Inﬂuence of plasma electrolytic oxidation on fatigue performance of AZ61 magnesium alloy q

The effect of plasma electrolytic oxidation (PEO) on the fatigue of AZ61 magnesium alloy has been inves- tigated under high cycle conditions in air and 3.5% NaCl solution. PEO employed an alkaline pyrophos-phate–silicate–ﬂuoride electrolyte, with an offset square waveform, a frequency of 50 Hz, and current density of 130 mA cm (cid:2) 2 . The PEO treatment led to reductions in the fatigue limit by 38% in air and by 56% in 3.5% NaCl solution. The reduction of the fatigue limit is attributed to cracking of the coating, roughness of the alloy/coating interface, corrosion and inﬂuences of Al–Mn particles. Authors. Elsevier The present investigates the effect of PEO on the fatigue behaviour of magnesium alloy in and in 3.5% aqueous solution. An alkaline


Introduction
Magnesium alloys are attractive for many applications because of their light weight compared with materials such as steels and aluminium alloys [1][2][3]. The alloys exhibit high specific strength, good castability and excellent shock resistance, and are also recyclable [4], but the corrosion resistance may be inadequate for some applications [1,5,6]. The corrosion resistance can be improved by reducing the concentrations of alloy impurities, especially iron, nickel and copper [7][8][9]. Protective treatments can also be applied, such as conversion coating, electroplating/electroless plating and anodizing [10]. Electrochemical processes are able to produce relatively thick, protective coatings [11], with much recent attention being given to plasma electrolytic oxidation (PEO) [12,13]. Ceramic coatings produced by PEO can provide high wear resistance and corrosion protection, the latter especially when the coating forms part of a paint scheme [13,14]. They are usually formed in aqueous electrolytes at high voltages, with the coating material being produced at locations of short-lived discharges [15]. Alkaline silicate or phosphate electrolytes, with various additives, such as borate [16] or fluoride [17], are often used. Fluoride ions may promote the formation of a protective barrier film [11,18], which can provide resistance to pitting corrosion. They may also promote sparking [11] and increase the wear resistance of the coatings [19].
For some applications, retention of fatigue strength after surface treatment is critical. Studies of fatigue of electrochemically treated magnesium alloys have yielded variable results. It has been reported that the Anomag process caused no change in the fatigue properties of AZ91 alloy [20]. In contrast, 7 and 15 lm thick coatings produced by the Keronite process caused 3% and 10% reductions respectively in the fatigue limit of a Mg-2%Al-1%Zn-0.2%Mn alloy [21]. Khan et al. reported a 30% reduction due to an unspecified anodic layer on AM60 alloy [22]. The corrosion fatigue of magnesium alloys has not been extensively studied. Further, the many influencing factors, such as alloy composition, production route, corrosion environment and fatigue conditions, make it difficult to draw general conclusions [23][24][25][26]. Unigovski et al. showed that extruded alloys were more sensitive to the action of 3.5% NaCl solution in comparison with die-cast alloys [24]. This was attributed to the high plastic deformation of the former alloys, which led to strain hardening, an increased chemical potential of the metal and mechanochemical dissolution [27]. Nevertheless, the fatigue life of extruded alloys was significantly longer than that of die cast alloys [24]. Khan et al. reported little difference in the fatigue life of bare AM60 alloy and with a 1 lm thick anodic coating when tested under humid conditions; a 15 lm thick coating resulted in a poorer performance, which was ascribed to defects in the coating.
The present study investigates the effect of PEO on the fatigue behaviour of extruded AZ61 magnesium alloy in air of 33% relative humidity and in 3.5% NaCl aqueous solution. An alkaline silicate-pyrophosphate-fluoride electrolyte was chosen based on previous work [28]. In accordance with the findings of Khan et al. [22,23] on the influence of coating thickness on fatigue, a relatively thin coating, of $5 lm thickness, was chosen for the fatigue examination.

Material and PEO treatment
AZ61 magnesium alloy was supplied by Magnesium Elektron Ltd. as extruded bars of 20 mm diameter. The chemical composition of the alloy is given in Table 1. Three types of specimen were employed for PEO: (i) flat specimens, cut transverse to the extrusion direction, of 20 mm diameter and 2.5 mm thickness for examination of the coating; (ii) fatigue specimens with a gauge section of length 15 mm and diameter 7 mm (see Fig. 1(a)); (iii) cylindrical specimens of length 50 mm and diameter 7 mm for assessment of the uniformity of coatings on a curved surface similar to that of the fatigue specimens. Specimens (ii) and (iiii) were machined with their longitudinal axes in the extrusion direction. The flat surfaces of specimens (i) were ground to a 1200 grit finish using SiC paper and cleaned in acetone in an ultrasonic bath. The specimens were then gripped by steel clips to provide electrical connection and covered with lacquer 45 (MacDermid plc.), leaving a working area of $7cm 2 [28]. Specimens (ii) and (iii) were mechanically polished to a 4000 grit finish using SiC papers. The cylindrical specimens were then gripped at one end by a steel clip and lacquer was applied to cover the clip and the end of the specimen, leaving an exposed length of 50 mm, with a working area of $11 cm 2 . The fatigue specimens were similarly gripped and lacquered leaving the gauge length and shoulders exposed, with a working area of $11 cm 2 .
PEO was carried out in 1 dm À3 of stirred electrolyte in a doublewalled glass cell, with water cooling that maintained the electrolyte temperature in the range 293-303 K. The counter electrode was a cylinder of type 304 stainless steel, with a length of 150 mm and a diameter of 90 mm. The specimens were placed at the centre of the counter electrode. The coatings were formed at a constant root mean square (rms) current density of 130 mA cm À2 and a frequency 50 Hz, using a square waveform with a positive to negative current ratio of À1.2. The electrolyte, which contained 10 g dm À3 sodium pyrophosphate (Na 4 P 2 O 7 Á10H 2 O), 11 g dm À3 sodium silicate solution (specific gravity 1.5), 2.5 g dm À3 potassium hydroxide (KOH) and 8 g dm À3 potassium fluoride (KF), was prepared using deionized water and analytical grade chemicals. The duration of the PEO treatment was 300 s. Following PEO, the specimens were rinsed with deionized water and dried in warm air. Fig. 1(b) shows the appearance of a fatigue specimen after PEO, with an off-white coating of uniform appearance on the gauge length.

Fatigue testing and specimen examination
The Younǵ s modulus (E), yield strength (YS), ultimate tensile strength (UTS), elongation and reduction of area of the AZ61 alloy were determined using a TiraTest 2300 tensile test machine. The average values from three tensile tests are given in Table 2. High-cycle fatigue tests were carried out on the bare and PEO-coated alloy using a PC controlled Amsler Vibrophore 422 machine, with a sinusoidal cycle of frequency 90 Hz, at room temperature in air of 33 ± 5% relative humidity and in 3.5% NaCl solution of pH 6.5. The ratio of the minimum stress to the maximum stress was 0. The solution was replenished during longer-term tests to maintain the pH within the range 6.5-7.2. For corrosion fatigue tests, a plastic cell containing naturally-aerated 0.5 dm À3 of 3.5% NaCl solution was attached to the specimens, which enabled the gauge length to be fully immersed in the solution. The fracture surfaces and gauge lengths of the specimens were examined using scanning electron microscopy (SEM), employing a Zeiss EVO50 instrument equipped with energy-dispersive X-ray (EDX) analysis facilities.
The adhesion/cohesion of the coating was evaluated by the scratch test method, using a Revetest system (CSM Instruments SA, Switzerland) equipped with a H-270 diamond indentor (200 lm diameter). Six scratch indentations were carried out under previously optimized conditions (linear progressive load mode 1-4 N, 4 N min À1 ). In order to aid in determination of location of spallation/delamination, an extended scratch length of 6 mm was employed. The scratch tracks were subsequently observed by SEM to determine the locations of the first coating failure and to understand the nature of the coating failure. During the scratch tests, the loading force and penetration depth were recorded and their respective values were correlated with the observed failure locations. The surface roughness of the coating was evaluated using a surface roughness tester (TR200, Timegroup Inc.) according to ISO standard [29]. Due to the presence of the open porosity in the outer layer of the coating, a measurement length for determination of the roughness (R a ) of 0.8 mm was used. In total, eight measurements were carried out in different directions.

PEO treatment
The surface of the coating formed following PEO treatment of a flat specimen for 300 s at 130 mA cm À2 is shown in the scanning electron micrograph of Fig. 2(a), which reveals a relatively uniform distribution of pores of size up to $5 lm and a network of fine cracks. Fig. 2(b) shows a cross-section of the specimen. The thickness of the PEO coating is 5.4 ± 0.7 lm. The alloy surface that was protected by lacquer during PEO can be seen at the left-hand side     pores of $5 nm diameter that are not resolved by conventional SEM [30]. A surface-connected porosity of 20% was estimated, with negligible occluded porosity. Fig. 4 shows the voltage-time response and the appearance of a cylindrical specimen during PEO. The voltage rose to $190 V in 25 s, when small, white discharges appeared. It then increased slowly to $200 V after 60 s, with large, white discharges becoming evident, first at the specimen edges and then elsewhere. Scanning electron micrographs of cross-sections of the cylindrical specimen at distances of 10, 25 and 40 mm from the bottom end are shown in Fig. 5. There were no significant differences in the coatings at the three locations, which were of thicknesses 5.2 ± 0.7, 5.8 ± 0.9 and 6.4 ± 0.8 lm at the respective positions. The thicknesses are similar to that of the flat specimen, namely 5.4 ± 0.7 lm, which produced a similar voltage-time response to that of Fig. 4.

Adhesion/cohesion of the coating
Scratch testing indicated that the coating failure mode was of a cohesive nature and that the initial micro-cracking developed at the interface between the outer and inner layers of the coating. Spallation and subsequent delamination of the outer layer was recorded under a load of 2.69 ± 0.10 N, which corresponds to a track length of approximately 3.4 ± 0.2 mm from the scratch start point. The lower layer of the coating did not detach from the substrate, as shown in Fig. 6. Instead, the layer was pushed deeper into the magnesium substrate, which was confirmed by SEM observations; no substrate could be observed in the track even at penetration depths of 31.5 ± 9.7 lm, which was the average of the depths recorded at the end of the scratch tracks. The failure mode observed is in agreement with the scratch testing model of a hard coating-soft substrate given by Bull and Berasetegui [31] and indicates that the cohesive strength of the coating layers is less than the adhesive strength between the coating and the substrate.

Fatigue tests
The results of fatigue tests are shown in Fig. 7. The tests that reached 10 7 cycles, marked with arrows, were terminated before fracture of the specimens. The data were fitted by minimizing the sum of squares of the deviations with two regression functions [32]: Kohout and V echet : where a, b, C and r 1 are parameters of the regression function, N f is the number of cycles and r is the maximum applied stress measured in MPa. The K and V function gave a better fit to data above 10 5 cycles, as shown for the uncoated specimen tested in air, and only this fitting is shown for other test conditions. Fig. 7 reveals the highest fatigue strength for the bare alloy tested in air and the lowest fatigue strength for the coated alloy tested in 3.5% NaCl solution. The bare alloy exposed to 3.5% NaCl and the coated alloy exposed to air exhibited similar, intermediate fatigue behaviours. The fatigue strengths predicted at 10 7 cycles are listed in Table 3 together with the parameters of the K and V function. For aluminium or magnesium alloys, which have no well-defined fatigue limit, it is common to report the fatigue strength at 10 8 cycles. However, the present fitted curves suggested that the fatigue strengths at 10 7 and 10 8 cycles differ by less than 2%. The fatigue limit of the bare alloy in air and 3.5% NaCl solution was 145.4 and 88.0 MPa respectively, indicating a reduction of 40% due to the NaCl solution. The corresponding values for the PEO-coated alloy were 91.5 and 63.8 MPa, indicating reductions of $40% and 56% in comparison with the bare alloy in air. Bhuiyan et al. [25] found an 85% reduction of the fatigue limit of bare AZ61 alloy in 5% NaCl solution, which exceeds the present reduction of 40%. However, Yerokhin et al., for a different magnesium alloy and PEO process, reported reductions of up to 10% for tests in air, compared with 40% for the present coating [21].

Fractographic analysis
Three specimens, from the start, middle and end of the each S-N curve respectively, were chosen for fractographic analysis. The fracture surfaces of the specimens are shown in the low magnification scanning electron micrographs of Fig. 8, which reveal an increase in the region of fatigue with decrease of the stress amplitude. Coated specimens tested in air at the two highest stresses had a greater number of crack initiation sits than the bare alloy. Several initiation sites also occurred on the bare and coated alloy in 3.5% NaCl solution at the highest stress amplitude. At relatively low stress amplitudes, only one initiation site was evident on both bare and coated specimens tested in either air or NaCl solution.
Crack initiation in the bare alloy tested in air often occurred near Al-Mn particles, which were typically of size between 1 and 15 lm. The backscattered electron micrographs of Fig. 9(a and b) shows examples for a test in air under stress amplitudes of 200 and 300 MPa, in which the particles are evident on the main fatigue crack ( Fig. 9(a)) and at secondary cracks on the gauge length ( Fig. 9(b)). The crack initiation is probably related to the stress enhancement that results from the large difference in hardness between the particles and the matrix [33][34][35]. The hardness of the particles has been determined previously as 948 ± 61 HV, compared with 64 ± 2 HV 0,1 for the matrix [28].
In the presence of the 3.5% NaCl solution, fatigue cracks initiated at regions of localized corrosion of the alloy, as shown by the example of Fig. 9(c and d) for a test at a stress amplitude of 120 MPa. EDX analysis at these locations revealed 60 at.% O, 26 at.% Mg, 7 at.% Al, 1 at.% Si and 6 at.% Cl, with the presence of oxygen and chlorine being associated with corrosion products. Other studies [23,24,26,36] have shown also that corrosion sites are favourable for fatigue crack initiation. Localized corrosion is promoted by corrosion cells formed by second phase particles and the matrix, which cause dissolution of the matrix adjacent to the particles [8]. Further, Bhuiyan et al. showed that cyclic loading of AZ80 alloy can enhance localized corrosion compared with the non-stressed condition [37]. The number of secondary cracks on the gauge length reduced with reducing stress amplitude for tests of the bare alloy both in air and in NaCl solution. At high stress amplitude, relatively large secondary cracks occurred over most of the gauge length whereas, under a low stress, such cracks were finer and confined mainly to the vicinity of the fracture surface.
For tests both in air and 3.5% NaCl solution, the fatigue crack surface had a coarse lamellar appearance near the initiation site, with secondary cracks that propagated along the lamellae, as revealed in Fig. 10(a) for a test in air at a stress amplitude of 160 MPa. Paired features, indicated by arrows in Fig. 10(b), on the fracture surface generated in NaCl solution, are possibly indications of twin boundaries [36]. The fracture surface close to the final fracture was of smoother appearance ( Fig. 10(c)), as observed in other work [36]. Secondary cracks and striations, the latter indicated by arrows, were observed in this area ( Fig. 10(d)).
Cross-sections of specimens were examined for tests carried out at a stress amplitude of 160 MPa in air and in 3.5% NaCl solution. Fig. 11(a) shows twins adjacent to the fatigue crack generated in 3.5% NaCl solution. The twins were particularly apparent from a depth of about 30% of the specimen diameter and were most abundant near the end of fatigue crack. Similar twinning was observed in the specimen tested in air. The occurrence of twinning in the region of fatigue crack propagation agrees with previous work [38,39]. Other studies have considered in detail the occurrence of twinning during fatigue testing. Zeng et al. found that crack initiation and propagation in extruded AM60 is related to the synergistic influences of slip bands, double twinning, intermetallic compounds and grain boundaries [35]. Shiozawa et al. observed crack growth along twin boundaries in an extruded AZ80Mg alloy, with the direction of propagation changing at grain boundaries [40]. Zhang et al. examined the formation of twins ahead of the crack tip during cyclic deformation of an extruded AZ61A alloy under different loading modes, finding the highest volume fraction of the residual twins under tension-compression, with almost every grain containing twins on reaching a strain amplitude of 1% [41].
The cross-section of Fig. 11(a) reveals that the alloy had a duplex grain size, with a non-uniform distribution of large grains, up to $40 lm size, and fine grains, $1 lm in size. This hindered     identification of the crack path, which appeared to be mainly transgranular, although the presence of intergranular regions could not be ruled out. The cross-section displays secondary cracks and decohesion of Al-Mn particles, as well as Mg 17 Al 12 at grain boundaries. The final fracture surfaces in all test conditions revealed a dimpled appearance, characteristic of a ductile failure. Al-Mn particles were observed inside the dimples. A representative example of the final fracture is shown for a coated specimen tested in air at a stress amplitude of 300 MPa (Fig. 11(b)).
In the case of tests of the coated alloy, spalling of the outer layer of the coating was observed. The spalling was confined to the region of the gauge length near the main fracture under low or medium stress amplitudes. In contrast, under a high stress amplitude, spalling of the outer layer affected most of the gauge length. An example of the spalling is provided by the photograph of a specimen tested at 180 MPa shown in the inset of Fig. 12(a), which reveals a slightly darker, adherent inner layer where the outer coating layer has detached. Fig. 12(a) also shows a micrograph of the boundary of the spalled region, which discloses narrow cracks, transverse to the loading axis, in the outer coating layer and occasional fine cracks in the inner layer. Fig. 12(b) shows a relatively rough fracture surface near a site of fatigue crack initiation, with a transition to a region of a smoother appearance, in a coated specimen tested in air at a low stress amplitude of 100 MPa, similar to observations for the bare alloy. Fatigue cracks on a specimen tested in NaCl solution at a low stress amplitude of 70 MPa appeared to initiate at a few closely-spaced sites, where beach-mark-like features were observed, as shown in Fig. 13(a). At low stress amplitudes, secondary cracking of the alloy on the gauge length was negligible, although a network of fine cracks in the coating was evident. In contrast, secondary cracks were relatively abundant at higher stress levels. Fig. 13(b) shows a secondary crack near an Al-Mn particle on the gauge length of a coated specimen tested in NaCl solution at a relatively high stress amplitude of 300 MPa. Under low stress amplitudes, a layer of corrosion product formed beneath the coating; Fig. 14 shows a specimen tested at 60 MPa for a time of $45 h, which displays a $2 lm thick layer.
The corrosive environment can reach the substrate through the open pores and cracks formed due to either the PEO process or the cyclic loading. Liang et al. [42] has previously suggested that the formation of magnesium hydroxide at the alloy/coating interface exerts a stress that can lift the coating from the substrate.
The degradation of fatigue properties of the AZ61 alloy after PEO treatment is probably due to a combination of factors, including stress concentration, that results from cracks in the coating layer [22], the roughness of the alloy/coating interface [22], and  corrosion beneath the coating, which enhance the stress locally in the alloy surface and assist crack initiation, leading to the higher number of initiation sites observed on the coated specimens compared with the bare alloy. Bhuiyan et al. [43] also considered that cracks developed during loading affect the residual stresses in the coating, which were assumed to be compressive, as measured in PEO coatings on aluminium alloys [44][45][46]. Lonyuk et al. indicated that compressive residual stresses in the coating retard fatigue crack initiation and therefore increase the fatigue life [46]. However, if the compressive residual stress is high, coating delamination and cracking may occur. They also found that cracks initiated in the substrate adjacent to the coating, mainly at locations where the coating was thicker.

Conclusions
1. PEO treatment of AZ61 magnesium alloy in the silicatepyrophosphate-fluoride electrolyte, forming a $5-6 lm thick coating, resulted in a reduction of the high-cycle fatigue limit in air from 145.4 MPa to 91.5 MPa, and a further reduction to 63.8 MPa in 3.5% NaCl solution. Experimental fatigue data could be best fitted by the Kohout and Věchet regression. 2. Fatigue cracks in the uncoated alloy tested in air initiated mainly at Al-Mn particles. The presence of sodium chloride solution caused corrosion of the alloy matrix around Al-Mn particles and the formation of corrosion pits, leading to stress concentration and fatigue crack initiation. 3. PEO enhanced stresses in the alloy surface, due to the roughness of the alloy/coating interface and stress transfer following cracking of the coating, which resulted in multiple crack initiation sites. Cyclic loading caused the outer coating layer to spall over an area of the gauge length that increased with the stress amplitude. In the presence of 3.5% NaCl solution, cracks and pores in the coating allowed the solution to reach the alloy, with corrosion then spreading under the coating leading to further degradation of the fatigue performance. 4. The fatigue crack propagation area had similar characteristic in all tested conditions, i.e. a coarse appearance with secondary cracks in the area close to initiation and a finer appearance in the area close to final fracture. Secondary cracks also occurred in the region of fatigue crack propagation. Dimples, containing Al-Mn particles, characterized the overload fracture.