Journal of
Electroanalytical Chemistry Journal of Electroanalytical Chemistry 578 (2005) 143–150 www.elsevier.com/locate/jelechem
Analysis of pitting corrosion behavior of pure Al in sodium chloride solution with the wavelet technique Chao Cai a
a,b
, Zhao Zhang b,*, Fahe Cao b, Zuoning Gao a, Jianqing Zhang b,c, Chunan Cao b,c
Key Laboratory of Energy Sources and Chemical Engineering, Department of Chemistry, Ningxia University, Yinchuan 750021, PeopleÕs Republic of China b Department of Chemistry, Zhejiang University, Hangzhou 310027, PeopleÕs Republic of China c State Key Laboratory for Corrosion and Protection of Metals, Institute of Metal Research, The Chinese Academy of Sciences, Shenyang 110016, PeopleÕs Republic of China Received 1 April 2004; received in revised form 9 December 2004; accepted 11 December 2004 Available online 5 February 2005
Abstract The potential electrochemical noise (EN) measurements in conjunction with the scanning electron microscopy (SEM) technique have been used to study the corrosion behavior of pure aluminum in neutral 3.0 wt% sodium chloride solution. EN information on the evolution of pitting corrosion over a period of five days has been obtained and analyzed with a wavelet transform technique. The results show that the EN signal was composed of a distinct type of events characterized by small scaling values, i.e., their time constants, and the wavelet transform technique is a useful alternative tool to overcome the limitations of FFT. Meanwhile, it is found that the energy distribution plots (EDP) can be used as ‘‘fingerprints’’ of EN signals, and good correspondence between the characteristics of EDP and the material corrosion type/severity has been obtained. Finally, the relationship between the EN features and the corrosion morphologies has been elucidated according to the ‘‘autoacceleration’’ and ‘‘cathodic protection’’ mechanisms of the corrosion process. Ó 2005 Elsevier B.V. All rights reserved. Keywords: Pure aluminum; Pitting corrosion; Electrochemical noise; Wavelet analysis
1. Introduction The measurements of electrochemical noise (EN), i.e., the spontaneous fluctuation of the potential and/or current generated during the corrosion process, have received considerable attention in recent years [1–4]. It is generally accepted that EN analysis can provide fundamental information about the nature of the corrosion process [5–10]. The analysis of EN data can be performed in the time domain by investigating its shape, *
Corresponding author. Tel.: +86 571 87952318; fax: +86 571 87951895. E-mail address:
[email protected] (Z. Zhang). 0022-0728/$ - see front matter Ó 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jelechem.2004.12.032
size and distributions in the time of potential or current transients observed during certain corrosion processes [11], or in the frequency domain by investigating the characteristics of PSD plots obtained by FFT or MEM techniques [12], or performed based on the statistical and chaos theory [13]. Recently, a new mathematical tool for data processing has been developed which is called wavelet analysis, and which has been applied in international fields of science and technology, for example, signal processing, image processing and some nonlinear subjects [14]. Also it can be used to evaluate and characterize the surface structure and material corrosion process [15–19].
144
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150
The aim of this work is to investigate the EN features of pitting corrosion of pure aluminum in sodium solution through the wavelet technique, and probes into the relationship between the characteristics of EDP and the corrosion process.
2. Basis of wavelet analysis In the orthogonal wavelet series approximation, the temporal signal sets Sn(t)(n = 1,2, . . . ,N) transform into amplitude coefficients, each basis function, Sj,k, Dj,k. . .D1,k. Thus, the time signal S(t) can be reconstructed by adding together contributing wavelets weighted by their corresponding coefficients: X X SðtÞ S J ;K /J ;k ðtÞ þ DJ ;K uJ ;K ðtÞ k
þ
X
K
DJ 1;K uJ 1;K ðtÞ þ þ
K
S J ;K ¼ Dj;K ¼
Z Z
X
D1;K u1;K ðtÞ;
K
SðtÞ/J ;K ðtÞ dt;
ð1Þ
SðtÞuj;K ðtÞ dt;
ð2Þ
where /J ;K and uJ ;K are the complex conjugates of the basis functions /J ;K and uJ ;K , respectively. J is often a small natural number which depends mainly on /, u and N, j = 1, 2, . . ., J, K = 1, 2, . . . ,N/2j, at the time, /J, K and uj, K are generated from a pair of functions through scaling, and translation follows: t 2J k J =2 J J =2 /J ;k ðtÞ ¼ 2 /ð2 t kÞ ¼ 2 / ; ð3Þ 2J t 2j k j=2 j j=2 ; uj;K ðtÞ ¼ 2 uð2 t kÞ ¼ 2 u 2j
ð4Þ
2j acts as the scale factor and 2jk as the translation parameter. Therefore, the wavelet coefficient measures the correlation or the agreement between the wavelet and the corresponding segment of the signal. Through compressing and expanding the wavelet, the signal can be studied at different resolutions and scales.
3. Experimental The rod specimens of pure aluminum (6.00 mm diameter and 20 mm length) were provided by the British Aluminum Corporation. Their composition was 99.99 wt% (mass fraction, hereinafter) pure aluminum, 0.002 wt% copper, 0.004 wt% iron and 0.003 wt% silicon. The specimens were connected, respectively, to copper wire by a screw at one cross section to accommodate
the specimen and provide electrical contact, then mounted in epoxy with another cross section exposed. Before testing, the exposed surface was polished using abrasive papers through 500- to 1200-grade and velvet, then rinsed in distilled water, degreased in acetone and ethanol and dried in air. Tests were carried out in the solution (1.5 dm3, 3.0 wt% NaCl, pH, 7.03) prepared using analytical reagents and de-ionized water at ambient temperature. During the above experiments, the specimens were removed and examined with optical microscopy to observe the differences in corroding morphologies. The EN was monitored as a function of time between the working electrode and a saturated calomel electrode (SCE, reference) using a Powerlab/4sp (made in Australia) apparatus, which was controlled by Chart4 software using the Windows 2000 operating system. This equipment allows resolutions of 1 lV for voltage signals and 1 pA for current signals. EN records, containing 1024 data points every time, were collected at 2.5 points/s. These conditions define a frequency window in which most usual corrosion processes can be detected. The analytical results for wavelet and the fast Fourier transform (FFT) techniques were obtained by specific data techniques. Matlab6.1 software and software compiled in our laboratory were used to apply the wavelet and FFT transform programs, respectively. The surfaces of the samples after long-time immersion were observed in a JEOL USA JSM-5510LV scanning electron microscope with a field emission gun operated at 20 kV.
4. Results and discussion 4.1. EN characters in the time domain Figs. 1–5 show the potential noise generated by pure aluminum immersed in 3.0% NaCl (mass fraction) solution after 182, 1865, 2727, 4235 and 7245 min. It can be seen that the EN presents different forms in different corrosion phases. After 182 min of immersion, the potential transients are characterized by a quick potential rise followed by a quick decay to the base noise (Fig. 1); both the EN signals during the rise and decay sections approach a linear state. A large amplitude (almost 20 mV) was apparent which persisted only for almost one sample cycle. With the prolongation of the immersion time, the potential, which drifted negatively, needed a longer time to be restored to the base potential (Figs. 2–4), i.e., the rehabilitation of a metastable pit is more difficult due to the autoacceleration of the corrosion process [20]. It can also be seen that the amplitude of potential noise in Figs. 2–4 is much larger than that in Fig. 1. The typical shape of the potential noise for pure aluminum after 4235 min of immersion in chloride
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150
145
E/V
-1.04
-1.02
-1.00 0
500
1000
points Fig. 1. Potential noise signal of pure aluminum after immersion in NaCl solution for 182 min.
-1.00
E/V
-0.98
-0.96
0
500
1000
points Fig. 2. Potential noise signal of pure aluminum after immersion in NaCl solution for 1865 min.
E/V
-1.02
-1.00
-0.98 0
500
1000
points Fig. 3. Potential noise signal of pure aluminum after immersion in NaCl solution for 2727 min.
146
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150
E/V
-1.08
-1.06
-1.04 0
500
1000
points Fig. 4. Potential noise signal of pure aluminum after immersion in NaCl solution for 4235 min.
E/V
-1.02
-1.00 0
500
1000
points Fig. 5. Potential noise signal of pure aluminum after immersion in NaCl solution for 7245 min.
The spectral power density plots (PSD) were obtained using the FFT technique. Fig. 6 shows variations of the relations for three parameters including the high frequency linear parameter (k), the white noise level parameter (PW) and the cut-off frequency (fc) during the whole pitting process.
Pw / dB V2 Hz-1 ; k / dB decade-1
4.2. The fast fourier transform (FFT) study
0.8 -20
0.6 -40 0.4
fc /Hz
solution is shown in Fig. 4. It can be seen that the potential transient was characterized by a quick potential rise followed by an exponential decay and the recovery process lasted for several decades of sample cycles. After long time immersion, the EN signals changed to the chaotic state characterized by a large number of transients appeared around the base potential, respectively (Fig. 5).
-60 0.2
-80 0.0
3
2.0x10
3
4.0x10
3
6.0x10
0.0 3 8.0x10
time / min
Fig. 6. Dependence of PW, fc and k on time. (h), Slope/dB decade1; (n), White noise/dB V2 Hz1; (s), cut-off frequency/Hz.
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150 0
logED or logES
From Fig. 6, it can be seen that pitting corrosion generally results in shallow slopes greater than 20 dB/decade, while more general corrosion exhibits spectra with slopes close to that value. During the time of pitting corrosion, the white noise level varied from 80 to 20 dB V2 Hz1. In addition, the cut-off frequency, which is regarded as the maximum frequency of the spectral power density, increases in the initial stage, then decreases with prolongation of the immersion time. The process increases again when general corrosion sets in. Therefore, it seems that some confusion exists between the different forms of corrosion. Consequently, no one form can properly describe the propensity and intensity of the overall corrosion process [21].
147
-2
-4
D1
D2
D3
D4
D5
D6
D7
D8
S8
Fig. 7. Energy distribution plots (EDP) corresponding to the potential noise of Fig. 1 (including S8).
4.3. The wavelet study
E¼
N X
S 2n
ðn ¼ 1; 2; . . . ; N Þ:
ð5Þ
n¼1
Then, the fraction of energy associated with each crystal can be calculated as: j
EDj ¼
N =2 1X 2 D E k¼1 j;k
ðj ¼ 1; 2; . . . ; J Þ;
ð6Þ
ðj ¼ 1; 2; . . . ; J Þ:
ð7Þ
j
ESj
N =2 1X 2 ¼ S j;k E
Since the chosen wavelet is orthogonal, the following equation is satisfied [23] E ¼ ESj þ
J X
EDj :
ð8Þ
j¼1
The plots of the relative energy accumulated by each crystal vs. the crystal name are referred to as the energy distribution plots (EDPs). The EDP corresponding to the signals in Fig. 1 is plotted in Fig. 7. It can be seen that the transients on the largest timescale prevail over those on small timescales in the original signal. During the whole pitting process of pure aluminum in 3 wt% sodium chloride solution, the EDPs maintain this feature as shown in Fig. 7. On the other hand, it is well known that the processes of metastable pitting and the initiation/repassivation of pits are much more rapid than those of adsorption and desorption; consequently, it is very important to probe the energy contribution from the D series crystals, since they are characterized by smaller timescales. However, from Fig. 7 it can be seen that the contribution of the smooth S8 coefficients to the overall signal is too
large to shelter the information about D series crystals; therefore, the EDP was re-plotted by discounting the contribution of S8 crystal from the ensemble signal energy. The replotted EDPs corresponding to the noise data in Figs. 1–5 are shown in Figs. 8–12. It can be seen that, with prolongation of the immersion time, the energy contribution of D crystals with the larger timescale increases. For example, as shown in Fig. 8, the maximum relative energy is defined at the position of the D1 crystal that occupies almost 60% of energy in the D series, and the contribution of Dj decreases with the increase of j. However, the contribution of Dj crystals on the medium timescale increases with prolongation of the immersion time (Figs. 9 and 10), and the maximum relative energy is defined finally at the position of the D8 crystal (Fig. 12). Because, during the whole corrosion period of pure aluminum in neutral sodium chloride solution, the rate of pitting initiation and repassivation process is much more rapid than that of pit growth, it is deduced rationally that the EN features corresponding to Figs. 1–5 and Figs. 8–12 characterize the evolution process of pits
0.6
0.4
ED
The noise shown in Figs. 1–5 was analyzed with the wavelet technique using the orthogonal db4 wavelet. The ensemble energy of the noise containing 1024 data points in this experiment is calculated as follows [22]
0.2
0.0
D1
D2
D3
D4
D5
D6
D7
S8
Fig. 8. Energy distribution plots (EDP) corresponding to the potential noise of Fig. 1.
148
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150 0.4
0.4
ED
ED
0.3
0.2
0.2
0.1
0.0
D1
0.0
D1
D2
D3
D4
D5
D6
D7
D8
Fig. 9. Energy distribution plots (EDP) corresponding to the potential noise of Fig. 2.
D2
D3
D4
D5
D6
D7
D8
Fig. 12. Energy distribution plots (EDP) corresponding to the potential noise of Fig. 5.
0.2
4.4. Relationship between the EN features and the corroding morphologies
ED
0.3
according to our previous study [24], the accumulation of relative energy on the largest D crystal indicates that the corrosion type will change from pitting corrosion to general corrosion.
0.1
0.0
D1
D2
D3
D4
D5
D6
D7
D8
Fig. 10. Energy distribution plots (EDP) corresponding to the potential noise of Fig. 3.
0.3
Because the EN originates from the corrosion process, some relationship must exist between the EN features and the severity of corrosion of the material. Encouraged by the above consideration, the experiments in this study were repeated several times and the morphologies corresponding to the EN features in Figs. 1–5 and in Figs. 8–12 were studied using the scanning electron microscopy technique (Figs. 13–16). It can be seen in Fig. 13 that the surface of the corroding pure aluminum is much rougher than that prior to exposure in NaCl solution (Fig. 17) and only a few pits can be observed. The uneven surface may originate from the
ED
0.2
0.1
0.0
D1
D2
D3
D4
D5
D6
D7
D8
Fig. 11. Energy distribution plots (EDP) corresponding to the potential noise of Fig. 4.
from the metastable to the stable state. In other words, the condition that the high relative energy accumulates at D crystals on the small, medium and large timescales indicates the initiation (metastable pitting), the propagation/repassivation and growth of pits, respectively. And
Fig. 13. SEM morphology of pure Al corroded in bulk electrolyte (3.0 wt% NaCl solution) for 182 min.
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150
Fig. 14. SEM morphology of pure Al corroded in bulk electrolyte (3.0 wt% NaCl solution) for 1865 min.
Fig. 15. SEM morphology of pure Al corroded in bulk electrolyte (3.0 wt% NaCl solution) for 2727 min.
149
Fig. 17. SEM morphology of pure aluminum prior to exposure in NaCl solution.
metastable pitting, i.e., the initiation and rehabilitation of metastable pits. With prolongation of the immersion time, some metastable pits develop into stable pits, which can act as local anodes and protect the area around them against attack from the aggressive ions in the solution. Meanwhile, because of the autoacceleration of pitting corrosion and the surface refining due to the thermodynamic effects, the number of small pits decreases and large pits can be observed on the corroding surface (Fig. 14), which results in the opposite change of the relative energy defined at the D crystals on the low and medium timescales, respectively (Figs. 9 and 10). With the propagation and the growth of stable pits, the protected area around these stable pits increases, resulting in pit initiation and growth mainly occurring in the large pits (Figs. 15 and 16), which produces the chaotic phenomenon (Fig. 5).
5. Conclusions
Fig. 16. SEM morphology of pure Al corroded in bulk electrolyte (3.0 wt% NaCl solution) for 7245 min.
(1) An electrochemical noise signal is composed of distinct types of events, which can be classified according to their time constants. Meanwhile, the EDP can be used as a fingerprint of the EN signal and can be used to characterize the corrosion process. For metastable pitting, the relative energy is mainly defined at D crystals on small timescales, while for pit propagation and pit growth, the relative energy is mainly defined at D crystals of medium and large timescales, respectively. (2) A good relationship between the EN features and the corroding morphologies has been observed during the pitting corrosion process of pure aluminum in neutral 3 wt% sodium chloride solution. The relationship, the evolution of corrosion type and severity have been interpreted according to the characteristics of corrosion processes such as
150
C. Cai et al. / Journal of Electroanalytical Chemistry 578 (2005) 143–150
autoacceleration of pitting corrosion and the protection of local anodes afforded to the area around them.
Acknowledgments The authors acknowledge the financial support of the National Natural Science Foundation of China (Project 20203015, Project 50499335) and the Natural Science Foundation of Ningxia (Project 2003E1001).
[8] [9] [10] [11] [12] [13] [14] [15] [16] [17]
References
[18] [19]
[1] P.C. Season, J.L. Dawson, J. Electrochem. Soc. 135 (1988) 1908. [2] G. Gusmano, G. Montesperelli, S. Pacetti, Corrosion 53 (1997) 860. [3] P.R. Roberge, Corrosion 7 (1994) 502. [4] K. Nisancioglu, J.K. Davanger, O. Strandmyr, J. Electrochem. Soc. 137 (1990) 69. [5] C. Gabrielli, M. Kedam, Corrosion 48 (1994) 794. [6] U. Bertocci, F. Huet, Corrosion 51 (1995) 131. [7] S.T. Pride, J.R. Scully, J.Electrochem.Soc. 141 (1994) 3028.
[20] [21] [22] [23] [24]
K. Hladky, J.L. Dawson, Corros. Sci. 22 (1982) 231. J.W. Isaac, K.R. Hebert, J. Electrochem. Soc. 146 (1999) 502. C. Uruchurtu, J.L. Dawson, Corrosion 43 (1987) 19. J.Q. Zhang, Z. Zhang, J.M. Wang, C.N. Cao, Acta Physicochem. 17 (2001) 651. P.C. Pistorius, Corrosion 51 (1995) 295. C.N. Cao, X.Y. Chang, H.C. Lin, J. Chin. Soc. Corrosion Protection (in Chinese) 9 (1989) 21. D. Sornette, Physica B 219–220 (1996) 320. J.A. Wharton, R.J.K. Wood, B.G. Mellor, Corros. Sci. 45 (2003) 97. A. Aballe, M. Bethencout, F.J. Botana, M. Marcos, R.M. Osuna, Electrochim. Acta 47 (2002) 1415. Janusz Smulko, Kazimierz Darowicki, Artur Zielinskl, Electrochem. Commun. 4 (2002) 388. A. Aballe, M. Bethencourt, F.J. Botana, M. Marcos, Electrochim. Acta 44 (1999) 4805. A. Aballe, M. Bethencourt, F.J. Botana, M. Marcos, J.M. San´chez-Amaya, Electrochim. Acta 46 (2001) 2353. C.N. Cao, The Principles of Corrosion Electrochemistry, Chemistry Industry Press, Peking, 1985, p. 328. F. Mansfeld, H. Xiao, J. Electrochem. Soc. 140 (1993) 2205. Z. Zhang, F.H. Cao, Y.L. Cheng, J.Q. Zhang, J.M. Wang, C.N. Cao, Trans. Nonferrous. Met. Soc. China. 6 (2002) 1206. A. Aballe, M. Bethencourt, Electrochem. Commun. 1 (1999) 266. Z. Zhang, J.Q. Zhang, Y.L. Cheng, F.H. Cao, C.N. Cao, Acta Metall. Sin. 15 (2002) 272.