International Journal of Fatiguelibrary.nmlindia.org/FullText/IJF32102411.pdf · Influence of microstructure on tensile properties and fatigue crack growth in extruded magnesium

International Journal of Fatigue 32 (2010) 411–419

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier .com/locate / i j fa t igue

Influence of microstructure on tensile properties and fatigue crack growthin extruded magnesium alloy AM60

Rongchang Zeng a,b,*, Enhou Han b, Wei Ke b, Wolfgang Dietzel c, Karl Ulrich Kainer c, Andrejs Atrens d

a Chongqing University of Technology, Xingsheng Rd. 4, Chongqing 400050, Chinab Institute of Metals Research, Chinese Academy of Science, Wencui Rd. 62, Shenyang 110016, Chinac GKSS-Forschungszentrum Geesthacht GmbH, Max-Planck-Strasse 1, D-21502 Geesthacht, Germanyd The University of Queensland, Division of Materials, Brisbane QLD 4072, Australia

a r t i c l e i n f o

Article history:Received 8 February 2009Received in revised form 16 July 2009Accepted 30 July 2009Available online 4 August 2009

Keywords:Magnesium alloyIntermetallic compoundsMicrostructureStrain-hardening exponentFatigue crack growth

0142-1123/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.ijfatigue.2009.07.021

* Corresponding author. Address: Chongqing Univeeng Rd. 4, Chongqing 400050, China. Tel./fax: +86 23

E-mail address: [email protected] (R. Zen

a b s t r a c t

The microstructure, mechanical properties and fatigue crack propagation (FCP) of extruded magnesiumalloy AM60 were investigated and compared with rolled AM60. The extruded AM60 has an inhomoge-neous microstructure characterized by a-matrix, b phases and Al–Mn precipitates and denuded zonesas well, whereas rolled AM60 has fine grains. The change in strain-hardening exponent of extrudedAM60 with strain rate is ascribed to inhomogeneous microstructure. In situ double twinning:ð1012Þ � ð0 112Þ and f1011g � f1012g occurred during FCP of extruded alloy. Its crack initiation andgrowth are related to slip bands, double twinning and intermetallic compounds. Small cracks resultedfrom oxide and intermetallic compounds in rolled AM60 may be responsible for oscillatory crack growthand crack arrest. Extruded AM60 has a slower FCP rate than rolled one.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Magnesium alloys have attracted interest as structural materi-als in the automobile industry due to their low density and ade-quate strength-to-weight ratio and specific stiffness [1]. Recently,there has been significant effort to improve their mechanical prop-erties. Grain refinement is one approach to increase yield strengthwith decreasing grain size according to the Hall–Petch relation-ship. Grain refined alloys could be produced by hot working byextrusion or rolling. Wrought magnesium alloys have mechanicalproperties superior to those of castings, particularly ductility andtoughness, and consequently are better for stressed components.Therefore, to facilitate the more widespread utilization of magne-sium alloys in structural applications, it is necessary to gain betterunderstanding of the relationship between fatigue and microstruc-ture in wrought magnesium alloys.

There has been some prior research on FCP of magnesium alloys[2–6]. Eisenmeier et al. [7] claimed that fatigue cracks in cast AZ91initiate from casting defects that act as stress raisers as did Mayer etal. [8] who reported that fatigue cracks initiate at pores in 98.5% ofthe samples of high-pressure die-cast AM60. In contrast, Shih et al.[9] and Zeng et al. [10] found that cracks of extruded AZ61 and AZ80initiated at inclusions near the specimen surface. Analogous results

ll rights reserved.

rsity of Technology, Xingsh-68665616.g).

were also observed on titanium alloys [11]. Previous work of Zengand co-workers [4,10,12] also revealed that fatigue crack growthis significantly influenced by the microstructure. Wrought materi-als do have sometimes heterogeneous microstructure features,which include lamellar structures, elongated grains, intermetalliccompounds and oxide inclusions; these can be at numerous lengthscales ranging from nanometers to millimeters [13]. These hetero-geneities may induce anisotropy in mechanical properties and mayinfluence fatigue. For instance, Xu et al. [6] reported that zonaldistributed Mg3Y2Zn3 in forged Mg–Zn–Y–Zr has an appreciableinfluence on the FCP rate. Moreover, fatigue crack initiation maybe caused by dislocation slip or twining. Nevertheless there is stilla lack of knowledge of the influence of twinning on fatigue.

The present study investigated the influence of microstructureon the strain rate sensitivity and fatigue crack initiation andgrowth of extruded AM60 and rolled AM60. Particular attentionwas placed on elucidating the microstructure features associatedwith the fatigue path in order to elucidate the intrinsic fatiguemechanism of magnesium alloys.

2. Experimental

2.1. Materials

The nominal composition of the AM60B is given in Table 1. Therolled reference was produced by rolling 18 mm extruded AM60

Table 1Nominal composition of magnesium alloy AM60, wt.%.

Al Mn Si Fe Cu Ni Mg

5.60–6.40 0.26–0.50 60.05 60.004 60.008 60.001 Balance

Fig. 2a. Microstructures of extruded AM60.

Fig. 2b. Microstructures of rolled AM60.

412 R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419

plates, which was fabricated by Guangling Magnesium IndustryScience and Technology Co., Ltd., in Beijing, China. The specimenswere pre-heated at 500 �C for 1 h, hot-rolled at 0.4 mm/pass, to10 mm with intermediate re-heating at 500 �C for 15 min. The roll-ing direction was parallel to the extrusion direction. The extrudedand the rolled materials are designated as AM60-F and AM60-R,respectively.

2.2. Metallography

The specimens for microstructure observations were machinedfrom extruded and rolled plates in three directions: longitudinal(L), long transverse (T) and short transverse (S). The specimens werepolished with fine grit emery papers, disc polished using 1 lm dia-mond powder, washed with acetone and etched in a solution of10 ml acetic acid, 4.2 g picric acid, 70 ml ethanol and 10 ml distilledwater. Microstructure characterization was carried out with anoptical microscope. A Philips ESEM � L30 FEG scanning electronmicroscope (SEM) equipped with energy dispersive spectroscopy(EDS) was employed to characterize fatigue crack surfaces.

2.3. Tensile tests

The specimens for tensile tests were cut from the extrudedplates along the extrusion direction (shown in Fig. 1). The plate testspecimens, conformed to Standard GB/T 16865-1997, with rectan-gle ends, width 6 mm, thickness 4 mm and gauge length 25 mm.The tensile tests were conducted using a MTS-858 Mini Bionixmechanical test machine at room temperature, using a clip-onextensometer with a gauge length of 25 mm. At least three speci-mens for each strain rate fractured in the gauge.

2.4. Fatigue tests

The single-notched (notch depth 3.5 mm) specimens of 32 mmby 8 mm cross-section and 150 mm length were machined alongthe extrusion direction. The schematic diagram of specimens canbe found somewhere [12]. For fatigue crack growth monitoring,specimens were polished to mirror finish using 1 lm diamondpowder for in situ microstructure observation. Constant loadamplitude fatigue crack propagation was studied using an Instron8500 fatigue machine in ambient air at 16–20 �C and 40–70% rela-tive humidity, at a frequency of 1 Hz and a load ratio of zero with asinusoidal waveform. The crack length, a, was measured using aQuesta traveling microscope. The stress intensity factor range DKwas calculated according to Ref. [10]. The threshold stress intensityfactor range, DKth, was obtained by a decreasing load method.

Fig. 1. Schematic diagram for the tensile specimens taken from the extruded platesalong the extrusion direction.

Fig. 3a. SEM micrograph of macro-segregation of Mg17Al12 (indicated by horizontalarrow) and Al–Mn particles (indicated by vertical arrows) in extruded AM60.

3. Results

3.1. Microstructure

The optical micrographs illustrating the grain structure ofextruded and rolled AM60 are presented in Figs. 2a and 2b, respec-tively. The microstructure of extruded AM60 was characterized by

Fig. 3b. SEM micrograph of the denuded zones without Mg17Al12 precipitates.

R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419 413

a-matrix, intermetallic compounds and twins as well as bandingelongated in the extrusion direction and a grain size ranging from10 lm to 150 lm with the average grain size of about 56 lm. Forrolled AM60, the microstructure consisted of well-defined grains(average grain size �17 lm) and twins, which appeared obviouslyon the S–L surface and shear bands as well.

Fig. 3a presents a typical SEM micrograph of the macro-segrega-tion of Mg17Al12 (indicated with a horizontal arrow) and Al–Mnparticles (indicated with vertical arrows) in extruded AM60. Therewere denuded zones, devoid of b phase particles at grain bound-aries, far from the regions rich in segregated intermetallic com-pounds in Fig. 3b. The smaller particles with square shapes inFig. 4 were identified by EDS to be composed of Al, Mn and a traceof silicon. The atomic ratio of Al:Mn was 1.14:1 indicating theseparticles were Al–Mn particles, originally from the cast alloy.

3.2. Tensile properties

Fig. 5 presents typical true stress–true strain curves forextruded AM60 at various strain rates at room temperature.

Fig. 4. Chemical composition of the Al–M

Surprisingly, the flow curve was most different at the middle strainrate of 0.147 s�1, whereas the other two curves at the lowest andhighest strain rates almost coincided. The ultimate tensile strength(UTS), yield strength (YS) and elongation to failure (EL) of extrudedAM60 in the longitudinal direction at various strain rates are pre-sented in Table 2. The scatter of the data may be attributed to theheterogeneity in microstructure. As the strain rate increased from5.3 � 10�3 s�1 to 1.47 � 10�2 s�1, the UTS and YS increased. How-ever, when the strain rate increased further up to 0.147 s�1, theUTS and YS decreased. It was also unexpected that the ductilityalso initially increased and later decreased with an increase instrain rate.

In order to evaluate the strain rate sensitivity, the strain-hard-ening exponent, n, was determined from the tensile curves. Thetrue flow stress, rT, obeys following equation:

rT ¼ kenT ð1Þ

where k represents the strength coefficient, eT is the true strain, andn is the strain-hardening exponent.

The strain-hardening exponent first decreased and then in-creased as the strain rate increased from 5.3 � 10�3 s�1 to1.47 � 10�1 s�1. The decrease in n value suggests obviously strainsoftening behavior and vice versa. It can be seen from Table 2 thatYS/UTS increases with an increase in strain rate, while ductilityincluding elongation to failure (EL) and reduction in area (RA) firstincrease and then decrease with increasing strain rate.

3.3. Fatigue crack propagation

The relationship between FCP, da/dN, and the cyclic stressintensity factor range, DK, for extruded AM60 and rolled AM60 ispresented in Fig. 6. Owing to its fine-grained microstructure, theFCP rate of rolled AM60 was considerably faster than that of ex-truded AM60. It is known that fine-grained material exhibits ahigher FCP [14] and a coarser microstructure or/and a more heter-ogeneous microstructure leads to a lower crack propagation rate.Since the plastic deformation zone in fine-grained material is nor-mally larger than the grain size, a reverse slip of the dislocationsduring unloading is often impossible and damage accumulation

n phase in AM60 identified with EDS.

Fig. 5. True stress–strain curves at various strain rates for extruded AM60. Fig. 6. FCP rate versus DK curves for extruded AM60 and rolled AM60.

Table 3Values of threshold stress intensity factor range, DKth, of the extruded materials.

Load ratio R 0 0.2DKth <2.24 <1.55

414 R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419

is large during cycling [15]. Furthermore, the characteristic plateaurange and subsequent acceleration in the curves indicate a de-crease in FCP rate of rolled AM60 at a DK of 9 MPa m1/2. This im-plies crack arrest in the high stress intensity factor range. Thisunusual phenomenon may be related to small crack initiationresulting from oxides. Chino [16] observed that the oxide contam-inants stimulated crack formation. And a fractograph will exhibitin later discussion.

The values (in Table 3) of the threshold stress intensity factorrange, DKth, of extruded AM60 reveal that the threshold decreasedwith an increase in stress ratio; this is similar to our previous work[4].

3.4. Fractography

The overall fatigue crack surfaces of extruded AM60 and rolledAM60 are presented in Figs. 7a, 7b, 8a and 8b. Rolled AM60 has asmooth fatigue crack surface, whereas extruded AM60 has a rela-tively rough fatigue crack surface attributed to the inhomogeneousmicrostructure. Fig. 7a shows a very rough fatigue crack surfacewith striations and tire tracks that are formed by being cyclicallyopened, shifted, and pressed together after the fatigue crack fronthas passed at a low DK. Cyclic straining of the fatigue crack sur-faces after the fatigue crack has passed causes slip band cracks.Fig. 7b is characterized by an extremely rough surface with clearindications of elongated shear dimples and Al–Mn particles on

Table 2Tensile properties for AM60 at various strain rates.

No. Strain rate (s�1) UTS (MPa) YS (MPa)

1 240 1562 0.147 237 1563 266 1864 273 1885 1.47 � 10�2 272 1656 267 1667 265 1408 5.3 � 10�3 240 1369 266 142

the bottom of the dimples at a high DK. It represents the FCP mech-anism of micro-void formation and coalescence at the fatigue cracktip. Micro-voids grow from the weak boundaries between Al–Mnparticles and the matrix. Often, their growth and coalescence pre-cludes macroscopic deformation and the absorption of largeamounts of energy. Under increasing stress conditions the free sur-faces grow into rounded voids, and under sufficient stress theycoalesce to form a local increment of fracture. Fig. 8a is character-ized by a very flat surface with many small embedded crackswhich may be related to oxides and Al–Mn precipitates. Oxide asan inclusion is normally found in metallic materials. Especially,magnesium alloys are more readily oxidized due to their higherchemical activity with oxygen. Therefore, a complete preventionof oxide inclusions is impossible. Because the rolled alloy washeated directly in air at high temperature before the rolling processfor some time, it is quite likely that the oxide film was rolled intothe interior. It has been demonstrated that in many cases oxideswere the nucleation sites of fatigue cracks in magnesium alloys. Fa-tigue cracks nucleated at surface oxide of AZ80 [13], at trapped oxi-des at near-surface of AZ91E [8]. Tearing voids also existed at a

YS/UTS n value EL (%) RA (%)

0.65 9.2 5.690.66 0.21 8.4 4.170.70 22.0 11.40.69 25.6 13.90.61 0.17 21.6 12.30.62 16.8 13.90.53 18.8 9.260.57 0.21 12.0 5.940.53 19.6 11.1

Fig. 7b. Fractographic observation showing Al–Mn particles on the bottom of sheardimples at a high DK level.

Fig. 8a. SEM micrographs showing numerous secondary macro-cracks near oxide(designated as white arrows) and Al–Mn particles (designated as black arrows) at alow DK level on flat fracture surface for rolled AM60.

Fig. 8b. SEM micrographs showing no apparent dimples at a high DK level on flatfracture surface for rolled AM60.

Fig. 7a. Fractographic observation showing tire tracks at a low DK level on thefracture of extruded AM60.

R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419 415

high DK in Fig. 8b. For rolled AM60 with a flat fatigue crack therewere no striations.

4. Discussion

Based on these experimental observations, the tensile proper-ties and FCP behavior of the wrought magnesium alloys was asso-ciated with the synergistic influence of microstructure anddislocation-microstructure interactions.

4.1. Influence of strain rate on tensile properties

It has been demonstrated that for magnesium alloys basal plane(0 0 0 1) has a tendency to be parallel to extrusion direction duringtheir extrusion processes [17,18]. Magnesium exhibits a strongpropensity for mechanical twinning because twinning has a lowercritical resolved shear stress (CRSS) than hc + ai pyramidal slip atroom temperature [19]. Twinning tends to occur in coarse grainsinstead of fine grains. Since the c-axis of the grains is predomi-nantly perpendicular to the extrusion or tensile direction, deforma-tion twinning tends to reorient basal planes to more favorableorientations which in turn lead to softening. Twining accommo-dates strain along the c-axis, and thus gives rise to a decrease inthe work hardening rate [19].

Our findings are practically identical with the results at twostrain rate ranges in literatures [19–21]. At a low strain rate rangebetween the strain rates of 5.3 � 10�3 s�1 and 1.47 � 10�2 s�1, ourexperimental results regarding UTS, YS and n are in agreementwith the findings from Jiang et al. [19] that the UTS and YS in-creased and n values decreased with increasing strain rate for ex-truded AM30 and AZ31B. The strain hardening behavior isattributed to softening induced by double twinning occurring insome ‘‘soft” grains or coarse grains [19,20]. While in the case ofthe extremely high strain rate of 1.47 � 10�1 s�1, it is very difficultfor twining to take place due to the fact that twins decrease withincreasing strain [22]. Our results in this scenario are in goodagreement with Takuda et al. [21] who observed that the n valueincreased with increasing strain rate for Mg–9Li–1Y at room tem-perature. Higher strain hardening may thus be a consequence ofthe numerous dislocation piles-up on the more fine-grain bound-aries and twinning boundaries as well as slip–twinning interaction[23]. In comparison with the Ref. [24] dealing with the influence ofstrain rate on the tensile properties of cast AM60, it can be con-cluded that tensile properties of extruded AM60 are more sensitiveto the strain rate.

416 R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419

It was unexpectedly to find that the ductility of extruded AM60was highest at the middle strain rate. In contrast, extruded AM30and AZ31B exhibit a decreasing ductility with increasing strain rate[13,19]. This may be attributed to the inhomogeneous microstruc-ture in which coarse grains and fine recrystallised grains co-exist inextruded AM60. The interaction of softening induced by twinningin coarse grains and hardening induced by grain boundaries in finegrains results in a higher ductility and a lower n value at the strainrate of 0.0147 s�1.

Although in the present study no direct observation on twin-ning after tensile tests was made, it can be found the support fromthe Ref. [25] that deformation persistent slip bands and twinningof as-extruded magnesium alloy are present near fracture surfaceafter tensile tests. Therefore, it can be postulated that the twinningin the heterogeneous microstructure may play a critical role in thisstrain rate sensitivity. Certainly, further observations are thusneeded to demonstrate this argument in later investigation.

4.2. Influence of slip bands and twins on FCP

Since the dominant deformation mechanisms in magnesium arebasal slip, secondary (prismatic and pyramidal) slip, and f1012gtension twinning and f1011g compression twinning [26], it islikely that fatigue crack nucleation and growth is controlled bythe two factors: dislocation slip and twinning.

In our earlier study [27], slip bands, near the manganese-richparticles and across the crack, as sites for micro-crack initiation;these appeared initially in individual grains with the same crystal-lographic system, probably basal planes on a pre-polished surfaceof extruded AM60. It has been also demonstrated that slip bandsappear in preferably oriented grains at room temperature in othermagnesium alloy such as cast AZ91 [3] and wrought magnesiumalloys MA2-1-T 2 and MA12-T 6 [2].

Anderson et al. [26] reported that basal slip occurs first at astress �0.5 MPa, followed by tension twinning and secondary slip,since basal slip cannot provide five independent slip systems.Twinning is thus the other particularly important cause of cracknucleation in coarse-grained magnesium alloys in a wide temper-ature range.

The prevailing crack propagation involves simultaneous oralternating flow along two slip systems. In the present work sec-ondary twining or double twinning also emerged much later thanslip bands, and crossed over the crack or appeared far from the

Fig. 9. Optical micrographs of two types of double twinning: A type-(1 0 1 2) � (0 1 1 2) twinning and B type-{1 0 1 1} � {1 0 1 2} twinning near Al–Mn particles and crack. Deflection and branches were also found on the surface ofthe extruded AM60.

crack on the pre-polished and etched surface of extruded AM60 in-stead of rolled AM60. Fig. 9 demonstrates that, near Al–Mn parti-cles and the fatigue crack for the extruded AM60 alloy, thereoccurred two types of secondary twinning: A� ð1012Þ � ð0112Þdouble twinning and B� f1011g � f1012g double twinning[28]. The matrix twins on either a f1011g or f1012g plane pro-duce a rotation of 38� about the same h1210i axis. The misorien-tation relationship between ð1012Þ and ð0112Þ planes isresponsible for the local maxima in the 50–60� range with a clus-tering of rotation axes near h0112i [19,28]. f1012g twinning oc-curred on two different f1012g planes in a considerable fractionof grains. A type secondary twinning is the predominant deforma-tion mode in tension-zero-tension fatigue crack propagation.

Once dislocations coming across the Al–Mn particles, they mayprogress by means of cross-slip. As can be seen from Figs. 10a and10b, the slip–twinning interaction appeared around the Al–Mnparticle (shown as C) on the way of dislocations slip between twinsA and B. As a result, slip changed direction after the Al–Mn particle.Therefore, it is probably that the crack deviates when the crack tipcomes across the Al–Mn phases.

Fig. 11 discloses a SEM observation on configurations of etcheddeformation twins produced after fatigue of extruded AM60. Theblack arrows show two thin parallel deformation twins. Fine scalesecondary twins are discernible. The white arrows show thestepped deformed twin interfaces due to the impingement and ter-mination of other twins. The dislocation slips and steps mayrespectively result in the striations and roughness in the fracturein Figs. 7a and 7b.

It is not the case for rolled AM60. There was no twinning butbranching and small cracks on pre-polished and etched specimenof rolled AM60 [27]. There may be reasons:

(i) The propensity for twinning decreases with decreasing grainsize. Twinning occurs preferentially in coarse grains due totheir lower CRSS. It is suggested that there exists a criticalgrain size below which deformation twinning at room tem-perature is suppressed [29]. For instance, this critical size inZK60 processed by warm equal channel angular extrusion(ECAE) is approximately 3–4 lm [29]. The grain size of rolledAM60 was smaller than that of extruded AM60. As a result,secondary twinning does not easily occur in rolled AM60.

(ii) Texture may have some impact on twinning. For rolledAM60, unidirectional rolling in the L-direction causes thebasal planes (0 0 0 1) to align parallel to the rolling plane,

Fig. 10a. Cross-slip occurs in twins A and B, due to interaction with Al–Mn particleC on the slip path.

Fig. 10b. The magnitude of Fig. 4a.

Fig. 12a. SEM morphology of extruded AM60 at crack tip at high-rate region(DK = 11.0 MPa m1/2). Micro-voids (indicated with horizontal arrows) occurred attriple grain boundaries in plastic deformation zone at an angle of 45� with crack.

Fig. 12b. Magnification of area A in Fig. 11a, crack tip is formed by coalescence ofmicro-voids (indicated with vertical arrows).

Fig. 11. SEM configurations of etched deformation twins produced after fatigue ofextruded AM60. The black and white arrows indicate deformation twins and thestepped deformed twin interfaces, respectively.

R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419 417

whereas, for extruded AM60, most of the grains were ori-ented with the basal planes parallel to the extrusion direc-tion and normal to the surface of the extrusion [15].

4.3. Influence of grain boundaries on FCP

GBs play a dual role in FCP. On one hand, crack branching, inmost cases, takes place at GBs. For example, Tokaji et al. [30] foundthat fatigue cracks initiated within grains (transgranular) or at GBs(intergranular) for rolled AZ31. Yue et al. [31] demonstrated that,for squeeze cast AZ91, transgranular fatigue cracks frequentlychanged direction at GBs. The amplitude of crack deviation wasmuch greater in coarse grain specimens. Moreover, micro-voidsare found to more readily evolve at the triple GBs. Figs. 12a and12b depicts the crack tip of extruded AM60 at a high DK of11.0 MPa m1/2. Micro-voids occurred at GBs in the plastic deforma-tion zone oriented at 45� to the direction of applied stress (Fig. 12a)because it was difficult for several grains to accommodate at higherDk. In addition, micro-voids, emanating in the interior of the grainat the crack tip (Fig. 11b), indicated that fatigue crack growth wasinduced by coalescence of micro-voids. This phenomenon is alsoobserved, using a transmission electron microscope (TEM), at thecrack front in extruded AZ80 alloy during FCP [12].

4.4. Influence of second-phase particles on FCP

The intermetallics Mg17Al12 and Al–Mn, just as GBs, have a cru-cial influence on the FCP behavior. Crack nucleation at secondaryparticles can lead either to de-cohesion of the particle–matrixinterface or to cracking of the inclusion. Both of these micro-crackshave been observed experimentally. These particles are obviously apotential source of void formation.

The other intermetallic compounds such as the manganese-richparticles in magnesium alloys may play a dual role: (1) promotingcrack nucleation through the formation of micro-voids and micro-cracks and (2) thwarting to some degree the FCP via deflection andbranching of the crack path.

Micro-cracks nucleated at Al–Mn particles on the fatigue cracksurface were discerned on both extruded AM60 (Fig. 13a) androlled alloy. An analogous result obtained by Uematsu et al. [32]proposed that Al–Mn intermetallics are the crack initiation sitesfor AZ80. Al–Mn particles facilitate accordingly crack initiation.

Al–Mn particles cause cyclic deflection and branching of thefatigue crack path depending on the grain size and on the

Fig. 13a. Fractographs revealing a micro-crack emanating at an Al–Mn particle(indicated with a horizontal arrow), which itself do not crack, on the fracturesurface of the extruded AM60 sample at DK of 4.9 MPa m1/2.

Fig. 13b. Fatigue crack deflection and branch (shown with horizontal arrows)caused by Al–Mn particles in the extruded AM60 alloy.

418 R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419

homogeneity of the microstructure (Fig. 13b). Because the Al–Mnparticles are too hard to be cracked, the fatigue crack has to deflector branch, even in cases where the fatigue crack runs across a clus-ter of these hard particles. Deflection, resulting from an Al–Mn pre-cipitate, at crack tip on the fatigue crack path is also apparently inFig. 12a. That means that, the particles may slow to some extentthe crack growth rate due to their small size. However, their influ-ence on FCP rate is very limited compared with microstructure.

Fig. 14. Crack growth profile, of extruded AM60, showing grain boundarie

4.5. Effect of crack closure on FCP

Crack closure, which develops compressive stresses, is not asstrong in rolled AM60 as in extruded AM60. f1012g twinning iscaused by tension and f1011g twinning results from compression.f1012g twinning could occur in both c-axis extension and com-pression [28]. As a matter of fact, no compressive stress was ex-erted in the fatigue machine, provided that the tests wereconducted under zero-tension loading. What gives rise tof1011g twinning? The answer is probably that crack closure pro-duces the compressive stress.

Surface roughness and oxidation of the fatigue crack are knownto produce crack closure [33]. Regardless of the closure mecha-nism, the effect is to reduce the strain amplitude at the crack tipso as to reduce the crack growth rate compared with a closure-freecrack. Large crack deflection and crack branching of extrudedAM60 happened (Fig. 14). At lower DK and a R ratio of zero, themaximum height of the fracture surface asperities reached88 lm, which is comparable to the average grain size of 56 lm.As can be seen in Fig. 14 that the fatigue crack advanced from leftto right and at a point ‘‘A” stopped propagating on the surfacewhile it continued inside the specimen. It grew around some mate-rial heterogeneity, e.g., grain boundaries r and s, and it reap-peared at B, C, . . ., G and continued propagation on the surfacebut on a shifted plane. The two fracture surfaces are locally con-nected (bridged) and their relative displacement is restrained,which has an obvious shielding on the crack tip. This similar phe-nomenon was also observed in coarse-grained magnesium [34].The FCP rate of extruded AM60 is much lower than that of rolledAM60 in the first stage, because of a roughness induced crack clo-sure (RICC) contribution to the decrease in crack growth of ex-truded AM60. The da/dN � DK curves (Fig. 6) show plateau,retardation, and subsequent acceleration in crack growth withincreasing cyclic stress intensity factor range. The explanation liesin following aspects: (1) crack deflection and crack branching atgrain boundaries and/or precipitates such as Al–Mn particles. Thecrack deflection from its normal growth plane reduces the effectivestress intensity factor at the crack tip, which decreases FCP rate. (2)The plastic deformation zone is smaller than the coarser grain size.(3) For the oscillatory crack growth phenomena of rolled AM60,small fatigue crack initiation (Figs. 9a and 11) consumes a part ofthe driving force. (4) Reduced fatigue crack growth may be partlyattributed to crack closure.

It should be mentioned that fatigue crack propagation of mag-nesium alloys is subject to the effect of environment, probablythe presence of moisture or humidity in laboratory air. It has beendemonstrated that laboratory air is a corrosive medium for magne-sium alloys and corrosion fatigue appears above a relative humid-ity of 80% [35,36]. However, in this study, the relative humidity inambient air is less than 80%; the influence of environment on FCPshould be insignificant.

s such as r and s, and branching such as A, B and C along the crack.

R. Zeng et al. / International Journal of Fatigue 32 (2010) 411–419 419

5. Conclusions

An investigation into the microstructure, mechanical propertiesand FCP behavior of extruded AM60 has been performed in compar-ison with rolled AM60. The followings are the salient observations:

(1) Extrusion of AM60 leads to a inhomogeneous grain structureconsisting of large elongated and small-recrystallised grainfractions. Macro-segregation of second-phase particles/con-stituent particles results in denuded zones that are depletedof particles at GBs.

(2) The mechanical properties of extruded AM60 illustrates thatextruded material is relatively sensitive to the applied strainrate, compared with cast AM60. The strain-hardening expo-nent, n, of extruded AM60 declines with increasing strainrate from 5.3 � 10�3 s�1 to 0.0147 s�1, whereas, the n valueincreases in the strain rate range of 0.0147 s�1 and0.147 s�1. Slip and twinning in the heterogeneous micro-structure are predominantly responsible for these findings.

(3) Two types of in situ double twinning: A-ð1012Þ � ð0112Þand B-f1011g � f1012g occurred during fatigue propaga-tion of the extruded AM60 alloy due to the localized coarsergrains in the microstructure. The FCP rate of rolled AM60 isfaster than that of extruded one due to its finer grains. Fati-gue crack initiation and advance is related to the synergisticinfluence of slip bands, double twinning, intermetallic com-pounds and grain boundaries for extruded AM60 and to oxi-des for rolled AM60. There is a difference between thefatigue crack surfaces: extruded AM60 has a rough fatiguecrack surface caused by cleavage and shear voids, whereasrolled AM60 develops a flat fatigue crack surface. Smallcracks resulted from oxide and/or intermetallic compoundson the fatigue crack surface of rolled AM60 may be respon-sible for oscillatory crack growth and crack arrest.

Acknowledgements

This work was supported by the National Hi-Tech Research andDevelopment Program under Grant No. 2001AA331050 and Project(CSTC, 2007AC4073, 2008BB0063 and 2009AB4008) for the finan-cial support. Thanks go to Guangling Mg Industry Science andTechnology Co., Ltd., for providing the tested materials.

References

[1] Zeng RC, Dietzel W, Zettler R, Chen J, Kainer KU. Microstructure evolution andtensile properties of friction-stir-welded AM50 magnesium alloy. TransNonferrous Met Soc China 2008;18:s76–80.

[2] Ogarevic VV, Stephens RI. Fatigue of magnesium alloys. Annu Rev Mater Sci1990;20:141–77.

[3] Potzies C, Kainer KU. Fatigue of magnesium alloys. Adv Eng Mater2004;6(5):281–9.

[4] Zeng RC, Han EH, Ke W. Fatigue and corrosion fatigue of magnesium alloys.Mater Sci Forum 2005;488–489:21–4.

[5] Lee SG, Patel GR, Gokhale AM. Inverse surface macro-segregation in high-pressure die-cast AM60 magnesium alloy and its effects on fatigue behavior.Scripta Mater 2005;52:1063–8.

[6] Xu DK, Liu L, Xu YB, Han EH. The fatigues crack propagation behavior of theforged Mg–Zn–Y–Zr alloy. J Alloy Compd 2007;431(1–2):107–11.

[7] Eisenmeier G, Holzwarth B, Hoeppel HW, Mughrabi H. Cyclic deformation andfatigue behavior of the magnesium alloy AZ91. Mater Sci Eng A 2001;319–321:578–82.

[8] Mayer H, Papakyriacou M, Zettl B, Stanzl-Tschegg SE. High cycle fatiguemechanisms in a cast AM60B magnesium alloy. Int J Fatigue 2003;25:245–56.

[9] Shih TS, Liu WS, Chen YJ. Fatigue of as-extruded AZ61A magnesium alloy.Mater Sci Eng A 2002;325:152–62.

[10] Zeng RC, Ke W, Han EH. Influence of load frequency and ageing heat treatmenton fatigue crack propagation rate of as-extruded AZ61 alloy. Int J Fatigue2009;31(3):463–7.

[11] Atrens A, Hoffelner W, Duerig TW, Allison JE. Subsurface crack initiation inhigh cycle fatigue in Ti6Al4V and in a typical martensitic stainless steel. ScriptaMater 1983;17:601–6.

[12] Zeng RC, Xu YB, Han EH, Ke W. Fatigue crack propagation behavior of as-extruded magnesium alloy AZ80. Mater Sci Eng A 2009;509:1–7.

[13] Nan ZY, Ishihara S, Goshima T. Corrosion fatigue behavior of extrudedmagnesium alloy AZ31 in sodium chloride solution. Int J Fatigue2008;30:1181–8.

[14] Horstemeyer MF, Yang N, Gall K, McDowell D L, Fan J, Gullett PM. High cyclefatigue of a die cast AZ91E-T 4 magnesium alloy. Acta Mater2004;52:1327–36.

[15] Mukai T, Watanabe H, Ishikawa K, Higashi K. Guide for enhancement of roomtemperature ductility in Mg alloys at high strain rates. Mater Sci Forum2003;419–422:171–6.

[16] Chino Y, Furuta T, Hakamada M, Mabuchi M. Fatigue behavior of AZ31magnesium alloy produced by solid-state recycling. J Mater Sci2006;41:3229–32.

[17] Kaya AA, Uzan P, Eliezer D, Aghion E. Electron microscopic investigation of ascast Z91D alloy. Mater Sci Technol 2000;16:1001–6.

[18] Wendt J, Hilpert M, Kiese J, Wagner L. Surface and environmental effects on thefatigue behavior of wrought and cast magnesium. In: Hryn J, editor.Magnesium Technology. The Minerals, Metals & Materials Society; 2001. p.81–5.

[19] Jiang L, Jonas JJ, Luo AA, Sachdev AK, Godet S. Twinning-induced softening inpolycrystalline AM30 Mg alloy at moderate temperatures. Scripta Mater2006;54(5):771–5.

[20] Lin XZ, Chen DL. Strain hardening and strain-rate sensitivity of an extrudedmagnesium alloy. JMEPEG 2008;17:894–901.

[21] Takuda H, Kikuchi S, Yoshida N, Okahara H. Tensile properties and pressformability of a Mg–9 Li–1 Y alloy sheet. Mater Trans 2003;44(11):2266–70.

[22] Barnett MR, Nave MD, Bettles CJ. Deformation microstructures and textures ofsome cold rolled Mg alloys. Mater Sci Eng A 2004;386:205–11.

[23] Zhu SQ, Yan HG, Xia WJ, Liu JZ, Jiang JF. Influence of different deformationprocessing on the AZ31 magnesium alloy sheets. J Mater Sci 2009.doi:10.1007/s10853-009-3513-9.

[24] Abbott T, Easton M, Song W. Mechanical behavior of cast magnesium alloys.Mater Sci Forum 2003;419–422:93–102.

[25] Liu WC, Dong J, Zhang P, Yao ZY, Zhai CQ, Ding WJ. High cycle fatigue behaviorof as-extruded ZK60 magnesium alloy. J Mater Sci 2009;44:2916–24.

[26] Anderson P, Caceres CH, Koikke J. Hall–Petch parameters for tension andcompression in cast Mg. Mater Sci Forum 2003;419–422:123–8.

[27] Zeng RC, Han EH, Liu L, Xu YB, Ke W. Effect of rolled microstructure on fatigueproperties of magnesium alloy AM60. Chinese J Mater Res 2003;17(3):241–5.

[28] Nave MD, Barnett MR. Microstructure and texture of pure magnesiumdeformed in plane-strain compression. Scripta Mater 2004;51:881–5.

[29] Lapovok R, Thomson PF, Cottam R. The effect of grain refinement by warmequal channel angular extrusion on room temperature twinning in magnesiumalloy ZK60. J Mater Sci 2005;40:699–708.

[30] Tokaji K, Kamakura M, Ishiizumi Y, Hasegawa N. Fatigue behavior and fracturemechanism of a rolled AZ31 magnesium alloy. Int J Fatigue 2004;26:1217–24.

[31] Yue TM, Ha H, Musson NJ. Cyclic deformation and fatigue behavior of themagnesium alloy AZ91. J Mater Sci 1995;30:2277–83.

[32] Uematsu Y, Tokaji K, Kamakura M, Uchida K, Shibata H, Bekku N. Effect ofextrusion conditions on grain refinement and fatigue behavior in magnesiumalloys. Mater Sci Eng A 2006;434:131–40.

[33] Suresh S. Fatigue of materials. Cambridge: Cambridge University Press; 1998.p. 609–10.

[34] Nicoletto G, Konecna R, Pirondi A. Fatigue crack paths in coarse-grainedmagnesium. Fatigue Fract Eng Mater Struct 2005;28:237–44.

[35] Tokaji K, Nakajima M, Uematsu Y. Fatigue crack propagation and fracturemechanisms of wrought magnesium alloys in different environments. Int JFatigue 2009;31:1137–43.

[36] Sajuri ZB, Miyashita Y, Mutoh Y. Fatigue characteristics of an extruded AZ61magnesium alloy. J Jpn Inst Light Metals 2002;52(2):161–6.