HAL Id: hal-00490666
https://hal.archives-ouvertes.fr/hal-00490666v2
Submitted on 9 Jun 2010
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Numerical computation of the Faradaic impedance of inlaid microdisk electrodes using a finite element
method with anisotropic mesh adaptation
Richard Michel, Claude Montella, Claude Verdier, Jean-Paul Diard
To cite this version:
Richard Michel, Claude Montella, Claude Verdier, Jean-Paul Diard. Numerical computation of the Faradaic impedance of inlaid microdisk electrodes using a finite element method with anisotropic mesh adaptation. Electrochimica Acta, Elsevier, 2010, 55, pp.6263-6273. �10.1016/j.electacta.2009.12.093�.
�hal-00490666v2�
Faradai impedane of
inlaid mirodisk eletrodes
using a nite element method
with anisotropi mesh adaptation
R. Mihel
1,⋆
, C. Montella
1
, C. Verdier
2
, J.-P. Diard
1
(1) Laboratoire d'Életrohimie et de Physiohimie des Matériaux et Interfaes, UMR 5631
CNRS+Grenoble-INP+UJF, Bât. PHELMA, 1130 Rue de la Pisine, B.P. 75, Domaine Uni-
versitaire, 38402 Saint Martin d'Hères, Cedex, Frane.
(2)Laboratoire de Spetrométrie Physique, UMR5588 CNRS-UJF, DomaineUniversitaire, 140
Avenue de la Physique, 38402 Saint Martin d'Hères, Frane.
⋆
orrespondingauthor.
Tel.: +33-4-76826547; fax: +33-4-76826630
e-mail: Rihard.Mihellepmi.grenoble-inp.fr
Abstrat
The Faradai impedane of a mirodisk eletrode inlaid in an insulating sur-
fae is revisited by numerial omputation using a nite element method (FEM)
with anisotropi mesh adaptation. New features of the numerial results, as om-
paredtopreviousworks,areanalyzed. Arstattrativefeatureisthatthediusion
impedanerelative to a mirodisk eletrode, evaluatedat theequilibriumpotential
of the eletrode, depends both on eletron-transfer and mass-transport kinetis, in
ontrast with the usual behaviour ofuniformly aessibleeletrodes. Next, thedo-
mainofvalidityoftheFleishmannandPonssemi-analytialformulationofdiusion
impedane is determined. Finally, the harateristi of impedane graphs, whih
arethe diusionresistane, theharateristi frequeny at theapexof theNyquist
diagramandtheimaginarypartofthediusionimpedaneatthisapex,arestudied
asfuntionsofadimensionlessparameterthatomparingthestandardrateonstant
of eletron transfer to themiroeletrode diusion onstant. Closed form approxi-
mationsareproposedfor allquantities inorder tohelptheanalysisofexperimental
data.
Keyword: Impedane, Miroeletrode, Mirodisk, Simulation, Finite element method,
Anisotropi mesh adaptation.
Despite the large literature dediated to the theory of eletrohemial impedane spe-
trosopy (EIS)[13℄ontheonehandandthetheoryofultramiroeletrodes(UMEs)[4,5℄
withappliationtosanningeletrohemialmirosopy[6℄ontheotherhand,onlyafew
papers have dealt with the theoretial derivation of the impedane of a mirodisk ele-
trode inlaidinaninsulatingsurfae. The analysispresented by FleishmannandPons [7℄
opened up the use of miroeletrodes to a impedane measurements. They alulated
the real and imaginary parts of the diusion impedane fromBessel's funtion integrals.
When ahieving this work, the numerial evaluation of Fleishmann and Pons formulae
was not an easytask. In order tomake the impedane alulationeasier, they presented
their numerial data in the form of tabulated funtions. Some additionalinformation is
available from the reent work by Navarro-Laboulais et al. [8℄. These authors derived
the theoretial formulation of the diusion resistane that is the low-frequeny limit of
the diusionimpedane. Theyevaluatednumeriallytheharateristidimensionlessfre-
queny at the apex of the Nyquist impedane graph. An algorithm for alulation of
the mirodisk impedane was outlined by these authors for implementation in omplex
non-linear least-squares tting (CNLS-Fit) programs.
An alternative approah for omputing the impedane of mirodisk eletrodes is based
on nite element analysis. The pioneering work on this topi was that of Ferrigno and
Girault [9℄, whih foused onthe axisymmetrireessed mirodiskgeometry. As a limit,
the inlaid disk eletrode was reovered when the reess depth tends towards zero. A
relatively good (qualitative) agreement was observed with the semi-analytial formula-
tion of Fleishmann and Pons. However, this alulation was limited to the reessed
mirodisk geometry to avoid the presene of a singularity in the neighbourhood of the
eletrode edge where the boundary ondition hanges from the Dirihlet type (uniform
onentration perturbation) on the disk to the homogeneous Neumann type (zero ux)
on the insulator (see [10℄ for analysis of this kind of loalsingularities). The singularity
refers to the onentration perturbation eld whih is not twie dierentiable. Gabrielli
ulationoftheeletrohemialimpedane ofaninlaidmirodiskeletrode usingCOMSOL
Multiphysis (formerly FEMLAB) software. These authors investigated the inuene of
the disk radius and the total eletrode radius (eletroative disk + insulating sheath)
on the impedane diagram. Their theoretial preditions were ompared to experimen-
tal data olleted froma 10µmdiameterPt mirodiskimmersed in a10 mM K3Fe(CN)6
+ 10 mM K4Fe(CN)6 + 0.5 M KCl aqueous solution. The impedane was measured
at the equilibrium potential of Pt eletrode. Very good agreement was found for the
impedane diagramsimulatednumeriallyusing Fleishmannand Ponsequations,aswell
as with the simulated FEM diagram. As disussed by Gabrielli et al. [11℄, Ferrigno and
Girault[9℄usedaDirihletboundaryondition atthedisk/eletrolyteinterfae(withthe
perturbationof interfaial onentrationof eletroative speies being diretly ontrolled
by the eletrode potentialperturbation), whileFleishmannand Pons [7℄used a uniform
non-homogeneous Neumann boundary ondition (i.e. the distribution of the diusional
ux perturbation is assumed to be uniform over the disk surfae). The more rigorous
treatmentproposedby Gabriellietal.[11℄ makesuse ofthe Fourier-Robinboundaryon-
dition that is the linearized formulation with respet to the eletrode potential of the
Butler-Volmerurrent-potentialharateristi.
Theaimofthisworkisanattempttorenethenumerialsimulationoftheimpedane of
inlaid mirodisk eletrodes using a nite element method (FEM) with anisotropi mesh
adaptation. In this artile, the omputation proedure is employed for modeling the
Faradai impedane of mirodiskeletrodes atthe equilibriumpotential.
The results presented here are foused on the harateristi quantities available from
impedanegraphs,whiharethediusionresistane,theharateristifrequenyobserved
at the apex of the Nyquist diagram, and the imaginary part of the impedane measured
atthis apex. Ofourse, thewholeimpedane diagramwillbeof majorimportaneforits
implementation in CNLS-Fit programs. However, the three quantities mentioned above
are suienttoomparethe aurayofthesemi-analytial(FleishmannandPons)and
inludingboth the meshadaptation strategy and some aspets of omputer implementa-
tion, is presented in Setion 3. Next, the inuene of omputational domain size on the
auray of omputed diusion impedane is investigated in Setion 4. The validity of
Fleishmann and Pons formulae is heked in Setion 5 by omparison with adaptative
FEM omputations. Finally, in Setion 6, the harateristi elements from the diusion
impedane graph are analyzed with respet to the dimensionless number that ompares
thestandardrateonstantofeletrontransfertothediusiononstantofmiroeletrodes.
2 Theory
2.1 Geometry
We onsider a mirodisk eletrode inlaid in an insulating surfae. The mirodisk radius
and the total eletrode radius (eletroative disk + insulating sheath) are respetively
denoted by re and rmax. The geometry of the devie is axisymmetri, so the alulation
domainΩisredued toa2-D meridiansetionofthe domainoupied by theeletrolyti
solution. This domain is skethed in Fig. 1 where (r, z) denote the usual ylindrial
oordinates. Its boundary Γ deomposes into Γs, the symmetry axis; Γe, the eletrode
surfae;Γi,theinsulatorsurfae;andΓb,thebulkeletrolyte. Thesepartsoftheboundary
Γ are dened asfollows:
Γs = {(r, z) ; r= 0 and 0< z < zmax} (1a) Γe = {(r, z) ; 0 < r < re and z = 0} (1b) Γi = {(r, z) ; re< r < rmax and z = 0} (1) Γb = {(r, z) ; (r=rmax and 0< z < zmax) or( 0≤r ≤rmax and z =zmax)}(1d)
Γb represents the part of the eletrolytewhih isloatedatasuient distane fromthe
mirodisk so that the inuene of the eletrohemial reation (2) ourring at the disk
surfaeanbenegletedonΓb. Thisassumptionallowstoimposethe boundaryondition
ofsemi-innitediusiononΓbwhile,inarigorousway, itshouldbeimposedataninnite
distane fromΓe. The validityof this assumptionmainlydepends onthe ratiormax/re as
will bedisussed inSetion 4.
2.2 Phenomenology: eletron-transfer and mass-transport pro-
esses
A one-step eletrohemialreation ours at the disk surfae and involves a n-eletron-
transfer proess (usuallyn = 1)between twosoluble speies(Oand R)atthe metal(e.g.
Pt)/eletrolyte interfae Γe:
O + ne −→←− R (2)
The reationrate is desribed by Butler-Volmer kinetis [3,13℄and, atthe eletrode po-
tentialE,it isexpressed in the usual way by:
v =k0 (exp (−αrξ) cO − exp (αoξ)cR ) with ξ= (n F/RT)
E−E0′
on Γe (3)
wherecX denotes theonentrationof speiesX = O,R,andthe eletron-transferproess isharaterizedbyitsstandardrateonstantk0,thestandard(formal)potentialE0′,and
the symmetry oeients αo and αr, with αo +αr = 1, the other symbols having their
usual meaning.
Duetothe quiesentsolutionandthepreseneofasuitablesupportingeletrolyte, migra-
tion and onvetion eets onmass-transportproesses an be negleted 1
, thusresulting
in apure diusion proess suh that the ux-vetor JX of speies X = O,R isgiven by:
JX =−DX∇cX in Ω (4)
where the diusion oeient DX is a onstant inthe presene of supporting eletrolyte, 1
Stritlyspeaking,mass-transportofeletroativespeiesbyunforedonvetionshouldbetakeninto
onsiderationusing themodel presentedby Amatoreet al.[14,15℄. Nevertheless, theseauthors showed
that theontributionofunfored onvetionanbe negletedforaneletroderadius re <25µm. This
and ∇cX is the onentration gradient whose omponents in ylindrial oordinates are the partialderivatives, ∂rcX and ∂zcX,of cX withrespet tor and z respetively.
2.3 Struture of the Faradai impedane
2.3.1 Steady-state problem
The rst step to analyse the Faradai impedane is to alulatethe steady-state regime
orresponding toastati value ES of the potentialimposed tothe eletrode. This regime
isharaterizedbythestationentrationscS,X,forspeiesX = O,R,whiharesolutions
of the following boundary value problem where the subsript 'S' stands for steady-state
onditions:
Given ES, nd, for X = O,R,cS,X : (r, z)∈Ω→cS,X(r, z)∈R suhthat 2
:
−DX∆cS,X= 0 in Ω (5a) DX∇cS,X·n=ǫXvS(ES, cS,O, cS,R) on Γe (5b)
DX∇cS,X·n= 0 on Γi (5)
cS,X =cbX on Γb (5d)
whereǫO =−1,ǫR= +1,cbX denotesthe bulkonentrationofspeiesX,nistheoutward unit vetor normal to Γ, and ∆ is the Laplae operator whose ylindrial expression is
given by ∆cS,X =∂rr2 cS,X+1r∂rcS,X+∂2zzcS,X. Thesteady-state reationrate vS in Eq. (5b)
omes from Eq. (3) where E and cX are respetively replaed by ES and cS,X. One the
problem(5)hasbeensolved,theinterfaialonentrationsareknownandthesteady-state
Faradai urrent isobtained by integration along the radial diretion(r) as follows:
IS,f =−2π n F Z
Γe
vSds =−2π n F Z re
0
vS(ES, cS,O, cS,R) r dr (6)
2
The boundary ondition on Γs takes on the formulation, r DX∇cS,X·n = 0, whih results from
the weak (variational) formulation of the boundaryvalueproblem used in the FEM framework. This
is straightforwardlysatisedbeauseofr = 0 onΓs. Hene, itis notneessaryto write thisboundary
onditioninEq. (5). Thesameremark appliesto theharmoniboundaryvalueprobleminEq. (9).
diusion boundary ondition cS,X(r, z)−→cbX when r orz −→ ∞.
2.3.2 Harmoni problem
In the seond step, a smallharmoni perturbation of the eletrode potential isimposed:
E(t) =ES+EH exp(jωt) (7)
where j = √
−1, ω = 2πf is the angular frequeny, f is the frequeny and EH is the
amplitude of perturbation. Under the above onditions, the permanent regime resulting
from the potentialperturbation leads tothe following onentrationelds:
cX(r, z, t;ω) =cS,X(r, z) +cH,X(r, z;ω) exp(jωt) (8)
The spatial part of the onentration elds perturbations cH,X(r, z;ω) are solutions of
the following boundary value problem where the subsript 'H' stands for the permanent
harmoni regime and where the notation ';ω' is used for highlighting the role of ω as a
parameter:
Given ES, EH and ω, nd, for X = O,R,cH,X : (r, z)∈Ω→cH,X(r, z)∈C suhthat
j ω cH,X−DX∆cH,X = 0 in Ω (9a) DX∇cH,X·n=ǫXvH(ES, cS,O, cS,R, EH, cH,O, cH,R) on Γe (9b)
DX∇cH,X·n= 0 on Γi (9)
cH,X = 0 on Γb (9d)
The harmoni perturbation of the reation rate vH results from linearization of the ex- pression in Eq. (3) aroundthe stati polarizationpoint (ES, cS,O, cS,R). Using Eq. (7) for
vH(ES, cS,O, cS,R, EH, cH,O, cH,R) = −k0 n F
R T gSEH − exp (−αrξS) cH,O + exp (αoξS) cH,R
(10)
where
gS =αr exp (−αrξS) cS,O + αo exp (αoξS) cS,R and ξS=n F
ES−E0′
/(R T) (11)
2.3.3 Harmoni perturbation of Faradai urrent and Faradai impedane
Onetheharmoniproblem(9)issolved,theharmoniperturbationsofinterfaialonen-
trationsare known and the permanentperturbationIH,f of the Faradaiurrentresulting
from Eqs. (7) and (8) an be derived by integration along the radial diretion (r) as
follows:
IH,f =−2π n F Z
Γe
vHds =−2π n F Z re
0
vH(ES, cS,O, cS,R, EH, cH,O, cH,R) r dr (12)
This expression isevaluatedfor eahvalue of the angularfrequeny ω. Theratio IH,f/EH
denes the Faradaiadmittane Yf(ω) asa funtionof ω. Its formulationresults diretly from Eq. (10). It an be derived as:
Yf(ω) = Gct+YdO(ω) +YdR(ω) (13a) Gct = k0 n2F2
R T 2π Z re
0
gS(r,0)r dr (13b)
YdO(ω) = −k0 n F EH
exp(−αrξS) 2π Z re
0
cH,O(r,0;ω)r dr (13) YdR(ω) = k0 n F
EH exp(αoξS) 2π Z re
0
cH,R(r,0;ω)r dr (13d)
whereGct istheeletron-transfer ondutane, andYdX denotesthe onentrationadmit-
tane relative to the speies X = O,R. Of ourse, the Faradai impedane is obtained
as
Zf(ω) = 1/Yf(ω) (14)
respondene EH −→ IH,f. So it is the admittane whih omes naturally from the above
equations and not the impedane.
3 Numerial resolution
3.1 Dimensionless formulation and limiting onditions
3.1.1 Dimensionless numbers
It is easy to show that the solutions cS,O and cS,R of the steady-state boundary value problem (5)satisfy the relation:
DOcS,O(r, z) +DRcS,R(r, z) =DOcbO+DRcbR for (r, z)∈Ω (15)
This property makes it possible to redue the problem (5) to the determination of one
onentration eld only. Arbitrarily,wehoose this onentration as cS,R and we use the
diusionoeientand bulkonentrationof speiesRasthereferenequantitiesforthe
denition of the dimensionlessvariablesr⋆ =r/re, z⋆ =z/re,t⋆ =t DR/r2e, c⋆S,X =cS,X/cbR
and c⋆H,X = cH,X/cbR where supersript '⋆' indiates a dimensionless variable or operator.
The dimensionless numbers resultingfromthis salingare:
Λ = k0re
DR
and u= ω r2e DR
(16)
3.1.2 Inuene of Λ: stati onditions
Λ ompares the standard rate onstant of eletron transfer to the diusion onstant of
miroeletrodes. It omes from the dimensionless form of the stati boundary ondition
(5b) writtenforthespeiesX = Rand ombinedtogether withEq. (15) inordertoelim-
inate cS,O. The resultingequation is expressed by the following Fourier-Robin boundary
∇⋆c⋆S,R·n⋆ = Λα(ξS)
c⋆NS,R(ξS)−c⋆S,R
on Γ⋆e (17a)
with:
c⋆NS,R(ξS) =
DOcbO
DRcbR
+ 1
1
1 + DDOR exp(ξS)
and α(ξS) = exp(αoξS) + DR
DO
exp(−αrξS)
(17b)
where c⋆NS,R(ξS) stands forthe dimensionless stati interfaialonentration for Nernstian systems.
When Λα(ξS) is very large, the Fourier-Robin boundary ondition (17a) leads to the Dirihlet boundary ondition c⋆S,R = c⋆NS,R(ξS) over the disk surfae Γe. In ontrast, at
verylowvaluesof Λα(ξS), theFourier-Robinboundaryondition leadstoauniformnon- homogeneous Neumann boundary ondition, i.e. assuming a uniform perturbation ux
over the disk surfae, exept ina smallneighbourhoodof the eletrode edge.
3.1.3 Inuene of Λ: harmoni onditions
WhenDO =DR,theperturbationsofonentrationseldssatisfycH,O(r, z;ω)+cH,R(r, z;ω) = 0 for (r, z)∈Ω, whih leads to the simpliedformulation of Eqs. (9b) and (10):
∇⋆c⋆H,R·n⋆ =−Λα(ξS)
c⋆NH,R+c⋆H,R
on Γ⋆e (18a)
with:
c⋆NH,R = ξH
exp (−αrξS) + exp (αoξS) gS
cbR
(18b)
where c⋆NH,R stands for the dimensionless harmoni interfaial onentration perturbation
for Nernstian systems and ξH = n F ER TH. Notethat gS, obtained from Eq. (11), is onstant
over the eletrode surfae when the eletrode impedane isalulated at the equilibrium
potential. The situation would be muh more intriate for impedane alulations per-
formed away from the equilibriumpotential.
So the system kinetis is still governed by Λα(ξS) under harmoni onditions. Two lim-
iting situations an be predited. When Λα(ξS) is large, the harmoni perturbations of onentration elds tend to satisfy Dirihlet onditions at the disk/eletrolyte interfae
likeinthe workof Ferrignoand Girault[9℄. When Λα(ξS) issmall, the harmonipertur-
bations of diusional uxes present approximately uniform values over the disk surfae,
and then, like in the work by Fleishmann and Pons [7℄, the Fourier-Robin boundary
ondition (18a) an be replaed by a uniform Neumann ondition on Γe. Both limiting
onditions willbenumerially veriedin Setion6.2.
At the opposite, when DO 6= DR, no simpliation of harmoni equations is possible, so FEM omputationsshould be performedwith two onentration perturbationelds.
In the present work, we use the same approah as Gabrielli et al. [12℄, i.e. the Fourier-
Robin boundary ondition dened by Eq. (9b) without any approximation, exept that
the steady-state potential of the eletrode is equal to its equilibrium potential, so ξS is
given by the Nernst equation: ξS = ln(cbO/cbR),therefore ξS= 0 whencbO =cbR.
3.1.4 Inuene of u
The dimensionlessangularfrequeny uomesdiretlyfromthe dimensionlessformofEq.
(9a):
j u c⋆H,X− DX
DR
∆⋆c⋆H,X = 0 in Ω⋆ (19)
Itompares theangularfrequenyω withthereiproalofdiusiontimeonstantre2/DR.
When u is very large, it follows from the partial dierential equation (19) that the on-
entration perturbations are vanishing inthe eletrolyte, exept in the immediate neigh-
bourhood ofthe disksurfae (due tothe boundaryondition (18a)),so aboundary layer
develops near Γ⋆e at high frequenies. Conversely, at low frequenies, the onentration elds perturbations extendfrom the interfae Γ⋆e intothe eletrolyteuntil they vanish on Γ⋆b.