Electrochimica Acta 51 (2006) 4505–4515

Interpretation of EIS data from accelerated exposure of coated metals

based on modeling of coating physical properties
B.R. Hinderliter, S.G. Croll, D.E. Tallman, Q. Su, G.P. Bierwagen ∗
Department of Coatings and Polymeric Materials, North Dakota State University, Fargo, ND 58105, United States
Received 23 June 2005; received in revised form 16 December 2005; accepted 20 December 2005
Available online 6 March 2006

Abstract
Electrochemical studies on the accelerated weathering of corrosion protective coatings systems in this laboratory have yielded time series data
in the low frequency portion of the spectrum that can be fit by a simple exponential decay function in time. A description of how this can be
empirically used for lifetime prediction for coatings has already been presented. Attempts have been made to use other parts of the EIS spectrum for
analyzing coating performance in accelerated exposure, but these have not yet yielded a predictive model for the effects of exposure on the structure
and performance of organic coatings. First, a simple model which considers only loss of coating thickness by UV-induced ablation is applied to
the interpretation of EIS time series results from coatings systems undergoing accelerated weathering. Next, more complete models which include
a diffusion front of electrolyte that moves with either Fickian or Type II diffusion kinetics are applied to the same problem. Incorporating these
models for degradation as a simple RC equivalent circuit representation of a coating gives predictions of EIS performance that have reasonable
agreement with experimental data. Further, using a two layer representation of coating changes due to water intrusion and known effective medium
theories for dielectric and resistivity changes due to water intrusion allow prediction of the changes in EIS spectra seen in immersion or other
accelerated testing of coating films. These results provide further evidence of the power of modeling physical processes in interpreting EIS data
and predicting coating protective lifetimes.
© 2006 Elsevier Ltd. All rights reserved.

Keywords: Electrochemical impedance spectroscopy (EIS); Organic coatings; Modeling; Corrosion protection; Accelerated testing; Lifetime prediction

1. Introduction The ability to predict a coating’s behavior gives confidence

Electrochemical impedance spectroscopy (EIS) is commonly
used to examine and rank the protective ability of corrosion
protective organic coatings [1–10]. The method, in general, is
used in a qualitative or semi-quantitative manner, due to large
variances between replicates and the difficulty in producing
identical coating samples. Equivalent circuit models using pas-
sive electrical engineering and physics circuit elements have
commonly been used to help interpret EIS studies of coating
systems [11–14]. Data analysis procedures use nonlinear, com-
plex least squares fit of these equivalent circuit models to the EIS
data. Apparent deviations of the circuit element fitted responses
have produced continuing debate over their physical interpreta-
tion [15–17].
∗ Corresponding author. Tel.: +1 701 231 8294.
E-mail address: gordon.bierwagen@nsdu.edu (G.P. Bierwagen).
that all relevant and necessary details of the system’s physi-
cal makeup are included. Understanding those specific coating
properties that impact the measurement, and their quantitative
importance, allows a more scientific design of coatings for
particular applications, improved predictive ability, and better
understanding of coating failure when it occurs. The inability to
predict, and thus interpret, in a physically reasonable and quanti-
tative way the EIS spectrum during coating degradation assures
that a great deal of work is left to be done, although some work
has been attempted in using electrochemical methods to model
cyclic accelerated testing of coatings [18].
One of the problems for a coating scientist in using standard
EIS data fitting and modeling techniques is that the tools of pre-
dicting and fitting the data are based solely on equivalent circuit
elements, and not based explicitly on solid state physics, elec-
trical engineering or metallurgical engineering. Two especially
informative references that discuss coated metal characteriza-
tion by EIS and the modeling of such data by standard circuit

0013-4686/$ – see front matter © 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2005.12.047
4506 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515

elements are in Refs. [19,20]. It can be very informative to exam-

ine coatings and their changes by EIS using analysis methods
that focus solely on the physical properties of the coating as
described by the simple Randles circuit description of a coating
[21]. An intact coating is frequently described by an equivalent
circuit that consists of a parallel plate capacitor with the elec-
trolyte solution as one electrode, the coating as the dielectric,
and the metal substrate as the second electrode. The Randles
equivalent circuit describes this simple behavior and consists of
a solution resistance in series with a parallel combination of a
capacitor and resistor. The inputs for describing the coating are
its resistive and capacitive properties.
Based on simple continuum theory [22], the capacitance and
resistance of a smooth coating are given as functions of the size
of the material and intrinsic properties. The capacitance, C, for
a parallel plate capacitor is given by
ε0 εR A
C=
d The second model is associated with a coating where water
where ε0 is the permittivity of free space (8.8542 ×
10−12 C2 /J/m), εR the coating dielectric constant, A the area of
the parallel plate capacitor, and d is the coating thickness. The
coating resistance per unit area is given by
ρd
R=
A
where R is the resistance, ρ the electrical resistivity (a mate-
rial constant), d the coating thickness, and A is the area. We
will describe and model data based solely on models for R and
C in which changes are only in the film thickness, film resis-
tivity or film dielectric constant. These variables will not be
treated as fitting parameters for general R and C circuit elements,
but as coating variables, based on the knowledge developed by
coating and polymer scientists for changes in coatings due to
well-described processes such as film thickness loss due to UV
degradation, film swelling, or due to aqueous diffusion. One
can describe multi-layer coatings as Randles circuits in series
to account for gradients in water and ion concentrations. The
present model homogenizes the water and coating into a single
material as the diffusion front advances.
2. Impedance analysis based on two models for coating
performance
Fig. 1. The diffusion of water into the coating changes the coating in the
diffusion-affected zone while the deeper level of the coating is nominally
unchanged and retains most of its original physical properties (labeled pristine
coating).
(1)
advances by diffusion into the coating. This model presented
in this paper will be called the diffusion-affected model, and
describes the changes in the coating layer(s) by at least one
water-swollen layer (see Fig. 1) in which Fickian or Type II dif-
fusion processes control the thickness of the region as a function
of time [27–29]. We will describe the coating changes for the
(2) second model by a Randles-like circuit with a water-polymer
composite resistance (see Fig. 2) in series with another Randles
circuit representing the unaffected portion of the coating.
2.1. Changes in coating behavior at low frequencies
The ablation model addresses the impact of the changes in a
coating’s effective thickness as a function of weathering assum-
ing the coating material is completely removed (no change in
material properties). The diffusion-affected model employs the
assumption that the resistance of the water affected coating is
negligible in the low frequency regime compared to the pristine
coating.
In previous publications we have noted that a log-linear decay
of coating resistance (equivalent to low frequency impedance
modulus) with time is observed in measurements of various
types of coating in accelerated exposure [10,17]. This behavior
can be modeled as a coating degrading only by surface ablation
due to exterior UV-exposure. Additionally, if the resistance of

The models for coating behavior applied in this paper are

based on the Randles circuit, a representation most often used
for a nominally intact coating [23]. Two models are investigated.
The first model is associated with a topcoat over a primer where
changes occur only in thickness by swelling or ablation. The
ablation model describes the effects of changes in a coating’s
effective thickness as a function of weathering, as a system dom-
inated by bulk changes of the coating. Monte Carlo simulations
and experimental measurements confirm that the average thick-
ness (or weight per area) of the topcoat undergoing mainly UV
degradation induced ablation decreases roughly linearly with Fig. 2. The regions of the coating are separated into components that model the
time [24–26]. diffusion-affected zone and pristine coating.
 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515 4507

the coating after water intrudes is very small compared to the resistance and capacitance:
remaining intact coating, and the low frequency response is dom- 
 
inated by the remaining intact coating, again we can reproduce 1 ρd ε0 εR A 2
log(|Z|) = log(R) − log 1 + ω
the log-linear decay of the low frequency impedance modulus 2 A d
[10,17]. These two proposed models will be tested for the full
1
EIS frequency range to see which model produces the most rea- = log(R) − log(1 + (ωε0 εR ρ)2 ) (9)
sonable EIS spectra. 2
In the following derivation, considering the solution resis- Using the standard expansion ln(1 + x) = x − 1/2x2 + O(x3 ) for
tance (generally small) detracts from the focus on the coating small x, one gets the form for log|Z| given in Eq. (10). (Also one
components, and therefore will be neglected. notes that that the RC factor loses dependence of the size, either
The impedance of a capacitor in parallel with a resistor is area or depth, of the coating.) For low enough frequencies using
1 R(1 − jωCR) only the first term in the expansion:
Z= = (3)

1
+ jωC 1 + (ωCR)2 ∼ ρd 1
R log(|Z|) = log = (ωε0 εR ρ)2 (10)
√ A 2 ln(10)
where, j = −1, and ω is the angular frequency. The phase
angle, θ, is given by The impedance is a weak function of the geometry at low fre-
quencies and a Bode plot of the impedance can reveal a small
θ = arctan(−ωCR) (4) slope at low frequencies (0.01–0.1 Hz), depending on the resis-
tivity and relative dielectric constant, diminishing at lower fre-
If the magnitude of the impedance is taken, then the log of both quencies when the second term in Eq. (10) becomes negligible.
sides of the equation gives the functional form that is visualized Monte Carlo predictions and/or diffusion front advance may
in a Bode plot: be incorporated into this low frequency approximation. To a

 first order, the loss of coating thickness (or weight per unit
R 2 2 2 1/2
log(|Z|) = log (1 + ω C R ) (5) area) [30,31] is linear with exposure time. Thus, for the abla-
1 + ω 2 C 2 R2 tion film thickness loss at a rate K1 (in units of thickness/time),
or we can write a simple expression for film thickness as function
of exposure time. Additionally, Monte Carlo models and statis-
log(|Z|) = log(R) − 21 log(1 + ω2 R2 C2 ) (6) tical calculations predict the roughness of the coating to vary
with the square root of exposure time in most cases. Then the
2.1.1. Continuum approximation based on homogeneous “average” thickness as a function of exposure time is then given
materials by Eq. (11):
Applying the continuum expression for the capacitance (Eq. √
d(t) = d0 − (K1 t + K2 t) (11)
(1)) and resistance (Eq. (2)) in Eq. (6) gives

 
  where d(t) is the effective thickness (or the portion of the coat-
ρd 1 ρd ε0 εR A 2 ing that has not been infiltrated by water), d0 the initial coating
log(|Z|) = log − log 1 + ω
A 2 A d thickness, and t is the time. K1 and K2 are the proportionality

 constants. These constants depend on the polymer, and its ero-
ρd 1 sion processes during weathering. K2 has units of (depth lost per
= log − log(1 + (ωρε0 εR )2 ) (7)
A 2 unit time)1/2 .
The diffusion-affected model describes Fickian (K2 ) and non-
which simplifies in the high frequency range to the capacitive
Fickian (Type II) diffusion (K1 ) which may also be considered as
response:
a means of effective thickness decrease [32]. If one assumes that

 the water swollen section of the film has a negligible resistance
ρd
log(|Z|) = log − log(ωρε0 εR ) with respect to the non-swollen “pristine” portion of the coating,
A

 the formulation (time dependence) of Eq. (11) is the same for
d diffusion driven resistivity loss of pristine material in the water
= log − log ω (8)
Aε0 εR swollen layer as it is for the ablation model. In this case, Type II
diffusion which has a constant velocity advances linearly with
Therefore, if the slope on a Bode plot varies from −1 for a rea- time [27,32–34], would be represented by the constant rate (K1 )
sonably flat layer, one of the material properties, the resistivity or term and Fickian diffusion, which advances according to the
the relative dielectric constant, of the material must be a function square root of time, would be represented by the second (K2 )
of the frequency. term. Thus, the functional form of the effective coating thick-
ness remains the same as a function of time, but the definition
2.1.2. Low frequency limit of K1 and K2 in Eq. (11) will be defined in terms of the diffusion
The simplest form for the low frequency limit is that constants and plasticizing rates (Type II) rather than the mate-
(ωRC)2  1, i.e. |Z| ≈ R. Using the same expressions for the rial loss parameters of film ablation. Thus, linear thickness loss
4508 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515

per unit time or advancement rate of Type II diffusion front is

described by K1 . K2 is the proportionality constant for decrease
in effective thickness of a coating due to roughening of the sur-
face, and is applicable to Fickian diffusion based advance of the
water front in the coating.
The low frequency resistivity of the water infiltrated coating
is assumed negligible compared to the pristine coating, i.e. the
resistance of the infiltrating solution is often 10 or more orders of
magnitude less than the pristine polymer. As long as the pristine
coating thickness remains a significant portion of the overall
coating thickness, the resistance of the water infiltrated layer can
be neglected. Thus, in the low frequency regime the Ablation and
Diffusion models give the same impedance response. When the
diffusion front advance leaves only a small amount of pristine Fig. 3. Experimental, low frequency impedance |Z| data is taken from acceler-
coating unaffected by water intrusion, the resistance of the water ated exposures of aircraft coatings. Used with permission from Ref. [12].
affected material can no longer be neglected. As the diffusion
front consumes most of the pristine coating, the resistance of negative; resulting in accelerated decrease in the impedance as
the water saturated coating needs to be added in series to the higher order terms in Eq. (12) become more significant. This is
resistance of the remaining pristine coating. exactly what we have seen in our studies of long term exposure
Incorporating Eq. (11) for thickness into the estimate of low of coatings, especially in ASTM D5894-96 exposure [17,35].
frequency impedance, Eq. (10), gives: Fig. 3, taken from Ref. [35], shows such experimental results.


√  Once the resistivity ρ and dielectric constant εR are measured
ρd0 K1 t K2 t
log|Z| = log 1− + using an independent method for both the pristine coating and
A d0 d0 water saturated coating, the only variable left in fitting the model
1 to measured data is the advance rate of the Type II diffusion front
− (ωε0 εR ρ)2 (12) (or rate of ablation). Thus, the model is more physical, can quan-
2 ln(10)
titatively match experimental data and reduces the number of
The last term of Eq. (12) is dependent only on material proper- fitting parameters compared to present methods based entirely
ties, not exposure time. on fitted equivalent circuit components. The one fitting parame-
Expanding the first term in Eq. (12) for short times gives a ter is the water/electrolyte-swollen front advance rate, which is
linear time dependence of impedance for Type II diffusion or itself a physical quantity of the system.
ablation (K2 = 0): Eq. (14) is fit to the data in Fig. 3 and plotted in Figs. 4 and 5

  √  for both Fickian and Type II diffusion. Fig. 3, which shows a
ρd0 1 K1 K2 t
log|Z| = log − t+ plot of the linear expansion calculated from Eq. (14), seems
A ln(10) d0 d0
reasonable over most of the coating’s effective life. However, as
1
− (ωε0 εR ρ)2 (13)
2 ln(10)
or re-ordered as

 
ρd0 1
log|Z| = log − (ωε0 εR ρ)2
A 2 ln(10)


√ 
1 K1 1 K2 1/2
− t− t (14)
ln(10) d0 ln(10) d0
The slope of the impedance at low frequencies should correlate
with the coating loss (or depth of coating penetration by signifi-
cant water) as a function of time determined by K1 . The intercept
on the plot should be defined by the geometry of the system, coat-
ing thickness, EIS area, and intrinsic dielectric properties of the
coating.
Thus, if one considers only surface ablation, with time con-
stant K1 , or Type II diffusion only, the plot of the log of the
impedance modulus versus exposure time is linear. With pre-
dominantly surface degradation via roughening, one will also
Fig. 4. Experimental low frequency impedance data taken from Fig. 3 for the
see a linear decrease with time of the log of the low frequency High Gloss Topcoat and fit to the low frequency impedance predictions based
impedance after a short transient due to the square root of time on advance of the water diffusion front. The dashed line represents the fitting
dependence. In both cases, the higher order terms are strictly based on Fickian diffusion and the solid line is based on Type II diffusion.
 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515
Fig. 5. Experimental low frequency impedance data taken from Fig. 3 for the
Self-Priming Topcoat and fit to the low frequency impedance predictions based
on advance of the water diffusion front. The dashed line represents the fitting
based on Fickian diffusion and the solid line is based on Type II diffusion.
the coating begins to fail, the decrease in impedance accelerates
as modeled in Figs. 4 and 5. The impedance of the initial coating
sets the intercept. The slope is simply the effective loss rate of the
coating, either by material loss or advance of the water saturation
front, divided by the full coating thickness.
Thus, a linear loss in film thickness due to coating ablation
or water saturation front advance from weathering without other
coating damage, is sufficient to give a log-linear plot of low
frequency impedance modulus versus exposure time. The decay
constant of the latter plot can be equated to the effective loss rate
of the coating, either by material loss or advance of the water
saturation front, divided by the full coating thickness. This decay
constant can be used to predict effective coating lifetime if the
low frequency impedance modulus of the coating at “failure” is
specified by the user [36].
2.2. Use of effective medium theory (EMT) models of the
swollen layer to model the entire EIS frequency spectrum
for the diffusion model
The above discussion represents the low frequency
impedance of the EIS spectrum for both models. In order to
simulate the entire EIS spectrum for the polymer coating in the
diffusion-affected model, the dielectric constant and the resistiv-
ity of the portion of the coating which contains water need to be
quantified. Two ways of looking at the water in saturated poly-
mers are suggested. It can be viewed as a set of isolated inclu-
sions that fill with water or interconnected diffusion/percolation
paths filled with water, like a conducting web, Fig. 6. Modeling
the electrical properties of mixtures of materials goes back to
the time of Maxwell and Raleigh [37]. These types of models
are now classified under the name effective medium theories and
are, in general, designed to cope with the low volume fraction
material as inclusions, not as percolation paths. The equivalent
circuit in Fig. 2 incorporates both of these cases as limits. As the
4509
Fig. 6. The electrical resistivity and relative dielectric constants for the diffusion-
affected zone are modeled in three ways: first as percolation paths which act
as channels to deliver water to the interface and advance the diffusion-affected
zone. The second model assumes a Maxwell type two-phase system where water
fills voids within the coating and coating stresses valve the polymer between
inclusions to result in water movement. The third model combines inclusions,
holding most of the water, with a few channels that connect the water inclusions.
interface resistance, the resistance of transport along the inter-
face, becomes negligible, the equivalent circuit becomes two
Randles circuits in series. Randles circuits in series model the
Maxwell approximation for spherical water inclusions and are
the first asymptote. The second asymptote is when the interface
has a very large resistance as would be the case of percolation
paths of water that are reasonably separated from one another.
Effective medium theory puts the conducting material in
a sphere of equivalent volume, surrounded by a matrix, usu-
ally the insulator. Thus, in the direct application of EMT to a
polymer/water system, insulating material always surrounds the
conducting sphere. The application of EMT has predominantly
been to solid inclusions, such as conducting solid spheres in
an insulating system, frequently a capacitor dielectric. In the
evolution of weathering or continuous immersion of coatings,
perhaps small low-density regions within the coating are filled
with water. However, in the water transport process the water
must diffuse from the surface to these inclusions and between
inclusions. This makes the Type II diffusion front more like
a continuous saturated polymer with, perhaps, islands of pris-
tine coating. In such a case, the saturated region of the polymer
may be considered the matrix material with enclosed unsatu-
rated polymer as the spherical inclusions. This view requires a
different description of the volume fraction of water and poly-
mer. One would need to define, identify and measure a fully
saturated region of the polymer.
For either the ablation or the diffusion model to be reason-
able, it must not only match the low frequency asymptote (which
they do), but must generate full spectra EIS for a considerable
portion of the coating lifetime. The parameters for the advance
of the diffusion front and ablation, generated in our modeling of
the low frequency EIS must give realistic full EIS spectra for that
time. Ablation alone, however, gives an excessive spread in high
frequency impedance, if it alone is considered as the cause for the
decrease in low impedance decrease. The high frequency (capac-
itive) region of the Bode plot often overlaps for EIS spectra
taken at various times during weathering, though ablation clearly
decrease the modulus of the impedance at high frequencies. This
4510 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515
is due to the large decrease in capacitance with film thickness
loss. However, if the change in impedance of the coating with
time is due to water infiltration, the water containing portion of
the polymer coating has an effective dielectric constant that lies
between that of the original coating and that of water. This breaks
the coating into two regions. The first, diffusion-affected zone
has electrical properties of the composite water/polymer system
and these properties are likely to vary with water concentration.
The dielectric constant of the water-affected, swollen coating
depends on how the water is distributed. This is where one can
use an EMT to estimate the resistivity and relative dielectric con-
stant. The water almost certainly exists as an inhomogeneous
distribution through the coating, which is beyond the simple
EMT approximations [38]. Finite element methods are currently
being employed to investigate these more detailed distributions,
and will be presented at a later time. What follows are EMTs
employed to get a first order estimate of the relative dielectric
constant and resistivity to be incorporated into the equivalent
circuit model and verify the reasonableness of the model in pre-
dicting the entire EIS spectra.
The relative dielectric constant and resistivity of the water
affected zone need to be calculated in order to estimate
the EIS response. We present below, in Eqs. (15)–(20), the
EMTs that have been used to describe composites such as our
[water + coating] systems. These are intended to be applicable
at inclusion volume fractions less than 10%.
The equation:
polymer
εeffective
R = εR (1 − φ)3/2
where εR is the dielectric constant of the superscripted material
and φ is the volume fraction of the secondary phase (water if the
water is assumed to remain a separate and distinct material) has
been suggested by Brueggman [37].
One can apply various effective medium theories. If one
assumes that the water is the spherically included phase and
that the saturated polymer is the continuous phase (neglecting
possible islands of pristine polymer) the following equations
results:
polymer
εeffective
R = εR (1 − φ) + εwater
R φ
This effective medium theory equation is a linear combination of
dielectric constants and is equivalent then to adding capacitors
in parallel:
1 1−φ φ “Maxwell” inclusions. It is noteworthy that positron annihilation
= polymer + water
εeffective
R εR ε R
this Eq. (17) is equivalent to adding the capacitance in series
(which implies that the resistivities would be added in parallel).
One can also develop a generalization of the above equations as
polymer n n 1/n
εeffective
R = [(εR ) (1 − φ) + (εwater
R ) φ]
with n = 1 giving Eq. (16) and n = −1 generating Eq. (17).
An approximation often used in the analysis of EIS data from
coatings was developed by Brasher and Kingsbury [39]:
polymer
log(εeffective
R ) = log(εR )(1 − φ) + log(εwater
R )φ
Eq. (20) gives the Maxwell approximation based on spherical
inclusions [40]:
water + 2εpolymer + 2φ(εwater − εpolymer )
polymer εR
εeffective
R = εR R
polymer
R R
εwater + 2ε − φ(ε water − εpolymer )
R R R R
(20)
Effective conductivity can also be calculated by equations hav-
ing the same form. The conductivity, the inverse of resistivity,
is inserted into the calculations in place of the relative dielectric
constant.
The effect of water within the polymer coating is much greater
on the resistivity than on the relative dielectric constant of the
water saturated material. This is due to the relative magnitudes of
the two intrinsic properties. Resistivity differs by nine orders of
magnitude between water and pristine coating, while dielectric
constant differs by one order of magnitude.
Estimates for resistivity and relative dielectric constant for
the diffusion-affected zone can be made. Water must transport
evenly across the exposed coating area through the diffusion-
affected zone (Zone II) towards the pristine polymer (Zone I)
interface in Fig. 6. Thus, the water diffusion front is parallel to
the plane of the coating. The volume fraction of water to use
in the equations above is the volume of absorbed water divided
by volume of the water affected zone, at any particular time.
These allow a test of the diffusion-affected model. The effective
medium theories give relative dielectric constants and resistiv-
ities as shown in Table 1. Although these estimations of the
(15) evolution of the coating and assignment of materials properties
are reasonable, more complete analysis is necessary.
Predictions for entire EIS spectra require knowledge of how
the water is entrained in the polymer. These early results from
modeling, employing EMTs, suggest that water tends to cluster
at low density locations within the coating, with some additional
percolation paths connecting a portion of these water rich regions
as shown in Fig. 6.
3. Application of EMT and equivalent circuit models
(16) The resistivity and relative dielectric constants derived for the
water/coating composite are applied to the equivalent circuits
described in the earlier section. The diffusion-affected model is
separated into two cases; the first has the water advancing in
percolation paths, while the second has water valving between
(17)
spectroscopy results are fitted well by assuming that the free vol-
ume is presented as spheres [41]. Emphasis will be placed on
Type II diffusion (K1 parameter) since it is common in polymer
systems, though it will be seen below that Fickian diffusion
(K2 parameter) also produces reasonable results. Both cases
described below assume a net advance of the diffusion front
(18)
with time and use the water front advance rates predicted by the
fits in Figs. 4 and 5. The coating imbibes water from the exter-
nal electrolyte, then the interface between the saturated layer of
coating and the intact coating moves inward towards the sub-
strate with exposure time, according to Type-II diffusion. In the
(19) figures that follow, no account is taken of the ablation of the
 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515 4511

Table 1
The range of resistivities and dielectric constants for various EMT approximations
Resistivity ( cm) Dielectric constant

0.05a 0.1a 0.2a 0.05a 0.1a 0.2a

Assuming that the volume fraction is the water phase as inclusions in polymer
Eq. (15) 3.0E + 04 1.0E + 04 3.7E + 03 8.6 9.3 11.0
Eq. (16) 6.6E + 03 3.3E + 03 1.6E + 03 8.4 8.8 9.8
Eq. (17) 1.9E + 11 1.8E + 11 1.6E + 11 11.7 15.4 22.8
Eq. (19) 7.3E + 10 2.6E + 10 3.5E + 09 9.0 10.1 12.7
Bruggeman’s equation 1.7E + 11 1.5E + 11 1.0E + 11 8.9 9.9 12.0
Maxwell’s equation 1.7E + 11 1.5E + 11 1.1E + 11 8.9 10.0 12.3
Assuming that the water effected phase is the matrix and unhydrated polymer the inclusions
Eq. (20) 9.7E + 03 4.8E + 03 2.3E + 03 10.6 13.3 19.0

Resistivity ( cm): water, 330; polymer, 2.00E +11. Dielectric constant: water, 82; polymer, 8.
a Water volume fraction.

coating, or of swelling of the saturated layer, nor is there a gra-
dient in water concentration with depth in the diffusion-affected
zone. These contributions can be added easily when data are
available to justify that level of detail. Thus, both models below
assume a coating of fixed thickness, composed of two homoge-
neous materials as shown in Fig. 1. The first region is a saturated
polymer that grows linearly in thickness. The second region is
the unaffected polymer, which loses thickness linearly as it is
consumed by the advance of the water diffusion front. The dif-
ference in the two cases below is how they view and account
for the resistivity and relative dielectric constant of the saturated
polymer. Experimental data which are known to involve water
intrusion effects have been obtained in this laboratory [42].
For the pristine coating, representative coating properties are
applied with resistivities approximately 1012  cm and relative
dielectric constant of five to eight. The coating thickness is taken
as 50 ␮m and the EIS cell area is assumed to be 5 cm2 . The resis-
tivity and dielectric constant of the water-modified coating are
very model dependent, especially with regards to the resistivity,
see Table 1. Future experimentation will provide further under-
standing of coatings and their degradation mechanism.
Fig. 7 shows changing EIS spectra calculated by assuming
that the saturated coating fails by having conducting electrolyte
paths to the interface with the intact coating (the percolation case
of the diffusion model). This is the mechanism when the inter-
face resistance is assumed infinite, where the electric field below

Fig. 7. Bode modulus and phase angle spectra from model calculations. The equivalent circuit in Fig. 2 is applied with the interface resistance infinite, simulating
percolation paths. The equivalent circuit is then numerically fit to the Extended Life Topcoat measurements of Fig. 3. The diffusion front advance rate is used to predict
EIS spectra for various times: (a) shows the impedance as a function of frequency for various times assuming a 0.2 ␮m/week movement of the diffusion-affected
zone interface with a saturation of 5% by volume of water; (b) shows the same as (a) but with a saturation of 20% by volume water.
4512 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515

Fig. 8. EIS Bode modulus spectra as predicted by model calculations when the interface resistance is set to 0. Part (a) shows results predicted by “Maxwell” type
averaging of both relative dielectric constant and resistivity. Parts (b)–(d) shows the calculated resistivity of inclusions combined with a few percolation paths (Fig. 6c)
which reduce the resistance () to various levels, listed in figure as R. The small impact on dielectric constant due to a small number of percolation paths is neglected.

the advancing percolation paths cannot communicate with the
electric field along the interface. Thus, resistance drops precip-
itously as water infiltrates. Fig. 7 employs the equivalent circuit
of Fig. 2 with the interface resistance removed since it is effec-
tively infinite. It was calculated with Eq. (16), which assumes
that the water saturating the coating can be approximated as
direct channels from the surface across the diffusion-affected
zone. This is the limiting case for the EMT since the resistivity
of the diffusion-affected zone is minimized. The capacitance of
the diffusion-affected zone is only slightly affected. Fig. 7 shows
the predicted EIS Bode modulus spectra for the ‘Extended Life
Topcoat’ assuming water saturation for the diffusion-affected
material occurs at 5% and 20%, respectively. The advance of
the diffusion-affected zone interface, and thus its volume, is
assumed to be linear with time (Type II diffusion) and it is
assumed to move at a rate of approximately 1% of the total
film thickness per week. The impact of the higher water at satu-
ration is to increase the spread in the high frequency regime of
the Bode plot. The matching predicted Bode phase angle spectra
for this situation are also given in Fig. 7.
If the interface resistance is zero, then we are assuming essen-
tially that the coating is in two distinct layers, each of which has
a finite resistance. Figs. 8 and 9(a) are constructed assuming
the water moves between discontinuous voids to the interface,
not maintaining a continuous path. Thus, the resistivity of the
diffusion-affected zone stays high. The Maxwell equation (Eq.
(20)) is used to calculate an effective dielectric, 10.0 (that of
water is assumed to be 82), and resistivity, 1.5 × 1011  cm, of
the saturated coating. The resistivity is then decreased, to model
the likely decreasing resistivity as water penetrates to increase its
concentration in the inclusions, Figs. 8 and 9(b)–(d). The results
of this modeling (see Fig. 6) give qualitative agreement with EIS
spectral measurements used for the low frequency impedance
modulus data.
Fig. 9(a)–(d) shows the Bode phase angle spectra from the
same model calculations as Fig. 8. One can see that we have clear
evidence of two time constants in parts (b)–(d) due to the two
coating regions, the water saturated region and pristine region
(Fig. 1). The presence of two time constants in the literature
concerning coated metal impedance spectra has almost always
been interpreted with two RC elements, with the second ele-
ment describing a pore resistance and a double layer capacitance
associated with a delamination area of the coating. These calcu-
lations, however, show this behavior may also occur in an intact
coating (dominated by bulk water transport, not defects) during
planar infiltration of water with no need to invoke a delamination
of the coating from the metal surface.
The Bode and phase angle spectra for the same set of data
used for the low frequency impedance fit for Fig. 4, taking only
the 10th weeks data are given in Fig. 10. The data is in reasonable
agreement given that the water/diffusion advance rate is taken
from the low frequency data and the water affected material is
homogenized via effective media theory approximations. The
Bode plot in Fig. 10 matches the trends in Fig. 8c, though the
low frequency plateau is much less flat in the measured data.
The measured phase angle spectra of Fig. 10 also show behavior
consistent with the model (Fig. 9b and c). The measured phase
angle has much broader peaks than the model, likely due to the
 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515 4513

Fig. 9. The EIS Bode phase angle spectra for the same conditions as given in Fig. 8. Spectrum (a) gives results predicted by “Maxwell” type averaging of both relative
dielectric constant and resistivity. (b)–(d) shows the combined effect of inclusions with a few percolation paths (Fig. 6c) which reduce the resistance to various levels,
listed in figure as R. The small impact on dielectric constant due to a small number of percolation paths is neglected.

homogenizing of the water in the model instead of having a water
concentration gradient. Another example of published EIS data
presents Bode and phase angle plots in Fig. 5 for a thick epoxy
polyamide-coated steel and phase angle in Fig. 2 for a thin epoxy
polyamide-coated steel [34] and shows the same trends as the
composite system in the present paper in Figs. 8 and 9.
Additional refinements will be added to the models as
justified by experimental data. The diffusion-affected, water-
Fig. 10. Bode and phase angle plots for the High Gloss (also used in Fig. 4)
topcoat is plotted every 10 weeks.
saturated polymer material is assumed here to be homogeneous
with respect to water volume fraction. It is reasonable to assume
the water volume fraction varies with depth within the diffusion-
affected zone. This gradient in the water concentration and thus
the resistivity and effective relative dielectric constant will be
incorporated into our model as better experimental data war-
rants further detail.
In summary, an equivalent circuit model with two physical
asymptotes is proposed to simulate a coating possessing a diffu-
sion front progressing into a coat (see Fig. 1). The first asymptote
(when the interface resistance is infinite and models percola-
tion paths that advance into the topcoat) is two serial capacitors
in parallel with two serial resistances (one is negligible). The
second asymptotic equivalent circuit model (when the interface
resistance is negligible and describes a nearly constant elec-
tric field at the water advance interface as occurs in Maxwell’s
approximation of spherical water inclusions) is a Randles cir-
cuit that describes the water-saturated coating in series with a
Randles circuit that describes the remaining pristine coating. In
both cases the volume of the water saturated coating increases
and the volume of the pristine coating decreases as the water
penetrates deeper into the coating. One can easily see that if
failure is identified with the saturated “front” reaching the metal
substrate, a lifetime prediction based on this model can easily
be made.
The transport of water with entrained ions to the substrate
metal should be restricted for limiting corrosion initiation and
growth. A topcoat is the major barrier to water infiltration, and
4514 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515
typifies the subject of this model. Whereas topcoats are very
resistive and are designed to have low water permeability, primer
layers are often designed to easily transport water and inhibitors
to the sites of corrosion. Modeling such a primer layer contin-
ues to be a part of our research. The design and quantification of
the barrier properties of the topcoat, separate from the primer,
remains an important aspect of corrosion prevention and coating
system lifetime prediction. Additional complications associated
with the cyclic wet–dry accelerated weathering make this model
limited to a net advance of the water front. Finite element models
are being developed to predict the water distribution through a
coating. The overall system impedance response to a frequency
dependent applied voltage applied to this water distribution
within a coating is simulated via finite element modeling as
a part of our continued research to simulate EIS response.
4. Summary and conclusions
Using a simple linear thickness loss with time from UV expo-
sure or loss of resistivity due to the advance of a water infiltration
front and inserting this into a simple Randles circuit description
for EIS measurements, one can predict a linear decrease with
time of the log of the low frequency impedance (equivalent to a
linear loss in the log of the film resistance with time). This agrees
qualitatively with data on weathering and accelerated exposure
of coating systems. Further, if we model the change in dielectric
properties as well as resistivity of coating films due to water
penetration in immersion or exposure to electrolyte using exist-
ing effective medium theories, we can predict and emulate many
of the qualitative features of the EIS spectra in the range from
low frequencies to 105 Hz that are seen in experimental systems,
including both good and poor performance corrosion protective
coatings. This is done using a two layer model for the coating
in exposure based on a water-saturated layer moving into the
film following Type II diffusion over a pristine coating layer
unaffected by water intrusion. In this case we insert the model
calculations into an equivalent circuit description using two Ran-
dles circuits in series to describe the two layers. This leads to
two time constants, but with an interpretation different from the
conventional reasoning that the only reason for a second time
constant is delamination. The physical parameter predictions
that are based on film thickness loss or dielectric and resistiv-
ity changes due to water intrusion use experimental values for
film loss or dielectric constant values from several coating sys-
tems as their starting point. This leads one to believe that more
complete quantitative descriptions of many experiments might
be predicted better by more realistic values of physical parame-
ters for coatings systems estimated by other measurements. This
will allow us to develop a more fundamental description of the
changes undergone by coatings to both accelerated lab exposure
as well as field exposure, allowing, perhaps, true lifetime predic-
tions for coatings in actual use. Further, the model can provide
formulation guidelines by allowing one to identify key physi-
cal parameters of the coating which control its performance in
exposure. As one might well expect, even this simple modeling
has shown that water “solubility” in the coating on exposure is
a key parameter for controlling coating performance.
This modeling that we describe in this paper is important to
coatings characterization by EIS as well as to the use of EIS
experimental data in the prediction of real-world coating per-
formance. The modeling that we have done differs significantly
from the standard emphasis on “modeling” that uses equivalent
circuit elements such as resistors, capacitors, constant phases
elements, etc., and focuses on the modeling of physical property
parameters of such elements, in this case resistivity and dielec-
tric constant that can then be used in resistor and capacitance
elements. This allows us to bring into EIS studies the work done
in coatings science and polymer science on changes in resistiv-
ity and dielectric behavior shown by polymers and pigmented
organic coatings upon exposure to UV and/or water/electrolyte.
The results we have shown give illustration to the value of such
modeling in interpreting and fitting experimental EIS data. Such
quantitative predictive capabilities will be crucial to the success-
ful rapid development and deployment of new coatings materials
for corrosion protection.
Acknowledgements
This work was performed with the support of the Air Force
Office of Scientific Research (AFOSR) under Grant F49620-04-
1-0398. The earlier work of this cited within the text and below
had various sponsors, the primary being the Office of Naval
Research (ONR) and AFOSR.
References
[1] G. Bierwagen, L. He, D. Tallman, Time–temperature effects in polymer
coatings for corrosion protection as analyzed by ELS, Macromolecular
Symposia, vol. 187, Presented at FATIPEC XXVI, Dresden, Germany,
September 9–11, 2002. Wiley-VCH, Weinheim, Germany, 2002, pp.
909–918.
[2] J.M. Hu, J.Q. Zhang, C.N. Cao, Determination of water uptake and
diffusion of Cl− ion in epoxy primer on aluminum alloys in NaCl solu-
tion by electrochemical impedance spectroscopy, Prog. Organic Coat. 46
(2003) 273–279.
[3] S. Duval, M. Keddam, M. Sfaira, A. Srhiri, H. Takenouti, Electrochem-
ical impedance spectroscopy of epoxy-vinyl coating in aqueous medium
analyzed by dipolar relaxation of polymer, J. Electrochem. Soc. 149
(2002) BS20-9.
[4] J.H.W. de Wit, Inorganic and organic coatings, in: P. Marcus, J. Qdar
(Eds.), Corrosion Mechanisms in Theory and Practice, Marcel Dekker,
New York, 1995, pp. 581–627 (Chapter 16).
[5] W. Funke, Corrosion tests for organic coatings—a review of their use-
fulness and limitations, J. Oil Chem. Assoc. 62 (1979) 63–67.
[6] H. Leidheiser Jr., Electrical and electrochemical measurements as pre-
dictors of corrosion at the metal organic coating interface, Prog. Organic
Coat. 7 (1979) 79–104.
[7] W.S. Tait, A discussion of the reliability of electrochemical impedance
spectroscopy data from coated metals div of polymeric materials: sci.
eng. preprints, Vol. Amer. Chem. Soc. National Meeting, Denver, April
1993, p. 101.
[8] J.R. Macdonald (Ed.), Impedance Spectroscopy, Wiley-Interscience, New
York, D.D. Macdonald and M.C.H. McKubrie, 1987, p. 301 (Chapter
4.3).
[9] G. Bierwagen, D. Tallman, J. Li, L. He, EIS studies of coated metals
in accelerated exposure, in: Proceedings of the Fifth International Sym-
posium on Electrochemical Impedance Spectroscopy, E1S 2001, June
2001 Marilleva, Italy Prog. Organic Coat. 46 (2003) 148–157.
 B.R. Hinderliter et al. / Electrochimica Acta 51 (2006) 4505–4515
[10] J. Kittel, N. Celati, M. Keddam, H. Takenouti, Influence of the coating-
substrate interactions on the corrosion protection: characterisation by
impedance spectroscopy of the inner and outer parts of a coating, Prog.
Organic Coat. 46 (2003) 135–147.
[11] I.D. Raistrick, J.R. MacDonald, D.R. Franceschetti, The electrical
analogs of physical and chemical process, in: J.R. MacDonald (Ed.),
Impedance Spectroscopy, John Wiley & Sons, New York, 2003, pp.
27–132.
[12] F. Mansfeld, H. Shih, H. Greene, C.M. Tsai, Analysis of EIS data for
common corrosion process, in: J. Scully, D.C. Silverman, M. Kendig
(Eds.), Electrochemical Impedance: Analysis and interpretation, ASTM
STP 1188, ASTM, Philadelphia, PA, 1993, p. 37.
[13] B. Normand, H. Takenouti, M. Keddam, H. Liao, G. Monteil, C. Cod-
det, Electrochemical impedance spectroscopy and dielectric properties of
polymer: application to PEEK thermally sprayed coating, Electrochim.
Acta 49 (2004) 2981–2986.
[14] J. Vogelsang, G. Strum, New interpretation of electrochemical data
obtained from organic barrier coatings, Electrochim. Acta 46 (2001)
3817–3826.
[15] G.W. Walter, A review of impedance plot methods used for corrosion
performance analysis of painted metals, Corros. Sci. 26 (1986) 681–703.
[16] G.P. Bierwagen, J. Li, N. Davis, D.E. Tallman, Thickness dependence
of electrochemical properties of organic coatings, in: Proceedings of the
Fifth Nuernberg Congress, vol. 1, Nuemberg, Germany, April 12–14,
1999. Vincentz Verlag, Hannover, Germany, p. 315.
[17] G. Bierwagen, J. Li, L. He, D. Tallman, Fundamentals of the measure-
ment of corrosion protection and the prediction of its lifetime in coating,
in: J. Martin, D. Ram (Ed.), Proceedings of the Second International
Symposium on Service Life Prediction Methodology and Metrologies,
Monterey, CA, November 14–17, 1999. ACS Symposium Series #805,
ACS Books, Washington, DC, 2001, pp. 316–350 (Chapter 14).
[18] J.H. Park, G.D. Lee, H. Ooshige, A. Nishikata, T. Tsuru, Monitoring
of water uptake in organic coatings under cyclic wet–dry condition,
Corrosion Sci. 45 (2003) 1881–1884.
[19] U. Rammelt, O. Reinhard, Application of electrochemical impedance
spectroscopy (EIS) for characterizing the corrosion-protective perfor-
mance of organic coatings on metals, Prog. Organic Coat. 21 (1992)
205–226.
[20] U. Hammelt, G. Reinhard, Characterization of active pigments in damage
of organic coatings on steel by means of electrochemical impedance
spectroscopy, Prog. Organic Coat. 24 (1994) 309–322.
[21] See, for example, G.P. Bimagen, Calculation of noise resistance from
simultaneous electrochemical voltage and current noise data, J. Elec-
trochem. Soc. 141 (1994) L155–L157.
[22] J.R. Macdonald (Ed.), Impedance Spectroscopy, Wiley-Interscience, New
York, 1987.
[23] G.P. Bierwagen, Calculation of noise resistance from simultaneous elec-
trochemical voltage and current noise data, J. Electrochem. Soc. 141
(1994) L155–L195.
[24] B.R. Hinderliter, S.G. Croll, Monte Carlo approach to estimating the
photo-degradation of polymer coatings, J. Coat. Tech. 2 (2005) 483.
4515
[25] S.G. Croll, A.D. Skaja, Spectroscopic absorption and effective dosage
in accelerated weathering of a polyester-urethane coating, J. Mater. Sci.
37 (2002) 4889–4900.
[26] G. Wypych, Handbook of Material Weathering, second ed., Chem Tec
Publishing, 1995.
[27] J. Crauk, The Mathematics of Diffusion, second ed., Oxford Science
Publications, 1975 (Chapter 11).
[28] A.S. Argon, R.E. Cohen, A.C. Patel, A mechanistic model of case II
diffusion of a diluent into a glassy polymer, Comput. Theor. Polymer
Sci. 9 (1999) 339–352.
[29] A. Friedman, G. Rossi, Phenomenological continuum equations to
describe case II diffusion in polymeric materials, Macromolecules 30
(1997) 153–154.
[30] E. Hoffman, A. Saracz, Weathering of paint films: IV. Influence of radi-
ation intensity on chalking of latex paints, J.O.C.C.A. 55 (1972) 101–
113.
[31] S. Hollande, J.-L. Laurent, Degradation process of an industrial ther-
moplastic elastomer polyurethane-coated fabric in artificial weathering
conditions, J. Appl. Polym. Sci. 73 (1999) 2525–2534.
[32] A.S. Argon, R.E. Cohen, A.C. Patel, A mechanistic model of case
II diffusion of a diluent into a glassy polymer, Polymer 40 (1999)
6991.
[33] J.R. Scully, Electrochemical impedance of organic-coated steel: corre-
lation of impedance parameters with long-term coating deterioration, J.
Electrochem. Soc. 136 (4) (1989) 979–990.
[34] M.M. Wind, H.J.W. Lenderink, A capacitance study of pseudo-Fickian
diffusion in glassy polymer coatings, Prog. Organic Coat. 28 (1996)
239–250.
[35] G. Bierwagen, D. Tallman, J. Li, L. He, EIS studies of coated metals
in accelerated exposure, Prog. Organic Coat. 46 (2003) 148–157.
[36] C. Schillcr, W. Strumz, The evaluation of experimental dielectric data
of barrier coatings by means of different models, Electrochim. Acta 46
(2001) 3619–3625.
[37] R. Meredith, C.W. Tobias, Conduction in heterogeneous systems, in:
C.W. Tobias, P. Delehay (Eds.), Advances in Electrochemistry and Elec-
trochemical Engineering, vol.2, Wiley-Interscience, NY, 1962, pp. 15–47
(Chapter II).
[38] G.K. van der Wel, O.C.G. Adan, Moisture in organic coatings—a review,
Prog. Organic Coat. 37 (1999) 1–14.
[39] D.M. Brasher, A.H. Kingsbury, J. Appl. Chem. 4 (1954) 62.
[40] R.E. De La Rue, C.W. Tobias, On the conductivity of dispersions, J.
Electrochem. Soc. 106 (1959).
[41] A.J. Hill, Positron annihilation lifetime spectroscopy to probe free-
volume effects in high temperature-polymers and composites, in: M.R.
Tant, J.W. Connell, H.L.N. McManus (Eds.), High-Temperature Proper-
ties and Application of Polymeric Materials of ACS Symposium Series,
vol. 603, 1995 (Chapter 5).
[42] G.P. Bierwagen, J. Li, L. He, L. Ellingson, D.E. Tallman, Consider-
ation of a new accelerated evaluation method for coating corrosion
resistance—thermal cycling testing, Prog. Organic Coat. 39 (2000)
67–78.