You are on page 1of 8

A study of localized corrosion susceptibility of different alloys by means of

Electrochemical Noise Analysis

G. Montesperelli1, G. Gusmano2

1
Dipartimento di Fisica ed Ingegneria dei Materiali e del Territorio - Università di Ancona
2
Dipartimento di Scienze e Tecnologie Chimiche - Università di Roma "Tor Vergata"

Abstract

The main literature regarding the application of Electrochemical Noise Analysis (ENA) to the study
of corrosion system is very controversial. Over the last decade ENA has received great attention
from researchers, but in many cases results are discording.
The aim of this paper is to show the effective advantage of the use of ENA for localized corrosion
monitoring of different systems. The setting of the main important parameters affecting the
accuracy of the method (such as instrument noise, direct current (DC) trend, noise sampling rate
and sampling duration) will be discussed.
As examples, the results of crevice and pitting corrosion test on AISI 430 stainless steel and on
ASTM C68700 Al brass in NaCl 3% solutions will be discussed. A comparison between ENA
signals coming from pitting and crevice test will be reported.
Finally, an example of the application of the Discriminant Analysis on ENA data in the frequency
domain will be given.

Keywords: Electrochemical Noise Analysis, Pitting corrosion, Crevice corrosion, Al brass,


Stainless steel, Corrosion monitoring.

Introduction

The use of Electrochemical Noise Analysis (ENA), i.e. the study of current and potential spontaneous
fluctuations of corroding electrodes, is very controversial.
From an analysis of main literature, it can be pointed out that ENA is a very powerful tool for
monitoring rapidly changing electrochemical processes such as passive film build up or breakdown
[1-3]. Moreover ENA works in free corrosion condition, without the need of applying an external
perturbation that could influence the onset the evolution of corrosion phenomena. For this reason
ENA is a very suitable method for studying localized corrosion phenomena.
Nevertheless a general quantitative correlation between ENA statistical parameters (standard
deviations and/or parameters in the frequency domain) and corrosion rate valid for every system has
not been found [4].
In the last years, the formal identity between electrochemical noise resistance (R n), defined as the
ratio of potential and current standard deviations of noise acquisitions, and polarization resistance
(Rp) has been accepted from a theoretical point of view [5].
Many discussions can be found in the recent literature regarding the experimental aspects of the
method and the elaboration of data. Recently some papers [6, 7] remarked the instrumental and
experimental factors that can affect the acquisition and the calculation of R n.
In this paper, some examples of ENA application on corroding electrodes are given. ENA data of
pitting and crevice corrosion tests of AISI 430 stainless steel and of C68700 Al-brass are reported.
Tests were carried out in NaCl solutions. In particular many attention has been given to the
possibility of discriminating the two different forms of corrosion by ENA.
This paper also examines the use of discriminant analysis to single out the ENA frequency bands
that contain most of the information in order to distinguish between two classes of behaviour (e. g.
corrosion attack or passivation).

Experimental
Tests were carried out on AISI 430 stainless steel and on ASTM C68700 Al brass.
A three-electrodes electrochemical cell was used for noise monitoring. Stainless steel electrodes
consisted of 120 by 25 mm plate, 0.7 mm in thickness. In order to induce a crevice corrosion attack,
a crevice former was attached to one electrode (working electrode). The crevice former resulted in
two 30 by 15 mm PVC blocks, with four grooves 3 mm wide and 1 mm deep. The crevice ratio
calculated as bare electrode area/crevice area (A c/Aa) was 5. The other electrodes were bare
stainless steel plates.
Al brass electrodes were obtained from pipe whose outside diameter was 19.05 mm. The crevice
former was realized with one 14 by 6 mm PVC block with four grooves 1.2 mm wide and 1 mm
deep, following the bend radius. The crevice ratio was 13.37.
Pitting tests were carried out on three bare electrodes.
Samples were prepared by abrading the specimen with 500 grit paper, washing in water then in
ethanol and drying in air.
Tests were carried out in a aerated stagnant solution of 3 % sodium chloride.
Potential and current noise signals were simultaneously recorded using a Solartron 1285
Potentiostat/Galvanostat.
The instrumental noise was measured following the experimental procedure described in [6].
The results obtained, in terms of current and potential average values ( ) and their standard
deviations (I, V), are summarized in table 1. In all the cases the intrinsic level of instrument noise
was negligible with respect to noise level measured during corrosion tests.

Table 1: Instrumental noise values

(A) 1.04 10-10


(V) 1.52 10-05
I (A) 6.87 10-12
V (V) 9.57 10-07

The choice of the sampling interval has been optimised by means of Discriminant Analysis on
preliminary tests, as already described [8] and as it will discussed later in results and discussion
section.
Data sets of 1024 readings were recorded at a sampling interval of 31.25 ms (sampling
frequency = 32 Hz) for stainless steel test. For Al-brass test a sampling interval of 250 ms
(sampling frequency = 4Hz) was used.
In some cases, a large DC trend was recorded during the period of noise acquisitions. It has been
demonstrated that [9,10] this DC trend influenced the statistical calculation of I, V. For this
reason, after data collection the dc linear trend was removed by linear regression removal method.
Data were transformed in the frequency domain through the Maximum Entropy Method (MEM)
and Fast Fourier Transform (FFT) algorithms performed by a dedicated software. All standard
deviation values reported and Power Spectrum Density (PSD) plots have been averaged through
more acquisitions.
Results and discussion

The mechanisms of crevice and pitting corrosion are well known [11-14] and, in the case of
samples in the active state can be sketched in four stages: passivation, activation, initiation,
propagation. The main difference in their mechanism is in the activation stage. In the case of
crevice corrosion, the anodic and cathodic areas are already fixed by the geometry of the system. In
pitting corrosion, during the activation stage a random metal surface differentiation occurs in anodic
and cathodic areas.
In figure 1, the current and potential signals for a crevice test on stainless steel, collected after two
hours of immersion are reported. These signals are characteristics of the first hours of immersion in
which a passive film is growing on metal surface. As a consequence of this, a depletion of oxygen
within the crevice occurs. Current and potential fluctuations are in a very narrow range (2.37 10 -7 A
and 2.11 10-4 V respectively) and they are going to be attenuated with immersion time. The standard
deviations assumed very low values (4.84 10-8 A and 7.07 10-5 V respectively).
After few hours, the current signal began to show some transients coupled with strong and
simultaneous drift in the potential signal (figure 2).
This behaviour is reported to be characteristic of localized corrosion of iron and iron-base alloys
[15]. The range of fluctuation sensitively increased (4.15 10-6 A and 2.43 10-3 V respectively) as
well as standard deviations (1.24 10-6 A and 1.01 10-3 V).

Figure 1: Current and potential signals for a Figure 2: Current and potential signals for a
crevice test on stainless steel after 2 hours of crevice test on stainless steel after 8 hours of
immersion (stage a). immersion (stage b).

The overall process can be easily followed by plotting the current and potential standard deviations
(I and V) as a function of immersion time. Figure 3 shows the I trend as a function of time. The
reported I values are the average of five acquisitions. The first four hours are characterized by the
already mentioned decreasing of I. In this stage, on the bare surface of the metal, a passive film
start to grow, by reaction with oxygen. Between 4 and 12 hours, the oxygen concentration inside
crevice begins to decrease thus creating the conditions for the activation of the surface. I is still
low, but some steep increases can be detected. Between 12 and 33 hours a sharp rise of standard
deviation can be detected thus characterizing the initiation stage.
The total time taken from immersion to reach this rising can be assumed to be the initiation time for
crevice corrosion in this condition.
The final stage is controlled by the permanent breakdown of the passive film in the anodic areas
inside the crevice
The error bars reported on figure 3, have been calculated the standard deviation of . It is worth
pointing out that the error bars are larger in the initiation step than during propagation. This is due
to the conflict between the passive film build up an its break down. V gave similar information.

Figure 3: Current standard deviation as a Figure 4: Noise resistance values for five
function of time for a crevice test on stainless different tests for a crevice test on stainless
steel. steel. Data are sorted by stage.

Results of noise resistance are shown in figure 4. Data coming from five different tests are reported.
The values are sorted by stage, and each point has been averaged within the values coming of the
same stage.
Rn permitted to well discriminate the first three stages of the mechanism and some uncertainty
remained in differentiating the onset from the propagation of the attack. As expected, R n showed a
slight increase moving from the first to the second stage due to the progressive passivation and then
a strong decrease during the onset of corrosion attack.
In the case of pitting test, analogous ENA signals were achieved, since the mechanism are very
similar.
It is interesting to compare data coming from pitting and crevice test. In figure 5, data from
different pitting and crevice tests are reported.

Figure 5: Comparison of current standard Figure 6: Current standard deviation as a


deviation for crevice and pitting tests on function of time for a pitting test on Al-brass.
stainless steel.
Data coming from the stage a (passivation) are very close, since I values are mostly determined by
the initial stage of metal preparation and the environmental condition, that are the same for both
test. During the stage b (activation) I differs, since the surface involved in crevice activation
(anodic area) is higher than that involved during pitting.
I of signals collected during stage c (initiation), are the same because the mechanisms involved at
this point are exactly the same.
In the stage d (propagation) I is higher for crevice than for pitting since the corroded area is larger
than the pitted one.
Pitting tests performed on Al brass lasted 28 days. In figure 6, the trend of current standard
deviation is reported. Also in this case is possible to identify the four stages even if the transition
from activation to initiation stage is put in evidence by a slight increasing of I from 1.92 10-8 A to
6.30 10-8 A.
A more clear discrimination of the different stages was obtained by the data transposition in the
frequency domain by FFT or MEM algorithms and by plotting the Power Spectral Density (PSD) as
a function of frequency. As an example, the PSD of stage a and c, are showed in figure 7. Many
differences can be observed., The output at frequency lower than 500 mHz is a factor 10 lower for
stage c than for stage a. Moreover curve related to stage a showed a sharp peak at a single
frequency (~1Hz) that become a wide peak in stage c. This seems to suggest that crevice corrosion
results in a defined cyclic fluctuation of noise characteristics of the electrode. Even though it has
been already reported a behaviour like this [16], our tests on a number of other systems does not
confirm this findings.
The frequency analysis can also give information on different type of corrosion. It has already been
reported that slopes of 20 dB/dec or less in the potential spectrum are indicative of localized
corrosion, while a roll-off of more than 20 dB/dec characterises a passive state [4,17].
During the first stage, a slope of 42.6 dB/dec was detected, thus confirming the tentative build up of
a passive layer. After 200 hours from immersion (i.e. at the beginning of stage b) the slope
decreased at 22.4 dB/dec and remained at that value until the end of the test. Slopes never assumed
values lower than 20 dB/dec.

Figure 7: Current PSD for a pitting test on Al Figure 8: discriminant function Dr calculated
brass. from current FFT data coming from stages a
and b for a pitting test on Al-brass.

Recently the Discriminant Analysis has been introduced as a simple criterion for the ability of ENA
to discriminate between two situations or two different stages of the same mechanism [8]. Equation
1 is able to quantify the information available at a single frequency, being ƒ 1 and ƒ2 the PSD at the
frequency  as calculated by FFT coming from two different time acquisitions [17].
(1)

Equation 2 shows the same equation expressed as percentage of total available information

(2)

where N is the number of frequency points (512) and JN(1,2) is given by:

(3)

Equation (1) will have a value of 0 if the two classes of time record are the same and greater than 0
if they differ. Equation (2) will have the same trend as (1) but its values are expressed as
percentage.
Since the highest and lowest noise signal frequency that can be resolved in the frequency domain
are given by the relationships (4):

(4)

Δt being the sampling interval and N is the number of acquisition in a noise recording run (1024 in
our tests), ƒmin and ƒmax can be optimised by changing N and Δt, and following the indication of
discriminant analysis.
An application of Discriminant Analysis is shown in figure 8 in which the function D r, calculated
from current FFT data acquired during the first two stages (a and b) are reported. The Discriminant
Analysis clearly shows that the main discriminatory information between stages a and b are
available at low frequencies. More than 90 % of these information can be achieved at frequencies
within 0.4 Hz.

Figure 9: discriminant function Dr calculated Figure 10: discriminant function Dr calculated


from current FFT data coming from stages b from FFT data coming from stages c and d for
and c for a pitting test on Al brass. a pitting test on Al brass.
The same analysis performed on data related to stage b and c, gave a very similar trend but a
contribution from intermediate frequency was found. A 60 % of information can be detected at
frequencies lower than 0.25 Hz , 11 % at frequencies between 0.6 and 1.1 Hz and 8.3 % between
1.2 and 1.4 Hz (figure 9). The application of discriminant analysis to data coming from stages c and
d (figure 10) showed a wide distribution of the meaningful frequencies. With respect to data related
to the other stages, the contribution at low frequencies markedly decreased while contribution at
intermediate frequencies (0.6-1.1 and 1.2-1.4 Hz) increased.

A comparison between pitting and crevice


tests on Al-brass can be done on the basis of
the analysis of PSD plot, showed in figure 11.
A first difference can be seen on the slopes.
Except the slope related to stage a, which
showed a very high value (48.9 dB/dec)
characteristic of a passive state, the other
values were lower than 20 dB/dec (17.6, 14.9
and 17.7 dB/dec respectively). In the case of
pitting, slopes remained always at values
higher than 20 dB/dec, probably due to the
very long duration of passive stage that
Figure 11: Current PSD for a crevice test on permitted the formation of a very stable
Al brass. passive film.

An other interesting difference can observed on Discriminant Analysis data. The distribution of
meaningful frequencies between stages a and b and stages b and c, are very similar: 77-80 % of
information in the range 0-0.3 Hz. Stages c and d are characterized by a wide distribution of
frequencies.

Conclusion

The ENA permitted an overall characterization of the localized corrosion (pitting and crevice) on
AISI 430 stainless steel and ASTM C68700 Al brass in 3% sodium chloride solution.
The ENA revealed the four stages of localized corrosion mechanism in all the cases. The time
recording analysis, revealed characteristic transients in the current signal, typical of this corrosive
attack. The analysis of the standard deviations, clearly evidenced the duration of each stage and
permitted to calculate the initiation time for the attack.
The values of PSD slopes give further information on the stability of the passive state.
A comparison of noise data for pitting and crevice tests has been given.
It has been shown an example of application of the discriminant analysis. This is a simple criterion
for the ability of ENA to discriminate between two situations and it can give useful information to
optimise the noise acquisition parameters (sampling frequency and duration) that influence the
sensitivity of the method.
References

1. Y. F. Cheng M. Wilmott, J. L. Luo: Corrosion Science 41 (1999) 1245.


2.F. Mansfeld, L.T. Han, C.C. Lee, G. Zhang: Electrochimica Acta 43 (1998) 2933.
3. P. R: Roberge, D. R. Lenard: J. Applied Electrochemistry 28 (1998) 405.
4. G. Gusmano, G. Montesperelli, S. Pacetti, A. Petitti, A. D'Amico: Corrosion 53 (1997) 860.
5. J.F. Chen, W.F. Bogaerts: Corrosion Science 37 (1995) 1839.
6. Y.J. Tan, S. Bailey, B. Kinsella: Corrosion 55 (1999) 469.
7. J. Goellner, A. Burkert, A. Heyn, J. Hickling: Corrosion 55 (1999) 476.
8. G. Gusmano, F. Marchioni, G. Montesperelli: Materials and Corrosion, 51, (2000) 537.
9. G. Gusmano, G. Montesperelli, S. Pacetti: A Study of the Factors Affecting Electrochemical
Noise of Carbon Steel in Natural Waters, Proceedings of the 8th European Symposium on
Corrosion Inhibitors (8 SEIC), Ann. Univ. Ferrara, N.S., Sez.V, Suppl. N.10, 1995, p. 561.
10. Y.J.Tan, B. Kinsella, S. Bailey, Corros. Sci. 38 (1996) 1681.
11. J.W. Oldfield, W.H. Sutton: Br. Corros. J. 13 (1978) 13.
12. J.W. Oldfield, W.H. Sutton: Br. Corros. J. 13 (1978) 104.
13. H. H. Uhlig, J. Gilman: Corrosion 19 (1963) 261t.
14. Y. Kolotyrkin, Corrosion 19 (1963) 261t.
15. C. Gabrielli, F. Huet, M. Keddam, R. Oltra: Corrosion 4 (1990) 266.
16. K. Hladky, J. L. Dawson: Corrosion Science 3 (1982) 231.
17. J.C. Uruchurtu, J.L. Dawson: Corrosion 43 (1987) 19
18. G.R. Dargahi-Noubary, P.J. Laycock: J. of Time Series Analysis 2 (1981) 71.

You might also like