E L S E V I E R Materials Science and Engineering A198 (1995) 225-230

MATERIALS SCIENCE &

ENGINEERING

A

Potential fluctuation during early stage of stress corrosion cracking of type-304 stainless steel in chloride solution

Hiroyuki Inoue, Hirohito Iwawaki, Koji Yamakawa College of Engineering, University of Osaka Prefecture, 1-1 Gakuen-cho, Sakai, Osaka, Japan

Abstract

Potential fluctuation of type-304 stainless steel during stress corrosion cracking (SCC) was measured in 25 wt.% MgC12 solution. The power spectrum distribution of the observed fluctuation was computed by fast Fourier transform (FFT). During the early stage of SCC (the period when small cracks propagate from a pit), the frequency distribution of the power spectrum had a clear peak at the range from 8 to 10 mHz. The correlation between the number of microcracks and potential fluctuations was examined: both of the numbers showed a linear relation. From this result, it was speculated that the fluctuation was generated by the rapid formation of a bare metal surface and its repassivation with the connection of the microcracks. To confirm this speculation, the change in the power spectrum when the stress increased at an early stage of SCC was investigated, because the propagation rate of an advancing crack can be varied by the intensity of the tensile stress. While a peak at higher frequency appeared with an increase in the tensile stress, the peak at the range from 8 to 10 mHz remained.

Keywords: Stress corrosion cracking; Type-304 stainless steel; Corrosion potential fluctuation; Frequency analysis; Neutral chloride solution

I. Introduction

Fluctuation of the open-circuit potential (or of the current under potentiostatic control) during pitting or stress corrosion cracking (SCC) has been investigated to get information about the dynamic change in the locally corroding electrode. Many studies on fluctuation have been published for pitting, but few reports on SCC [1-8]. Less attention has been paid to the poten- tial fluctuation during SCC, probably because the am- plitude of the fluctuation is smaller than that during pitting. The smallness of the amplitude makes it difficult to obtain data, and to confirm the relationship between the signal and a physical phenomenon.

The first paper on the fluctuation of the electrochem- ical signals during SCC was written by Newman and Sieradzki in 1983 [1]. They observed oscillations of the current under potentiostatic control during SCC of ~-brass. From the results of scratch tests and scanning electron microscope (SEM) images of the fracture sur-

0921-5093/95/$09.50 © 1995 - - Elsevier Science S.A. All rights reserved SSD! 0921-5093(94)04527-5

face, they speculated that the electrical charge passed during each current fluctuation was equal to that gener- ated by each crack advance. Also, some studies about potential fluctuation during SCC have been published [2-5]. The time-series data of potential fluctuation were transformed to a frequency-domain representation - - a power spectrum distribution - - by use of the fast Fourier transform (FFT) or the maximum entropy method (MEM). The computed results were plotted on log- log scales. Then, the relation between physical changes in the specimen and the power spectrum at low frequencies was investigated. It was concluded that crack initiation and specimen rupture gave the highest power. On the power spectrum curves obtained by this method, there is a low frequency plateau. As Loto and Cottis stressed [5], this plateau is an experimental arti- fa~ct, not the result of a physical phenomenon. Thus, analyses at low frequency are not reliable.

Yamakawa and Inoue [6] divided the observed poten- tial fluctuation signals during SCC into two compo-

226 H. lnoue et al. / Materials Science and Engineering A 198 (1995) 225-230

nents: a signal that shifts to a more noble potential and the one which shifts to a less noble potential. They then investigated the difference of the mean spectrum power for each type of signal, and found a good correlation with the elongation rate of the specimen. Wells et al. [7-8] investigated potential fluctuations and current pulse during the intergranular SCC of sensitized type- 304 stainless steel. The frequency of the potential fluc- tuation was proportional to the degree of sensitization as indicated by the Electrochemical Potentiokinetie Re- activation (EPR) test. They assumed the fluctuation was caused by the initiation of cracks on the sensitized grain boundaries. Current transients were measured on a specimen coupled through a zero-resistance ammeter to a large cathode of annealed type-304 stainless steel. They confirmed that the number of current transients was roughly equal to the number of cracks found by SEM observation. They concluded that measured cur- rent pulses were associated with the nucleation of the microcracks.

SCC tests are generally carried out under more severe conditions than those actually employed, in order to finish the experiments within a reasonable time. How- ever, in many cases, a material that has an excellent performance in accelerated tests ruptures in a shorter time upon actual usage. It was speculated that the morphology and the mechanism of SCC changed when the accelerated testing conditions were used, therefore giving rise to the above mismatch. If it were possible to get a result about S C C tests from the data at an earlier stage, the experimental time could be substantially shortened. In other words, tests could be done under milder conditions, which would be close to that of actual employment, which needs too long an experi- mental time with the method currently in general use. The observation of potential or current fluctuation makes it possible to detect the nucleation or propaga- tion of a crack in the earlier stage of SCC. Now, it appears that the nucleation of a crack can be detected by the same methods as reviewed above. However, these methods are insufficient for the evaluation of SCC sensitivity. It is necessary not only to detect the nucle- ation of the cracks but also to assess the propagation rate directly, just after it has been nucleated.

We have measured the potential fluctuation of type- 304 stainless steel during SCC in chloride solution and computed a power spectrum of observed data by the use of the FFT. During the early stage of SCC, the frequency distribution of the power spectrum had a clear peak. This suggests that the potential fluctuation occurs at regular intervals. It is expected that the poten- tial fluctuation synchronizes with each step of the crack advance - - a connection of a microcrack in the vicinity of the advancing crack tip - - and there is a possibility that the propagation of a crack can be detected dynam- ically from an analysis of the potential fluctuation. So,

we examine the correlation between the number of microcracks and potential fluctuation, and discuss the obtained results.

2. Experimental procedure

The material used in this study was type-304 stainless steel purchased from Japan Stainless Steel Association. This steel is a rolled sheet 2 mm thick and solution- annealed. The composition of the material was: 0.06C, 0.58Si, 0.82Mn, 0.29P, 0.002S, 8.75Ni, 18.29Cr, 0.14Mo, 0.14Cu in mass% . This was machined to a flat tensile specimen, with a 20 mm gage length and a 4 x 2 mm 2 cross-sectional area, and was finished with emery paper of 2000 grit. The specimen was annealed at 1373 K for 600 s to remove a residual strain induced by machining. It was sensitized at 923 K for 10.8 ks (3 h) in nitrogen gas and then cooled in air. After these heat treatments, the specimen was slightly etched in HF + HNO 3 solu- tion to remove an oxide film, and then it was electro- polished in H3PO 4 + CrO 3 solution, followed by acid cleaning, and ultrasonic cleaning in acetone. 5 mm of the gage length was exposed to the test solution and the remainder was sealed with polysiloxane.

The testing environment was 25 mass% MgC12 solu- tion. The solution was exposed to air throughout the testing and its temperature was kept at 353 + 0.5 K. The solution was poured into the cell after the load was applied, and it was not agitated during the experiment. A constant tensile stress of 245 MPa was applied to the specimen with a spring-type tensile test apparatus. The fluctuation of the corrosion potential with respect to the saturated Ag/AgC1 reference electrode was mea- sured with an accurate digital voltmeter; the 1.2 V range was used and the least significant digit of this range was 1 ~tV. The sampling interval was 0.5 s, and the data were stored in a memory bank in the voltmeter successively. When the number of stored data reached 4096, these data, which were dealt with as one data block, were transferred to a desktop computer through an IEEE-488 bus, then saved on a floppy disk. This sequence of potential measurement was repeated every 2400 s.

The intensity of the inherent noise of the measure- ment system was assessed from the time record of potential difference between two reference electrodes inserted in one cell. A noise was observed synchronized with making or breaking of the heater circuit, which was inserted in a glass tube and placed inside the testing cell to control the solution temperature. The amplitude of the observed noise was 10 gV or below. The faradaic impedance of the Ag/AgCI electrode is higher than that of the specimen, therefore oscillation signals which exceeded 10 gV in amplitude were judged to be valid in this measurement system.

H. Inoue et al. / Materials Science and Engineering A 198 (1995) 225-230 227

A power spectrum of corrosion potential was ob- tained to estimate the frequency of fluctuation using FFT. The Hanning window was employed for a win- dow function. Calculation was done for every data block, including the 4096 points of potential data as mentioned above. For the data sampled every 0.5 s, the highest frequency of the computed power spectrum was I nz and the frequency resolution of the spectrum was 1/2048 Hz. Before the execution of the FFT, d.c. and a drift component included in the original data were removed by use of the least-squares method; the devia- tion of each measured data value from the regression line is employed as data for the FFT computation.

-310

T ~ e (26.4k~/div.)

Fig. 2. Corrosion potential change during the initiation stage to the early stage of SCC.

3. Results and discussion

3.1. Macroscopic corrosion potential change throughout an SCC test

In this section, the focus is on the macroscopic change of corrosion potential throughout the SCC test. Fig. 1 shows the example of a typical corrosion poten- tial change, from 7.2 ks after the solution was poured to 1.5 ks before the specimen ruptured. After the com- pletion of the test, the ruptured surface was observed by SEM. The specimen was ruptured by one crack and any other macrocracks were not regarded on it. During period (A), potential rose to the peak, then oscillated a few times, and finally dropped to the lower potential. In other tests performed under the same experimental conditions, the potential reached a peak at 10 to 20 ks after the pouring of the solution. It is speculated that the formation of a stable pit brought about the poten- tial drop. Throughout period (C), the corrosion poten- tial gradually decreased as one macro crack was propagated further, which could be observed by the naked eye.

During period (B), the corrosion potential kept a steady value. This potential plateau was usually ob- served in the other experiments done under the same

;;,,.

~=.ml

:= I=1

0

-270

-320

-370

I

¢,(A) I I I I

. . . . ( c )

' ( B ) ' ~ * ~ ~ , ~

I I I I I

Time (70ks/div.)

Fig. 1. Example of macroscopic corrosion potential change through- out an SCC test. (Periods (A)-(C): see text.)

experimental conditions. A few small cracks were ob- served by SEM on the surface of the specimen when the experiment terminated in this stage. These cracks gener- ated from one pit and propagated almost perpendicular to the direction of the applied stress and were hardly visible any longer with the naked eye. We refer to this stage as the "early stage of SCC", and potential fluctu- ation observed in this period was analyzed and is discussed in this paper.

3.2. Frequency characteristics in the early stage of SCC

Fig. 2 shows the changes of corrosionpotential dur- ing the initiation stage to the early stage of SCC. Potential measurements started from 3.6 ks after the pouring of the solution and terminated at the time corresponding to the right-hand end of this figure. Fig. 3 displays the power spectrum of the potential fluctua- tion shown in Fig. 2. These curves are calculated by averaging the six power spectra, which are computed from the sequentially saved six data blocks. Each sec- tion in Fig. 2, marked with letters (A) to (I), is con-

0)

4.0 " .....

! 5 15 25 35

F r e q u e n c y (mlRz)

Fig. 3. Power spectrum of the potential fluctuation shown in Fig. 2.

228 H. Inoue et al. / Materials Science and Engineering A 198 (1995) 225 230

O~ O~

K,

Fig. 4. Typical example of a SEM image of the surface area near a crack tip.

structed with six data blocks, so the power spectra are obtained in each section, and the letters written on the right side of each curve in Fig. 3 correspond to those of this figure. The frequency range represented in this figure is 5 to 35 mHz. The spectrum curves under 5 mHz are obviously affected by the drift components included in the data. In the frequency range over 35mHz to 1Hz, no peaks are detected and the intensity of the power spectrum decayed monotonically.

All power spectra shown in Fig. 3 have a clear peak frequency within 8 to 10 mHz. The same results were obtained in other experiments carried out under the same experimental conditions. However, when many cracks propagated from some pits simultaneously, the obtained power spectrum curves became complicated and a clear peak did not appear. The clear peak of the power spectrum indicates that the potential fluctuation occurs at a regular interval, and this type of potential fluctuation during SCC testing has not been reported before. In the following section, the origin of this signal will be discussed.

3.3. Surface morphology in the vicinity of the crack tip

The experiment terminated at an early stage, and then surface observation was carried out by SEM. The surface trace of the propagated crack was gradually corroded with time, but in the vicinity of the advancing crack tip, it was expected that some information about crack propagation was preserved. Fig. 4 shows a typical example of an SEM image of the surface area near a crack tip. The coarse slip lines were seen on the figure and were inclined at about 45 ° to the direction of the applied stress. From this micrograph, it can be sup- posed that a crack propagates on the surface of a specimen as follows. First, isolated microcracks are generated along coarse slip lines in front of the crack tip, where stress is concentrated. Then these micro-pits are coupled successively and the crack propagates fur-

ther in this manner. It is estimated that the former process is controlled by corrosion along the coarse slip line and the latter is ruled by the mechanical crack advance. Therefore, the latter process induces a poten- tial fluctuation, because a relatively large area of bare metal surface is generated rapidly and then repassi- vated. Also, it is speculated that when the process occurs at the inside of the specimen, a detectable poten- tial fluctuation is not generated, because of the effect of the IR drop of the solution. If the above assumption is correct, it is speculated that the number of connected microcracks will be equal to the sum of the observed potential fluctuations.

3.4. The correlation between the number of microcracks and that of fluctuations

The average length of a microcrack was measured to estimate the number of microcracks included in the specimens. Nine experiments were terminated at an arbitrary time in the early stage of SCC and the surface of the tested specimens was observed by SEM. First, it was observed that the average microcrack length for each adequate area of SEM image where the trace of microcracks was well preserved was in the vicinity of the crack tip of the specimens. Four to twelve data values were taken for every specimen, which were sampled from different areas of the specimen. A histogram of the average microcrack length shown in Fig. 5. The average value of data displayed in this graph is 2.12 ~tm. The number of microcracks included in each specimen was computed by dividing the whole length of the surface crack-propagation trace by the average value of microc- rack length. Also, the number of potential fluctuations, from the potential dropping to the experiment termina- tion, was estimated from the peak frequency of the power spectrum of every data block: we 'multiplied the value of the each peak frequency by the total measure- ment time of each data block. The correlation between the number of microcracks and that of fluctuations is shown in Fig. 6. From this figure, it is judged that the number of microcracks is equal to the sum of the observed potential fluctuations. So it is speculated that the connection of microcracks is a signal source for the potential fluctuation in the early stage of SCC.

3.5. A shift of the peak frequency by the change of tensile stress in the middle of the SCC test

The propagation rate of the advancing crack in- creases with the intensity of the tensile stress. In the early stage of SCC, if the crack propagates with the connection of microcracks on the surface of a specimen and the potential fluctuation occurs when the micro- cracks connect each other, as speculated above, it is expected that increasing the stress will lead to the

H. lnoue et al. / Materials Science and Engineering A 198 (1995) 225 230 229

30

T e s t A T e s t B T e s t C T e s t D T e s t IE T e s t F T e s t G T e s t H T e s t I

( 3.6ks, 5 a r e a s ) ( 4.2ks, 5 a r e a s ) ( 9.0ks, 13a reas ) ( 39 .6ks , 7 a r e a s ) ( 48.6ks, 5 a r e a s ) ( 66.0ks, 14a reas ) ( 73 .2ks , 7 a r e a s ) ( 73 .2ks , 11areas ) ( 75.0ks, 18a reas )

o 20 I= = =T

r,., 10

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

L e n g t h of M i e r o e r a c k s (~Zm)

Fig. 5. Average microcrack length of each evaluated area. The termination time and the number of evaluated areas of each test are shown within the parentheses.

(k)

4.0

Q0

0 5 45 85 125

Frequency (mHz)

Fig. 7. Power spectrum of the potential fluctuation from 1.2 ks to 25.2 ks after the stress increase (245 to 294 MPa).

Figs. 7 and 8 show the change of the power spectrum from 1.2 ks to 25.2 ks after the increase in the stress intensity. The former is the result when the stress was increased from 245 to 294 MPa and the latter is that when the stress was changed from 245 to 343 MPa. These curves are computed from the sequentially saved

decreasing of the occurrence interval of the potential fluctuation. The interval can be evaluated from the peak frequency of its power spectrum, as shown in Fig. 3. So, to confirm the above speculation, the stress was increased at the early stage of SCC and the shift of the peak frequency of the power spectrum was investigated.

W I:1

ts00 o

m

=1 tD 0 ~u 500

o

ID

0 Im

I I I I I I I I

- t o

- , ' b ~ l l l J l J

500

I I I I I I I

s • s. S

o o

s

s °S O

I I I I I I I I I I

1000 1500

Number of Microcracks

a"

=

0.5

~h 0 5 165 325 485

Frequency (mHz)

a

m/div. )

Fig. 6. Correlation between the number of microcracks and that of Fig. 8. Power spectrum of the potential fluctuation from 1.2 ks to fluctuations. 25.2 ks after the stress increase (245 to 343 MPa).

230 H. lnoue et al. / Materials Science and Engineering A 198 (1995) 225 230

a',

r~ 245MPa

0 (7.2ks/div.) 5 15 25 35

F r e q u e n c y (mHz)

Fig. 9. Power spectrum of the potential fluctuation before and after the stress increase.

data block. In Fig. 7, a clear peak is shown around 45 mHz from 1.2 ks to 8.4 ks (the curves (a) to (c) in Fig. 7) after the increasing of the stress. This type of peak is also shown in Fig. 8. Its frequency is higher than that shown in Fig. 7 (curves (1) to (9) in Fig. 8), though the intensities of the spectrum peaks shown in Fig. 8 are lower than those shown in Fig. 7. When a constant stress of 245 MPa was applied to the specimen, as the result shown in Fig. 3, the frequency of the clear peak was within 8 to 10 mHz. It can be seen that the peak shifts to higher frequency with an increase of the stress, but the lower peak also remained as displayed in Figs. 7 and 8.

Fig. 9 is the change in the power spectrum at low frequency before and after the stress increase. The analyzed potential data are common with the data presented in Fig. 8. These curves are calculated by the averaging of the three power spectra, which are com- puted from the sequentially saved three data blocks. The averaged data of the curves (c),(d) and (e) in Fig. 8 are the curve (~) in Fig. 9. The curves (~') and (r/) in Fig. 9 correspond to the averages of the curves (f) to (h) and (i) to (k) in Fig. 8, respectively. A clear peak is seen before and after the increasing of the stress. The fre- quency of the peak is about 8 mHz and it does not change with the changing of the stress. As shown in Figs. 7 and 8, the higher-frequency peak disappeared

after a while. It also cannot be explained if the higher peak is brought by increasing of the propagation rate. When the stress is increased, the strain rate of the specimen is also increased. So, there is a possibility that a peak which has a higher frequency is not generated by acceleration of the crack propagation, but it is induced by the increase in the strain rate.

4. Conclusion

Potential fluctuation of type-304 stainless steel during SCC was measured in 25 mass% MgC12 solution and its power spectrum was obtained by FFT. The power spectrum of the observed fluctuation had a clear peak, and its frequency was within the range from 8 to 10 mHz. Its source was speculated to be the rapid forma- tion of a bare metal surface and its repassivation, with the connection of the microcracks which had been generated along coarse slip step lines in front of the crack tip on the surface of the specimen. To confirm this mechanism, the stress was increased in the middle of the SCC tests and the shift of the peak frequency of the power spectrum was investigated. A new peak appeared at a higher frequency, while a peak at the lower frequency also existed before and after the change of the tensile stress. These results were not explicable by the above-speculated mechanism, so it was concluded that the evidence about the correlation between the signal and the phenomenon could not be obtained within the extent of the results investigated in this paper.

References

1 R.C. Newman and K. Sieradzki, Scr. Metall., 17(1983) 621. 2 C.A. Loto and R.A. Cottis, Corrosion, 43 (1987) 499. 3 K. Yamakawa, N. Kajita, M. Murakami and T. Hirayama, J.

Soc. Mater. Sci. Jpn., 37 (1988) 43. 4 C.A. Loto and R.A. Cottis, Corrosion, 45 (1989) 136. 5 C.A. Loto and R.A. Cottis, Corrosion, 46 (1990) 12. 6 K. Yamakawa and H. lnoue, Corros. Sci., 31 (1990)

503. 7 D.B. Wells, J. Stewart, P.M. Scott and D.E. Williams, Corros.

Sci., 33 (1992) 39. 8 D.B. Wells, J. Stewart, P.M. Scott and D.E. Williams, Corros.

Sci., 33 (1992) 73.