1 Introduction

Weight reduction is considered a primary design metric for increasing fuel efficiency and reducing automotive and aircraft component costs.[1] High-strength and lightweight aluminum alloys, exemplified by the 6000-series grades, have been used successfully as structural automotive components. However, additional significant weight and process cost reductions could be achieved if magnesium alloys become the foremost replacement for heavier materials.[2]

Contemporary Mg alloys possess the highest strength-to-weight ratio among all structural metals, with a density of 1.74 Mg/m3. One ongoing major barrier to the industrialization of Mg alloys in automotive applications is their relatively low resistance to fatigue and corrosion service conditions.[3–6] These low fatigue and corrosion resistances of Mg alloys are mainly due to the material’s physical properties. The Mg alloys are remarkably reactive and exceptionally susceptible to forming microstructural singularities during shaping processes.[3] To overcome the detrimental effects of these microstructural singularities, current material selection, manufacturing, and design optimization methods resort to overdesigning. The resulting increased safety factors counteract the original attractive attributes of Mg alloys that have a high stiffness-to-weight ratio.

To fully achieve the broad benefit of these magnesium alloys, the mechanisms of fatigue and corrosion must be rigorously linked to microstructural properties. Corrosion resistance has been the primary concern in the recent research work topics.[4–6] However, too little effort has been deployed in identifying microstructural explanations for fatigue crack growth mechanisms. Moreover, the results reported by various authors have sometimes been controversial and only treated Mg-based alloys separately.[7–17] A study of the fatigue crack growth mechanisms of the AM50 Mg alloy was recently reported by the present authors.[17]

In this article, we quantify the microstructural defects to fatigue crack incubation, microstructurally small cracks (MSC), physically small cracks (PSC), and long cracks (LC) for four commercially dominant high-pressure die-cast (HPDC) magnesium alloys: AM50, AM60, AZ91, and AE44. A full description of the incubation, MSC, PSC, and LC regimes was reported in Reference 18. We performed scanning electron microscope (SEM) fractographic analyses on specimens tested in a fully-reversed strain control condition at a single strain amplitude. The microstructural defects relevant to fatigue crack growth are identified and analyzed using backscatter electron mode and electron probe microanalyses (EPMA).

2 Experimental

For each alloy, plates with dimensions of 100 × 150 × 3 mm were cast in a permanent mold by a high-pressure die-casting technique. Flat dog-bone-shaped samples having a 25.4-mm gage length, a 10-mm gage width, and a 3-mm thickness were extracted from the center of the plate for fatigue testing. This implies that the two large faces of the rectangular specimens exhibited the original skin microstructure at the free surface. Only the sections through the thickness (the 3-mm-thick edges) were subjected to the milling process. Actually, we believe that the machining had a minor effect on crack incubation, because in all specimens, the cracks started at shrinkage or cast pores emerging at the free surface. The skin surface was not subjected to any post-surface treatment. As such, we did not perform X-ray diffraction (XRD) analyses. The skin microstructure generated by the high die pressure had a noticeable effect on the mechanical properties of the material.[19]

Constant amplitude fatigue tests were performed with an MTS servohydraulic machine (MTS Systems Corporation, Eden Prairie, MN) in strain control under a condition of R = −1 (fully reversed), at a frequency of 0.5 Hz, and at multiple strain amplitudes. The tests were conducted at room temperature in laboratory air. Fractographic analyses were performed using a JEOLFootnote 1 field-emission gun (FEG)-SEM.

For the AZ91 alloy, we also performed fatigue tests on round samples, but only at a strain amplitude of 0.3 pct. These samples were cast directly into a round shape, and no post machining or sample preparation was performed after casting.

The microstructure was examined by means of an FEG-SEM on polished samples using either secondary electron imaging (SEI) or backscatter electron imaging (BSEI). The specimens were polished to a mirror-surface finish using standard metallographic equipment. Water contact was avoided, to prevent the formation of surface hydroxides; thus, grinding and polishing were conducted using a mixture of 1 part glycerine and 3 parts ethanol. The grain size of each alloy was measured by the conventional linear intercept method, using optical microscopy (OM). In this approach, a test pattern of lines of known length was randomly laid over a plane section, and the individual intercept lengths across several grain boundaries were measured. An average grain size was calculated by dividing the length of a segment by the total number of intersections occurring within the grain boundaries.

The EPMA were also performed to reveal the nature of the phases in the matrix detected at high magnification with the BSEI technique. The quantitative EPMA chemical composition results of all four alloys are reported in Table I, in comparison with those provided by the industrial supplier General Motors Worldwide (GMW). The element contents were averaged over ten analyzed windows with a surface area of 50 × 50 μm.

Table I Quantitative Chemical Compositions (in Weight) for All Four Alloys Obtained by Means of EPMA in Comparison with Those Provided by General Motors (Material Supplier)
Full size table

3 Microstructure of alloys

In order for the fractographic analyses to be relevant, we characterized the microstructure of the four alloys. The FEG-SEM micrographs and X-ray mapping analyses using EPMA are shown in Figures 1 through 8. Several observations can be made from these figures.

Fig. 1
figure 1

Backscatter FEG-SEM micrographs of a polished AM50 sample cut from the grip section of a fractured fatigue specimen showing (a) an overall view of the microstructure depicting shrinkage pore clusters and an Al-rich eutectic layer fully occupying the dendrite cell boundaries and large presolidified grains and (b) a higher-magnification image revealing the formation of the beta phase and Mn-rich intermetallics within the Al-rich eutectic layer

Full size image
Fig. 2
figure 2

EPMA X-ray maps of a mirror-finish AM50 sample surface, shown in (a) backscattered mode, revealing (b) a Mg-concentration contour, (c) an Al-concentration contour delineating the Al-rich eutectic layer, beta phase, and Mn-rich inclusions, (d) a Mn-concentration contour showing the size and distribution of the Mn-rich inclusions, (e) a Si-concentration contour revealing the association of this element with the Al-rich eutectic region, and (f) a Zn-concentration contour showing no localization of this element in the AM50 alloy

Full size image
Fig. 3
figure 3

FEG-SEM micrographs of a polished AM60 sample cut from the grip section of a fractured fatigue specimen showing (a) an overall view of the microstructure depicting the Al-rich eutectic layer fully occupying the dendrite cell boundaries, with black and white arrows indicating beta-phase and Mn-rich particles, respectively, and (b) a higher-magnification image revealing more abundant formation of the beta phase and Mn-rich phase inside the Al-rich eutectic layer than in the AM50 alloy, with the Mn-rich particles correlating with the beta phase precipitated in the form of strings

Full size image
Fig. 4
figure 4

EPMA X-ray maps of a mirror-finish AM60 sample surface, displayed in (a) backscattered mode, revealing (b) a Mg-concentration contour, (c) an Al-concentration contour delineating the Al-rich eutectic layer, beta phase, and Mn-rich inclusions, with the Al-rich eutectic layer being thicker and more consistent and containing more and larger beta-phase and Mn-rich particles, (d) a Mn-concentration contour showing the size and distribution of the Mn-rich inclusions, (e) a Si-concentration contour, and (f) a Zn-concentration contour

Full size image
Fig. 5
figure 5

FEG-SEM micrographs of a polished AZ91 sample cut from the grip section of a fractured fatigue specimen showing (a) the Al-rich eutectic layer taking a completely divorced form, with a large and continuous presence of the beta phase and (b) large casting pores reaching several millimeters in size

Full size image
Fig. 6
figure 6

EPMA X-ray maps of a mirror-finish AZ91 sample surface shown in (a) backscattered mode, revealing (b) a Mg-concentration contour; (c) an Al-concentration contour revealing that the Al-eutectic rich phase takes a completely divorced form in which massive particles of the beta-phase precipitate, (d) a Zn-concentration contour revealing that zinc correlates with the beta phase, suggesting that the massive particles could correspond to an (Al,Zn) x Mg y phase, (e) a Mn-concentration contour, and (f) a Si-concentration contour

Full size image
Fig. 7
figure 7

Backscatter FEG-SEM micrographs of a polished AE44 sample showing (a) an overall view of the microstructure depicting shrinkage pores and localization of the Al11RE3 phase within the interdendritic eutectic region and (b) a higher-magnification image illustrating the platelet-like shape of the Al11RE3

Full size image
Fig. 8
figure 8

EPMA X-ray maps of a mirror-finish AE44 sample surface, shown in (a) backscattered mode, showing (b) an Mg-concentration contour, (c) an Al-concentration contour, (d) a Ce- concentration contour revealing the nature of the Al/Ce-rich platelets, (e) a Mn-concentration contour revealing no presence of the Mn-rich particles consistently found in the AM50, AM60, and AZ91 alloys, and (f) a Si-concentration contour showing no presence of this element in the AE44 alloy

Full size image

For the AM50, AM60, and AZ91 alloys, Al enrichment took place to different extents in the α-Mg dendrite cells (Figures 1 through 6). These Al enrichment gradients comprised a bulky Al-poor dendrite core, followed by a eutectic Al-rich solid solution band at the dendrite cell boundary. Within this Al-rich eutectic band, β-Mg17Al12 phase precipitated in lamellar morphology. This phase was shown to have a cubic crystal structure.[20] The presence of the β compound increased as the Al content increased in the alloys. Thicker strips of both β-phase and Al-rich eutectic solid solution bands were observed in the AM60 alloy than in the AM50 alloy (Figures 1 through 3). For the AZ91 alloy, the Al-rich eutectic took a completely divorced form, with massive and continuous precipitation of the β phase.[21]

For the AM50, AM60, and AZ91 alloys, round particles rich in Al and Mn were observed in the Al-rich eutectic phase. These particles were correlated with the presence of β-phase particles and reached over 10 μm in size. The Mn-rich inclusions were demonstrated to present in the form of a Mn-poor round phase overlaid by smaller hexagonal Al8Mn5 particles.[17] The hexagonal Al8Mn5 particles were revealed to have a hexagonal crystallographic structure.[20]

For the AE44 alloy (Figures 7 and 8), extensive concurrent enrichment of Al and mischmetal (RE) occurred as fine platelets decorating the interdendritic eutectic. This enrichment may correspond to the Al11RE3 phase reported in Reference 20. In marked contrast to the other alloys, the β-Mg17Al12 and Al8Mn5 were very scarce in the AE44. These noticeably lower contents of the β-Mg17Al12 and Al8Mn5 compounds can be explained by the higher affinity of Al to the RE elements that have significantly lower solubility in Mg, compared to those of Al and Mn.

For the AM50 and AM60 alloys, the dendrite cells were relatively small and had an average linear intercept size of 9 μm. However, large presolidified grains with an average size of 50 μm were also observed.[22] For the AZ91 alloy and the AE44 alloys, the average linear intercept dendrite cell size dropped below 5 μm.

4 Fatigue life

The strain amplitudes vs cycles to failure are presented for all alloys in Figure 9(a). Figure 9(b) provides the average cycle to failure for each strain amplitude, in an effort to statistically illustrate the difference in fatigue lives among the four alloys. The main point from the fatigue life data is that the AM50 and AM60 alloys showed a better fatigue resistance than did the two other alloys. The lowest fatigue lives were consistently recorded for the AE44 alloy. The AE44 alloy was designed to present the best creep resistance, owing to the lamellar form of the Al11RE4 phase that forms in the interdendritic eutectic.[21] The fatigue life variability increased when the strain amplitude decreased, and this increase in variability intensified for the high-cycle fatigue regime.

Fig. 9
figure 9

Fatigue test results of AM50, AM60, AZ91, and AE44 magnesium alloys tested using rectangular specimens at a frequency of 0.5 Hz with an R ratio of −1 showing (a) strain-life data for each specimen and (b) data on average cycles to failure illustrating that AM50 consistently exhibited a better fatigue resistance followed immediately by AM60, and that the AE44 showed the lowest fatigue life followed immediately by AZ91

Full size image

In an effort to observe the relevant fatigue crack growth mechanisms associated with each regime, we selected the 0.3-pct strain amplitude condition for our analyses. The representative samples extracted for that purpose are shown in Figure 10 and are designated as sample Z1 (N f  = 2736) and Z2 (N f  = 5685) for two AZ91 alloy fatigue lives, E1 (N f  = 6034) for one AE44 alloy fatigue life, N1 (N f  = 11649) for one AM60 alloy fatigue life, and M1 (N f  = 1371) and M2 (N f  = 4520) for two AM50 alloy fatigue lives.

Fig. 10
figure 10

Strain-life data of AM50, AM60, AZ91, and AE44 magnesium alloys tested using rectangular samples at a 0.3-pct strain amplitude and a frequency of 0.5 Hz with an R ratio of −1; the arrows indicate the samples subjected to fractography analyses

Full size image

5 Fractography and discussion

A detailed fractographic analysis of fatigue crack growth mechanisms in the AM50 alloy was reported in a different article.[17] We briefly restate the relevant results previously stated for this alloy. These results served as a comparison for understanding the mechanisms controlling the fatigue crack growth in the three other Mg alloys.

5.1 Fatigue Crack Incubation

For the AM50 alloy, the fracture surfaces close to the initiation region in samples M1 and M2 are presented in the micrographs of Figures 11 and 12, respectively. For both samples, shrinkage pore clusters near the free surface were observed to nucleate the dominant fatigue crack linked to the final failure. These shrinkage pore clusters exhibited the same size and morphology in both samples. However, the overall aspect of the dendrite cells seemed to be smoother in sample M1 than in sample M2. This is due to a greater connectivity of the interdendritic shrinkage pores in the shrinkage pore cluster of sample M1 (Figures 11 and 12). This difference accounts for the greater fatigue life of sample M2 over sample M1, since smaller cracks will encounter fewer barriers during their propagation.

Fig. 11
figure 11

FEG-SEM fracture surfaces of sample M1 (N f  = 4520) showing (a) an overall view of the fracture surface delineating the fatigue crack propagation and (b) a higher magnification of Region A, which corresponds to the crack propagation in a shrinkage pore cluster

Full size image
Fig. 12
figure 12

FEG-SEM fracture surfaces of sample M2 (N f  = 1379) showing (a) an overall view of the fracture surface delineating the fatigue crack propagation and (b) a higher magnification of Region A, which corresponds to the crack propagation in a shrinkage pore cluster

Full size image

For the AM60 alloy, (sample N1 with N f  = 11649), the crack incubated at a shrinkage pore cluster located near the free surface (Figure 13). A similar nucleation phenomenon occurred in sample M3 (N f  = 14171) of the AM50 alloy reported in Reference 17. However, the pore cluster was two orders of magnitude smaller in sample M3[17] than in sample N1. This difference explains the greater fatigue life for sample M3. The fatigue crack, however, might have propagated more slowly in sample N1 than in sample M3, yet not slowly enough to overshadow the difference caused by the size of the initial fatigue crack.

Fig. 13
figure 13

FEG-SEM fracture surfaces of sample N1 (N f  = 11649) cut from an AM60 alloy plate showing (a) an overall view of the fracture surface and (b) a higher-magnification view of the region, including the surface shrinkage pore that nucleated the main fatigue crack

Full size image

For the AZ91 alloy, the fracture surface of specimen Z1 (N f  = 5685) shows that the main fatigue crack nucleated at an agglomeration of large casting pores (Figure 14). These large casting pores were partly contained within the original skin layer of the cast plate and extended into the main bulk microstructure. As typically shown in Figure 5, the AZ91 plates typically exhibited a higher content of large casting pores as compared to the other alloys. The authors recently demonstrated through X-ray computed tomography (XRCT) that AM50 and AM60 exhibited a lower volume of large casting and shrinkage pores than did AZ91 and AE44 cast under the same conditions.[23] This explains the very low fatigue durability of the AZ91 alloy compared to the AM50 and AM60 alloys. The large number of casting pores in specimen Z1 induced a rough crack propagation path through the coalescence of fatigue cracks that nucleated at neighboring casting and shrinkage pores (Figure 14). Figure 15 shows the fracture surface of specimen Z2, which had the lowest fatigue life among all the tested AZ91 samples. Here again, the main fatigue crack nucleated at a casting pore that was located within the original skin layer of the cast plate. This pore, however, was significantly more elongated in the crack propagation direction.

Fig. 14
figure 14

FEG-SEM fracture surfaces of sample Z2 (N f  = 5685) cut from an AZ91 alloy plate showing (a) an overall view of the fracture surface and (b) a higher magnification of the region, including the casting pore that nucleated the main fatigue crack. Note that the casting pore extends along the sample free surface

Full size image
Fig. 15
figure 15

FEG-SEM fracture surfaces of sample Z2 (N f  = 2736) cut from an AZ91 alloy plate showing (a) an overall view of the fracture surface and (b) a higher magnification of the region, including the casting pore that incubated the main fatigue crack. Note that the casting pore extends perpendicular to the sample free surface

Full size image

For the AE44 alloy, we investigated the E1 specimen, which had the highest fatigue life among all the tested samples. Figure 16 revealed that the fatigue crack nucleated at a large surface shrinkage pore with a pancake-like shape that extended along the sample free surface. This shrinkage pore is within the original skin microstructure of the cast plate. The length and width of this pore were approximately 1.3 and 0.14 mm, respectively. As can be observed by comparing Figures 1 and 7, the size of the interdendritic shrinkage pores was inversely correlated with the dendrite cell average size. Hence, the beneficial effect of the fine microstructure on fatigue resistance was counterbalanced by the increased propensity to interdendritic shrinkage pore coalescence. The fatigue life of specimen E1 was below the lowest fatigue life limit of the AM50 and AM60 alloy specimens.

Fig. 16
figure 16

FEG-SEM fracture surfaces of sample E1 (N f  = 6085) cut from an AZ91 alloy plate showing (a) an overall view of the fracture surface and (b) a higher magnification of the region, including the large surface shrinkage pore that nucleated the main fatigue crack. Note that the pore extends along the sample free surface

Full size image

5.2 Fatigue Crack Propagation

5.2.1 The AM50 alloy

In Reference 17, we observed that, at very low crack tip driving forces, the MSC showed preferential propagation at the Al-rich eutectic phase/α-Mg dendrite cell interface (Figure 17(a)). The MSC were susceptible to following important meandering paths, depending on the size and morphology of the dendrite cell ahead of the crack tip. At an intermediate stage of fatigue life corresponding to approximately to 100 to 150 μm of crack propagation, the driving force of the PSC seemed to be sufficient to deflect the crack into the Al-rich eutectic. In this eutectic, the PSC advanced by a debonding mechanism between fine persistent slip bands formed by a dislocation pile-up phenomenon (Figure 17(b)).[20] The effect of the β-phase particles at this important stage of the fatigue crack propagation can be both beneficial and detrimental. The β phase could be detrimental because this phase is brittle and susceptible to fracture ahead of the crack tip and, as such, provides weak resistance and easy possible paths for crack propagation. A potential beneficial effect is that the fractured particles attract and force the PSC to remain in the eutectic layer, which appears to be more resistant to fatigue crack propagation.[17] The PSC would encounter more intact particles within the eutectic. As a result, the PSC would be forced to meander around these particles, thus reducing the crack tip opening displacement. In the long crack fatigue regime, the effect of the associated widespread plasticity induced a mixed interdendritic-transdendritic crack propagation mode. The interdendritic propagation was predominantly due to the fragility of the β phase. The transdendritic propagation involved several tens of dendrite cells, which resulted in a sawtooth pattern (Figure 17(c)). The tendency for the transdendritic propagation was greater near shrinkage pores.

Fig. 17
figure 17

FEG-SEM fracture surfaces of sample M4 (N f  = 7269) reported in Reference 17 and sample M1 (N f  = 4520) showing (a) a dendrite cell close to the initiation site covered with fine striations revealing microstructurally small crack propagation between two cohesive dendrites along the Al-rich eutectic layer/dendrite cell interface, (b) physically small crack propagation through the Al-rich eutectic layers leaving the entire dendrite volume, but with a flat surface, and (c) a high-magnification view of a shear band pattern crossing several tens of dendrite cells, as evidenced by the high density of individual interdendritic shrinkage pores

Full size image

5.2.2 The AM60 alloy

For the AM60 alloy, the small crack propagation occurred in a way that was similar to that observed for the AM50 alloy. At the edge of the shrinkage pore cluster, where the fatigue crack incubated, smooth dendrite cells were observed to be covered with very fine striations, as illustrated in Figure 18. The forefront of the main fatigue crack was very rough, illustrating a meandering through the dendrite cell, following the Al-eutectic rich layer/dendrite cell interface. Some dendrite cells seemed, however, to be more flat, due to very fine persistent slip bands emerging at the free surface. This phenomenon is a characteristic of the intermediate stage of propagation observed for the AM50 alloy. In this stage, the corresponding PSC cut through the Al-rich eutectic layer. For the AM60 alloy, the early stage of the crack propagation seemed to be characterized by a lesser tendency toward the interfacial small crack advance, as compared to that of the AM50 alloy. This is probably due to a more pronounced presence of the brittle phase that fractured at smaller driving forces and that, as a result, inclined the crack to deflect rapidly into the eutectic layer. Within this eutectic, the higher concentration of the β-phase particles compared to the AM50 alloy implies that more particles would fracture in the AM60 alloy. As a result, for the AM60 alloy, the effect of the β phase is more detrimental than it is beneficial, as compared to the effect for the AM50 alloy.

Fig. 18
figure 18

FEG-SEM fatigue fracture surfaces of sample N1 (N f  = 11649) cut from an AM60 alloy plate and showing (a) an overall view of the region surrounding the shrinkage pore cluster nucleating the main fatigue crack and (b) a higher magnification of the region delineated in (a) depicting round and smooth dendrite cells covered with fine striations characteristic of the microstructurally small crack propagation at the Al-rich eutectic layer/dendrite cell interface

Full size image

After approximately 100 μm of propagation (vs 200 to 300 μm for sample M3[17]) away from casting and shrinkage pores, the fatigue crack exhibited a relatively flat and smooth surface covered by coarse persistent slip bands, twins, and cleavage-like bands, as shown in Figure 19. These crack propagation mechanisms corresponded to the long fatigue crack growth regime. In the AM50 alloy, this stage was characterized by a mixed crack propagation mode with a preponderance of the interdendritic mode. This interdendritic long crack propagation mode left the surface mostly covered by fractured particles in the Al-rich eutectic. In the AM60 alloy, however, despite the higher content of the brittle β-phase particles, the LC showed significantly less tendency for interdendritic propagation, and the fracture surface remained more flat and homogeneous.

Fig. 19
figure 19

FEG-SEM fracture surfaces of sample N1 (N f  = 11649) showing (a) transdendritic shear nodules in the long crack fatigue regime spreading over several tens of dendrites and (b) a cleavage-type fracture in the long crack fatigue regime spreading over several dendrite cells

Full size image

5.2.3 The AZ91 alloy

Figures 14 and 15 revealed that the large incubated cracks in the AZ91 alloy induced a noticeable acceleration of the onset of the long fatigue crack propagation regime. Also, the presence of coarse interdendritic β-phase nodules at the expense of the Al-rich eutectic layer seemed to increase stress concentrations at the dendrite cell boundaries. This is evinced by the localization of persistent slip bands at the vicinity of β-phase strings (Figure 20). Moreover, an important number of large casting and shrinkage pores existed at the fracture surface (Figures 14 and 15).

Fig. 20
figure 20

A FEG-SEM fracture surface of sample Z2 (N f  = 5685) cut from an AZ91 alloy plate showing the localization of persistent slip bands at the vicinity of an interdendritic beta-phase string

Full size image

The long fatigue crack propagation regime in the AZ91 alloy was mainly characterized by a mixed transdendritic-interdendritic mode. Figure 21 shows that the interdendritic propagation was substantiated by the presence of smooth β-phase particles at the fracture surface. For the same specimen, Figure 22(a) shows many “deep” cracks in the direction of the main fatigue crack propagation. These deep cracks are in-plane cracks that emerged at the fracture surface, which is an atypical phenomenon. They were bordered with fine striation-containing nodules revealing the creation of subsidiary small cracks that propagated over a shorter range (Figure 22(b)). The deep cracks were, in all likelihood, the result of damage induced by coalescence mechanisms between relatively large secondary cracks evolved from casting and shrinkage pores. In marked contrast to the AM50 and AM60 alloys, we did not observe in the AZ91 alloy large transdendritic persistent slip band nodules around casting and shrinkage pores extending over several tens of dendrite cells. Rather, interdendritic propagation exposing smooth β-phase strings accompanied by the presence of fine slip bands and striations occurred in the surrounding areas of bulk shrinkage and casting pores (Figure 21). This interdendritic propagation revealed that the enhanced porosity and β-phase precipitates in the AZ91 alloy promoted damage mechanisms that are characteristic of the low-cycle fatigue regime. The absence of the large shearing nodules in the AZ91 alloy may be due to more than the pronounced presence of casting pores.

Fig. 21
figure 21

FEG-SEM fracture surfaces of sample Z1 (N f  = 2736) cut from an AZ91 alloy plate and showing (a) smooth beta-phase precipitates revealing interdendritic and microstructurally small-crack-propagation-related mechanisms at the vicinity of a shrinkage pore cluster encountered 600 μm away from the crack nucleation site and (b) a higher magnification of the region highlighted in (a) depicting fine striations around an elongated beta-phase precipitate

Full size image
Fig. 22
figure 22

FEG-SEM fracture surfaces of sample Z1 (N f  = 2736) cut from an AZ91 alloy plate showing (a) an overall view of the region surrounding the casting pore nucleating the main fatigue crack and containing a deep crack and (b) a higher magnification of the region delineated in (a) depicting striations of a small crack that most likely incubated from the transversal in-plane deep crack

Full size image

We actually performed fatigue tests on an AZ91 alloy that was directly cast into the form of round fatigue test specimens and that exhibited lesser porosity as compared to that observed in the AZ91 based plates. The fractography analyses on these specimens showed no presence of extra large shearing nodules analogous to those encountered in the AM50 and AM60 alloys (Figure 23). This finding suggests that the interdendritic β-phase precipitates act as barriers for dislocation pile-up transmission between adjacent dendrite cells.[20] Note that, for these specimens, the main fatigue crack initiated on inclusion clusters formed by surface macrosegregation phenomena, as illustrated in Figure 23.

Fig. 23
figure 23

FEG-SEM fracture surfaces of a round and cast AZ91 sample N1 showing (a) an overall view of the fatigue crack propagation surface illustrating surface segregating particles (arrows) that originated the main crack and (b) a higher magnification of the region delineated in (a) representing a typical aspect of the fracture surface associated with the long crack growth regime, in which no extra large persistent slip bands nodules were observed

Full size image

5.2.4 The AE44 alloy

Figure 24 reveals that the fatigue crack nucleated at the vicinity of a surface shrinkage pore. Afterward, the fatigue crack totally propagated in a transdendritic mode over the entire fracture surface. The persistent slip/twin bands exhibited a thickness that spanned a range from 100 nm to a few micrometers. The main fatigue crack cut straight through the persistent slip/twin bands and left visible striations at the beginning of propagation (Figure 25), but no striations were detected after approximately 200 μm of crack extension. This is probably due to the fatigue crack propagation rate becoming greater than the size of a typical persistent slip/twin band. As with the AZ91 alloy, the persistent slip/twin bands were confined within areas in the range of the dendrite cell size, and no extra large shearing band nodules similar to those observed in the AM50 and AM60 alloys were observed. For the AZ91 alloy, we have suggested that the lamellar β phase localized at the dendrite cell boundaries acts as a barrier to dislocation migration across the dendrite cells. Similar mechanisms seemed to have operated in the AE44 alloy, in view of the interdendritic eutectic lamellar phase Al11RE4. This phase may have played a major role in blocking dislocation movement. One important result of our observations is that the fatigue crack advanced in a very soft material with higher rates than those observed for the AM50, AM60, and AZ91 alloys. Indeed, the relatively low percentage of Al in this alloy, combined with the higher tendency of rare earth elements to attract Al only in the interdendritic regions, resulted in poorer Al solid solution Mg dendrite cells than those in the three other studied alloys (Figure 26).

Fig. 24
figure 24

FEG-SEM fracture surfaces of sample E1 (N f  = 6034) showing (a) a higher magnification of the region delineated in Fig. 16(a) and exhibiting total coverage of the fracture surface by persistent slip/twin band nodules that are the same size as the dendrite cells and (b) a higher-magnification view of the region delineated in Fig. 16(a) showing that the thick shear/twin bands were activated in the same direction as that of the Al11RE4 platelet axis

Full size image
Fig. 25
figure 25

FEG-SEM fracture surfaces of sample E1 (N f  = 6034) showing (a) an overall view close to the nucleation region and (b) a higher magnification of the region delineated in (a) displaying fine striations corresponding to the initial stages of fatigue crack propagation

Full size image
Fig. 26
figure 26

Quantitative Al concentration profile for all investigated alloys obtained using EPMA. The distance between two adjacent analyzed points was 1 μm

Full size image

6 Summary

The effects of microstructure on fatigue crack growth mechanisms were studied for AM50, AM60, AZ91, and AE44 Mg alloys. The proposed fatigue mechanisms are summarized in Table II. We distinguish two main conclusions: four main stages observed for each alloy and the structure-property relations for fatigue crack incubation and small crack growth.

Table II Qualitative Summary of Microstructural Effects on Fatigue Crack Growth over the Four Distinct Regimes
Full size table

6.1 Multistage Fatigue Regimes

The observed fatigue crack propagation mechanisms in the AM50 and AM60 Mg alloys showed four fatigue life stages:

  1. 1.

    an incubation stage;

  2. 2.

    a microstructurally small crack stage, in which the crack advanced at the Al-rich eutectic/dendrite cell interface;

  3. 3.

    a physically small crack stage, in which the crack advances within the eutectic; and

  4. 4.

    a long crack regime, in which the crack propagation was transdendritic or interdendritic.

In the AZ91 Mg alloy, the coalescence of several cracks was observed because of the high porosity levels. This coalescence was promoted by the fracture of the β-phase. The β-phase was present in the form of small particles within the cell matrix, but formed continuous strings at the cell boundaries. We observed, however, that the small cracks propagated around the β-phase particles, and this regime may predominate when less severe porosity is present. In the AE44 Mg alloy, the propagation was completely transdendritic. This transdendritic propagation arose because a relatively soft matrix and a very thin Al11RE4 lamellar phase offered little resistance to even small crack growth.

6.2 Effect of Structure-Property Relationships on Fatigue Crack Growth

The lowest fatigue resistance was recorded for the AE44 alloy; the next lowest was recorded for the AZ91 alloy. The AM50 alloy showed the best fatigue durability, followed by the AM60 alloy.

We first compared the fatigue lives of the AE44 and AZ91 alloys, because these two alloys showed large pore size and content and smaller dendrite cell size, as compared to the AM50 and AM60 alloys. Although the AE44 alloy had a smaller pore size and content, it showed a lower fatigue resistance than did the AZ91 alloy, probably because of the lower content of Al in the AE44 matrix, augmented by the effect of rare earth elements, which combined with Al in the eutectic regions. These two mechanisms contributed to the reduction of the solid solution strengthening effect of Al.

The pore size and content in the AE44 and AZ91 alloys were greater than those in the AM50 and AM60 alloys. The effect of porosity on fatigue life was preponderant, because it counterbalanced the beneficial high Al content, especially in the AZ91 alloy. The effect of large pores is primarily on the incubation life, because of their higher stress concentration. The other noticeable effect of large pore size is the substantial reduction of the MSC regime life, because a larger pore size will induce an initial crack of greater length. Similarly, higher pore content is detrimental to fatigue resistance, because it promotes easier paths for crack propagation in the LC regime.

However, the high Al content in the AZ91 alloy was revealed to promote precipitation of continuous strings of the brittle β phase and to eliminate the Al-rich eutectic layer. The Al-rich eutectic layer was observed to be beneficial during the small crack regime, for the AM50 and AM60 alloys.

The lower Al content in the AM60 alloy compared to that in the AZ91 alloy gave rise to an Al-rich solid solution layer in the dendrite cell boundaries with less uniformly dispersed precipitation of the β phase. The less uniformly dispersed precipitation of the β phase reduced the pinning effect of the dendrite cell boundaries that operated during the primary recrystallization following the casting process. As a result, the grains observed in the AM60 alloy were coarser than those in the AZ91 alloy. The coarser grains reduced the propensity of the interdendritic shrinkage pores to coalesce. These effects contributed to the increased fatigue resistance of the AM60 alloy as compared to AZ91 and AE44 alloys. The effect of the β-phase particles at the PSC stage of the fatigue crack propagation can be both beneficial and detrimental. The detrimental effect was correlated with the brittle nature of this phase, which fractured ahead of the crack tip and provided weak paths for the crack advance. The beneficial effect occurred because the fractured particles attracted and forced the fatigue crack to remain in the eutectic layer and meander around the particles.

The beneficial effects of the β-phase particles were maximized for the AM50 alloy in that the Al-rich solid solution in Mg was thicker and contained fewer particles. These characteristics offered a relatively better fatigue performance for the AM50 alloy than for the AM60, AZ91, and AE44 alloys.