Journal of The Electrochemical Society, 146 (11) 4183-4188 (1999) 4183

S0013-4651(98)05-026-5 CCC: $7.00 © The Electrochemical Society, Inc.

Treatment of the Impedance of Mixed Conductors

Equivalent Circuit Model and Explicit Approximate Solutions
J. Jamnik*,a and J. Maier*
Max-Planck Institute für Festkörperforschung, 70569 Stuttgart, Germany

The electrochemical impedance of a mixed conductor with two charge carriers sandwiched between two equal electrodes is mod-
eled phenomenologically. A simple analytical approximation is found which applies in a very broad materials parameter window
(arbitrary carrier mobilities and concentrations) as long as linear response is guaranteed and the sample thickness is great compared
to the Debye length. The approximation is applicable to reversible, selectively blocking as well as to partially blocking electrodes.
The validity of the approximation is justified by comparison with the exact solution of Nernst-Planck-Poisson set of equations as
given, e.g., by Macdonald. The analytical approximation describes the typical Warburg response as well as the single-carrier behav-
ior as limiting cases. Finite interfacial resistances (not perfectly blocking or not perfectly reversible electrodes) lead to additional
semicircles and/or to distorted Warburg responses.
© 1999 The Electrochemical Society. S0013-4651(98)05-026-5. All rights reserved.

Manuscript submitted May 11, 1998; revised manuscript received April 9, 1999.

It is well known that the impedance response of a material with component partial pressure, the questions posed deserve careful con-
(i) two different mobile speciesb and (ii) contacted with ideal selec- sideration.
tively blocking electrodes consists of a high frequency semicircle The second problem we tackle is the impact of nonideal elec-
reflecting the dielectric relaxation of the bulk followed by Warburg trodes. In spite of possible special precautions (glass encapsulation,
diffusion (Fig. 1a). In fact, in aqueous electrochemistry this is fre- use of inert metals or pure ionic conductors) it is still probable, if not
quently the case and spectra of type (a) (Fig. 1a) are observed. In the rule, that the electrode will exhibit a leak for carriers which are
contrast to liquid electrolytes, in the solid state often only one ionic supposed to be blocked, and/or a small charge-transfer resistance for
species is mobile. On the other hand, we also have electronic carri- carriers which are supposed to cross the boundary reversibly. In the
ers to consider.c Since also for this charge carrier combination selec- absence of geometrical complications18-20 (which we also assume to
tively blocking electrodes can be prepared2,3 (using, e.g., Pt or C be absent in this paper) the electrode nonidealities were theoretical-
which is blocking for ions and reversible for electronsd) one expects ly studied by Macdonald 9,11 using the continuum model (Nernst-
to measure responses of type (a), just as in liquid electrochemistry.
Indeed, Warburg diffusion is observed,4,5 but not at all as a rule. As
demonstrated, e.g., in studies of materials with low electronic con-
ductivity,6 measured spectra are often found to be of the type (b),
which is exactly the behavior expected when neglecting the elec-
trons a priori. On the other hand, we can conceive each solid as a
mixed conductor. If we let, in the well-known Warburg ansatz,7
seon/sion approach 0, the predicted spectrum is still of type (a) (458
rise) but of course with a very extended low frequency part. How to
explain this apparent inconsistency, or more rigorously formulated,
how to describe the transition from type (a) to type (b)? Which are
the material parameters that determine the crossover?
Although very detailed treatments of two-carrier systems exist in
liquid and solid electrochemistry7-14 as well as in the field of mixed
conducting polymers,15,16 and although deviations from the 458
Warburg response have been predicted for thin systems,10,12 the
above questions remained, to the best of our knowledge, substantial-
ly unanswered. Since the impedance spectroscopy of mixed conduc-
tors with partially blocking electrodes is one of the key techniques
for discriminating between the ionic and electronic conductivity and
for determining the component chemical diffusion coefficient,4,17
and since in the solid state the electron concentration can be varied
over several orders of magnitude by simply changing temperature of

* Electrochemical Society Active Member.

a On leave from the National Institute of Chemistry, SI-1000 Ljubljana, Slovenia.
b A rigorous discussion of what is meant by “two different mobile species” is diffi-

cult and in the literature often avoided or presented incorrectly. Under the assumption of
local equilibrium the term “two different mobile species” designates more precisely two
“conservative ensembles”1 of carriers and traps, which, by definition, have no external
sources or sinks. If local equilibrium does not prevail, one has to consider the detailed
kinetics of all defects (carriers and traps).
c Although ionic and electronic carriers in general may recombine to form a neutral
Figure 1. Complex plane impedance of a two-carrier system with selective-
component within the sample we exclude this complication here but allow for the pres-
ence of completely dissociated (i.e., shallow), immobile impurities.
ly blocking electrode: (a) idealized spectra exhibit finite length Warburg dif-
d Whenever we want to be specific, we discuss the electrodes being blocking for ions fusion (cell with liquid electrolytes and solid state mixed conductors5), (b)
and reversible for electrons. Addressing the opposite case only the index ion and eon idealized spectra of nearly single carrier system (solid electrolyte with very
have to be interchanged in the corresponding equations. low electronic conductivity6).

Downloaded on 2014-07-16 to IP 129.130.252.222 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 4184 Journal of The Electrochemical Society, 146 (11) 4183-4188 (1999)
S0013-4651(98)05-026-5 CCC: $7.00 © The Electrochemical Society, Inc.

Planck-Poisson equations). Although the model has been studied in accuracy and validity of the much simpler closed form approxima-
depth, here we cite the author of Ref. 11, “all its implications have tion Zapprox, which we develop in the next section.
not yet been fully explored.”
In this paper we present a closed form approximation for the im- A Novel Equivalent Circuit Model
pedance of mixed conductors which is almost as general as the con- The circuit.—We have found that for macroscopically thick sam-
tinuum model but sufficiently transparent to answer the above ques- ples (L >> LD) the continuum impedance model, Zexact(v), can be
tions. The approximation is derived on the basis of a novel equiva- well approximated by the equivalent circuit shown in Fig. 2a. At
lent circuit model, which, in the limit of macroscopically thick sam- high frequencies the transport is described by ionic and electronic
ples, approaches the continuum model in a very broad range of ma- resistances, Rion and Reon, and the dielectric capacitance of the sam-
terial parameters. In contrast to the equivalent circuits existing in the ple C` ; ee0 A/L. At low frequencies the chemical capacitance Cchem
literature,13,14 both types of impedance responses (Fig. 1a and b) as is of importance. It describes the capability of the sample (MX) to
well as the mixed cases which also include electrode effects can eas- store chemical energy, i.e., the changes of the component chemical
ily be predicted from it. Perhaps more importantly, the materials potential, mX, due to the stoichiometry variations (i.e., variations of
parameters on which the type of spectra depend critically, can be the X content in MX). Under local electroneutrality, the expression
identified and expressed in a closed form. To remain on the safe side, for Cchem ; (zionzeonq)2AL(∂mX /∂cX)21 as proposed by Maier17 can
we explore the continuum model as well and use it for the verifica- be used. In the case of dilute charge carriers the derivative ∂mX /∂cX
tion of analytical approximations throughout the paper. can easily be evaluated22 which in the simplest situation (without
traps) yields
Continuum Model
21
The continuum model9,10,13 as we use it in this paper, is based on q2  1 1 
Cchem 5 AL  2 eq 1 2 eq  [5]
the following assumptions: a slab of laterally homogeneous material kT  z ion cion zeon ceon 
sandwiched between two identical electrodes positioned at x 5 0 and
x 5 L, with a geometric area of A, is considered. We assume two In contrast to the electrostatic capacitances, Cchem is proportional to
charge carriers (of concentrations cj) both being mobile but having no the sample volume. As seen from Eq. 5, Cchem can be further decom-
sources or sinks within the material (e.g., a trap, that is to say a free j
posed into two chemical capacitances, C chem 5 q2/kT ALz 2j c eq
j , con-
mixed conductor with one ionic and one electronic charge carrier). nected in series. A treatment of chemical capacitance which applies
The continuity equations for the carrier flux densities, Jj, then read also to space-charge regions will be given elsewhere.23
The transport across the mixed conductor/electrode interfaces is
∂Jj ∂c j
52 [1] modeled by a parallel combination of the charge-transfer resistance
∂x ∂t R⊥j and the interfacial capacitance C ⊥j , which are both defined for
each charge carrier individually. On the equivalent circuit level these
The symbol t denotes time. We further assume that the flux densities interfacial parameters are independent fit parameters and may take
are composed of drift and diffusion terms according to Nernst- account of different realistic complications, e.g., finite ion size ef-
Planck transport equation fects, equilibrium space-charge layers adjacent to the electrodes, ad-
∂c j s j ∂f sorption of intermediates. On the other hand, in the situation assumed
J j 5 2Dj 2 ⋅ [2] by the continuum model, R⊥j and C⊥j , are well defined. Interfacial
∂x z jq ∂ x

The symbol Dj denotes time (i.e., frequency v) invariant self-diffu-

sion coefficient which is related to the carrier mobility, uj, via the
Nernst-Einstein relation. The electrical conductivity of the carrier
with the charge number zj is denoted by sj, whereas the symbols q
and f denote the absolute value of the electronic charge and the elec-
tric potential, respectively. The coupling between carriers is provid-
ed by Poisson’s equation

∂ 2f
∑ zc
q
52 j j [3]
∂x 2 ee 0
j

The symbols e0 and e denote the permittivity of free space and, in

our model, the frequency independent relative dielectric constant.
We assume elementary electrode kinetics which we take into
account by Change-Jaffe boundary conditions21
Jj 5 7kj(cj 2 c jeq) [4]
The minus sign applies at x 5 0 and the plus sign at x 5 L. The
superscript “eq” refers to the equilibrium value. The rate constants kj
are vanishing in the limit of completely blocking electrodes and are
very high in the limit of completely reversible electrodes (of course,
they have different values for different carriers). We neglect any
equilibrium space-charge layers adjacent to the electrodes. Nonethe-
less the Debye length LD ; !w ee0kBT/(q2w ojz 2j cjeq) is a decisive para-
meter describing the screening of nonequilibrium electrode charge.
Figure 2. (a) Proposed equivalent circuit for a mixed conductor with thick-
The symbol kB stands for the Boltzmann constant and T for the ness much greater than Debye length. The quantities, Rion, Reon, and Cchem,
temperature. which all refer to the overall sample, are weighted by 1/n, n being the num-
Since the solution of Eq. 1-4, i.e., the impedance Zexact(v; cj, uj, ber of transmission line unit cells. (b) Equivalent circuit of the high frequen-
zj, kj, L), is very involved,9 we do not write it here explicitly. As cy impedance Z` defined by Eq. 8a. (c) Equivalent circuit of the low fre-
already mentioned in the introduction we use it mostly to control the quency impedance Z0 (see Eq. 8b).

Downloaded on 2014-07-16 to IP 129.130.252.222 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 146 (11) 4183-4188 (1999) 4185
S0013-4651(98)05-026-5 CCC: $7.00 © The Electrochemical Society, Inc.

resistances are inversely proportional to microscopic rate constants, 1 1 1

kj, which appear in Chang-Jaffe boundary conditions, according to ; ⊥
1 ⊥ [8b]
Z0 Rion 1 2 Z ion Reon 1 2 Z eon
R⊥j 5 kBT/(q2z j2c eq 21 ⊥
j )kj . The interfacial capacitances, C j , depend on
the individual screening lengths, L D, j, stemming from the jth carrier respectively. Graphically, Z` and Z0, are depicted in Fig. 2bc. We see
that the fluxes can only switch between electronic and ionic rails at
A ee 0 kB T
C j⊥ ; ee 0 LD, j ; [6] high frequencies.
LD, j q 2 z 2j c eq
j The approximative impedance Zapprox depends on the same set of
material parameters as Zexact. Figure 3 compares Zapprox and Zexact
The definition of interfacial capacitances via individual Debye calculated for two samples of different thicknesses (2LD and 20LD).
lengths is necessary in order to describe situations in which charge As we see, the approximation in the case of the thicker sample is
carrier densities differ significantly (a rather usual situation in the very good. We should stress that the slight deviations of Zapprox from
solid state). Even though approximate (Debye length in Eq. 6 refers Zexact are not due to discretization (finite number of elements in the
to a single-carrier system) this description is quite precise at least for equivalent circuit), but due to an ad hoc division of the sample into
cion >> ceon or cion << ceon case. The maximum error occurs for the electroneutral bulk and charged interfacial layers.
cion < ceon but does not exceed 41%. There it is expected that the
total interfacial capacitance will be ~ L21 Ideal selectively blocking electrodes.— If the electrodes are ide-
D which differs by a factor
of !w2 (we assume an univalent case for simplicity) from the individ- ally blocking for one of the carriers (e.g., ions) and completely
ual Debye lengths (Eq. 6). We comment on this point again in the reversible for the other one (i.e., electrons) Eq. 7 simplifies to
section on Comparison with the continuum model. k 
The impedance of the proposed equivalent circuit shown in Zapprox  ion r 0 5 R`
Fig. 2a can be derived in a straightforward manner. If we skip the  keon 
contribution of C` for the purpose of clarity, define interfacial
impedances Z ⊥j ; R⊥j /(1 1 ivR⊥j C ⊥j ), and express the product ivL2
~ ( Reon 2 R` ) tanh
(Rion 1 Reon)Cchem by the chemical diffusion coefficient D 5 4D
L /(Rion 1 Reon)Cchem, a simple exercise yields
2 1 [9]
ivL2 1 ⊥ ivL2
Zapprox(v) 5 Z` 1 1 iv ( Rion 1 Reon ) Cion tanh
4D 2 4D
 Rion 1 Reon  ivL2 The symbol R` ; RionReon/(Rion 1 Reon) denotes the resistance of
1 1 ⊥ ⊥ 
tanh the high frequency semicircle. Further simplification for low fre-
 2( Z ion 1 Z eon )  4D quencies yields
[7]
ivL2 Rion 1 Reon ivL2 k 
1 tanh R 2 R`
4D ⊥ ⊥
2( Z ion 1 Z eon ) 4D Zapprox  ion r 0, v r 0 < R` 1 eon [10]
 keon  1 1 ivt low
The high and low frequency impedances, Z` and Z0, are defined as
corresponding to a semicircle in a complex plane. Since Eq. 10
Rion Reon Z⊥ Z⊥ strictly applies only in the limit of low frequencies, the inverse of the
Z` ; 1 2 ⊥ ion eon⊥ [8a] relaxation time, tlow ; (Rion 1 Rion)(1/2 C⊥ion 1 1/12 Cchem), is just an
Rion 1 Reon Z ion 1 Z eon approximation for the peak frequency of the spectra. At any rate, the
relaxation time, tlow is determined by the chemical, Cchem, as well as
by the electrostatic capacitance, C⊥ion. The numerical factor of one-
half appearing in the front of C⊥ion is due to “serial” connection of the
two electrodes, while the factor one-twelfth reflects a nonuniform
chemical polarization of the sample (e.g., in the middle of the sam-
ple the polarization is completely vanishing). As we see from Eq. 10,
the nature of the low frequency response (chemical or electrostatic
relaxation) depends on the ratio C⊥ion/Cchem. Indeed, in extreme situ-
ations Eq. 9 simplifies

⊥ k 
Cion >> Cchem : Zapprox  ion r 0 < R`
 eon
k 
Reon 2 R`
1 [11a]
1 ⊥
1 1 iv ( Rion 1 Reon )Cion
2

⊥ k 
Cion << Cchem : Zapprox  ion r 0 < R`
 keon 

ivL2
tanh
4D
1 ( Reon 2 R` ) [11b]
ivL2
4D
Figure 3. Comparison of Zexact (solid lines) with Zexact (symbols). For clari-
ty only real parts are shown in Bode plot. Parameters used in calculations: yielding a perfect low frequency semicircle (Eq. 11a) or the well-
ceon/cion 5 1026, ueon/uion 5 103, kion r 0, keon r `, zion 5 1 5 2zeon. The known finite length Warburg impedance (Eq. 11b). There is a funda-
relative error for L 5 2LD case if D 5 0.8. For the thicker sample L 5 20LD mental difference between both extremes. In the Warburg case
it is much lower D 5 0.1. (Eq. 11b) the relaxation is governed by the chemical capacitance and

Downloaded on 2014-07-16 to IP 129.130.252.222 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 4186 Journal of The Electrochemical Society, 146 (11) 4183-4188 (1999)
S0013-4651(98)05-026-5 CCC: $7.00 © The Electrochemical Society, Inc.

the corresponding relaxation time is proportional to L2 (Cchem ~ L). a frequency dependent e or s does not always have to be assumed to
In the opposite case (Eq. 11a) the relevant capacitance C⊥ion is inde- explain distorted and depressed semicircles (examples of the contin-
pendent of the sample size and the relaxation time is proportional to uum model prediction of the frequency dispersion are given by Mac-
L. In this case an electrostatic relaxation occurs coherently within donald10,11). Figure 5a shows the effects due to deviation from elec-
the whole sample. Thus no frequency dispersion takes place (a per- tronic reversibility and Fig. 5b shows the effects due to deviation
fect semicircle). from an ideal blocking of ions. In the former case the dc resistance
Impedance spectra for both extreme cases as well as for one sit- (the total electronic resistance) increases. In the latter case ionic
uation in between are shown in Fig. 4. For this figure and also for all leakage lowers the dc resistance.
forthcoming figures we always calculated Zexact and Zapprox although The situation is more peculiar in case that Cchem >> C⊥ion. As
only one of them is displayed. The reason is that from the complex shown in Fig. 6a nonidealities with respect to electronic reversibility
plane plots it is difficult to distinguish between Zexact and Zapprox if lead to an additional semicircle which is squeezed between the high
the “relative error” D (see Eq. 12), which quantifies the deviation of frequency bulk response and the low frequency Warburg diffusion as
Zapprox from Zexact, is below 0.01 (already for D 5 0.1 the difference assumed by Maier.25 A very different situation appears if the elec-
is not significant even in Bode plots, see Fig. 3). The actual “relative trode is not entirely blocking for the nominally “blocked” carrier
error” D is always given in the figure captions. (Fig. 6b). In this case no additional semicircle, but a distortion of the
The impedances in Fig. 4 were calculated for different ratios of low frequency part of the spectra occurs, i.e., a kind of “depression”
ionic and electronic charge carrier concentrations while keeping the of the ideal Warburg response. These effects are in agreement with
other parameters fixed (ueon/uion 5 1, zeon/zion 5 21, and L/LD 5 the continuum model results calculated for binary electrolytes.11
2?104). The low frequency part of the spectra is only shown for dia- It is worth mentioning here, that the approximation for the low
grammatic reasons. At the low frequency side of the spectra the frequency relaxation time, tlow, as derived for ideal selectively
impedance starts with the dc cell resistance, which is the resistance blocking electrodes (Eq. 10) does not apply in general. In other
due to the unblocked carriers, in our case Reon. At high frequencies, words, in the case of nonideal electrodes tlow is not necessarily pro-
the blocking capacitance, C⊥ion, is short-circuited so that both carri- portional to L2. Nevertheless, one can derive from Eq. 7 analytical
ers can conduct. The resistance is given by the parallel combination approximations for some specific cases, e.g., if the electrode is com-
of ionic and electronic resistances R21 21 21
` ; Rion 1 Reon. At equal con- pletely reversible for electrons but not entirely blocking for ions (sit-
centrations of both carriers the chemical capacitance is much higher uations shown in Fig. 5b and 6b). Then tlow < [(Rion 1 Reon)21 1
than the electrode one (in our case Cchem 5 7?103C⊥ion ).Thus the (R⊥ion)21]21(1/2C⊥ion 1 1/12Cchem).
finite length Warburg behavior is obtained, as expected (Eq. 11b).
With lowering the electron concentration, the chemical capacitance Comparison with the continuum model.—The validity range of
decreases (Eq. 5), while the electrode capacitance remains nearly Zapprox can be verified numerically by means of the continuum
unaffected (Eq. 6). For, Cchem 5 20C⊥ion (dashed curve), the Warburg
response is already distorted while for Cchem 5 2?1022C⊥ion a perfect
semicircle appears (dash-dotted) curve, see Eq. 11a).
Real electrodes.—In reality it is rather difficult to prepare ideal
selectively blocking electrodes. One may expect that the electrodes
nominally “reversible” for electrons, will still represent a certain
interfacial resistance while the (same) electrodes “blocking” for
ions, will allow for a certain ionic leakage. Although both effects are
comprised by Eq. 7, we consider them separately in the following.
We first assume that Cchem 5 2C⊥ion (Fig. 5). First, we see that in all
the spectra shown in Fig. 5 the low frequency part is not an ideal
semicircle. The reason is a rather low chemical capacitance (when
compared to C⊥ion), which was chosen on purpose, just to show that

Figure 5. Influence of imperfect electrodes on impedance spectra of the sam-

Figure 4. The low frequency response of a two-carrier system with ideal ple with comparable electrode and chemical capacitances (Cchem 5 2C ion ⊥ ).
selectively blocking electrodes according to Eq. 7. The transition from a typ- (a) Impact of poor electronic reversibility (electrode is perfectly blocking for
ical mixed conductor spectra (solid line) to the spectra of a single carrier sys- ions). (b) Impact of imperfect blocking of ions (electrode is perfectly
tem (dashed-dot line) is obtained by varying the electronic carrier density. reversible for electrons). Impedances were calculated according to Eq. 7.
Impedances are normalized such that the lowest and the highest frequency Parameters used in calculations: L 5 2?104LD, ceon/cion 5 1024, ueon/uion 5
points of the three curves coincide. Other parameters used in calculations: 103, zion 5 1 5 2zeon. In Fig. 5 Zapprox is shown. Comparison with Zexact
L 5 2?104LD, ueon/uion 5 10, zion 5 1 5 2zeon. Comparison with Zexact yields showed the best accuracy in (keon 5 `) cases (D < 1024). In other situations
the relative error D 5 1024 2 6?1024. the error was larger (D < 1023).

Downloaded on 2014-07-16 to IP 129.130.252.222 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 146 (11) 4183-4188 (1999) 4187
S0013-4651(98)05-026-5 CCC: $7.00 © The Electrochemical Society, Inc.

(ceon/cion, ueon/uion). The high frequency response at (1025, 1025)

and the low frequency response at (105, 105) are only shown for dia-
grammatic reasons (in the impedance representation they are actual-
ly vanishing). Along the border line between the dashed and
undashed region, the deviation of the finite Warburg response
(Eq. 11b) from the continuum model is D 5 0.33. The criterion
based on the ratio 1/2C⊥ion/1/12Cchem results in the vertical borderline
located at ceon/cion 5 1023.5 (not shown in the plot). The difference
between both borderlines stems from the fact that the
1/ C⊥ /1/ C
2 ion 12 chem criterion was defined on the basis of the low fre-
quency approximation (Eq. 10) while the relative error (Eq. 12) takes
into account the entire frequency range.
In Fig. 7b the influence of interfacial rate constants is examined.
In this case the concentration and mobilities of both carriers are
fixed (we choose the parameter values which refer to the central
point in Fig. 7a, i.e., ceon/cion 5 1 and ueon/uion 5 1) while the rate

Figure 6. Influence of nonideal electrodes on impedance spectra of a mixed

conductor exhibiting Warburg diffusion (Cchem >> C ion ⊥ ). (a) Impact of poor
electronic reversibility (electrode is perfectly blocking for ions). (b) Impact
of imperfect blocking of ions (electrode is perfectly reversible for electrons).
Impedances were calculated according to Eq. 7. Parameters used in calcula-
tions: L 5 2?107LD , ceon/cion 5 1024, ueon/uion 5 103, zion 5 1 5 2zeon. Error
analysis yields: D < 6?1024 2 4?1023.

model impedance Zexact. We quantify the mismatch between Zapprox

and Zexact as the relative error

v i 5N  Re( Z 2
exact 2 Zapprox ) 
∑
1
D;  
N  Re( Z exact ) 
v i =1 
1
 Im( Z exact 2 Zapprox )  
2 2
1   [12]
 Im( Z exact )  

with the weighting denominators also used in the fitting of equiva-
lent circuit models to measured impedance data.26 The summation
runs over the frequencies 1/100 tlow21 # v # 100t 21 (t ; R C ),
i ` ` ` `
equidistantly in logarithmic scale.
Figure 7a shows the results for a materials parameter window
extending over ten orders of magnitude in the carrier concentration
ratio (abscissa) and over ten orders of magnitude in the carrier
mobility ratio (ordinate). The sample thickness (L 5 2?104LD) and
the interfacial rate parameters (kion r 0, keon r `) are kept constant
over the entire range. The upper shadowed triangle of the window
refers to materials with dominant electronic conductivity and the
lower triangle to materials with dominant ionic conductivity. The
finite Warburg impedance (Eq. 11b) applies in the dashed region
only; there the agreement with the continuum model is good (D <
0.33, in some parts of the dashed region even D < 1026). On the Figure 7. Validity ranges and characteristics of impedance spectra for differ-
other hand, the closed form approximation which we propose in this ent materials parameter windows. (a) Carrier concentrations and mobilities are
paper Eq. 7 (C` was added to Eq. 7 in parallel, before the compari- varied, (b) interfacial rate constants are varied. In both cases the calculations
son was made) approximates the continuum model over the entire were done for sample thickness L 5 2?104LD and univalent charge carriers zion
5 1 5 2zeon. The insets show impedance spectra for different (ceon/cion,
range with D < 0.01. ueon/uion) and different (kion, keon). Two insets (Fig. 7a and b) with dotted back-
The insets in Fig. 7a show schematically the impedance spectra ground (Fig. 7a and b) refer to the same (overlapping) parameter set. The War-
pertaining to different parts of the materials parameter window bury part of the spectra has been emphasized (bold line) for clarity.

Downloaded on 2014-07-16 to IP 129.130.252.222 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 4188 Journal of The Electrochemical Society, 146 (11) 4183-4188 (1999)
S0013-4651(98)05-026-5 CCC: $7.00 © The Electrochemical Society, Inc.

parameters are varied over four orders of magnitude. Here the agree- for the “unblocked” carrier an additional semicircle in the complex
ment between Zapprox and Zexact was not as good as in the case above. plane (Fig. 6a) appears while insufficient blocking of the “blocked”
The largest error, D 5 0.24, appears in the totally blocking situation carrier leads to a distorted Warburg response (Fig. 6b).
(keon 5 1022, kion 5 1022), and the smallest D 5 0.02 in the selec-
tively blocking cases, e.g., around coordinates (1062, 1072). The The Max-Planck Institut für Festkörperforschung assisted in meeting the
detailed analysis shows that the error is due to electrode capacitance publication costs of this article.
(Eq. 6), which appears to be too high by a factor of !w 2. That is exact-
ly what we expected for a cion < ceon case (see the section on The References
circuit). In view of the applicability range of Zapprrox which spans 1. J. Maier and G. Schwitzgebel, Phys. Status Solidi B, 113, 535 (1982).
over many orders of magnitudes for a variety of materials parame- 2. M. H. Hebb, J. Chem. Phys., 20, 185 (1952).
ters (carrier concentrations, mobilities, and the interfacial rate con- 3. C. Wagner, Z. Electrochem., 60, 4 (1956).
stants), this inaccuracy is still tolerable. 4. I. Yokota, J. Phys. Soc. Jpn., 16, 2213 (1961).
5. R. Andreaus and W. Sitte, J. Electrochem. Soc., 144, 1040 (1997).
The insets in Fig. 7b show the calculated impedance spectra (the 6. W. I. Archer and R. D. Armstrong, Specialist Periodical Report, p. 157, The Chem-
scale is the same for all insets) referring to different (kion, keon). The ical Society, London (1980).
insets located along the diagonal (the dashed line) refer to equal rate 7. D. R. Franceschetti, J. Macdonald, and R. P. Buck, J. Electrochem. Soc., 138, 1368
constants for ionic and electronic carriers (electrodes are not selec- (1991).
8. J. E. B. Randles, Discuss. Faraday Soc., 1, 11 (1947).
tive). As expected no Warburg diffusion is observed in this regime. 9. J. R. Macdonald, J. Chem. Phys., 58, 4982 (1973).
The electrodes are most selective at the opposite corners (1062, 10. J. R. Macdonald, J. Chem. Phys., 61, 3977 (1974).
1072). Let us start from one of those points, e.g., at 1022, 102). Mov- 11. J. R. Macdonald, Electrochim. Acta, 37, 1007 (1992).
ing up, the ionic blocking weakens yielding a distorted Warburg 12. S. Gottesfeld and S. W. Feldberg, J. Electroanal. Chem., 194, 1 (1985).
13. R. P. Buck, J. Electroanal. Chem., 210, 1 (1986).
response (see also Fig. 6b), while moving to the left the electronic re- 14. D. R. Franceschetti, Solid State Ionics, 70/71, 542 (1994).
versibility weakens yielding additional semicircle (see also Fig. 6a). 15. W. J. Albery, C. M. Elliott, and A. R. Mount, J. Electroanal. Chem., 288, 15 (1990).
16. M. A. Vorotyntsev, L. I. Daikhin, and M. D. Levi, J. Electroanal. Chem., 364, 37
Conclusions (1994).
A general closed form approximation (Eq. 7) for the impedance 17. J. Maier, Solid State Phenom., 39-40, 35 (1994).
of mixed conductors is presented. It corresponds to an easily inter- 18. B. Sapoval, J.-N. Chazalviel, and J. Peyriere, Phys. Rev. A, 38, 5867 (1988).
19. L. Nyikos and T. Pajkossy, Electrochim. Acta, 35, 1567 (1990).
pretable equivalent circuit (Fig. 2). The decisive materials parameter 20. J. Fleig and J. Maier, J. Electroceram., 1:1, 73 (1997).
which determine whether or not the impedance spectrum is of War- 21. H.-C. Chang and G. Jaffe, J. Chem. Phys., 20, 1071 (1952).
burg type is the sample’s chemical capacitance. Only if it is large 22. J. Maier, J. Am. Ceram. Soc., 76, 1212 (1993).
when compared to the electrode capacitance can the Warburg diffu- 23. J. Jamnik and J. Maier, In preparation.
24. J. Jamnik and J. Maier, Ber. Bunsenges. Phys. Chem., 101, 23 (1997).
sion appear. 25. J. Maier, Z. Phys. Chemie Neue Folge, 140, 191 (1984).
The impact of nonideal electrodes on Warburg diffusion can be 26. J. R. Macdonald, in Impedance Spectroscopy-Emphasizing Solid Materials and
formulated as follows: if the electrodes are not perfectly reversible Systems, J. R. Macdonald, Editor, p. 183, Wiley, New York (1987).

Downloaded on 2014-07-16 to IP 129.130.252.222 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).