Sunteți pe pagina 1din 18

794

SCIENCE
CORROSIONOCTOBER 1992
Review of Applications of Impedance
and Noise Analysis to Uniform
and Localized Corrosion

C. Gabrielli and M. Keddam*


0010-9312/92/000189/$3.00/0
1992, National Association of Corrosion Engineers

Submitted for publication February 1992; in revised form, March 1992.


* UPR15 du CNRS, Physique des Liquides et Electrochimie, Tour 22, 4
place Jussieu, 75252 Paris Cedex 05, France.
ABSTRACT
Steady-state techniques are very limited when they are used
for analyzing corrosion processes for uniform or for localized
phenomena. Non-steady-state techniques give valuable
information concerning these processes. The measurement
of the electrochemical impedance and the analysis of the
electrochemical noise are particularly fruitful. In this paper,
examples are given in the case of active and inhibited
corrosion of iron, corrosion of coated metal, anodized
aluminum alloys, and stainless steel. The principles of
measurement techniques and analysis are also described.
KEY WORDS: electrochemical noise, impedance, localized
corrosion, uniform corrosion
INTRODUCTION
The corrosion of metals is an electrochemical
phenomenon. Since no electrochemical method
directly measures the corrosion rate, its evaluation is
based on a certain number of hypotheses. Therefore,
knowledge of the corrosion mechanism is often the
prime necessity to verify the validity of the hypotheses
adopted. The metal dissolves through a charge
transfer at the electrochemical interface that occurs at
the end of a succession of more or less coupled
elementary phenomena: (1) transport of reactive
species in the bulk of the solution or through a layer
often associated with chemical reactions in the bulk
phase; (2) adsorption of the reactive species on the
electrode; and (3) electrochemical and chemical
interfacial reactions.
Adsorption and reactions take place on the
electrode surface, but mass transport is a
homogeneous phase phenomenon that has to be
carefully controlled. The aim of the electrochemist is to
be able to study each elementary phenomenon that
leads to corrosion in isolation from the others.
Therefore, he has to use a technique able to extract
the data that allow these phenomena to be separated.
Some techniques that are able to characterize the
state of the surface or adsorbed species on the
interface require the use of vacuum techniques (such
as LEED and Auger). Therefore, they cannot be used
for an in situ analysis of the electrochemical interface.
Other techniques using electromagnetic waves
(optics: ellipsometry or x-rays: EXAFS) are used in the
study of the electrochemical interface, but they can
hardly be applied when an alteration as dissolution of
the surface has occurred. Hence, electrical methods
are often the only possible recourse for in situ
investigations.
In fact, the use of electrical quantities allows a
kinetic study to be done, which permits dissection of
the couplings between the elementary phenomena by
controlling the reaction rates. This enables the
monoelectronic steps in the reaction mechanisms to
be distinguished and the often unstable reaction
intermediates involved in these reactions to be
795
SCIENCE
CORROSIONVol. 48, No. 10
counted. If these techniques do not allow a real
identification of the bondings and the reaction
intermediates from a chemical point of view, they give
information on the kinetics of the reaction mechanism
governing the behavior of the electrochemical
interface and some characterization of these
intermediates.
In the study of the electrical quantities, steady-
state plots, impedance, and noise analysis form a
hierarchy of investigation techniques that give more
and more detailed information. In spite of their relative
complexity, the last two techniques are often used
because steady-state plots give information only on
the rate-determining step. When corrosion occurs in
several steps or in a random manner, this is not
sufficient to understand its reaction mechanism.
The use of the impedance techniques rests on
principles analogous to those that justify relaxation
methods used at equilibrium state in chemical kinetics.
Disturbing the reaction from the steady state by
applying a perturbation to the electrochemical system
obliges the system to relax to a new steady state. As
the various elementary processes change at different
rates, the response can be analyzed to dissect the
overall electrochemical process.
The choice of technique depends on whether one
is trying to establish a reaction mechanism, i.e.,
testing a model, or determining kinetic parameters of a
known or at least commonly assumed mechanism.
Some transient techniques are extensively used
because they are well suited for extracting kinetic
parameters when the mass transport is the rate-
limiting step. In some very favorable cases, several
techniques may be of comparable use. However,
when complex heterogeneous reactions interact with
mass transport, time analysis of the transients will lead
to very poor results in trying to extract a reaction
mechanism, and a frequency analysis is more
efficient.
On the other hand, many semimacroscopic
phenomena related to corrosion are random by nature
such as pitting, abrasion, gas evolution. This special
property of the processes makes the classical
deterministic techniques (steady-state techniques for
I-E plotting or sine wave techniques for impedance
measurements) difficult to perform. Even when the
processes seem to be in a steady-state regime (e.g.,
uniform corrosion), nonlinearity problems often
remain. This point constrains to use a perturbing
signal of very low amplitude that is hard to detect.
If corrosion is regarded as a stochastic process,
the corrosion current (potential) at a given potential
(current) is a random signal; therefore, it is called
electrochemical noise. Then, time and frequency
processing techniques, especially devised for random
signals, can be applied.
In the case where the elementary current
transients (pit nucleation and growth, bubble growth,
and departure) are not too numerous to be discernible
due to the random events, a statistical counting can be
done on the time series displayed by the current to
evaluate a mean appearance rate (e.g., pitting rate,
evolution rate). However, in the case closer to the
cases of practical interest, where many events occur
at the same time so that their current transients
overlap, only a spectral analysis of the random signal
will allow the characteristic parameters of the process
to be attained.
In the case of uniform corrosion where the
exciting signal can lead to an alteration of the process
studied, a simple listening of the metal/electrolyte
interface, obtained by analyzing the natural
fluctuations of the corrosion potential, allows the
process to be investigated.
The potentialities of the deterministic approach of
the corrosion process through impedance analysis of
the mean corrosion potential (or current) and of the
probabilistic approach through spectral analysis of the
random part of the corrosion potential (or current) will
be reviewed here.
The experimental arrangements will be described
for impedance measurements that use a transfer
function analyzer and for noise spectral analysis that
is based on a spectrum analyzer.
As impedance technique can only be used, in
principle, with stationary processes, it can be easily
used for investigating generalized uniform corrosion.
However, for localized corrosion, which is usually
random, its application is more complicated. Various
examples of uniform corrosion studies will be covered
(such as iron, stainless steel, and aluminum), and the
possibility of impedance technique in localized
corrosion situations will be discussed.
Noise spectral analysis can be used when the
corrosion current (or the potential) is completely or
partly random. This is obviously the case for localized
corrosion due to a mechanical breakdown (e.g.,
abrasion) or a chemical breakdown (e.g., by Cl

addition). However, this is also the case at the


corrosion potential when a concomitant gaseous
evolution occurs (e.g., iron dissolution in acidic
medium where hydrogen simultaneously evolves).
EXPERIMENTAL ARRANGEMENTS
The experimental arrangements for impedance
measurements and spectral analysis of the
electrochemical noise will be successively described.
Impedance Measurements
1,2
An impedance, like any complex quantity, is
characterized by a real and an imaginary part or by
796
SCIENCE
CORROSIONOCTOBER 1992
the combination of a modulus and a phase shift. The
choice of one or another form is not unimportant. The
correspondence between the real part and the
imaginary part is a one-to-one relationship for a
physical system, according to the Kramers-Kronig
relationship. Theoretically, it should be enough to
measure one component throughout the whole
frequency spectrum. However, usually both parts are
measured since it increases the measurement
accuracy. With the other form, the correspondence
between the modulus and the phase is not always a
one-to-one relationship because the Bode relationship
(equivalent to the Kramers-Kronig relationship for the
modulus and the phase shift) does not apply to
systems of nonminimal phase shift. The constraints
placed on the measuring instrument by the frequency
range to be explored, by the conditions of linearity,
and by the problem of signal-to-noise ratio are
reviewed.
Frequency Range of Interest
Experimental results obtained during the last 20
years show that the time constants involved in
electrochemical processes are often much longer than
1 s. Taking into account the presence of the double-
layer capacitance, the interfacial impedance must be
measured over a wide frequency range extending
from several tens of kHz down to the subacoustic
region. The low frequency limit is often reached only
for frequencies of the order of 10
2
or 10
3
Hz. In this
range, the impedance varies between its high
frequency-limiting value, corresponding usually to the
electrolyte resistance, and its low frequency-limiting
value, equal to the slope of the steady-state
polarization curve at the polarization point at which the
measurement is carried out.
Respect for Conditions of Linearity
The electrode impedance is defined only for a
small amplitude of the perturbing signal. The
maximum amplitude to be used is dependent on the
polarization point and the frequency. It depends also
on the accuracy of the measurements determined by
the impedance measurement device. For an
electrochemical system, the maximum admissible
amplitude is usually smaller at low frequency than at
high frequency.
It is possible to ascertain whether the amplitude
!E of the sine wave signal respects the conditions of
linearity by measuring, for example, the modulus of
the impedance as a function of !E. When !E is
increased and experimental values start varying, one
is no longer in the linear domain. Measurement of
higher harmonics, using a spectrum analyzer, can also
be used as a test. To respect the conditions of
linearity, it is necessary to work at low amplitude. On
the other hand, impedance measurements should be
carried out over a wide frequency range. This
produces a difficult problem with respect to signal-to-
noise ratio, which controls in large part the overall
accuracy of measurements.
Signal-to-Noise Ratio
The noise associated with the signal to be
measured may arise from a number of different
origins, which may, however, be classified into three
main categories: noise from the control circuit due to
the semiconductors and resistances, noise associated
with parasitic radiations and with ripple in the power
supply, and noise due to the electrochemical cell itself.
All such noises limit the measurement accuracy,
especially when very low amplitude signals are used.
Transfer Function Analyzer
The general scheme of operation is shown in
Figure 1. The input signal, S(t), representing the
response of the cell to the applied signal, x(t) = x
o
sin
("t) produced by a sine wave generator, is correlated
with two synchronous reference signals in phase and
in quadrature with x(t) so that the following expression
can be calculated:
Re =

1
T
S(t) sin(" t) dt
O
T
(1)
Im =
1
T
S(t) cos(" t) dt
O
T
(2)
where
S(t) = x
0
K(") sin(" t #) + A
i
sin(i" t #
i
) + n(t)
$
i
(3)
is the sum of the fundamental, harmonics, and noise.
K(") exp #(") is the transfer function of the system. T
is the duration of the measurement. In practice, T is
equal to an integral number of periods; therefore, the
integrals of the harmonics are equal to zero.
If the duration of the measurement T (the
integration time) were infinite, all components of S(t)
except the fundamental would give an integral equal to
zero and particularly the noise component. This device
is, therefore, equivalent to a filter having an ideally
narrow bandwidth that only allows the passage of the
component whose frequency is that of the generator.
In such cases:
Re =
T%&
lim

1
T
S(t) sin(" t) dt = x
0
K(") cos #
O
T
(4)
797
SCIENCE
CORROSIONVol. 48, No. 10
Im =
T%&
lim

1
T
S(t) cos(" t) dt = x
0
K(") sin #
O
T
(5)
From there, two quantities whose values are
proportional to the real Re and imaginary Im parts of
the signal are obtained. In practice, the measurement
duration T cannot be infinite; therefore, the integral of
the noise is not equal to zero, and the equivalent filter
has a bandwidth whose value is not zero. These
nonzero quantities depend on the integration time and
lead to a measurement error at a frequency f that is
equal, in the case of a galvanostatic control, to:
2
e
2
(_Z_) '
1
N

f
nn
f
a
2
/2

1
R
2
+
1
_Z_
2
(6)
where Z is the impedance of the cell, R is the stan-
dard resistor used to measure the current, N is the
number of averaging,
nn
is the power spectral
density of the noise, a is the amplitude of the perturb-
ing signal, and z is the estimated value of z.
Experimental Arrangement
for Impedance Measurement
In Figure 2, two examples of the experimental
arrangement used for electrochemical impedance
measurements are shown. The main device of the
impedance measurement itself is a transfer function
analyzer with a built-in sine wave generator and with
two inputs, X and Y. The metal/electrolyte interface is
polarized by means of a regulating device that can be
either a potentiostat (Figure 2[a]) or a galvanostat
(Figure 2[b]) in the case of the rest potential (I=0). A
summation circuit $ allows the sine wave perturbing
signal coming from the generator of the transfer
function analyzer to be added to the bias. If the
working electrode is connected to the ground in a
galvanostatic regulation mode, no modification of
electrical connection will be needed when the mode of
regulation is changed. The results of the impedance
measurements are recorded under various forms.
The transfer function analyzer synthesizes a sine
wave voltage signal and carries out the correlation
calculation numerically. A numerical processing allows
very high stability to be obtained at extremely low
frequencies compared to an analogue one and, thus,
is particularly adapted for electrochemical systems.
The sine wave generator of this analyzer can be
frequency programmed between 10
5
and 20 x 10
6
Hz,
fixing the maximum and minimum analyzed frequency
(f
min
, f
max
) and the number of frequencies to be studied.
The frequency for each impedance measurement is
automatically stepped. This greatly improves the
FIGURE 1. Principle of operation of the transfer function
analyzer. Re: Real part of the impedance; Im: imaginary part of
the impedance; x(t): perturbing signal; and S(t): electrochemical
cell response.
(
(
facility of the procedure and decreases the total time
of measurement. Clearly, automatic recording of
results is generally needed. Through a digital-to-
analogue converter, the impedance diagram can be
directly traced on an X-Y plotter. Alternatively,
numerical data can be delivered for each
measurement in a printed read-out form. It is also
possible to store all of the results in buffer memories
and send them to a computer that will analyze the
results when the whole spectrum is obtained. When a
routine experimental procedure is carried out, one
may control the running of the transfer function
analyzer by means of a microcomputer.
The regulating device (e.g., electrochemical
interface) is equipped with two differential amplifiers D
1
and D
2
for current and potential signals with offset to
eliminate the steady-state polarization. Then, the sine
wave current and potential output signals of these two
amplifiers are fed to the transfer function analyzer.
Power Spectral Density Measurement
The characterization up to the second order of
stationary random signals can be done either in the
time domain (correlation function of the signal) or in
the frequency domain (power spectral density [PSD]
of the signal that is the Fourier transform of the
correlation function). In the frequency range used in
electrochemistry (10
3
10
5
Hz), the latter is much more
accurate and easier to interpret.
Principle of the Measurement
As the parasitic noises coming from the mains,
the electromagnetic radiations, the electronics of the
amplifiers, and the voltage control (potentiostat) can
be, in some cases, as high as the useful
798
SCIENCE
CORROSIONOCTOBER 1992
electrochemical noise, a special two-channel
measurement arrangement has to be used (Figure 3).
The PSD measurements are based on a cross-
spectral analysis carried out on a Fourier analyzer.
The signals of interest X and Y at the outputs of the
two measurement channels are sampled by an
analog-digital converter at a sampling frequency F
s
(0.2 Hz to 100 kHz) during the acquisition time T. The
samples are stored in memory blocks of size M (64-
4096). The analog-digital conversion is done in 12 bits
(including one for the sign); therefore, the input
dynamic of the converter is 20 log 2
11
= 66 dB that is
largely sufficient in general.
A wired fast Fourier transform (FFT) algorithm
computes:
X
T
(m!f) = !t x(n!t) exp (2j
m n
M
); 0 ' m '
M
2
$
n=0
M1
(7)
where j
2
= 1, !t = 1/F
s
is the sampling period, and
!f = 1/T is the frequency resolution. The estimation of
the cross spectrum of X and Y, (m!f), where T is
the acquisition time of an elementary spectrum, is
derived through the average of N elementary spectra:

)
xy
T
(m!f) =
1
N

1
T
Y
T,i
(m!f) X
T,i
*
(m!f)
$
i=1
N
(8)
where Y
T, i
(m!f) and X
T, i
(m!f) are the i concomitant
samples of X and Y at the frequency m!f, and *
means complex conjugate. The spectrum is obtained
for M/2 frequencies ranging between 0 and F
max
=
F
s
/2.
The measurement time for N averages is the sum
of:
the total acquisition time of the signals X and Y
t
a
= NT
NM
2F
max
(9)
the computing time
t
c
= 0.22
NM
1024
(10)
as the calculation of an elementary cross spectrum of
1,024 points needs 220 ms.
The errors arising from the elementary spectrum
determination originate in general from:
Quantification: This error is usually negligible
due to the accuracy of the analog-digital converter
(12 bits);
Computing: This error is negligible as the
analyzer computes the Fourier transform with 16-bit
words (0.1% error) and as the computation of the
elementary spectra is done in double precision
(32 bits);
Estimation: This error, due to the finiteness of
the measurement time NT, is often preponderant and
is such as:
'
1
N
1 + 1/ + 1/(2)
2
(11)
where is the signal-to-noise ratio.
Experimental Arrangement
To diminish some of the measurement errors,
special arrangements are used (Figure 3). First, the
potentiostat and the differential amplifiers are battery
powered in order to eliminate the influence of the
mains. Second, the potentiostat, the electrochemical
cell, the amplifiers, and the batteries are protected
against radiated parasitic noise by a shielding cage.
Finally, a high-pass filter and a low-pass filter are
FIGURE 2. Experimental arrangements for impedance measurements (a) potentiostatic mode; (b) galvanostatic mode.
(a) (b)
)
xy
T
799
SCIENCE
CORROSIONVol. 48, No. 10
inserted in each measurement channel to eliminate
the drift on the one hand and the spectrum aliasing
error on the other hand.
In practice, even after high-pass filtering, the
measurement validity is ascertained by a systematic
monitoring of the signal stationarity during the
measurement by means of an oscilloscope.
On a practical point of view, the PSD is calculated
through a FFT algorithm. If the current time series is
sampled each !t the maximum frequency that can be
analyzed is:
f
max
=
1
2!t
(12)
If the FFT is performed over M samples, the
lowest frequency that can be analyzed is:
f
min
=
1
M!t
(13)
The dual measurement channels allow the
parasitic noises n
x
and n
y
, coming from the amplifiers,
filters, and shunt resistors R, to be eliminated by using
a cross-spectrum analysis. As an example, in
potentiostatic mode (Figure 3[a]), X = G(Ri + n
x
) and
Y = G(Ri + n
y
), the cross spectrum is then equal to:
XY* = G
2
R
2

i
(f) (14)
and is proportional to the current PSD of interest as
n
x
and n
y
are independent from each other and
independent from i(t). However, this is valid only for
an infinite number of averages of elementary spectra
XY,* but the experimentalist is able to average over a
finite number N of elementary spectra; therefore, the
estimation error is equal to:
'
1
N
(15)
for a perfect signal-to-noise ratio (N%& in Equation
[11]).
To obtain a reasonable accuracy, between 10 and
100 averages are necessary, which can lead to very
long measurement times as this time is equal to M N!t
FIGURE 3. Experimental arrangements for noise spectral analysis (a) potentiostatic mode; (b) galvanostatic mode. n
x
, n
y
: voltage
parasitic noises of the measurement channels X and Y with gain G.
800
SCIENCE
CORROSIONOCTOBER 1992
I
c
= I
c
exp
2.303
b
c
E
Similarly, for a true equilibrium potential (I = 0), the
impedance is

1
Z
F
=
1
R
p
= 2.303
I
a
b
a

I
c
b
c
(20)
and at zero overall current, I
a
= I
c
= I
corr
, the relation-
ship between the impedance and the corrosion
current is such as:

1
R
p
= 2.303 I
corr
(b
a
+ b
c
)
b
a
b
c
(21)
In this case, the polarization resistance is identical
to the charge transfer resistance, and the Faradaic
impedance is frequency independent. Basically, the
frequency dependence of Z
F
can be expressed by
assuming that I
a
and/or I
c
are dependent not only on E
but also on one or more parameters, each of which
does not instantaneously follow E. As an example, if
only one parameter p
a
is involved in the anodic
component:

!I
!E
=
1
Z
F
=
!I
a
!E
p
a
+
!I
c
!E
p
a
+
!I
a
!p
a
E

x
!p
a
!E
(22)
In contrast with Equation (18), only the first two
terms on the right-hand side of Equation (22) give a
frequency independent contribution to Z
F
. The third
term is a function of the dynamic behavior of p
a
vs E, it
is time dependent, and it is controlled by an equation
of the general form:

d!p
a
dt
= *
a
!p
a
+
a
! (23)
If !E and, consequently, !p
a
are subjected to a
sinusoidal time variation of small amplitude, from
Equation (23) the frequency behavior of !p
a
is:
j"!p
a
= *
a
!p
a
+
a
! (24)
or

!p
a
(")
!E
=

a
j" *
a
(25)
At high frequencies, p
a
is frozen (!p
a
/!E = 0), i.e.,
it can no longer follow the variation of the electrode
potential in the way it does at sufficiently low
frequencies, when the steady state is achieved:
(e.g., carrying out a measurement down to 0.01 Hz
needs 10,000 s to have a 10% accuracy). Given the
poor steady state of the electrochemical systems, the
spectral analysis is difficult to perform down to very
low frequencies.
APPLICATION OF IMPEDANCE
TECHNIQUE TO CORROSION STUDIES
Through examples, it will be shown that
impedance techniques are particularly suitable for
estimating the corrosion rate, which is equal to:
x =
dW
dt
=
AI
a
zF
(16)
where W is the weight loss of the metal, A is the
atomic weight, z is the valence of dissolution, and l
a
is the anodic current of dissolution. In the past, I
a
was
often determined by measuring the polarization
resistance (impedance at f = 0). However, it has been
shown that impedance measurements over the whole
frequency range can give more accurate results.
4-7
Various corrosion processes have been investigated
by these techniques. Most of the studies have been
done on uniformly corroding surfaces, but localized
corrosion and stress corrosion cracking have begun
to be analyzed.
Estimation of Rate of Uniform Corrosion
In this type of corrosion, the dissolution of the
metal is uniform all over the surface of contact with the
electrolyte. The overall current I flowing through the
interface is:
I = I
a
+ I
c
(17)
where I
c
is the cathodic current. In the case of
spontaneous corrosion, the net measurable current is
I = 0; it is not possible to determine directly the
dissolution current I
a
to obtain x (Equation [16]).
If I
a
and I
c
are assumed to be potential dependent
and are able to follow instantaneously any variation of
this potential, then differentiating Equation (17) gives:

!I
!E
=
1
Z
F
=
!I
a
!E
+
!I
c
!E
=
1
R
p
(18)
where E is the electrode potential corrected for ohmic
drop.
Moreover, if I
a
and I
c
obey Tafel laws of slopes b
a
and b
c
I
a
= I
a
exp
2.303
b
a
E (19)
801
SCIENCE
CORROSIONVol. 48, No. 10
anodic and cathodic reactions, the relationships
between R
p
and I
corr
on one side and R
ct
and I
corr
on the
other. It is seen that I
corr
can be related to R
ct
in two
cases of interest (partial diffusion control and multistep
electron transfer) where the usual polarization
resistance technique would fail. The only case where
R
p
must preferably be used is that for passive
electrodes.
Similarly, for R
p
, the value of the technique is
primarily to indicate changes of corrosion rate with
respect to variables in the corrosive environment.
Values of the Tafel slopes were successfully used for
the corrosion of iron in acid media,
9
where both anodic
and cathodic reactions take place through a two-step
reaction. Usually, the Tafel slopes of the steps that are
not rate-determining cannot be easily obtained, and a
calibration procedure, based for instance on weight
loss data, must be used. The experimental data
presented below illustrate the relevance of impedance
measurements to corrosion rate estimation under
various conditions.
Active Corrosion of Iron
Anodic and cathodic reactions occur in two
consecutive, irreversible steps. Faradaic impedance
measurements provide the values of $
+
(l/b
a
i
) and
$
j
(l/b
c
j
) that are involved in the relation between R
ct
and I
corr
(relation 4, Table 1). A good agreement was
found between weight loss experiments and the
absolute value of the corrosion rate deduced from R
ct
for either a high-purity iron (0.8 mg x cm
2
x h
1
) or a
pure iron of industrial origin (Holzer type) (2 mg x
cm
2
x h
1
) (Figure 4)
9
.

"%0
lim

!p
a

(
"
)
!E

=

!p
a
!E
st

=

a
*
s
(26)
where (!p
a
/!E)
st
gives the steady-state value of
!p
a
/!E.
More generally, Equation (22) becomes:

1
Z
F
=
1
R
ct
+
!I
a
!p
a

a
j" *
a
(27)
Equation (27) accounts for the general features of
the frequency dependence of Z
F
stated above,
namely:
for "%& Z
F
= R
ct
(28)
and for "%0
1
Z
F
=
1
R
ct
+
!I
a
!p
a

!p
a
!E
st
=
1
R
p
At finite frequency values, Z
F
is frequency
dependent, according to Equation (27). Consequently,
exploring a narrow potential domain around the
corrosion potential does not eliminate, as it is
sometimes believed, the complications arising from
surface changes with potential. From the viewpoint of
the estimation of the corrosion rate, it is worth
considering the derivation given above in a number of
particular cases. It can be shown that Equation (27) is
able to provide a relationship between I
corr
and R
ct
instead of R
p
.
8
Obviously, both relationships tend to
become identical in those cases where R
ct
% R
p
.
Table 1 gives, under different assumptions for the
Relation to Corrosion Current I
corr
TABLE 1
Theoretical Relationships Between Corrosion Current and Charge Transfer and Polarization Resistances
(A)
Kinetics Control
Anodic Reaction Cathodic Reaction Polarization Resistance R
p
Charge Transfer Resistance R
ct
One step, tafelian electron One step, tafelian electron
transfer transfer
One step, tafelian electron Purely diffusional
transfer
One step, tafelian electron Mixed control, partly Complicated equation, depends
transfer diffusional upon the degree of control
by diffusion
n
a
irreversible tafelian, n
c
irreversible tafelian, Complicated equation,
consecutive steps(ba
i
) consecutive steps (ba
j
) depends upon the whole
set of rate constants
Passive dissolution One irreversible tafelian Complicated equation, depends
transfer on the passive area upon the kinetics of
dissolution and passivation
(A)
Under various anodic and cathodic kinetics (from <8>).
b b
2.303 (b b ) I
a c
a c corr
+
b b
2.303 (b b ) I
a c
a c corr
+
b
2.303 I
a
corr
b b
2.303 (b b ) I
a c
a c corr
+
b b
2.303 (b b ) I
a c
a c corr
+
b
2.303 I
a
corr
1
2.303
1
n
1
b
1
n
1
b
I
a a
i
c c
j
j 1
n
i 1
n
corr
c a
+
,
-
.
/
0
1
= =
$ $
802
SCIENCE
CORROSIONOCTOBER 1992
Inhibited Active Corrosion of Iron
Corrosion rate estimations based on R
ct
, R
p
, and
double-layer capacitance C
d
measurements were
compared to weight loss data for a Johnson-Matthey

iron in 1 M H
2
SO
4
as a function of an inhibitor
(propargylic alcohol) concentration (Figure 5). Figure
5(a) shows the impedance diagram at the corrosion
potential for 0 and 10 mM inhibitor concentrations. As
in the case of uninhibited iron corrosion (Figure 4), two
inductive loops are found at low frequencies. They
were proved to originate from the potential
dependence of the surface coverages by the inhibiting
species (hydrogen or hydrogen-bonded organic
compound) at lower frequencies and by the anodic
intermediate species FeOH at higher frequencies,
respectively. Relative corrosion rates with respect to
the uninhibited conditions can be expressed in terms
of the dimensionless inhibiting efficiency:
H =
100 (x
o
x)
x
o
(29)
where x
o
is the corrosion rate in the absence of an
inhibitor and x the corrosion rate in the presence of
an inhibitor; x is either directly measured by weight
loss (H
dir
) or calculated from the charge transfer
resistance R
ct
, the double-layer capacitance C
d
, or
the polarization resistance R
p
that are assumed
inversely proportional to x. As an example, the
inhibiting efficiency H
C
d
calculated from the change of
the double-layer capacity depends on a standardiza-
tion coefficient K
l
such as:
H
C
d
= 100 K
l

C
do
C
d
C
do
(30)
where K
l
is chosen so that H
C
d
= H
dir
at a 0.5 mM
concentration of propargylic alcohol.
As shown in Figure 5(b), the values of H obtained
from R
ct
agree with those of H
dir
, except at very high
concentrations where the corrosion rate estimated
from R
ct
is higher than the weight loss data. The
values of H
R
p
can be considered as acceptable, but,
beyond a 2 mM inhibitor concentration, they predict a
negative inhibition (i.e., an acceleration of the
corrosion) in complete disagreement with direct
measurements. According to the theoretical derivation
given above, this discrepancy is related to the
increasing size of the low frequency inductive arc as
the inhibitor concentration is increased, resulting in a
corresponding difference between R
p
and R
ct
. As for
the values of H
C
d
, they show the same type of error on
the inhibiting efficiency obtained from H
R
p
at high
inhibitor concentrations. Impedance techniques were
also used in the case where inhibition is caused by a
FIGURE 4. Impedance measurement of a corroding electrode.
Holzer-type iron in deoxygenated 1 M, H
2
SO
4
; disc electrode
rotation speed = 1,600 rpm, A = 0.2 cm
2
.
8
FIGURE 5. (a) Impedance measured 1 h after immersion for
Johnson-Matthey, 0.2 cm
2
disc in H
2
SO
4
, 2 N (diagram A) and
in H
2
SO
4
, 2 N + 10 mM p.a. (diagram B).
10
(b) Inhibiting
efficiency, with respect to the inhibitor (propargylic alcohol)
concentration, determined by: (1) weight loss, (2) R
ct
measurements, (3) R
p
measurements, and (4) double-layer
capacitance measurements (negative inhibiting efficiency
indicates the acceleration of corrosion).
8
(a)
(b)

Trade name.
803
SCIENCE
CORROSIONVol. 48, No. 10
three-dimensional protective layer in which the
inhibitor is incorporated.
11
Corrosion of Coated Metal
Not one of the steady-state techniques was found
ideally suitable as a corrosion rating test for painted
metals.
12
Faradaic impedance measurements on
painted iron (Figure 6[a]) showed that the metal is
corroded in the same way as when it is unprotected
but on a much reduced area at the place where the
paint layer present pore-like flaws. The insulating part
of the paint layer leads to a characteristic capacitive
behavior at high frequencies.
Corrosion rates calculated from R
ct
data were
compared to weight loss for a number of specimens
protected with epoxy paint from 0 to 80 m in
thickness (Figure 6[b]). Coated thin foils of high-purity
iron were exposed to attack in a 0.5 M H
2
SO
4
solution
for two to four days. The rate of corrosion continuously
increases during the exposure, and R
ct
was measured
as a function of time. The average calculated
corrosion rate was compared to weight loss at
different coating thicknesses. A very good correlation
is found in Figure 6(b): the thicker the layer, the better
the protection.
13,14
Anodized Aluminum Alloys
The resistance against corrosion of aluminum-
based alloys protected by an anodic oxide layer is
usually evaluated by a salt spray test that needs a
long time of exposure. The impedance method is
faster and has the same degree of reliability.
15,16
The
anodized sample to be tested is immersed in a 3%
NaCl solution, buffered at pH 4 and deoxygenated by
nitrogen bubbling. The impedance is measured at its
free corrosion potential. Figure 7(a) shows the
impedance diagram. Two capacitive contributions can
be resolved when the high frequency behavior is
enlarged (Figure 7[b]). This shorter time constant
probably arises from the dielectric property of the
oxide layer. A correlation between corrosion
susceptibility and impedance behavior must
reasonably be expected in the low frequency domain.
In fact, a good correlation was found between the low
frequency resistance R
LF
(Figure 7[a]) and the
exposure time at which the first pit opened. The
impedance technique being nondestructive,
impedance and salt spray tests can be applied
successively to the same sample so that a highly
significant comparison can be made. No Faradaic
model of corrosion is available in the case of
aluminum, and R
LF
can hardly be dealt with in terms of
either R
ct
or R
p
. Therefore, the value of the test must
be founded on a large number of experiments. Figure
7(c) gives, as an example, the selection made among
83 samples from the same batch on the basis of
impedance measurement or salt spray test. A pass
in the salt spray test is recorded if no pits have
appeared after 300 h of exposure to the spray. The
pass/fail threshold value of R
LF
(5.5 M2 x cm
2
) was
empirically determined. A similar correlation was found
between the two techniques for different compositions
of aluminum alloys and oxidizing baths.
15
Stainless Steel
The corrosion of stainless steel occurs in the
passive range. As shown in Table 1, the quantity used
for estimating the corrosion rate in the passive range
(a)
(b)
FIGURE 6. (a) Corrosion of iron coated with epoxy paint.
6
Impedance measurement of a corroding electrode: Armco

iron
(A = 60 cm
2
) covered by 40 m epoxy paint (ICI 5802022). (b)
Estimated corrosion rate by R
ct
measurement is compared to
the results of the weight loss measurement. Iron sheet covered
by various thicknesses of epoxy paint film; parameter is film
thickness in m; 0 corresponds to the bare specimen.

Trade name.
804
SCIENCE
CORROSIONOCTOBER 1992
is the polarization resistance R
p
. This is only an
assumption for stainless steel because the actual
dissolution mechanism is not known. However,
experimental results have shown the relevance of
such a choice.
17
As shown in Figure 8, the processes
involved in the corrosion process of stainless steel are
so slow that the low frequency limit, R
LF
, can often be
estimated only by extrapolation of a best-fitted half
circle for the experimental values.
Localized Corrosion
In contrast with uniform corrosion, localized
corrosion occurs in a few small places (pits, crevices,
cracks). From there, the corroded surface can either
be analyzed as a whole, taking into account several
corroding places, or each corroding place can be
analyzed in isolation. These approaches have been
investigated.
Pitting Corrosion
Concerning the impedance measured on
stainless steel in a chloride solution, which is known to
cause pitting if the potential is raised sufficiently in the
anodic range, the onset of pitting is marked by a
decrease in the modulus of the impedance, especially
in the low frequency range (Figure 9).
18
This special
shape has been interpreted by considering the
frequency dependence of the spatial current
distribution.
19
In the low frequency limit, the pit can be
regarded as a small conducting disk embedded in an
insulating plane. At high frequencies, the whole
electrode surface (both pit and passive surface) is a
perfect conductor (primary current distribution) due to
the double-layer capacitance. Therefore, the
electrolyte resistance of the whole electrode depends
on frequency.
Impedance measurements have been performed
for a pre-existing hole, made mechanically before
immersing the sample in a H
2
SO
4
, 2 N solution.
20
In
Figure 10(a), the polarization curves give the anodic
behavior of the electrode for various geometries of the
initial hole. Taking into account a coupling through
diffusion between dissolution and passivation, a model
explains the mean features observed on both the
polarization curve and the impedance diagrams
(Figure 10). However, it seems that the consideration
(a)
(b)
FIGURE 7. Corrosion of an anodized aluminum alloy.
15
(a) Impedance measurement of a corroding electrode: 2024T4 alloy anodized
in chromic bath (A = 50 cm
2
) in 0.18 M NaCl + 0.0115 M CH
3
COOH +0.02 M CH
3
COONH
4
. Impedance given for a unit surface area.
(b) Enlarged diagram in the high frequency range of the diagram (a). (c) Comparison of impedance measurement and salt spray test.
(c)
805
SCIENCE
CORROSIONVol. 48, No. 10
of metal cation hydrolysis and migration is necessary.
Hence, the diffusion in the occluded cell is responsible
for maintaining iron in an active state at a potential for
which it should be passive.
Stress Corrosion Cracking
Cracking under anodic dissolution control belongs
to a particular class of localized corrosion systems in
which the actively corroding site is located at the
bottom (crack tip) of a narrow electrolyte path. This
occurs also when thick layers of conducting and
porous corrosion products grow on the metal surface.
The sine wave response reflects the attenuation of the
alternating signal as a function of its penetration depth
parallel to a conducting wall. This situation has been
modelled by a one-dimensional transmission line,
which is essentially similar to the cylindrical pore
model developed for the potential distribution in
porous electrodes.
From a general point of view, it has been
attempted to show that, especially in the case of low-
carbon steels in carbonate/bicarbonate solutions at
70C, impedance measurements can give similar
information about the electrochemical reactions to that
obtained by the application of elastic stresses.
21
Measurements of the impedance of the electrode
in the active-passive transition and measurement of
stress corrosion at constant strain rate have been
performed on the same material. It has been shown
that the susceptibility to stress corrosion cracking for
low-carbon steel is directly related to a time constant,
, deduced from impedance data measured in the
passivation range.
Given an impedance diagram, such as that
depicted in the inset of Figure 11, if R
1
and R
2
are the
values of the intersection of the semicircle
approximating the low frequency passivating loop with
the real axis and "
o
* is the characteristic angular
frequency of this loop (inverse of the time constant
equal to the angular frequency of the maximum of the
loop), the value of is equal to:
3 =
1
"
o
*

R
0
+ R
1
R
0
where
1
R
0
=
1
R
2

1
R
1
(31)
The time constant can be related to the
incremental passivation. In Figure 11, it can be seen
that the maximum of corresponds closely to the
maximum susceptibility of the stress corrosion
cracking when it varies with respect to the potential.
22
APPLICATION OF NOISE ANALYSIS
TO CORROSION STUDIES
23
For a given polarization potential (or current), the
electrolysis current (or the rest potential) can be totally
FIGURE 8. Impedance measurement of stainless steel
corroding electrode, Fe 17Cr 16Ni 5.5Mo 2.7Cu 0.03C in
aerated 1.8 M H
2
SO
4
. Solid line: experimental results; broken
line: best-fitted half circle.
17
or partially random. When it is totally random (e.g.,
pitting), impedance techniques can hardly be used,
and the analysis of the current fluctuations is the only
way to reach information in the processes involved.
However, when it is only partially random (e.g.,
uniform corrosion), the two techniques can be used.
The observed random electric signals related to
corrosion are generally a succession of current (or
potential) pulses that have to be analyzed. Therefore,
the experimental results will be preceded by a short
mathematical background to outline the main
properties of the stochastic point processes that are of
basic importance in this topic.
Power Spectrum
of Stochastic Pulses Sequences
24,25
During corrosion sequences of elementary current
(or potential), transients are often observed (Figure
FIGURE 9. Complex plane impedance plot of 316 stainless
steel (UNS S31600) in 3% NaCl polarized at 55 mV vs SCE.
18
806
SCIENCE
CORROSIONOCTOBER 1992
FIGURE 10. (a) Polarization curves (obtained by using negative output impedance regulation) of a fixed iron electrode with artificial
pit in H
2
SO
4
, 2 N. The schematic section of the electrode shows the shape taken by the initial artificial pit (diameter d, depth h) after
dissolution: (1) flat electrode; (2) d= 2 mm, h = 2.5 mm; and (3) d= 1 mm, h = 3 mm. The flat electrode; (2) d = 2 mm, h = 2.5 mm; and
(3) d = 1 mm, h = 3 mm. The flat part of the electrodes (2) and (3) are covered with inert coating. (b) Impedance measurements for
the electrodes described previously. Diagrams (a) to (f) correspond to points (a) to (f) on the polarization curves.
20
12). These sequences can be modeled by assuming
the nucleation of the elementary current transient i
j
(t)
at times
j
, which form a point process. Therefore, the
observed current I(t) can be written under the form:
I(t) = i
j
(t 3
j
, u
j
)
$
j
N(t)
(32)
where N(t) is the counting process associated with
the point process and i
j
(t) is a random function that
can depend on another parameter u
j
(amplitude, time
constant).
The PSD of a random process (t) of zero mean
value is the Fourier transform of the correlation
function:
)
4
(") = e
i " t
< 44
3
> dt
&
+&
(33)
The process defined by:
807
SCIENCE
CORROSIONVol. 48, No. 10
4(t) = 5 (t 3
j
$
j
N(t)
(34)
will play a special role as it is only characteristic of
the sequence of arrival times of the point process
(e.g., nucleation times of the pits).
The power spectral density of such pulse
sequences have been calculated, assuming that the
elementary events i(t) are deterministic.
If F(") is the Fourier transform of the current
transient i(t)
F(") = i(3) e
i " 3
d3
&
+&
(35)
The mean and the PSD of the global current l(t),
called a filtered point process, are given by:
<I> = F(0) <> (36)

I
(") = F(")
2
S{,"} (37)
where <> and S{,"} are the mean and the PSD of
the process (t). In the case of the point processes
involved in corrosion modeling, S{,"} can be sup-
posed to be:
(1) Poisson process with intensity :
S{,"} = (38)
In this case, it can be shown that when the current
transient i(t) shows a sudden birth or a sudden death,
the high frequency limit of the PSD varies like 1/f
2
,
whatever the shape of the transient. A slow birth or a
slow death give rise to a 1/f
4
trend.
(2) Doubly stochastic process (i.e., a Poisson pro-
cess with a random intensity [t]):
S{,"} = +

(") (39)
where and

(") are the mean and the PSD of (t).


(3) Renewal process (the interval distribution is
imposed):
S{4,"} =
2
<4>
1 + 2 Re
6(")
1 6(")
(40)
where Re means real part, and 6(") is the Fourier
transform of the probability distribution of the interval
times.
(4) Triggered Poisson process (a Poisson process
triggers another Poisson process that forms bunches
of points):
FIGURE 11. (a) Stress corrosion cracking (SCC) study for steel
GCSl in solution 0.325 M NaHCO
3
+ 0.325 M Na
2
CO
3
(pH 9.7);
disc electrode rotation speed = 270 rpm. Variation in susceptibility
to SCC vs electrode potential (- - -) and corresponding current-
voltage curve (_ _ _). (b) SCC study for steel GCSl in solution
0.325 M NaHCO
3
+ 0.325 M Na
2
CO
3
(pH 9.7); disc electrode
rotation speed = 270 rpm. Variation of in the active-passive
transition (- - -) and the corresponding current voltage curve
(_ _ _).
22
FIGURE 12. Observed total current l(t) represented as the sum
of elementary current transients i
j
(t) born at time
j
.
If the elementary transients are independent of their
duration and of the bunch duration, the PSD of l(t) is:
)
i
(") =
2*(7+8)
78 {(7+8)
2
+ "
2
}
* F(")
2
(41)
where * is the convolution product.
808
SCIENCE
CORROSIONOCTOBER 1992
Study of Localized Corrosion
Through Electrochemical Noise
The kinetics of repassivation following the film
breakdown are regarded as determining for the
corrosion resistance of passive metals against more or
less localized attacks.
26
Two examples of
characterization of the breakdown and healing
processes have been given in the cases of a
mechanical and chemical depassivation of the metal.
Mechanical Depassivation
27,28
A mechanical abrasion of the passive surface of
iron in sulfuric acid medium is obtained by an
impinging jet of SiC particles in suspension in the
electrolyte. The resulting passive current is shown in
Figure 13. The particle impacts at the surface occur at
random in time and space, and the visible damage is
restricted to small spots (1 m
2
). The concentration of
SiC particles is estimated by means of a
phototransistor that measures the optical absorption of
the light by the particle suspension in the electrolyte.
If the impact times are assumed to follow a doubly
stochastic Poisson law, whose rate has a mean and
a PSD

(f), the PSD of the current under abrasion


from Equations (37) and (39) is equal to:

I
(f) = { +

(f)} x I(f)
2
(42)
where I(f) is the Fourier transform of the elementary
current transient due to one impact.
The comparison of
i
(f) (Figure 14, curve A) and

(f), obtained from optical absorption, demonstrates


that the 1 Hz cut-off frequency is due to the
fluctuations of the abrading particle density. The
spectrum of the current, corrected for the
fluctuations by taking into account the optical
response (Figure 14, curve B), shows only one cut-off
frequency in the high frequency range that can be
related to the repassivation kinetics. Its 1 kHz value
(i.e., a 0.16 ms time constant) demonstrates that the
repassivation time is very short and can be related to
the electrochemical characteristic of the whole
surface.
Chemical Depassivation
19,20
Investigations of chemical breakdowns of the
passive layer on iron in sulfuric acid medium were
performed after triggering by addition of Cl

ions.
29-32
The current noise and the mean passive current have
been continuously recorded from the chloride addition.
The current time recording (Figure 15) shows six
principal peaks, according to a direct surface
observation that exhibits six principal pits.
33,34
The
exponential-like current growth is followed by a steep
repassivation. Whenever the falling current returns to
its initial value, the noise spectrum coincides again
with that of a passivated electrode, indicating that pit
repassivation is achieved. The current noise
measured during the current growth reveals a great
pitting activity due to a large amount of small
secondary pits of polygonal shapes (r 9 10 m), which
can be observed inside the large pit (e.g., r = 400 m).
It has been noticed that the density of the small pits is
much greater inside the large pits than on the
remaining electrode surface. This may be related to
the strongest aggressivity of the solution inside the
principal pit. The secondary pits are responsible for
the current fluctuations detected from the
electrochemical noise spectrum (Figure 16), which
increase when the pitting activity increases. A cut-off
frequency (60 Hz) and a low frequency plateau are
FIGURE 14. PSD of (a) total current fluctuations during abra-
sion; (b) corrected for particle concentration fluctuations (mean
current l = 20 A; E = 0.3 V/sulfate-saturated electrode [SSE];
particle size: 100 m).
FIGURE 13. Passive current fluctuations during abrasion (low
particle concentration, particle size: 200 m).
809
SCIENCE
CORROSIONVol. 48, No. 10
FIGURE 16. PSD of the current after chloride addition
(mean current density j = 30 A x cm
2
).
FIGURE 15. Current-time recording after chloride addition
(c = 3.10
3
M); Fe in 0.5 M H
2
SO
4
, E = 0.2 V/SSE, disc diameter
5 mm).
detected. This can be accounted for by assuming a
random nucleation of the secondary pits following a
Poisson law. Large pits grow by coalescence of the
secondary ones and by general attack.
It is noticeable that, in the case of iron-base
alloys, the current PSD often increases when
frequency decreases and does not show a white noise
low frequency limit that is easier to interpret.
Especially when the elementary current transients are
clearly related to a finite electric charge, this 1/f
behavior has to be ascribed to the point process itself:
either to a nonstationary pit generation rate or to some
special distribution. In the later case, similarly to any 1/
f noise in other physical or chemical domain, the
interpretation is not clear at all, and this special
behavior can lead to problems even in the counting
procedure.
Study of Uniform Corrosion
Through Electrochemical Noise
10,35
The corrosion rate of a metal immersed in an
aggressive medium has been investigated in the case
of the corrosion of iron in acidic medium by analyzing
the random changes of the corrosion potential.
At low anodic current density, the dissolution of
iron in sulfuric acid medium takes place with hydrogen
evolution. The measured total current is the sum of the
iron dissolution current and the hydrogen evolution
current that are indistinguishable. The hydrogen
evolves through bubbles, which causes fluctuations in
the measured current or voltage. The time recording of
the electrode potential under galvanostatic control is
given in Figure 17 for an electrode facing upwards. An
optical observation shows that the slow increase is
due to the growth of a hydrogen bubble, although the
steep jump is due to the bubble detachment.
FIGURE 17. Fe dissolution in 2 N sulfuric acid medium.
Voltage-time recording (mean anodic current density J = 3.7
mA x cm
2
, disc diameter 5 mm).
The elementary voltage transient can be
approximated by a linear function of time such as:
v(t-s,u) = k.(t-s) (t-s,u)
where (t-s,u) = 1 if t-s < u (43)
(t-s,u) = 0 if t-s 9 u
where s is the random birth time of the bubble that is
assumed to follow a Poisson law N(t) of intensity ,
and u is the random life time that is assumed to be
exponentially distributed with parameter .
The voltage increase related to H
2
bubble
evolution is supposed to be such as:
810
SCIENCE
CORROSIONOCTOBER 1992
V(t) = v(t-s
i
,u
i
)
$
i=1
N(t)
Hence,
<V(t)> =
*k
7
2
and c
V
(f) =
2* k
2
(3 7
2
+ 4 :
2
f

7
2
(7
2
+ 4 :
2
f
2
)
2
(44)
In Figure 18, an example of the PSD of the
random fluctuations of the corrosion potential at a free
electrode
V
(f) is given. The measurements are made
between 10
2
to 10
2
Hz. The lower frequency is
imposed by the natural drift of the system that is of the
order of 10 m V x mn
1
. Above 10
2
Hz, the noise of
interest is buried in the parasitic noise coming from the
regulation.
8
A careful inspection of the PSD of the
voltage fluctuations shows that below a frequency of
the order of 10
1
Hz, the electrochemical noise can be
considered as white, i.e., the PSD seems independent
of the frequency and equal to
V
(0).
As the experimental PSD of the voltage
fluctuations is in rather good agreement with the PSD
predicted by the theoretical model (Figure 19), the
characteristic parameters , k, and <V> can be
extracted. They are given in Table 2 for two types of
iron immersed in the sulfuric acid solution at pH 0
together with the charge transfer resistance obtained
from impedance measurements. It can be noticed that
when the corrosion rate of iron increases (i.e., the
charge transfer resistance decreases), the parameter
<V> increases.
FIGURE 18. PSD of the voltage fluctuations corresponding to
Figure 17.
FIGURE 19. Calculated PSD from Equation (44).
The voltage fluctuations related to bubble
evolution are due to a screening effect of the hydrogen
bubble that grows on the electrode surface. Therefore,
the mean of the corrosion potential fluctuations<V>
can lead to a measure of the volume of hydrogen that
evolves on the iron surface. As at the corrosion
potential, the hydrogen current is equal to the
dissolution current of iron; an increase of <V> means
an increase of the corrosion current of the metal. This
analysis of the fluctuations of the corrosion potential
can lead to an effective way to estimate the corrosion
rate without perturbing the electrochemical system.
REFERENCES
1. C. Gabrielli, M. Keddam, Electrochim. Acta 19, 355 (1974).
2. D.D. Macdonald, Transient Techniques in Electrochemistry (New York,
NY: Plenum Press, 1977).
3. C. Gabrielli, F. Huet, M. Keddam, Electrochim. Acta 31, 1025 (1986).
4. C. Gabrielli, Identification of Electrochemical Processes by Frequency
Response Analysis, Technical Report no. 4/83 (Farnborough, United
Kingdom: Schlumberger, 1981).
5. C. Gabrielli, Use and Applications of Electrochemical Impedance Tech-
niques, Technical Report (Farnborough, United Kingdom:
Schlumberger, 1990).
6. C. Gabrielli, M. Keddam, H. Takenouti, Treatise on Materials Science
and Technology, Vol. 23, Corrosion: Aqueous Processes and Passive
Films, J.C. Scully, ed. (London, United Kingdom: Academic Press,
1983).
7. J.R. Macdonald, Impedance Spectroscopy (New York, NY: John Wiley
and Sons, 1987).
8. I. Epelboin, C. Gabrielli, M. Keddam, H. Takenouti, Proc. of the ASTM
Symp. Progress in Electrochemical Corrosion Testing, F. Mansfeld, U.
Bertocci, eds. (Philadelphia, PA: ASTM, 1981).
9. I. Epelboin, M. Keddam, H. Takenouti, J. Appl. Electrochem. 2, 71
(1972).
10. C. Gabrielli, F. Huet, M. Keddam, A. Macias, Proc. of the Symp. Sur-
faces, Inhibition and Passivation, M. Cafferty, R.J. Brodd, eds. 86-7
(Pennington, NJ: The Electrochem. Soc., 1986), p. 507.
11. W.J. Lorenz, F. Mansfeld, Electrochim. Acta 31, 467 (1986).
12. G.W. Walter, Corrosion Sci. 26, 39 (1986).
TABLE 2
k <V> R
Ct
Metal s
1
V x s
1
V 2
Iron I 0.715 37 52 50
Iron II 0.357 55 154 11.5
811
SCIENCE
CORROSIONVol. 48, No. 10
13. L. Beaunier, I. Epelboin, J.C. Lestrade, H. Takenouti, Surf. Technol. 4,
237 (1976).
14. G.W. Walter, J. Electroanal. Chem. 118, 259 (1981).
15. J.J. Bodu, M. Brunin, I. Epelboin, M. Keddam, G. Sertour, H. Takenouti,
Aluminio 46, 277 (1977).
16. F. Mansfeld, M.W. Kendig, J. Electrochem. Soc. 135, 828 (1988).
17. C. Gabrielli, M. Keddam, H. Takenouti, V.Q. Kinh, F. Bourelier,
Electrochim. Acta 24, 61 (1979).
18. M.G.S. Ferreira, J.L. Dawson, Passivity of Metals and Semiconduc-
tors, M. Froment, ed. (Amsterdam, The Netherlands: Elsevier, 1983),
p. 359.
19. R. Oltra, M. Keddam, Corrosion Science 28, 1 (1988).
20. I. Epelboin, C. Gabrielli, M. Keddam, H. Takenouti, Z. Physikalische
Chem. NF98. 215 (1975).
21. R.D. Armstrong, A.C. Coates, Corrosion Sci. 16, 423 (1976).
22. R.D. Armstrong, A.C. Coates, J. Electroanal. Chem. 50, 303 (1974).
23. C. Gabrielli, F. Huet, M. Keddam, H. Takenouti, Proc. of the 8th Euro-
pean Meeting of Corrosion 2, 37 (Paris, France: CEFRACOR,1985).
24. R.L. Stratonovich, Topics in the Theory of Random Noise, 1 (New York,
NY: Gordon and Breach, 1963).
25. C. Gabrielli, F. Huet, M. Keddam, R. Oltra, Advances in Localized Cor-
rosion, H.S. Isaacs, U. Bertocci, J. Kruger, S. Smialowska, eds., Intern.
Conf. Series, 9 (Houston, TX: NACE, 1991), p. 93.
26. N. Sato, J. Electrochem. Soc. 123, 1197 (1976).
27. C. Gabrielli, F. Huet, M. Keddam, R. Oltra, J.C. Colson, Proc. Symp.
Equilibrium Diagrams and Localized Corrosion, R.P. Frankenthal, J.
Kruger, eds., 84-9 (Pennington, NJ: The Electrochemical Society,
1984), p. 522.
28. C. Gabrielli, F. Huet, M. Keddam, J.C. Colson, R. Oltra, Proc. 9th In-
tern. Meeting on Metallic Corrosion 3, 427 (1984).
29. D.E. Williams, C. Westcott, M. Fleishman, J. Electrochem. Soc. 132,
1796 (1985).
30. D.E. Williams, C. Westcott, M. Fleishman, J. Electroanal. Chem. 180,
549 (1984).
31. U. Berocci, Y.X. Ye, J. Electrochem. Soc. 131, 1011 (1984).
32. G. Okamoto, K. Tachibana, S. Nishiyama, T. Sugita, Passivity and Its
Breakdown On Iron and Iron Base Alloys, USA-Japan seminar, R.N.
Staehle, H. Okada, eds. (Houston, TX: NACE, 1976), p. 106.
33. C. Gabrielli, F. Huet, M. Keddam, R. Oltra, C. Pallotta, Passivity of
Metals and Semiconductors, M. Froment, ed. (Amsterdam, The Neth-
erlands: Elsevier, 1983), p. 293.
34. C. Gabrielli, M. Keddam, M. Krarti, C. Pallotta, 166th Meeting of the
Electrochemical Society, New Orleans, Ext. Abstract (1984), p. 288.
35. C. Gabrielli, F. Huet, M. Keddam, J. Appl. Electrochem. 15, 503 (1985).

S-ar putea să vă placă și