Materials Science and Engineering A 419 (2006) 218–224

Study on the initiation stage of stress corrosion of an aluminum

alloy using the subjective speckle technique
Hui-Ji Shi ∗ , Chunyi Ruan, Xide Li
Key Laboratory of Failure Mechanics, Chinese Ministry of Education, Department of Engineering Mechanics,
Tsinghua University, Beijing 100084, PR China
Received in revised form 22 December 2005; accepted 22 December 2005

Abstract
The aim of this paper is to investigate the initiation stage of stress corrosion of an aluminum alloy. Three-point bending experiments are carried
out under displacement-controlled mode. The tested specimens are Al2024-T3 alloy with thickness of 2 mm which are immersed in the 3.5%
sodium chloride solution in the experiments. The solution temperature was controlled from 20 to 80 ◦ C and the exerted stresses were 0, 174 and
348 MPa, respectively, to study the influences of the temperature and the stress in the stress corrosion process. The subjective speckle method
has been employed to observe the processes of the stress corrosion in microregions by means of measuring the local surface roughness of the
specimens. Treatments of employing both average intensity and an autocorrelation coefficient of the subject speckle fields confirmed that the
subjective speckle method is effective to characterize the stress corrosion initiation process of the tested aluminum alloy. The experimental results
show that the temperature influences the rate of corrosion distinctly and meanwhile the rate of the corrosion also depends on the stress state
considerably.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Stress corrosion; Aluminum alloy; Subjective speckle technique; Roughness; Average speckle intensity; Autocorrelation

1. Introduction experimental methods using optical, electrochemical and pho-

toelectron chemical techniques to obtain local information on
Conventional research on stress corrosion has concentrated precursor sites and vulnerable areas on metal surfaces. Jin and
on the propagation stage of failure, that is, to determine the frac- Chiang [14] introduced a digital speckle pattern technique to
ture life [1], critical stress and critical stress intensity factor [2], reveal pitting corrosion areas as small as 5 mm × 5 mm wherein
rate of crack propagation [3], etc. Pre-cracking or notched sam- a cluster of pits with 0.3 mm in diameter and 0.1 mm in depth
ples are usually required to investigate stress corrosion crack was measured. An artificial stress corrosion crack was used to
(SCC) and models emphasizing the cooperative effects of cor- demonstrate the validity of this method for detection of stress
rosion and plasticity have been proposed in these cases [4]. corrosion cracking. In the work of Toh et al. [15], a laser speckle
However, experimental investigation of the initiation stage of correlation technique was applied to measure the surface rough-
stress corrosion has not been studied as much because of lim- ness, which ranged from 1.6 to 50 ␮m. Instead of moving the
ited observation methods and difficulties in characterizing under laser beam, the specimen was rotated to achieve angular speckle
very low corrosion levels. correlation in the far-field plane. The above mentioned methods
In the literature, some experimental techniques have been have been successfully employed to measure corrosion rates, and
used to observe the corrosion processes, such as thermal to locate corrosion and precursor sites. However, much work is
wave microscopy [5,6], fluorescence microscopy [7,8], micro- still required to find a technique for precise observation of the
ellipsometry [9,10], differential reflectometry [11] and photo- initiation stage of the stress corrosion.
acoustic spectroscopy [12]. Sukamto et al. [13] described several In this paper, a highly sensitive and non-destructive optical
method, subjective speckle method was developed to investigate
the surface roughness changes during the corrosion process. The
∗ Corresponding author. Tel.: +86 10 6277 2731; fax: +86 10 6278 1824. method combines scatter and statistical speckle techniques to
E-mail address: shihj@mail.tsinghua.edu.cn (H.-J. Shi). observe the progression of corrosion with real time. A long work-

0921-5093/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2005.12.028
 H.-J. Shi et al. / Materials Science and Engineering A 419 (2006) 218–224 219

ing distance microscope is introduced into the subjective speckle

system to microregion measurement. The roughness range of the
tested specimens is from 0.03 to 0.65 ␮m. The effects of the tem-
perature and applied stress on the initiation of stress corrosion
are investigated, and results of the roughing caused by the stress
corrosion process are presented.

2. Methods

2.1. Principle of experiments

The speckle technique makes use of the random collection

of bright and dark spots (speckles) formed in space when a dif-
Fig. 1. Illustration of the subjective speckle measuring system.
fusely reflecting object is illuminated by coherent light (laser)
or when an object’s surface is naturally reflective [16,17]. In
objective speckle situation, the random elementary waves prop- sity distribution or the self-correlation coefficient of the speckle
agate freely in the space but in subjective speckle or sometimes patterns [20].
called imaging plane speckle, an imaging system is introduced To the digital speckle intensity field, the average intensity of
to form the speckle image of the imaged surface [18]. In this the observed can be defined by
case, the speckle size is essentially defined by the imaging aper-
N
M
i=1 j=1 I(i, j)
ture. Generally, the micro profile of the scattering surface is Ī = (2)
considered to produce corresponding geometrically related to NM
the phase changes of the scattering light. Therefore, microscopic where I(i, j) is grey value of speckle pattern at point (i, j), and
properties of the measured surface, like pitting corrosive surface N and M the observation points in the microregion. Obviously,
represented by roughness, can be transferred to investigation the if a sufficiently large area of the speckle field is taken, Eq. (2)
properties of the speckle pattern producing when the object sur- is valid for describing the average intensity distribution of the
face is illuminated with the laser beam. In fact, this speckle speckle pattern.
pattern is a complex coded which carriers of the tested surface Similarly, the autocorrelation coefficient C12 can be used to
information. The surface change by physical mechanism (corro- calculate the speckle pattern changes during the stress corrosion.
sion, for example) with time can be inferred from a comparison The autocorrelation coefficient is defined by [21]:
of the scattering speckle patterns at different states. I1 I2  − I1 I2 
In our investigation, an optical system using the subjective C12 = 1/2
(3)
speckle method was set up to study the initiation stage of stress [(I1  − I1 2 )(I22  − I2 2 )]
2

corrosion. In this system, a He–Ne laser with a wavelength of where I1 is the intensity of the original speckle intensity and I2
0.6238 ␮m illuminates angularly the specimen surface on x–y is the corroded speckle intensity.
plane. A long working-distance microscope, called LDM with To calculate the autocorrelation coefficient in one dimension,
a field-of-view from Ø 0.5 mm to Ø 10 ␮m and a center spatial such as in x direction, the relationship between I1 and I2 is:
resolution of 1.5 ␮m, was introduced as an imaging system to
observe the surface microregion of the specimen. The lens of I2 = I1 (x −
x) (4)
LDM was placed at 100–150 mm from the specimen surface and
provide 0.2 mm × 0.2 mm view field. Comparing the patterns where
x is position offset in the observation field. The curve
obtained by illuminating the specimen’s surface with laser beam of C12 is obtained when the value of
x is gradually increased
at 30◦ , 45◦ and 60◦ , respectively, it was found that the pattern from zero to half of the observation field. C12 is equal to one
generated at the 30◦ angle was the clearest and most sensitive. when
x = 0, then the coefficient decreases with the increase of
An image recording unit, charge-coupled devices (CCD) and a
x. The value of
x is called half-breadth when the C12 is equal
Matrox digital image analysis system were employed to record to 1/e (namely 0.3678). We used this value as a representative
and digitize the speckle patterns at the imaging with the spatial value to distinguish different autocorrelation curves caused by
0.3 ␮m/pixel. An illustration of the experimental setup is shown the corrosion process. Thus, the relationship between the half-
in Fig. 1. breadth and the surface roughness can be determined by using
Assuming the complex amplitude of the speckle pattern can these curves.
be represented by U(x, y), then the intensity distribution of the
speckle field is as follows [19]: 2.2. Calibration of the optical test system

I(x, y) = |U(x, y)|2 . (1) Conventional corrosion parameters such as the crack propa-
gation rate or critical stress intensity are not suitable for charac-
The relationship between the surface roughness and the terizing the initiation stage of stress corrosion. During corrosion
speckle field can be obtained by investigating the average inten- testing, it was found that the roughness of the test surface
220 H.-J. Shi et al. / Materials Science and Engineering A 419 (2006) 218–224

Fig. 2. Micrographs of roughness standard pieces of aluminum alloys 2024-T3. The corresponding roughness Ra are (a) 0.0248 ␮m, (b) 0.0302 ␮m, (c) 0.0364 ␮m,
(d) 0.0473 ␮m, (e) 0.0719 ␮m, (f) 0.1565 ␮m, (g) 0.288 ␮m, (h) 0.4445 ␮m and (i) 0.6465 ␮m, respectively.

increases rapidly with time due to the initialization of stress test pieces made from the same 2024-T3 aluminum alloy were
corrosion. The optical method is widely employed to measure used to standardize the optical test system.
surface roughness [22], and the roughness of a surface was used The nine standard roughness pieces were prepared careful by
here to represent the degree of corrosion. Roughness standard manual polishing with SiC paper and their surface roughness

Fig. 3. Subjective speckle patterns of the roughness standard pieces: (a) roughness 0.0248 ␮m, (b) 0.0302 ␮m, (c) 0.0364 ␮m, (d) 0.0473 ␮m, (e) 0.0719 ␮m, (f)
0.1565 ␮m, (g) 0.288 ␮m, (h) 0.4445 ␮m and (i) 0.6465 ␮m.
 H.-J. Shi et al. / Materials Science and Engineering A 419 (2006) 218–224 221

Fig. 4. Average intensity vs. the roughness of the standard pieces. Fig. 6. Half-breadth of autocorrelation coefficient vs. the roughness of the stan-
dard pieces.
was measured using a contact surface roughness contourgraph.
The tested averaged roughness, Ra , were 0.0248, 0.0302, 0.0364,
0.0473, 0.0719, 0.1565, 0.288, 0.4445 and 0.6465 ␮m. Optical
micrographs of these roughness standard pieces are shown in
Fig. 2.
The roughness standard pieces are tested in the subjective
speckle test system, the corresponding speckle patterns are
shown in Fig. 3. From these speckle patterns, the average inten- Fig. 7. Three-point bending specimens with 2 mm thick, 12 mm wide and
110 mm long.
sity and autocorrelation coefficient can be calculated using Eqs.
(2) and (3), respectively.
A curve of the average intensity versus the surface roughness (refer to Fig. 5). The relationship between the half-breadth and
is shown in Fig. 4. It shows that the relationship between the roughness can be expressed as the following formula:
average intensity and surface roughness is almost linear, and 
can be expressed as the following formula: −2.9439 + 215.07644r for r < 0.0473 ␮m
w=
14.52343 − 8.84285 e (−r/0.2164) for r ≥ 0.0473 ␮m
Ī = 142.381 + 164.9r (5) (6)
where r is the surface roughness and Ī is the average intensity. where w is the coefficient of the half-breadth. Therefore, the
The autocorrelation coefficients of the speckle patterns for relationship between the surface roughness and Ī and C12 can
the standard pieces can also be calculated using Eqs. (3) and be obtained using Eqs. (5) and (6).
(4). Some C12 curves are also shown in Fig. 5. The half-breadth
versus roughness curve is shown in Fig. 6 for each roughness
standard pieces by calculating the
x values when C12 = 0.3678 2.3. Testing procedure

Three-point bending specimens, Al2024-T3 flat strips of

2 mm thick, 12 mm wide and 110 mm long, were prepared for
stress corrosion testing (Fig. 7). Before testing, the surfaces of
the specimens were mechanically polished to a mirror with fine
diamond paste and then the fine fabric.

Table 1
The chemical composition of Al2024-T3 (mass fraction, %)
Material Al2024-T3
Si 0.5
Fe 0.5
Cu 3.8–4.9
Mn 0.3–0.9
Mg 1.2–1.8
Cr 0.1
Zn 0.25
Al Rest
Fig. 5. Autocorrelation coefficient vs. the roughness of the standard pieces.
222 H.-J. Shi et al. / Materials Science and Engineering A 419 (2006) 218–224

Fig. 10. Average intensity vs. corrosion time under 22, 40, 60 and 80 ◦ C. The
Fig. 8. Illustration of the testing system. An autoclave and a three-point bending
applied stress during the experiments is kept on 174 MPa.
loading unit are contained in this system.

3. Results and discussion

The chemical composition of Al2024-T3 is listed in Table 1.
The stress corrosion experiments were carried out using a Three point bending experiments were conducted to compare
displacement-controlled three-point bending setup. The length the effects of temperature and stress on the initiation stage of
between the two load-bearings was 100 mm. Fig. 8 shows the stress corrosion. First, a series of stress corrosion experiments
testing system which contains an autoclave and a three-point were performed at different temperatures of 22, 40, 60 and 80 ◦ C,
bending loading unit. The specimen was supported at both ends while the stress was constant at 174 MPa in the mid-length of the
and a bending load was exerted at the center of the bending specimens. Then a stress corrosion experiment of 348 MPa and
specimen using a screw. In this loading configuration, the max- 60 ◦ C was performed, followed by a series of simple corrosion
imum stress occurs at the mid section of the specimen and experiments without the presence of stress at 40 and 60 ◦ C.
decreases linearly to zero at the ends. The bending unit was For tests conducted at 22 ◦ C, the patterns of specimens
pre-calibrated to the specimen using the strain gauge. During obtained before and after 6 h of stress corrosion are compared
the testing, the surface stresses of the mid section of the spec- in Fig. 9. The horizontal and vertical bright lines are chiseled
imens were set at 174 and 348 MPa, respectively. The 3.5% lines on the specimen, which are used as locating marks for
sodium chloride solution in the autoclave was used as the cor- observation.
rosion solution to immerge the loading unit and the specimens.
The temperature of the corrosion solution was controlled by 3.1. Effect of the temperature
the temperature adjustment system of the autoclave and the
corrosion environment with flowing solution is designed to pre- The average intensity method and autocorrelation coefficient
vent forming of the oxidation film on specimen surface. The method were used to investigate the effect of the temperature on
temperature of the solution was set at 22, 40, 60 and 80 ◦ C, stress corrosion. In this situation, the tress was kept on 174 MPa
respectively, and the optical observation (imaging) interval was and the average intensities of all patterns were calculated for
1 h. different temperatures. Fig. 10 presents the average intensity

Fig. 9. Images of the subjective speckle patterns: (a) before the stress corrosion and (b) after 6 h stress corrosion under the room temperature.
 H.-J. Shi et al. / Materials Science and Engineering A 419 (2006) 218–224 223

Fig. 13. Roughness vs. the corrosion time calculated by the average intensity
Fig. 11. Half-breadth vs. corrosion time under 22, 40, 60 and 80 ◦ C. The applied method (solid lines) and the autocorrelation coefficient method (dashed lines)
stress during the experiments is kept on 174 MPa. under 0, 174 and 348 MPa. The corrosion temperature is kept on 60 ◦ C during
the experiments.

versus the corrosion time and Fig. 11 the half-breadth versus the
corrosion time. agreement and can be used to calculate the amount of the sur-
Using Eq. (5), the roughness versus the corrosion time by face roughing which corresponds to the stress corroding.
the average intensity method is shown in Fig. 12 (solid lines).
The results show that when the corrosion time increases, the 3.2. Effect of the applied stress
roughness value increases, but the rate of roughing decreases.
Meanwhile, with the corrosion temperature increases, the rate The effect of the applied stress on the stress corrosion was
of roughing increases obviously. investigated. Both the average intensity and self-correlation
The change of the roughness versus the corrosion time with coefficient methods were used to calculate the roughness ver-
different temperatures using Eq. (6) is shown in Fig. 12 with sus the corrosion time. The tests were performed at a constant
dashed lines. It is found that the results of autocorrelation coeffi- temperature of 60 ◦ C and the applied stresses were 0, 174 and
cient method are about 20% less than that of the average intensity 348 MPa, respectively.
method. It appears that the difference between two methods is Using Eq. (5), the curves for the roughness versus the cor-
induced by different pattern treatment. To the average method, rosion time at different levels of applied stress were obtained
the total speckle patterns are used to calculate the average inten- (shown as continuous lines in Fig. 13). Note that the trends of
sity distribution. But in autocorrelation coefficient calculation the two curves are very similar at the applied stresses of 174
only a part of the speckle patterns are used. However, the results and 348 MPa. However, they are very different from the pure
of two methods indicate that both methods are in reasonable corrosion test results where no stress is imposed. It is likely that
this trend only exists when the applied stress exceeds the thresh-
old value. Comparing the pure corrosion and stress corrosion
curves, the corrosion initialization stage was more pronounced
and rapid for specimens tested under stress corrosion conditions.
The autocorrelation coefficient method also gives the sim-
ilar results, shown as the dashed lines in Fig. 13. The same
conclusion is that the existence of stress highly influences the
developing of the roughness and increases the corrosion rate.
An obvious difference of the roughness is given between the
stresses 174 and 348 MPa using the autocorrelation coefficient
method.
Fig. 14 shows the roughness versus the corrosion time under
0 and 174 MPa, and a constant temperature 40 ◦ C.

3.3. Interaction between the temperature and the stress

In order to explain the effect of interaction between the tem-

Fig. 12. Roughness vs. the corrosion time calculated by the average intensity perature and the stress, the roughness versus the corrosion time
method (solid lines) and the autocorrelation coefficient method (dashed lines) without stress and temperature at 40 and 60 ◦ C are shown in
under different temperatures. Fig. 15. Compared with Fig. 12, in general corrosion, the trends
224 H.-J. Shi et al. / Materials Science and Engineering A 419 (2006) 218–224

to characterize the properties of stress corrosion in their initial

developing stage. When the stress is kept on a given value, the
change of the roughness has the similar trend during the corro-
sion while under the different temperatures. However, the rate of
roughening increases rapidly with the increase of the corrosion
temperature.
When the temperature is kept under a constant, the rate of
pure corrosion is lower than that of in the stress corrosion. The
interaction between the temperature and the stress increases the
corrosion rate.

Acknowledgements

We are grateful for the financial support provided by

the Chinese National Natural Science Foundation (Grant
Fig. 14. Roughness vs. the corrosion time calculated by the average intensity
method (solid lines) and the autocorrelation coefficient method (dashed lines)
Nos. 10472049, 10232030 and 50371042), and the Special
with under 0 and 174 MPa, and a constant temperature 40 ◦ C. Fund for Major State Basic Research Projects (Grant No.
2004CB619304).

References

[1] R. Nishimura, K. Yamakawa, Nucl. Eng. Des. 182 (2) (1998) 165–
173.
[2] J. Hu, R.S. Luo, C.K. Yao, L.C. Zhao, Mater. Chem. Phys. 70 (2001)
160–163.
[3] K. Saito, J. Kuniya, Corros. Sci. 43 (9) (2001) 1751–1766.
[4] D. Delafosse, J.P. Chateau, Mater. Sci. Eng. A 234–236 (1997) 889.
[5] T. Nguyen, A. Rosencwaig, Appl. Surf. Sci. 24 (1985) 57.
[6] T. Nguyen, A. Rosencwaig, Mater. Charact. 28 (1992) 291.
[7] N.E. Cipollini, J. Electrochem. Soc. 129 (1982) 1517.
[8] R.C. Engstrom, S. Ghaffari, H. Qu, Anal. Chem. 64 (1992) 2525.
[9] K. Sugimoto, S. Matsuda, Y. Ogiwara, K. Kitamura, J. Electrochem.
Soc. 132 (1985) 1791.
[10] C.C. Streinz, J. Kruger, P.J. Moran, J. Electrochem. Soc. 141 (1994)
1126.
[11] R.E. Hummel, R.J. Smith, Corros. Sci. 30 (1990) 849.
Fig. 15. Roughness vs. the corrosion time under temperatures 40 and 60 ◦ C. No
[12] M. Seo, Z.C. Jiang, N. Sato, Werkstoffe Korros. 39 (1988) 583.
stress is loaded.
[13] J.P.H. Sukamto, W.H. Smyrl, N. Casillas, M. A1-Odan, P. James, W.
Jin, L. Douglas, Mater. Sci. Eng. A 198 (1995) 177.
of the curves are smooth. There was a little difference between [14] F. Jin, F.P. Chiang, Mater. Eval. 55 (1997) 813–816.
the curves at 40 and 60 ◦ C. It reveals that the effect of the temper- [15] S.L. Toh, C. Quan, K.C. Woo, C.J. Tay, H.M. Shang, Opt. Laser Technol.
ature on the rate of stress corrosion is small. When stress was 33 (2001) 427–434.
applied, the differences were obvious, particularly at 22 and [16] Y. Song, R. Kulenovic, M. Groll, Z. Guo, Opt. Commun. 139 (1997)
40 ◦ C. In general, the interaction between the temperature and 24.
[17] E. Kolenovic, W. Osten, W. Juptner, Opt. Commun. 171 (1999) 333.
the stress appears to accelerate the corrosion process of Al2024- [18] K.D. Hinsch, T. Fricke-Begemann, G. Gulker, K. Wolff, Opt. Lasers
T3. The physical explanation for this process is complicated and Eng. 33 (2000) 87.
we don’t discuss it in this paper. [19] J.W. Goodman, in: J.C. Dainty (Ed.), Statistical Properties of Laser
Speckle Patterns in Laser Speckle and Related Phenomena, Springer,
4. Conclusion Berlin, 1984, p. 9.
[20] D.J. Whitehouse, in: P.K. Rasrog (Ed.), Photomechanics, Springer,
Berlin, 1999, pp. 413–459.
In this paper the stress corrosion behavior of Al2024-T3 was [21] F. Jin, F.P. Chiang, Res. Nondestr. Eval. 10 (1998) 63–73.
investigated using the subjective speckle method. Both aver- [22] H.C. Kandpal, D.S. Mehta, J.S. Vaishya, Opt. Lasers Eng. 34 (2000)
age intensity and autocorrelation coefficient methods were used 139–148.