You are on page 1of 10

Materials Science and Engineering A 517 (2009) 334343

Contents lists available at ScienceDirect

Materials Science and Engineering A


journal homepage: www.elsevier.com/locate/msea

Effect of strain ratio and strain rate on low cycle fatigue behavior of AZ31 wrought magnesium alloy
S. Begum a , D.L. Chen a, , S. Xu b , Alan A. Luo c
a b c

Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, Ontario M5B 2K3, Canada CANMET-Materials Technology Laboratory, Natural Resources Canada, 568 Booth Street, Ottawa, Ontario K1A 0G1, Canada General Motors Research and Development Center, Warren, MI 48090, USA

a r t i c l e

i n f o

a b s t r a c t
Magnesium alloys are increasingly used in automotive and aerospace industries for weight reduction and fuel economy improvement. Low cycle fatigue (LCF) behavior of these alloys is an important consideration for the structural applications. The objective of the present investigation was to identify inuences of strain ratio and strain rate on cyclic deformation characteristics and fatigue life of an AZ31 extruded alloy. As the strain ratio decreased, stronger cyclic hardening rate, more asymmetric hysteresis loop, smaller stress amplitude, lower mean stress, and higher initial plastic strain amplitude were observed due to increasing compressive stresses. This was considered to be associated with the twinning during cyclic deformation in the compressive phase, and detwinning in the tensile phase. The residual twins acting as barriers to dislocation slip and pile-up were considered to be the main cause for the occurrence of cyclic hardening. Fatigue life increased with decreasing strain ratio and increasing strain rate. Fatigue crack initiation occurred at the specimen surface due to the presence of larger grains near the surface, and fatigue crack propagation was characterized by a mixture of striations and dimple-like ductile fracture features. 2009 Elsevier B.V. All rights reserved.

Article history: Received 5 January 2009 Received in revised form 27 March 2009 Accepted 24 April 2009 Keywords: Magnesium alloy Cyclic deformation Low cycle fatigue life Strain ratio Strain rate

1. Introduction The increasing energy consumption and global climate change have spurred research and development activities in lightweight materials for transportation applications. Recently, potential applications of wrought magnesium alloys have attracted a great deal of research interest in these alloys [15]. Due to lightweight, high strength-to-weight ratio and high specic stiffness, wrought magnesium alloys are among the most promising structural materials. However, wrought magnesium alloys with a hexagonal closepacked (hcp) structure generally show a strong anisotropy or tensioncompression asymmetry [6]. The limited number of slip systems plays an important role in the orientation dependence of the mechanical properties [7], which enhance the nucleation of twins to assist the plastic deformation [8,9]. For AZ31B alloy, an Al, Zn rich magnesium alloy with little content of Fe, Ni, and Cu, numerous work has been done on the tensile and compressive properties but only a limited amount of work has been reported on the cyclic deformation behavior [1015]. A major concern while using this alloy as structural components is the tensile and compressive yield asymmetry. Monotonic tests are unable to envisage the effect of

stress, strain, and deformation mode of the alloy subjected to alternating stresses. In order to examine the combined effect of these parameters it is essential to investigate the low cycle fatigue (LCF) resistance. The objective of this paper is, therefore, to evaluate the cyclic deformation behavior and fatigue fracture mode with particular emphasis on the effects of strain ratio and strain rate as no such work on this alloy has been reported in the literature. The cyclic stress response, change of mean stress, evolution of plastic deformation, fatigue life, and failure characteristics are presented in this paper. 2. Test material and experimental procedure The AZ31 material used in the present investigation was in an extruded form, with a nominal composition listed in Table 1. The alloy was extruded at about 370 C with a plate thickness of 7 mm. The plate contains two bars of about (30 mm 25 mm) size along the entire length of the plate. During the manufacturing process the applied extrusion speed was 50.8 mm/s and the extrusion ratio was 6. After extrusion the alloy was air cooled. Sub-sized fatigue samples of about 140 mm in length were machined in the extrusion direction. The samples had a gauge length of 25 mm and a width of 6 mm (following ASTM E8 standard). The thickness of the samples was kept the same as the asreceived plate (7 mm). The specimens were ground progressively

Corresponding author. Tel.: +1 416 979 5000x6487; fax: +1 416 979 5265. E-mail address: dchen@ryerson.ca (D.L. Chen). 0921-5093/$ see front matter 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2009.04.051

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343 Table 1 Chemical composition (wt%) of the extruded AZ31 magnesium alloy used in the present investigation. Al 3.1 Zn 1.05 Mn 0.54 Fe 0.0035 Ni 0.0007 Cu 0.0008 Mg Balance

335

along the loading direction with 1/0, 2/0, 3/0, 4/0 emery papers to achieve a smooth and consistent surface. For the strain controlled pullpush type fatigue tests the ASTM standard E606 was followed as a guide. All the tests were performed at room temperature using a computerized Instron 8801 fatigue testing system via Fast Track Low Cycle Fatigue (LCF) program. To study the effects of mean strain and mean stress on the LCF behavior of AZ31 alloy, ve different strain ratios, Rs = 0.5, 0, 0.5, 1, and 2, were used at a given total strain amplitude of 0.4% and a constant strain rate of 1 102 s1 . Different strain rates of 1 103 , 1 102 and 8 102 s1 were further applied to evaluate the effect of strain rate on the fatigue life of the alloy. After fatigue tests scanning electron microscope (SEM) was used to examine the fatigue crack initiation sites and identify the mechanism of fatigue crack propagation under the above applied conditions. 3. Results and discussion 3.1. Microstructure, texture and tensile properties The microstructural characterization and tensile test results of the extruded AZ31 magnesium alloy have been reported earlier in [15]. Non-uniform grain sizes along the thickness of the specimen were observed, with larger grains of an average size of about 150 m appeared at both top and bottom surfaces of the specimen. Smaller grains with an average size of about 6 m appeared in the middle of the specimen. Some Mn- and Al-containing particles were also observed [15]. The test material had an yield strength of 201 MPa, ultimate tensile strength of 264 MPa, and percent elongation of 15.2% obtained at a strain rate of 1 102 s1 [15]. The AZ31 extruded magnesium alloy was observed to con 4) 2 2 4 1 , as shown in Fig. 1, where tain a main texture of (1 1 2 the pole gures were determined using an X-ray diffractometer D8 DISCOVER with GADDS software. Due to the presence of the crystallographic texture, the extruded magnesium alloy exhibited a strong tensioncompression asymmetry in the longitudinal/extrusion direction. As presented in [16], the compressive yield strength was lower than a half of the tensile yield strength. The strain life of this alloy fatigued at a xed strain ratio (R = 1) and a constant strain rate of 1 102 s1 and the relevant fatigue parameters have been reported in [15]. 3.2. Effect of strain ratio The strain ratio was dened as: Rs = min max

Fig. 1. Pole gures of the extruded AZ31 magnesium alloy selected in the present study. (a) 102, (b) 110, and (c) 103.

(1)

To reveal the potential effect of strain ratio on the fatigue characteristics of AZ31 alloy, different strain ratios were applied while

keeping other parameters unchanged. The obtained fatigue life is given in Table 2. It is seen that the fatigue life increased with decreasing strain ratio or mean strain under constant strain amplitude and strain rate. Evolution of stress amplitude during cyclic deformation is shown in Fig. 2. It is seen that the material exhibited higher cyclic hardening at the lower (or more negative) strain ratio than at the higher strain ratio, and at Rs = 0.5 the cyclic hardening became much weaker and almost linear in the semi-log coordinate. The stress amplitude also increased with increasing Rs value under the same strain amplitude and strain rate. This was related to the higher mean strain (Table 2), in agreement with the results reported in [12] where the absolute value of high mean strain cor-

Table 2 Test parameters and fatigue life under different strain ratios at a strain amplitude of 0.4% and a strain rate of 1 102 s1 . Strain ratio (Rs ) 0.5 0 0.5 1 2 Mean strain (mean ) (%) 1.2 0.4 0.133 0 0.133 Maximum strain (max ) (%) 1.6 0.8 0.533 0.4 0.267 Minimum strain (min ) (%) 0.8 0 0.267 0.4 0.533 Number of cycles to failure (Nf ) 3545 4090 4234 4294 7189

336

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343

Fig. 2. Stress amplitude vs. the number of cycles for different strain ratios, Rs , at a given strain amplitude of 0.4% and a constant strain rate of 1 102 s1 .

responded to the high stress amplitude. The evolution of cyclic stress response during cyclic deformation was an important characteristic in LCF process, and it was mainly governed by the cyclic stability of the internal microstructural features related to dislocation multiplication [1719]. Furthermore, twins acting as barriers to dislocation slip also facilitated the evolution of cyclic stress [17,2025]. Fig. 3 shows the variation of plastic strain amplitude with the number of cycles for different applied strain ratios, Rs . It is seen that at Rs = 0.5 and 0 there existed a sudden drop of the plastic strain amplitude from the rst cycle to the second cycle, and then it remained almost constant until failure. As the Rs decreased from 0.5 to 0, the magnitude of the sudden change in the plastic strain amplitude decreased. Such an abrupt drop was again related to the high positive mean strain (Table 2), which caused a large amount of plastic deformation in the initial tensile phase of the rst cycle, as seen in Fig. 4(a). At negative strain ratios the abrupt change was

Fig. 3. Plastic strain amplitude vs. the number of cycles for different strain ratios, Rs , at a given strain amplitude of 0.4% and a constant strain rate of 1 102 s1 .

absent (Fig. 3) due to the absence of yielding in the tensile phase of the rst cycle (Fig. 4(a)). However, it is clearly seen from Fig. 4(a) that the compressive yielding occurred much earlier than the tensile yielding, i.e., the compressive yielding stress was only about one third the tensile yielding stress, consistent with the results from the individual tensile and compressive tests [16]. The amount of the compressive plastic deformation increased with decreasing strain ratio. The asymmetry or skewness of the hysteresis loops could be better seen from the second cycle (Fig. 4(b)), where the hysteresis loop at Rs = 2 was most skewed, which became less skewed with increasing strain ratio, and eventually it became almost symmetric at Rs = 0.5. The reason for the asymmetry of hysteresis loops will be discussed later. The shift of the hysteresis loops at Rs = 0.5 and 0 in Fig. 4(b) was due to the control of strain limits (i.e., the xed values of the maximum strain max and minimum strain min given in Table 2), which compelled the occurrence of a large amount of plastic deformation in the tensile phase of the rst cycle (Fig. 4(a)). The skewness of the hysteresis loops almost disappeared at the halflife of the fatigue process (Fig. 4(c)). The changes in the hysteresis loops corresponded to the changes in the plastic strain amplitude in Fig. 3, where the plastic strain amplitude decreased with increasing strain ratio and with increasing number of cycles. The plastic strain amplitude became nearly constant after about 750 cycles for all the strain ratios. The stronger decrease in the plastic strain amplitude at the lower strain ratio in Fig. 3 reected a stronger cyclic hardening characteristic as shown in Fig. 2. The variation of the hysteresis loops with the strain ratio could also be expressed as a minimum-to-maximum stress ratio, c / t , vs. the strain ratio (Fig. 5), where the absolute values of the compressive minimum stress c and the tensile maximum stress t were used. The stress ratio at the rst cycle decreased drastically with increasing strain ratio, while the extent of change in the stress ratio at the mid-life cycle became much smaller. It should be noted that the change in the stress ratio in Fig. 5 indeed indicated the mean stress relaxation or cycle-dependent relaxation [26]. As expected, the stress ratio between the rst cycle and the mid-life cycle was almost the same at Rs = 1. However, the stress ratio at the mid-life cycle became lower when the strain ratio was smaller than 1, and higher when the strain ratio was larger than 1. The change in the strain ratio also led to different mean stresses, as shown in Fig. 6, where the mean stress was plotted as a function of the ratio of the number of cycles to the fatigue life, N/Nf , for different strain ratios. It is seen that the initial mean stress decreased for Rs > 1 and increased for Rs 1, and the stabilization basically occurred after about 510% of fatigue life for the applied strain ratios, except for the strain ratio of 0.5 where the mean stress decreased continuously. The unstability in the mean stress at Rs = 0.5 might be related to the initial big over-stretching (Fig. 4(a)), which also resulted in almost linear and weaker cyclic hardening (Fig. 2). The nearly stable and positive/tensile mean stress over the majority of fatigue life decreased from 50 to 25 MPa with decreasing strain ratio from Rs = 0 to Rs = 2. This was in agreement with the results reported by Goodenberger and Stephens [27] on AZ91E-T6 cast magnesium alloy. The presence of tensile mean stresses was known to reduce fatigue life, and the compressive mean stresses extended the fatigue life, according to Morrow mean stress correction [7,26]. Normally, for materials exhibiting symmetric hysteresis loops the mean stress would affect more the fatigue life of components in the highcycle regime since the large cyclic plastic strains in the low-cycle regime would lead to the mean stress relaxation. However, for the extruded AZ31 magnesium alloy tested at a given total strain amplitude with different strain ratios, while the cyclic plastic strain caused a certain extent of mean stress relaxation (Fig. 5), quite big, and almost stable tensile mean stresses were indeed observed to increase with increasing strain ratio (Fig. 6). Consequently, the

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343

337

Fig. 4. Hysteresis loops for different strain ratios, Rs , at a given strain amplitude of 0.4% and a constant strain rate of 1 102 s1 . (a) The rst cycle, (b) the second cycle, and (c) the mid-life cycle.

Fig. 5. Stress ratios of the rst cycle and the mid-life cycle as a function of the applied strain ratio.

Fig. 6. Mean stress vs. the ratio of the number of cycles to the fatigue life for different strain ratios.

338

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343

fatigue life increased as the strain ratio decreased, as shown in Table 2. Furthermore, the increase of fatigue life with decreasing strain ratio was also associated with the lower cyclic stress amplitude and stronger cyclic hardening characteristics at the lower strain ratio (Fig. 2), in spite of the initial higher plastic strain amplitude (Fig. 3). While the effect of cyclic stress amplitude on the fatigue life was straightforward (i.e., the lower the stress amplitude, the higher the fatigue life), the relationship between cyclic hardening and fatigue life needs to be stated. Normally, materials with higher monotonic strain hardening exponent exhibited cyclic hardening and materials with larger values of cyclic strain hardening exponent had longer low cycle fatigue lives [7]. It has been reported that the strain hardening exponent of solder affected the fatigue life under temperature cycling as a result of thermal expansion mismatch, representing a fatigue condition equivalent to the strain controlled low cycle fatigue [7,28]. A higher strain hardening exponent corresponded to a lower cyclic strain range in the solder, thus increasing the fatigue life [28]. Furthermore, the stronger cyclic strain hardening corresponded to a longer fatigue initiation life in metal metrix composites [29]. These results reported in the literature corroborated the fact that the stronger cyclic hardening would promote a longer low cycle fatigue life. The weaker and nearly linear cyclic hardening (Fig. 2) and relatively high and unstable mean stress (Fig. 6) would be the main reasons for the short fatigue life of the extruded AZ31 magnesium alloy tested at Rs = 0.5. The obvious asymmetry or skewness of the hysteresis loops at lower strain ratios as shown in Fig. 4 needs to be addressed. It is known that two characteristic deformation mechanisms occurred during plastic deformation of magnesium at room temperature: sliding and twinning [30,31]. In the plastic deformation of wrought magnesium alloys twinning played an important role [6,810,13,14,17,3050]. This was mainly attributed to the existence of strong crystallographic texture (Fig. 1). In a magnesium single crystal basic mechanical twinning occurred when the basal plane was in compression or the prism plane was in tension [30,31]. During compression the occurrence of twinning induced the compressive yield stress to be smaller than the tensile yield stress. Cyclic deformation behavior of common magnesium alloys at room temperature was mostly dominated by the formation and change of twins as well [10,14,15,18,30,32,33,44,48,49]. In the cyclic deformation process twinning started in the compressive loading phase and detwinning (or dissolution of twins) occurred in the tensile loading phase [6,14,15,18,33,38,44,49,50]. A typical example on the formation of twins near the fracture surface after cyclic deformation is shown in Fig. 7, where a large number of twins could be seen in some larger grains. Due to the mechanical twinning the extruded hexagonal close-packed magnesium alloy exhibited tensioncompression yielding asymmetry (Fig. 4) and the material deformed primarily by slip in one direction and by twinning in the other [33,38]. This was due to the fact that the dominant basal slip mode had only limited slip systems which could not meet the requirement of ve independent slip systems based on the von Mises criterion for an arbitrary homogeneous straining. In particular, the basal plane slip could not accommodate straining along the c-axis. Besides the dislocation slip with c + a Burgers vectors, being a very hard deformation mechanism [44,51], twinning was considered to be the only active deformation mode that could provide straining along the c-axis at room temperature [44]. In-situ neutron diffraction during cyclic loading in tension and compression of extruded bar revealed that detwinning resulted in complete texture reversal during initial cycles, but eventually fatigued resulting in some residual twin component [33]. The variation of cyclic hardening characteristics in the AZ31 magnesium alloy with decreasing strain ratio (Fig. 2) could be understood to be associated with the formation of residual twins during cyclic deformation [1315]. Wu

Fig. 7. Typical SEM micrographs showing the formation of twins near fracture surface. (a) A lower magnication image, (b) a magnied view of area B enclosed by the dashed line box, and (c) a magnied view of area C enclosed by the solid line box.

et al. [18] observed that in a wrought magnesium alloy ZK60A residual twins formed after the act of twinning and detwining in each cycle, and with increasing number of cycles the volume fraction of residual twins increased, leading to an increasing hardening rate. As seen from Fig. 4, with decreasing strain ratio leading to more asymmetric hysteresis loops, more twins would be expected to form in the compressive phase due to the larger negative/compressive stress, and subsequently more residual twins would also be likely after the occurrence of detwining in the tensile phase, due to the decreasing maximum tensile stress. Therefore, the stronger cyclic hardening rate at the lower strain ratio (Fig. 2) was attributed to the accumulation of more residual twins as cyclic deformation

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343 Table 3 Variation of fatigue lifetime with the strain rate applied at a strain amplitude of 0.4% and a strain ratio of 1. Strain rate (s1 ) 1 103 1 102 8 102 Number of cycles to failure (Nf ) 2467 4294 7268

339

proceeded, since the residual twins would act as barriers to the dislocation slip and pile-up. On the other hand, at the high strain ratio of Rs = 0.5, as the introduction of high mean strain (Table 2) resulted in an insufcient compressive stress, the twinning was not anticipated, leading to the disappearance of skewness of the hysteresis loop (Fig. 4) and the weaker and nearly linear cyclic hardening (Fig. 2). 3.3. Effect of strain rate Three different (low, middle, and high) strain rates were applied to examine the strain rate effect on the fatigue life of the extruded AZ31 magnesium alloy. The test results obtained at the strain rates applied are listed in Table 3. It is seen that a higher strain rate corresponded to a longer fatigue life. Fig. 8 shows the stress evolution during the strain controlled low cycle fatigue tests where the stress amplitude vs. the number of cycles was plotted for different strain rates. The stress amplitude evolved under the applied high, middle, and low strain rates was almost the same within the experimental error (with an initial stress amplitude of about 117123 MPa). It was, however, clear that cyclic hardening occurred at all the three strain rates. While the cyclic hardening rates were nearly the same (i.e., almost parallel) for the tests carried out at the strain rates of 1 102 and 8 102 s1 , the cyclic hardening rate was somewhat lower for the sample tested at the low strain rate of 1 103 s1 . This might be related to the time dependence of cyclic deformation. As mentioned above, a certain degree of cycle-dependent stress relaxation occurred. If more time in a cycle were given, corresponding to a lower strain rate, more cycle-dependent relaxation would be expected, giving rise to a lower cyclic hardening rate. Fig. 9 shows the variation of the plastic strain amplitude with the number of cycles at different strain rates. Similarly, the magnitude

Fig. 9. Plastic strain amplitude vs. the number of cycles at a strain amplitude of 0.4% and different strain rates.

of the plastic strain amplitudes under the high, middle, and low strain rates was roughly the same within the experimental error (with an initial plastic strain amplitudes of about 0.090.1%). The curves of the plastic strain amplitude vs. the number of cycles at the strain rates of 1 102 and 8 102 s1 were approximately parallel with a slightly steeper slope, and the curve obtained at the low strain rate of 1 103 s1 was somewhat atter. All of these changes corresponded well to those in the stress amplitude at different strain rates shown in Fig. 8. Therefore, the cyclic hardening characteristics could be well represented by either the increase of the cyclic stress amplitude (Fig. 8) or the decrease of the plastic strain amplitude (Fig. 9) with increasing number of cycles in a strain controlled low cycle fatigue test. This has been discussed in our previous publications [1315]. It is also seen from Fig. 9 that the plastic strain amplitude rst decreased and then suddenly changed its direction from decrease to increase prior to failure. Such an abrupt change has been found to represent the onset of fatigue crack initiation [13]. It appeared that the fatigue crack initiation stage before the turning point became shorter at the faster strain rate. However, the process represented by an increase in the plas-

Fig. 8. Stress amplitude vs. the number of cycles at a strain amplitude of 0.4% and different strain rates.

Fig. 10. Mean stress vs. the ratio of the number of cycles to the fatigue life at a strain amplitude of 0.4% and different strain rates.

340

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343

tic strain amplitude after the turning point, corresponding to the fatigue crack propagation stage prior to the nal rapid failure [13], lasted much longer with increasing strain rates from 1 103 to 8 102 s1 (the X-axis in Fig. 9 was in the log scale). Therefore, the total fatigue life consisting of crack initiation life and propagation life was mainly dependent on the crack propagation stage in the varying strain-rate tests. The variation in the strain rates also led to some changes in the mean stress, as shown in Fig. 10, where the mean stress was presented as a function of the ratio of the number of cycles to the fatigue life, N/Nf . It is seen that the initial mean stresses for all the three strain rates were approximately the same, about 2530 MPa, but they changed as cyclic deformation progressed. At the high strain rate of 8 102 s1 , the mean stress increased rapidly within about

10% of the fatigue life, and then it remained basically constant. The range within which the mean stress increased became about 20% of the fatigue life at the intermediate strain rate of 1 102 s1 , while the mean stress increased slowly and continuously at the low strain rate of 1 103 s1 . Such changes gave rise to the largest difference in the mean stresses occurring at the about 10% of the fatigue life, after which the difference became gradually smaller although the slightly higher mean stress arising from the higher strain rate could still be seen. 3.4. Examination of fracture surfaces Fig. 11 shows an overall view of fracture surfaces of samples fatigued at Rs = 0.5, 0, 0.5, 1 and 2, respectively, containing

Fig. 11. Low magnication SEM images showing an overall view of fracture surfaces of specimens fatigued at a strain amplitude of 0.4% at a strain ratio of (a) 0.5, (b) 0, (c) 0.5, (d) 1, and (e) 2.

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343

341

Fig. 12. SEM micrographs of fracture surfaces near the crack initiation of specimens fatigued at a strain amplitude of 0.4% at a strain ratio of (a) 0.5, (b) 0, (c) 0.5, (d) 1, and (e) 2.

fatigue crack initiation, propagation, and nal fast fracture regions. It is seen that fatigue crack initiation basically occurred from the specimen surface. In these low magnication images the river line patterns were observed in all samples which were irregular and broken and owed along the crack propagation direction like a wave form. The size of the crack propagation area varied largely with the applied strain ratio; the lower the strain ratio, the larger the propagation area. The sample tested at a strain ratio of Rs = 2, Fig. 11(e), exhibited the largest propagation area compared with the samples tested at other strain ratios, due to the lowest maximum tensile stress (Fig. 4) and the smallest mean stress (Fig. 6). This corresponded to the longest fatigue life compared to other samples (Table 2). Fatigue crack propagation was dominated by the stress intensity factor range, K, and only pulsating tension basically determined the propagation rate [7]. As seen in Fig. 2, the stress amplitude at Rs = 2 test was the lowest of all the tests, which led to a delayed crack initiation and a lower value of stress intensity factor range for the crack propagation, thus increased the fatigue life. Fig. 12 shows the images near the crack initiation sites of the fatigued samples. Crack initiation with varying orientations in

different grains was observed. Crack growth was predominantly transgranular or intercrystalline [52,53]. Such a cracking feature varied with the applied strain ratio, with the highest percentage occurred at Rs = 0.5 sample compared to the other samples. The initial stage of crack growth in the hexagonal close-packed alloys like Mg and Ti has been reported to exhibit striation-like features [52]. Indeed, fatigue crack propagation was mainly characterized by a mixture of fatigue striations and dimple-like ductile fracture features observed at higher magnications as shown in Fig. 13(a)(d) for Rs = 0.5, 0, 1, and 2 tests. The striation spacing became larger with increasing strain ratio. In addition, some secondary cracks could also be seen. It is known that the occurrence of fatigue striations was caused by a repeated plastic blunting-sharpening process in face-centered cubic materials due to the slip of dislocations in the plastic zone ahead of the fatigue crack tip [7,54]. The formation of the fatigue striations in the extruded hexagonal closepacked magnesium alloy was anticipated to be associated with twinning in the compressive phase and detwinning in the tensile phase [6,1315,18,33,38,44,49,50]. Further studies on the formation mechanisms of the fatigue striations in magnesium alloys are needed.

342

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343

Fig. 13. SEM micrographs of fracture surfaces in the fatigue crack propagation region of specimens fatigued at a strain amplitude of 0.4% at a strain ratio of (a) 0.5, (b) 0, (c) 1, and (d) 2.

4. Conclusions 1. Stronger cyclic strain hardening was observed to occur at the lower strain ratio than at the higher strain ratio. This was related to the formation of twins during cyclic deformation in the compressive phase, and detwinning in the tensile phase. The residual twins acting as barriers to dislocation slip and pile-up were the main cause for the occurrence of cyclic hardening. The decrease of cyclic hardening capacity at the higher strain ratio would be a result of fewer twins due to a smaller or insufcient compressive stress. 2. In the initial stage of cyclic deformation, the hysteresis loops were more asymmetric or skewed with decreasing strain ratio because of the presence of larger compressive stresses. The skewness of hysteresis loops at the lower strain ratios basically vanished at the half-life of fatigue process. 3. While the initial stress ratio decreased drastically with increasing strain ratio, this decrease became modest at the half-life. A certain extent of mean stress relaxation or cycle-dependent relaxation was observed especially when the strain ratio was not equal to 1 or the strain rate was low. 4. Fatigue life of the extruded AZ31 magnesium alloy increased with decreasing strain ratio or mean strain and with increasing strain rate, as a combined consequence of the cyclic hardening rate, stress amplitude, plastic strain amplitude, and mean stress. 5. While fatigue crack initiation stage seemed to be shorter, the crack propagation stage was observed to be much longer. Thus a longer total fatigue life at the higher strain rate was obtained. 6. Fatigue crack initiated at the surface of the test samples due to the presence of larger grains near the specimen surface. With decreasing strain ratio, the size of crack propagation area became larger. Fatigue crack propagation was characterized by a mixture of striations and dimple-like ductile fracture features.

Acknowledgements The authors would like to thank the Natural Sciences and Engineering Research Council of Canada (NSERC) and AUTO21 Network of Centres of Excellence for providing nancial support. This investigation involves part of CanadaChinaUSA Collaborative Research Project on the Magnesium Front End Research and Development (MFERD), and nancial support from CANMET-MTL is acknowledged. One of the authors (D.L. Chen) is also grateful for the nancial support by the Premiers Research Excellence Award (PREA), Canada Foundation for Innovation (CFI), and Ryerson Research Chair (RRC) program. Dr. K. Sadayappan, Dr. J. Lo, Dr. W. MacDonald and Dr. J. Jackman are gratefully acknowledged for their support and continuous encouragement while performing this investigation. The authors would also like to thank Dr. Xichen Sun of Chrylser for providing the pole gures in this study, and Messrs. A. Machin, Q. Li, J. Amankrah, D. Ostrom, and R. Churaman for easy access to the laboratory facilities of Ryerson University and their assistance in the experiments. The authors would also like to thank Prof. S.D. Bhole, Prof. N. Atalla, Prof. S. Lambert, Prof. H. Jahed, Prof. Y.S. Yang, Mr. R. Osborne, Dr. X.M. Su and Mr. L. Zhang for helpful discussion. References
[1] E.A. Nyberg, A.A. Luo, K. Sadayappan, W.F. Shi, Adv. Mater. Processes 166 (2008) 3537. [2] H.E. Friedrich, B.L. Mordike, Magnesium Technology Metallurgy, Design Data Applications, Springer-Verlag, Berlin, Heidelberg, Germany, 2006, pp. 63310, 499627. [3] A.A. Luo, JOM 54 (2002) 4248. [4] A. Stalmann, W. Sebastian, H. Friedrich, S. Schumann, K. Droder, Adv. Eng. Mater. 3 (2001) 969974. [5] K.U. Kainer, MagnesiumAlloys and Technologies, WILEY-VCH Verlag Gmbh & Co., KG, Weinheim, Cambridge, 2003.

S. Begum et al. / Materials Science and Engineering A 517 (2009) 334343 [6] J.P. Nobre, U. Noster, M. Kornmeier, A.M. Dias, B. Scholtes, Key Eng. Mater 230232 (2002) 267270. [7] G.E. Dieter, Mechanical Metallurgy, McGraw-Hill series, NewYork, USA, 1986. [8] D.W. Brown, S.R. Agnew, M.A.M. Bourke, T.M. Holden, S.C. Vogel, C.N. Tome, Mater. Sci. Eng. A 399 (2005) 112. [9] Y.N. Wang, J.C. Huang, Acta Mater. 55 (2007) 897905. [10] U. Noster, B. Scholtes, Z. Metallkd. 94 (2003) 559563. [11] U. Noster, B. Scholtes, Mater. Sci. Forum 419422 (2003) 103108. [12] S. Hasegawa, Y. Tsuchida, H. Yano, M. Matsui, Int. J. Fatigue 29 (2007) 18391845. [13] X.Z. Lin, D.L. Chen, Mater. Sci. Eng. A 496 (2008) 106113. [14] S. Begum, D.L. Chen, S. Xu, A.A. Luo, Metall. Mater. Trans. A 39A (2008) 30143026. [15] S. Begum, D.L. Chen, S. Xu, A.A. Luo, Int. J. Fatigue 31 (2009) 726735. [16] S. Xu, W.R. Tyson, R. Bouchard, R. Eagleson, Mater. Sci. Forum 618619 (2009) 527532. [17] M.H. Yoo, J.R. Morris, K.M. Ho, S.R. Agnew, Metall. Mater. Trans. A 33A (2002) 813822. [18] L. Wu, A. Jain, D.W. Brown, G.M. Stoica, S.R. Agnew, B. Clausen, D.E. Fielden, P.K. Liaw, Acta Mater. 56 (2008) 688695. [19] M.H. Yoo, S.R. Agnew, J.R. Morris, K.M. Ho, Mater. Sci. Eng. A 319321 (2001) 8792. [20] M.R. Barnett, Z. Keshavarz, A.G. Beer, D. Atwell, Acta Mater. 52 (2004) 50935103. [21] G.F. Quan, Key Eng. Mater. 345346 (2007) 717720. [22] R. Gehrmann, M.M. Frommert, G. Gottstein, Mater. Sci. Eng. A 395 (2005) 338349. [23] N. Munroe, X.L. Tan, H.C. Gu, Scripta Mater. 36 (1997) 13831386. [24] A. Serra, D.J. Bacon, R.C. Pond, Metall. Mater. Trans. A 33A (2002) 809812. [25] L. Jiang, J.J. Jonas, R.K. Mishra, A.A. Luo, A.K. Sachdev, S. Godet, Acta Mater. 55 (2007) 38993910. [26] N.E. Dowling, Mechanical Behavior of Materials, third edition, Pearson-Prentice Hall, New Jersey, 2007, pp.649652, 728736. [27] D.L. Goodenberger, R.I. Stephens, J. Eng. Mater. Technol. 115 (1993) 391397. [28] M. Pecht, Integrated Circuit, Hybrid and Multichip Module Package Design Guidelines: A Focus on Reliability, Wiley-IEEE, 1994, pp. 204207. [29] Q. Zhang, D.L. Chen, Int. J. Fatigue 27 (2005) 417427. [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54]

343

H. Zenner, F. Renner, Int. J. Fatigue 24 (2002) 12551260. A. Staroselsky, L. Anand, Int. J. Plast. 19 (2003) 18431864. S.M. Yin, H.J. Yang, S.X. Li, S.D. Wu, F. Yang, Scripta Mater. 58 (2008) 751754. D.W. Brown, A. Jain, S.R. Agnew, B. Clausen, Mater. Sci. Forum 539543 (2007) 34073413. A.N. Chamos, S.G. Pantelakis, G.N. Haidemenopoulos, E. Kamoutsi, Fatigue Fract. Eng. Mater. Struct. 31 (2008) 812821. S.M. Yin, F. Yang, X.M. Yang, S.D. Wu, S.X. Li, G.Y. Li, Mater. Sci. Eng. A 494 (2008) 397400. F. Yang, S.M. Yin, S.X. Li, Z.F. Zhang, Mater. Sci. Eng. A 491 (2008) 131136. H. Somekawa, N. Maruyama, S. Hiromoto, A. Yamamoto, T. Mukai, Mater. Trans. 49 (2008) 681684. M. Krauss, B. Scholtes, Materialwiss. Werkstofftech. 38 (2007) 802807. J. Koike, Mater. Sci. Forum 449452 (2004) 665668. R.C. Zeng, E.H. Han, L. Liu, Y.B. Xu, W. Ke, Chin. J. Mater. Res. 17 (2003) 241246. T.S. Shih, W.S. Liu, Y.J. Chen, Mater. Sci. Eng. A 325 (2002) 152162. M.J. May, R.W.K. Honeycombe, Inst. Met. J. 92 (1963) 4149. R.W. Armstrong, G.T. Horne, Inst. Met. J. 91 (1963) 311315. L. Wu, S.R. Agnew, D.W. Brown, G.M. Stoica, B. Clausen, A. Jain, D.E. Fielden, P.K. Liaw, Acta Mater. 56 (2008) 36993707. S. Xu, V.Y. Gertsman, J. Li, J.P. Thompson, M. Sahoo, Can. Metall. Q. 44 (2005) 155166. M.M. Myshlyaev, H.J. McQueen, A. Mwembela, E. Konopleva, Mater. Sci. Eng. A 337 (2002) 121133. H.J. McQueen, M.M. Myshlyaev, A. Mwembela, Can. Metall. Q. 42 (2003) 97112. M.A. Gharghouri, G.C. Weatherly, J.D. Embury, J. Root, Phil. Mag. A 79 (1999) 16711695. E.C. Oliver, M.R. Daymond, P.J. Withers, Mater. Sci. Forum 490491 (2005) 257262. X.Y. Lou, M. Li, R.K. Boger, S.R. Agnew, R.H. Wagoner, Int. J. Plast. 23 (2007) 4486. S.R. Agnew, O. Duygulu, Int. J. Plast. 21 (2005) 11611193. V.V. Ogarevic, R.I. Stephens, Annu. Rev. Mater. Sci. 20 (1990) 141177. K. Tokaji, M. Kamakura, Y. Ishiizumi, N. Hasegawa, Int. J. Fatigue 26 (2004) 12171224. C. Laird, Fatigue Crack Propagation, ASTM STP 415, 1967, pp.131168.

You might also like