated with p = 0 . 27 and N = 100 nodes. The algorithm to built the network is detailed in [22] . The shaded area marks the instability region for the Brusselator model, as obtained under the CGL normal form representation. Here, A = 1 . 55 , B c = 1 + A 2 , D φ= 0 . 6 and D ψ = 4 . 9 . The perturbative parameter is = 0 . 07 . For the case at hand, A c = 1 . 48 . Panel (b): the
real part of the dispersion relation (magenta triangles) for the same choice of network and parameters as in panel (a). The black line originates from the continuous theory. Panel (c): pattern relative to species φobtained by numerical integration of system (15) . Panel (d): pattern relative to species φ
calculated using the CGL approximation.
oscillators.FixthevaluesofDφ andDψ.Then,thestabilityofthesynchronousstateisultimatelycontrolled byA,thesole parameterthatonecanfreelytune.WewilldenotewithActhevalueofA(ifitexists),abovewhichstabilityiseventually
lost.Moreover,wewillsettheparametersc1andc2 soastoenforcestabilityonasymmetricsupport,namely1+c1c2<0.
TheinstabilitythatmaterializesforA>Acishenceindirectlyreflectingtheasymmetryoftheimposedcouplings.
TheresultsoftheanalysisisdepictedinFig.1(a)and(b).Theshaded greyareaintheplane (
(α)Re ,
(α)Im)delimitsthe
regionofinstability,asitfollowsinequality(14).SymbolsrepresenttheeigenvaluesoftheLaplacianmatrixinthecomplex plane andfollowthespecificchoice ofthenetwork operated.Asexplained inthecaptionofFig.1,we hereworkwitha balancedNewman-Wattsgraph,generatedbyfollowingtherecipediscussedin[22].FortheselectedvalueofA(largerthan thecriticalvalueAc=1.48),twoeigenvaluesoftheLaplacianmatrixprotrudeinsidetheshadedregion,thustriggeringthe
instability.Thisisindeedatopologydriveninstability,asitcannotrealizeifthesamechoiceofparametersisoperatedforan homologoussystemhostedonasymmetricsupport.Toarguealongthisline,considerasymmetricnetworkofcouplings:the spectrumoftheLaplacianmatrixisnowreal(
(α)Im =0)andshouldhencefallontherealaxisofFig.1(a).Thegreyregion wheretheinstabilitymaterializes,intersectstherealaxisonlyintheorigin,wherethetrivialzeroeigenvalue(corresponding totheuniform,fullysynchronized,limit cyclestate)sits.ForthechosenvaluesofA andB,hencec1 andc2,itistherefore
necessarytoactivateanimaginarycomponentof
(α)Im,toinvadetheregionofthecomplexplanedeputedtotheinstability. Tostate it differently,forthesame choiceofparameters, theinstability sets inonlyifa suitable degreeof asymmetryis enforcedinthenetworksofconnections,whileitisboundtofadeaway otherwise.ThisscenarioisconfirmedbyFig.1(b) wherethe (magenta online)triangles representthereal partofthe dispersionrelation,
λ
Re, asafunction of−(α)Re . Also
inthiscase, the dispersionrelation ispunctuallypositive incorrespondenceoftwo specificentries −
(α)Re .The solid line
curve representsthehomologousdispersionrelationwhenthediffusive couplingisinstead declinedonasymmetric(and continuous)support.
PatternsemergingfromtheaforementionedinstabilityaredisplayedinFig.1.Panel(c)showspatternsrelativetospecies
φ
obtainedvia a numericalintegration of system(15),while panel(d) originates fromthe CGLE (8). The degreeof corre-spondenceissatisfactoryandtestifiesaposterioriontheadequacyoftheproposedapproximation.F. Di Patti et al. / Commun Nonlinear Sci Numer Simulat 56 (2018) 447–456 453
Fig. 2. Panel (a): eigenvalues of the Laplacian for a balanced Newman–Watts [26] network generated with p = 0 . 27 and N = 100 nodes. The algorithm to built the network is detailed in [22] . The shaded area marks the instability region for the Brusselator model with A = 1 . 35 < A c , D φ= 0 . 6 and D ψ = 4 . 9 . The perturbative parameter is = 0 . 07 . Panel (b): the real part of the dispersion relation (magenta triangles) for the same choice of network and parameters as in panel (a). The black line originates from the continuous theory. Panel (c): pattern relative to species φobtained by numerical integration of system (15) . Panel (d): pattern relative to species φcalculated using the CGL equation.
Fig.2refers toA<Ac:thesynchronizedlimit cyclesolutionispredictedtobestable,followingtheCGLapproximation.
Directintegrationoftheoriginalreactiondiffusionsystemconfirmsthisconclusion.Thedegreeofcorrespondencebetween originalandapproximatedschemesis,alsointhiscase,satisfying.
4. Conclusions
Synchronizationis a fascinatingfield of investigation that hasapplications ranging frombiology to computer science, justto namefew examples.Whenthisphenomenon occurson complexnetworks,thestabilityofthesynchronizedstates isstronglyaffectedbythetopologyofthegraph.Whenthenetworkisdirected,external nonhomogeneousperturbations candestabilize synchronousoscillations alsoifthe parametersare setto valuesthat fallin theregion ofstabilityforthe dynamicsformulatedonasymmetricsupport.
Startingfromthesepremises,wehereconsideredareactiondiffusionsystemdefinedonadirected(andbalanced)graph. Thesystemismadeto evolveinthevicinityofasupercritical Hopfbifurcationandweakdiffusive couplingsareassumed toholdbetweenadjacentnodes.BygeneralizingtheworkofNakao[20]tothesettingwhereasymmetryinthecouplings isenforced,andadapting themultiple-scales approachdiscussed inKuramoto [18]tothecaseofa discrete,network like support,wederivedanormalformequationtoapproximatethelocalevolutionoftheunderlyingreaction-diffusionsystem. ThisisaComplexGinzburg-Landauequation(CGLE)whosecoefficientsclearlydependontheparametersofthemodel,but alsoonthetopologicalcharacteristicsofthehostingnetwork.TheCGLE canbeeffectivelyemployedtoprobethestability oftheperiodicuniformsolutionandworkout theconditionsfortheonsetofthesymmetrybreakinginstabilitythat even-tuallydestroythesynchronizedregime. NumericaltestsperformedfortheBrusselatormodel,confirmtheadequacyofthe proposed approximation.Patterns obtained underthe CGLE schemeresemblecloselythose displayedupon integration of theoriginalreaction-diffusionsystem.
Acknowledgments
ThisworkhasbeensupportedbytheprogramPRIN2012fundedbytheItalianMinisterodell’Istruzione,dellaUniversità edellaRicerca(MIUR).D.F.acknowledgesfinancialsupportfromH2020-MSCA-ITN-2015projectCOSMOS642563.Thework
ofT.C.presentsresearch resultsoftheBelgianNetworkDYSCO(DynamicalSystems,Control,andOptimization),fundedby theInteruniversityAttractionPolesProgramme,initiatedbytheBelgianState,SciencePolicyOffice.
AppendixA. Thesolvabilitycondition
WeconsideralinearoperatorA,andtwocomplexvectorsu(t)andb(t)ofthesamelength.AccordingtotheFredholm theorem,thelinearsystemAu
(
t)
=b(
t)
issolvableifv(
t)
,b(
t)
=0forallvectorsv(t)solutionofA∗v(
t)
=0,whereA∗is theadjointoperatorsatisfyingA∗y,x=y,Ax∀
x,y.Theangularbracketsdenotethescalarproductthatwe heredefine asv(
t)
,b(
t)
=2π/ω00 v†
(
t)
b(
t)
dt,thesymbol†standingfortheconjugatetranspose.With referenceto Eq.(6), thefirst requirement ofthe Fredholm theoremconsists in findingv(t) such that
(
∂
/∂
tI2− L0)
∗v(
t)
=0.Bypartial integration,andrecalling thatL0 isa realmatrix,we findthat(
∂
/∂
tI2− L0)
∗=−(
∂
/∂
tI2N+L0)
T.Asa consequence,the systemtobe solvedis−
(
∂
/∂
tI2+L0)
Tv(
t)
=0.Inanalogy withEq.(7),wesearchv(t)intheform v(
t)
=U∗0eiω0t forsomevectorU∗0.Substitutingthisansatz intotheprevious equation,wefindL0TU∗0=−i
ω
0U∗0.Tosimplifycalculation,itisconvenienttonormalizeU∗0sothat
(
U∗0)
†U 0=1.Having defined U∗0, one can explicitly write down the solvability condition
U∗0eiω0t,B(ν)j
(
t,τ
)
=0. Since B(ν)j(
t,τ
)
turns out to be periodic functions of period 2
π
/ω
0, it is appropriate to express them in the form B(ν)j(
t,τ
)
=+∞
l=−∞
B(ν)j
(
τ
)
l
eilω0t. If we multiply this series by
(
U∗0eiω0t
)
† we again obtain periodic functions that, wheninte-grated over the period 2
π
/ω
0 give zero. The only exception holds for l=1 which gives U∗0eiω0t, B(ν)j(
τ
)
1 eiω0t= 2π/ω0 0(
U∗0)
† B(ν)j(
τ
)
1dt.Theintegranddoesnotdependontimetandthereforetheintegraliszeroonlyiftheintegrand
itselfisidenticallyequaltozero.Forthisreasonthesolvabilityconditionreducesto
(
U∗0)
†B(ν)j
(
τ
)
1=
0
∀
j.AppendixB.DetailsofthecoefficientsoftheGLfortheBrusselatormodel
InthissectionweexplicitlyderivethecoefficientsoftheCGLequationfortheBrusselatormodel. TheJacobianmatrixassociatetosystem(15)evaluatedattheequilibriumpointreads
L 0=
B− 1 A2 −B −A2 .Settingthe systemat thebifurcation point means requiringL0 to havea pairof imaginaryeigenvalues, which translates
intotrL0=0.ThisrelationimpliesaconstrainonthecriticalvalueoftheparameterBwhichmustbechosenasBc=1+A2.
Withthisassumption,L0 canberewrittenas L 0=
A2 A2
−
(
1+A2)
−A2anditspairofimaginaryeigenvaluesare
λ
0=±iω
0=±iA.TherighteigenvectorsofL0andLT0correspondingto,respectively,eigenvaluesiAand−iAaregivenby
U 0=
1 −1+ i A U ∗0=12 1+iA iA .Toproceedwiththemultiscaleanalysis,we perturbBc asB=Bc+
2Bc andweimmediatelyfindtheentriesofmatrix
L1 L 1=
(
1+A2)
1 0 −1 0andthevalueof
σ
whichturnsouttobe(
1+A2
)
F. Di Patti et al. / Commun Nonlinear Sci Numer Simulat 56 (2018) 447–456 455
TheoperatorsM0andN0aregivenby
(
M0xy)
1=−(
M0xy)
2 =1+A2 A x1y1+A(
x1y2+x2y1)
(
N0xyz)
1=−(
N0xyz)
2 =1 3(
x1y1z2+x2y1z1+x1y2z1)
where(· )idefinestheithcomponentofatwo-dimensionalvector.Thesepreliminaryquantitiesallowustocalculate
M0U 0U 0=
⎛
⎜
⎝
1 A− A + 2i −1 A+A− 2i⎞
⎟
⎠
M0U 0U ¯0=⎛
⎜
⎝
1 A− A −1 A+A⎞
⎟
⎠
(B.1) and N0U 0U 0U ¯0=⎛
⎜
⎜
⎝
−1+ i 3A 1− i 3A⎞
⎟
⎟
⎠
.FromEq.(B.1)wecanmakeexplicitV0andV2 V 0= 2
(
A2− 1)
A3 0 1 V 2=− 1 3⎛
⎜
⎜
⎝
−4 A+2 1 A2− 1 i −1 A 1 A2 − 5 +22 A− A i⎞
⎟
⎟
⎠
andthus M0U ¯0V 2=⎛
⎜
⎜
⎜
⎝
−1+ 1 A2+i 2 3 A2− A4− 1 A3 +1− 1 A2− i 2 3 A2− A4− 1 A3⎞
⎟
⎟
⎟
⎠
M0U 0V 0=⎛
⎜
⎜
⎝
2(
A2− 1)
A2 −2(
A2− 1)
A2⎞
⎟
⎟
⎠
.NoticethatourequationsforU0andV2 areslightlydifferentfromthosereportedin[18].Despitethisminordifference,we
endupwithsamevaluesofdandg.Finally,wecanwritedownthetwolastcoefficientsoftheCGLequation: d= 1 2[Du+Dv− iA
(
Du− Dv)
] g= 12 A2+2 A2 +i 4A4+4− 7A2 3A3 . References[1] Murray JD . Mathematical biology II: spatial models and biomedical applications. third ed. Springer–Verlag; 2003 . [2] Ball P . Shapes. Oxford: Oxford University Press; 2009 .
[3] Karsenti E . Self-organization in cell biology: a brief history. Nat Rev Mol Cell Biol 2008;9:255–62 .
[5] Buhl J , Sumpter DJT , Couzin D , Hale JJ , Despland E , Miller ER , et al. From disorder to order in marching locusts. Science 2006;312:1402–6 . [6] Bonabeau E , Theraulaz G , Deneubourg J-L , Aron S , Camazine S . Self-organization in social insects. Trends Ecol Evol 1997;12:188–93 . [7] Pikovsky A , Rosenblum M , Kurths J . Synchronization. Cambridge: Cambridge University Press; 2001 .
[8] Glass L . Synchronization and rhythmic processes in physiology. Nature 2001;410:277–84 . [9] Buck J . Synchronous rhythmic flashing of fireflies. II.. Q Rev Biol 1988;63:265–89 .
[10] Rohden M , Sorge A , Timme M , Witthaut D . Self-organized synchronization in decentralized power grids. Phys Rev Lett 2012;109:064101 . [11] Dörfler F , Chertkovb M , Bulloa F . Synchronization in complex oscillator networks and smart grids. PNAS 2013;110:2005–10 .
[12] Arenas A , Díaz-Guilera A , Kurths J , Moreno Y , Zhoug C . Synchronization in complex networks. Phys Rep 2008;469:93–153 . [13] Osipov G , Kurths J , Zhou C . Synchronization in oscillatory networks. Berlin, Germany: Springer; 2007 .
[14] Barrat A , Barthélemy M , Vespignani A . Dynamical processes on complex networks. Cambridge: Cambridge University Press; 2008 . [15] Boccaletti S , Latora V , Moreno Y , Chavez M , Hwang D-U . Complex networks: structure and dynamics. Phys Rep 2006;424:175–308 .
[16] Boccaletti S , Bianconi G , Criado R , del Genio CI , Gómez-Gardeñes J , Romance M , et al. The structure and dynamics of multilayer networks. Phys Rep 2014;554:1–122 .
[17] Newman MEJ . Networks: an introduction. Oxford: Oxford University Press; 2010 .
[18] Kuramoto Y . Chemical oscillations, waves, and turbulence. New York: Springer-Verlag; 1984 . [19] Cross MC , Hohenberg PC . Pattern formation outside of equilibrium. Rev Mod Phys 1993;65:851–1123 .
[20] Nakao H . Complex Ginzburg–Landau equation on networks and its non-uniform dynamics. Eur Phys J Spec Top 2014;223:2411–21 . [21] Asllani M , Challenger JD , Pavone FS , Sacconi L , Fanelli D . The theory of pattern formation on directed networks. Nat Commun 2014;5:4517 . [22] Contemori S , Di Patti F , Fanelli D , Miele F . Multiple-scale theory of topology-driven pattern on directed networks. Phys Rev E 2016;93:032317 . [23] Fredholm EI . Sur une classe d’equations fonctionnelles. Acta Math 1903;27:365–90 .
[24] Nakao H , Mikhailov AS . Diffusion-induced instability and chaos in random oscillator networks. Phys Rev E 2009;79:036214 . [25] Di Patti F , Fanelli D , Miele F , Carletti T . Benjamin-Feir instabilities on directed networks. Chaos, Solitons Fractals 2017;96:8–16 . [26] Newman MEJ , Watts DJ . Scaling and percolation in the small-work network model. Phys Rev E 1999;60:7332–42 .